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 |
|---|---|---|---|---|---|
Microsatellite expansions are the leading cause of numerous neurodegenerative disorders . Here we demonstrate that GGGGCC and CAG microsatellite repeat RNAs associated with C9orf72 in amyotrophic lateral sclerosis/frontotemporal dementia and with polyglutamine diseases , respectively , localize to neuritic granules that undergo active transport into distal neuritic segments . In cultured mammalian spinal cord neurons , the presence of neuritic GGGGCC repeat RNA correlates with neuronal branching defects , and the repeat RNA localizes to granules that label with fragile X mental retardation protein ( FMRP ) , a transport granule component . Using a Drosophila GGGGCC expansion disease model , we characterize dendritic branching defects that are modulated by FMRP and Orb2 . The human orthologs of these modifiers are misregulated in induced pluripotent stem cell-differentiated neurons ( iPSNs ) from GGGGCC expansion carriers . These data suggest that expanded repeat RNAs interact with the messenger RNA transport and translation machinery , causing transport granule dysfunction . This could be a novel mechanism contributing to the neuronal defects associated with C9orf72 and other microsatellite expansion diseases .
Expansions of short tandem nucleotide repeat sequences termed 'microsatellite repeats' cause various devastating dominantly inherited neurodegenerative disorders , including spinocerebellar ataxias , Huntington’s disease , and the myotonic muscular dystrophies ( expansion of CAG , and CUG and CCUG repeats , respectively ( Orr and Zoghbi , 2007 ) ) . Most recently , the GGGGCC repeat expansion in the C9orf72 gene has been shown to be associated with amyotrophic lateral sclerosis/frontotemporal dementia ( ALS/FTD ) ( DeJesus-Hernandez et al . , 2011; Renton et al . , 2011 ) . How microsatellite repeat expansions occurring both within coding and non-coding segments of the affected genes cause neuronal degeneration remains a central question in the field . Microsatellite repeat RNAs are thought to induce neurodegeneration through multiple distinct mechanisms ( Narayan et al . , 2014; Nelson et al . , 2013 ) . These include both loss and gain of function in the encoded protein ( Blum et al . , 2013 ) ; however , a number of disease-associated expanded microsatellite repeats , like ( GGGGCC ) n , occur in non-coding sequence , suggesting that the RNA product may be toxic ( Belzil et al . , 2012 ) . Nuclear toxicity has been proposed to be a disease mechanism mediated either by expanded repeat RNA present in nuclear foci , or by expanded repeat RNA-encoded repeat associated non-ATG ( RAN ) translated peptides ( Haeusler et al . , 2014; Kwon et al . , 2014; Zu et al . , 2011; Jovicic et al . , 2015; Zhang et al . , 2015; Freibaum et al . , 2015 ) . However , the RNAs generated from these loci commonly have high structural context ( Napierala and Krzyzosiak , 1997; Sobczak et al . , 2003; Michlewski and Krzyzosiak , 2004; Fratta et al . , 2012; Reddy et al . , 2013 ) , which is a striking feature of cis-acting localization signals that target messenger RNAs ( mRNAs ) to specific subcellular sites where they can then undergo local translation ( Hamilton and Davis , 2007; Martin and Ephrussi , 2009; Holt and Schuman , 2013 ) . Therefore , we hypothesized that such disease-associated RNAs might interact with the mRNA localization and/or translation machinery with deleterious consequences . Here we show that expanded microsatellite repeat RNAs , including the GGGGCC repeat RNA associated with ALS/FTD , become localized to granules in neurites of mammalian neurons in culture . Such neuritic GGGGCC RNA-positive granules are also present in iPSNs from GGGGCC expansion carriers . This subcellular localization is shared among many expanded repeat RNAs associated with human disease that bear high structural content , including CAG , CUG , and CCUG repeat RNAs . We further show by detailed analysis that at least two of these RNAs—GGGGCC and CAG—become localized to dynamic RNA-granules in neurites . Detailed focus on the GGGGCC repeat RNA revealed neuritic branching defects and suggests the expanded microsatellite repeat RNA may interfere with transport granule function . These data indicate that this property may contribute to the degenerative effects conferred by expanded GGGGCC RNA and additional expanded microsatellite repeat RNAs associated with a wide class of human neurological disorders .
To explore the idea that expanded repeat RNAs may be localized in neurons , we initially focused on the expanded GGGGCC repeat associated with ALS/FTD ( DeJesus-Hernandez et al . , 2011; Renton et al . , 2011 ) . The GGGGCC RNA repeat is highly structured , assuming both G-quadruplex and stem-loop conformations ( Fratta et al . , 2012; Reddy et al . , 2013 ) . We analyzed its localization in iPSNs derived from two C9orf72 hexanucleotide expansion carriers ( carrier 1 , line #5; carrier 2 , line #11 ( Almeida et al . , 2013 ) ) . These neurons contain RNase sensitive nuclear GGGGCC foci specifically in carrier samples , and not in control-derived samples ( Almeida et al . , 2013 ) . We confirmed that iPSNs contained nuclear GGGGCC RNA foci , but also found that 78 ± 12% SD ( carrier 1; n=25 neurons ) and 75 ± 11% SD ( carrier 2; n=23 ) of iPSNs that contained nuclear GGGGCC RNA foci also contained neuritic GGGGCC RNA particles by in situ hybridization ( Figure 1A , L ) . The GGGGCC RNA particles were detected both proximally and distally at over 45 μm from the cell body in neurites and were , in some cases , lined up , consistent with possible association with a cytoskeletal track ( Figure 1A–B ) . In addition , GGGGCC repeat RNA particles were detected in the cell body in nearly all iPSNs that also contained GGGGCC RNA nuclear foci ( Figure 1C , also Almeida et al . , 2013 ) . We did not detect GGGGCC RNA in control iPSNs , indicating that the non-expanded repeat is either present below the detection level or not stably expressed in wild-type iPSNs . These data thus suggested that endogenous expanded GGGGCC microsatellite repeat RNA was localized to particles in neurites , in addition to localization elsewhere in the cell . 10 . 7554/eLife . 08881 . 003Figure 1 . The GGGGCC repeat and other microsatellite RNA repeats with high secondary structure content are neuritically localized . ( A–B ) GGGGCC repeat RNA is neuritically localized in discrete granules in human iPSNs derived from C9orf72 GGGGCC repeat expansion carriers . ( A–C ) GGGGCC repeat RNA ( red ) was detected with a ( GGCCCC ) 4 antisense probe . Neuritic GGGGCC RNA granules ( arrowheads ) were observed ( A ) proximally , ( inset in A ) as linear arrays , and ( B ) distally . ( C ) GGGGCC RNA granules in the cell body . ( A–C ) iPSNs from carrier 2 are shown . Neurites and cell body are outlined ( dotted line ) . ( D–E ) CAG repeat RNA is localized to neuritic granules in primary rat spinal cord culture . In situ hybridization of primary rat spinal cord neurons transfected with ( CAG ) 100 RNA . The ( CAG ) 100 RNA construct was not MS2-tagged ( see Figure 1—figure supplement 2A and Materials and methods ) . Neuritic RNA granules ( white; arrowheads ) were detected with a ( D ) ( CUG ) 8 antisense but not with a ( E ) ( CAG ) 8 sense probe . Distributions shown are representative of two biological replicates . ( F–K ) Primary rat spinal cord neurons transfected with NLS-CP-GFP ( green; arrowheads ) and ( F ) LacZ-MS2 , ( G ) ( GAA ) 100-MS2 , ( H ) ( CAG ) 100-MS2 , ( I ) ( GGGGCC ) 48-MS2 , ( J ) ( CUG ) 100-MS2 , or ( K ) ( CCUG ) 100-MS2 ( see Figure 1—figure supplement 2A and Materials and methods for construct details ) . Whereas ( F ) the control RNA LacZ-MS2 , or ( G ) an expanded repeat RNA without secondary structure , ( GAA ) 100-MS2 , did not show GFP accumulations , ( H–K ) the other expanded repeat RNAs conferred punctate GFP staining indicative of RNA enrichment in neuritic RNA granules . See Figure 1—figure supplement 1E for the pattern of expression of the MS2 alone control , which lacked neuritic puncta . DsRed ( magenta ) was coexpressed to outline the neurons . ( L ) Quantitation of iPSNs with neuritic GGGGCC repeat RNA granules . Nuclear foci positive ( + ) neurons were defined as having ≥5 ( carrier 2 ) or ≥1 ( carrier 1 ) nuclear GGGGCC RNA foci . ANOVA p-value = 0 . 00033 . ( M ) Primary rat spinal cord neurons were transfected as in ( F–K ) and the percentage of neurons with neuritic repeat RNA particles was determined . At right , neurons within the mixed culture with the large morphology characteristic of motor neurons were scored for neuritic particles . ANOVA p-value = 1 . 1×10–7 . ( L–M ) Numbers indicate the total number of neurons scored from a minimum of ( L ) two or ( M ) three biological replicates . Averages ± standard deviation are given . See Materials and methods for statistical analysis . Post-hoc: ****p < 2 . 5×10–6 , ***p < 0 . 00042 , **p < 0 . 0085 , *p < 0 . 023 . ( D ) A single confocal section or ( A–C , E–K ) Z-series projections . Bars: 10 μm . DAPI: blue . See also Figure 1—figure supplement 1–2 . ANOVA , analysis of variance; DAPI , 4' , 6-diamidino-2-phenylindole; CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; NLS , nuclear localization signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 00310 . 7554/eLife . 08881 . 004Figure 1—figure supplement 1 . Neuritic localization of ( CAG ) 100 RNA by in situ hybridization . ( A–B ) In situ hybridization of primary rat spinal cord neurons expressing ( CAG ) 100 RNA . The ( CAG ) 100 construct was not MS2-tagged , but contained a leader sequence and translation reporter tags ( see Figure 1—figure supplement 2A ) . ( A ) ( CUG ) 8 antisense or ( B ) ( CAG ) 8 sense probes were used for RNA detection ( white; arrowheads ) . ( B ) Magenta indicates a transfected cell . ( C–D ) Detection of nuclear CAG repeat RNA foci . Primary rat spinal cord neurons expressing ( C ) CP-GFP fused to a NES ( NES-CP-GFP; green ) , or ( D ) NES-CP-GFP and ( CAG ) 100-MS2 RNA . ( E-F ) Primary rat spinal cord neurons were transfected with NLS-CP-GFP ( green ) and ( E ) MS2 or ( F ) LacZ-MS2 . ( E ) MS2 RNA ( green ) was not enriched in cellular processes ( 0/52 neurons scored had neuritic MS2 RNA granules ) . ( F ) LacZ-MS2 was not enriched in cellular processes ( 0/65 neurons scored had neuritic LacZ-MS2 RNA granules ) , see quantitation in Figure 1M , and high magnification of neurite in Figure 1F . ( G–K ) Primary rat spinal cord neurons were transfected with NLS-CP-GFP ( green;arrowheads ) and ( G ) ( GAA ) 100-MS2 , ( H ) ( CAG ) 100-MS2 , ( I ) ( GGGGCC ) 48-MS2 , ( J ) ( CUG ) 100-MS2 , or ( K ) ( CCUG ) 100-MS2 . Whereas an expanded repeat RNA , ( GAA ) 100-MS2 , without secondary structure was not enriched , the other expanded repeat RNAs showed punctate GFP staining , indicative of granules . The MS2 , ( GAA ) 100-MS2 , ( CAG ) 100-MS2 , ( GGGGCC ) 48-MS2 , ( CUG ) 100-MS2 , and ( CCUG ) 100-MS2 constructs all contained a leader sequence , a translation reporter tag , and an MS2 tag ( see Figure 1—figure supplement 2A and Materials and methods ) . Therefore , in addition to serving as a control for the MS2 tag and structured RNA , the MS2 and ( GAA ) 100-MS2 constructs also serve as controls for the leader and translation reporter tags . ( A ) A single confocal section , or ( B–K ) confocal Z-series projections . Distributions shown are representative of ( A–B ) two biological replicates , or ( C–K ) a minimum of three biological replicates . ( A–B , F–K , and top panels of C–E ) Bar: 15 μm . ( C–D , E ) Bottom panels show neuronal processes at high magnification . Bar: 10 μm . ( E–K ) DsRed ( magenta ) was coexpressed to outline the neurons . DAPI: blue . CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; DAPI , 4' , 6-diamidino-2-phenylindole; NES , nuclear export signal; NLS , nuclear localization signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 00410 . 7554/eLife . 08881 . 005Figure 1—figure supplement 2 . Microsatellite repeat and control expression constructs . ( A ) Constructs used to express control ( top three constructs ) and microsatellite repeat ( bottom nine constructs ) RNA in transfected primary rat spinal cord neurons are shown . With the exception of ( CAG ) 100 , which was cloned into pcDNA , all constructs , including ( CAG ) 100-MS2 , were cloned into the pGW expression vector . The three control constructs were MS2 , ( GAA ) 100-MS2 and LacZ-MS2 . ( Red box ) A leader sequence that includes 6 stop codons , 2 in each reading frame . ( Green box ) A repeat sequence . ( Blue box ) Three tags ( FLAG , HA , and Myc , one in each reading frame ) to detect RAN translation . ( Yellow box ) Twelve MS2 repeats ( the CP binding site to allow detection via CP-GFP binding ) . ( B ) Constructs used to express control and microsatellite repeat RNA in Drosophila da neurons . Control UAS construct ( top , UAS-DsRed ) and the experimental UAS construct ( bottom , UAS- ( GGGGCC ) 48 ) are diagrammed . The UAS- ( GGGGCC ) 48 construct has the same 5’ leader as the constructs described above in ( A ) . CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 005 To see if the finding of neuritic localization was a shared property of microsatellite repeats with high secondary structure , we examined a CAG repeat RNA . We expressed an RNA consisting of 100 repeats of the CAG trinucleotide , ( CAG ) 100 , in primary rat stage E14 mixed spinal cord neurons ( Mojsilovic-Petrovic et al . , 2006 ) and probed the RNA localization by in situ hybridization . Expanded CAG repeat RNA assumes stem-loop secondary structure ( Michlewski and Krzyzosiak , 2004 ) , confers neurodegeneration ( Li et al . , 2008 ) , and has been noted to assemble into nuclear foci ( Ho et al . , 2005; Li et al . , 2008; Wojciechowska and Krzyzosiak , 2011 ) . As with the GGGGCC microsatellite repeat , we detected the expanded CAG repeat RNA in discrete particles in neurites , and in the cell body ( Figure 1D–E , Figure 1—figure supplement 1A–B ) . We also noted nuclear foci as previously described ( Figure 1—figure supplement 1A–D ) . These data indicated that two distinct expanded microsatellite repeat RNAs—a GGGGCC repeat and a CAG repeat—both highly structured , are incorporated into particles in neurites . Localization of these RNAs to neurites has not been noted previously . To further assess whether this subcellular localization to RNA particles in neurites may be a property common to highly structured expanded microsatellite repeat RNAs , we utilized the bipartite MS2 system ( Bertrand et al . , 1998 ) to examine additional repeat RNAs , as well as a series of control RNAs . We tagged the RNAs with 12 MS2 stem-loops that are recognized by coat-binding protein , which is fused to a nuclear localization signal and to green fluorescent protein ( NLS-CP-GFP ) . When expressed alone , the NLS-CP-GFP signal was predominantly nuclear ( not shown ) . Co-expression of NLS-CP-GFP with either of two control RNAs , the MS2 ( Figure 1—figure supplement 1E ) or LacZ-MS2 ( Figure 1F , Figure 1—figure supplement 1F ) RNAs , produced results similar to NLS-CP-GFP alone: there was no enrichment of signal in cellular processes ( see Figure 1—figure supplement 2A for construct details ) . We then examined the localization of an expanded repeat RNA that does not assume stem-loop secondary structure , the GAA repeat associated with Friedreich’s ataxia ( Sobczak et al . , 2003 ) . ( GAA ) 100-MS2 did not alter the distribution of the GFP reporter , indicating this repeat RNA without secondary structure was not localized to neurites ( Figure 1G , Figure 1—figure supplement 1G ) . RNA particles were neuritic in <0 . 5% of neurons in cultures expressing NLS-CP-GFP with control LacZ-MS2 or ( GAA ) 100-MS2 RNA ( Figure 1M ) . In contrast , neurons transfected with NLS-CP-GFP and ( CAG ) 100-MS2 had an RNA distribution like that of ( CAG ) 100 by in situ hybridization ( compare Figure 1H with 1D , and Figure 1—figure supplement 1H with Figure 1—figure supplement 1A ) , with 94 . 4 ± 9 . 6% SD ( n=3 cultures , 87 neurons total ) of cotransfected neurons containing particles in neurites ( Figure 1M ) . Next , we examined ( GGGGCC ) 48-MS2 RNA and found it neuritically localized in 21 . 1 ± 3 . 7% SD ( n=4 cultures , 147 neurons total ) of all neuron types in the cultures ( Figures 1I , M , Figure 1—figure supplement 1I ) , and in 66 . 6 ± 33 . 3% SD ( n=3 cultures , 9 neurons total ) of large neurons with a morphology characteristic of motor neurons ( Figure 1M , at right; also Figure 5—figure supplement 1A ) . The RNA repeat expansions associated with myotonic dystrophy types I and II—CUG and CCUG , respectively—are also highly structured RNAs that assume stem-loop conformation ( Napierala and Krzyzosiak , 1997; Sobczak et al . , 2003 ) . Indeed , ( CUG ) 100-MS2 and ( CCUG ) 100-MS2 RNA particles were also present in neurites in over 75% of the transfected neurons ( Figure 1J-K , M , see also Figure 1—figure supplement 1J–K ) . Thus , in contrast to the control RNAs ( MS2 , LacZ-MS2 , and ( GAA ) 100-MS2 ) , multiple microsatellite RNA repeats ( CAG , GGGGCC , CUG and CCUG ) with high structural context became localized to RNA particles in neurites , by independent detection methods and in a variety of neural systems . Disease severity and age of onset in patients with trinucleotide repeat expansion disorders ( e . g . CAG and CUG ) correlates with increasing repeat number ( Orr and Zoghbi , 2007 ) . Therefore , we examined the dependence of particle formation on repeat number for the MS2-tagged CAG and GGGGCC RNAs in mixed rat spinal cord neurons , focusing on neurons that contained at least one neuritic RNA particle . The fraction of primary arbors that had particles containing RNAs of 20 , 40 , 70 , and 100 CAG repeats were 0 . 08 ± 0 . 07 SD , 0 . 20 ± 0 . 06 SD , 0 . 45 ± 0 . 07 SD , and 0 . 86 ± 0 . 10 SD , respectively ( Figure 2A–D ) . These data indicate that , for CAG repeat RNA , there is repeat length specificity for neuritic localization as the prevalence of neuritic particles was highly correlated with increasing repeat number . In contrast , the fraction of primary arbors with ( GGGGCC ) 3-MS2 RNA particles ( Figure 2E ) was higher ( 0 . 69 ± 0 . 12 SD ) than the fraction with ( GGGGCC ) 48-MS2 particles ( 0 . 24 ± 0 . 04 SD ) ( Figure 2D ) , and the percentage of neurons in the mixed culture with neuritic particles was also higher for ( GGGGCC ) 3-MS2 ( 46 . 4 ± 22 . 7% SD [n=3 cultures , 110 neurons total] ) , than for ( GGGGCC ) 48-MS2 ( 21 . 1 ± 3 . 7% s . d . ( n=4 cultures , 147 neurons total ) . These data indicate that three GGGGCC units , at the low end of non-expanded C9orf72 alleles ( DeJesus-Hernandez et al . , 2011; Gijselinck et al . , 2012; Renton et al . , 2011; van der Zee et al . , 2013 ) , are sufficient to confer neuritic localization , and that targeting information is retained in expanded GGGGCC repeat RNA . These data may also suggest that expanded GGGGCC repeat RNA is less efficiently incorporated into RNA granules , or that arbors with expanded GGGGCC repeat RNA had degenerated ( hence a lower fraction of arbors with particles ) . 10 . 7554/eLife . 08881 . 006Figure 2 . Particle formation dependence on CAG and GGGGCC RNA repeat number . ( A–C , E ) Rat mixed spinal cord neurons were transfected with NLS-CP-GFP ( green; arrowheads ) and ( A ) ( CAG ) 20-MS2 , ( B ) ( CAG ) 40-MS2 , ( C ) ( CAG ) 70-MS2 , or ( E ) ( GGGGCC ) 3-MS2 . ( CAG ) 100-MS2 and ( GGGGCC ) 48-MS2 were transfected as above and are shown in Figure 1H and Figure 1—figure supplement 1H , and in Figure 1I and Figure 1—figure supplement 1I , respectively . See Figure 1—figure supplement 2A and Materials and methods for construct details . ( D ) Quantitation of the fraction of primary arbors containing ≥1 GFP granule . Parentheses indicate the total number of primary branches counted in three biological replicates . A minimum of 15 transfected neurons were scored for neuritic GFP granules in total . Averages ± standard deviation are given . Data are representative of a minimum of three biological replicates . Confocal Z-series projections . DsRed ( magenta ) was coexpressed to outline the neurons . Bottom panels: High magnification of neuronal processes . Bars: top panels , 15 μm; bottom panels , 10 μm . DAPI: blue . DAPI , 4' , 6-diamidino-2-phenylindole; CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; NLS , nuclear localization signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 006 The microsatellite repeat RNAs were present in particles not only close to the neural cell body , but also distally in neuronal processes . This raised the possibility that the RNA-particles in the neurites were being actively transported along the length of the projections . To examine this in detail , we explored the dynamics of the localized RNA particles by performing time-lapse imaging . This approach showed that both ( GGGGCC ) 48-MS2 and ( CAG ) 100-MS2 particles were undergoing anterograde , retrograde , and bidirectional movement , both proximally as well as distally along neurites ( Figure 3A–B , Videos 1–3 ) . Similar to previous reports for transport ribonucleoprotein particles ( RNPs ) and consistent with velocities for dynein or kinesin-mediated transport ( Kiebler and Bassell , 2006 ) , the mean average velocity of uninterrupted unidirectional movement was 1 . 06 μm/s for ( GGGGCC ) 48-MS2 , and 1 . 30 μm/s for ( CAG ) 100-MS2 , and the average max velocity was 1 . 40 μm/s for ( GGGGCC ) 48-MS2 , and 1 . 85 μm/s for ( CAG ) 100-MS2 ( Table 1 ) . By contrast , particles that underwent corralled movements had a mean average basal velocity of 0 . 12 μm/s ( Table 1 ) . We could not detect motile particles above background in neurons expressing control RNAs LacZ-MS2 or MS2 . These data indicate that the microsatellite repeat RNAs could be assembling into mRNA transport granules that are dynamic along the neuronal projections . 10 . 7554/eLife . 08881 . 007Figure 3 . ( GGGGCC ) 48 and ( CAG ) 100 RNA assembles into neuritic transport particles and neuritic ( GGGGCC ) 48 RNA correlates with branching defects . ( A–B ) Rat spinal cord neurons were transfected with NLS-CP-GFP and ( A ) ( GGGGCC ) 48-MS2 or ( B ) ( CAG ) 100-MS2 , and the trajectory of motile RNA particles along neuronal processes was captured by time-lapse microscopy . The location of an RNA particle ( arrowheads ) at indicated time points is shown in individual frames . ( A ) The uninterrupted , unidirectional anterograde particle run originates 22 μm from the cell body ( which is outside of the shown frames ) , and ends 18 μm further away ( see Video 1 ) . ( B ) The uninterrupted , unidirectional anterograde particle run originates 67 μm from the cell body ( which is to the left and outside of the shown frames ) , and ends 44 μm further away ( see Videos 2 and 3 ) . Larger stationary particles are also seen . ( C–G ) Cultured primary rat spinal cord neurons with neuritically localized ( GGGGCC ) 48-MS2 RNA have fewer primary branches . ( C–G ) Tracings depicting the cell body and primary branches of rat mixed spinal cord neurons expressing either NLS-CP-GFP and ( C ) ( GAA ) 100-MS2 , ( E-F ) ( GGGGCC ) 48-MS2 , ( G ) ( GGGGCC ) 3-MS2 , or ( D ) NES-CP-GFP and ( GGGGCC ) 48-MS2 RNA ( see Figure 1—figure supplement 2A and Materials and methods for construct details ) . ( H ) Neurons were transfected as in C–G and the number of primary branches were scored in neurons that had ( D ) nuclear RNA particles , ( E ) somatic ( but not neuritic ) RNA particles , or those that had ( F ) neuritic RNA particles , as indicated . Repeat construct and GGGGCC repeat RNA localization affected branch number ( p < 0 . 0001 , ANOVA ) . For individual comparisons by post-hoc Tukey’s multiple comparisons test: ****p < 0 . 0001; ***p < 0 . 001; **p <0 . 01; N . S . p > 0 . 05 . ( I ) Neurons with neuritic ( GGGGCC ) 48-MS2 RNA did not have significantly higher expression level , as determined by ImageJ measurement of mean fluorescence intensity of the neural cell bodies ( see Materials and methods ) , than neurons with somatic ( GGGGCC ) 48-MS2 RNA . N . S . p > 0 . 07 . ( J ) Expression level did not affect branch number ( p = 0 . 2331 , ANOVA ) . Individual comparisons did not reach significance ( p value range: 0 . 2865 to > 0 . 9999 ) . ( H , J ) Only neurons with cell bodies >20 μm and with >2 primary branches were included . Standard deviations are given . Bars: A , 5 μm; B , 10 μm; G , 100 μm . ANOVA , analysis of variance; CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; NES , nuclear export signal; NLS , nuclear localization signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 00710 . 7554/eLife . 08881 . 008Figure 3—figure supplement 1 . Nuclear ( GGGGCC ) 48-MS2 foci in transfected rat primary spinal cord neurons . ( A ) Primary rat spinal cord neurons transfected with NES-CP-GFP ( green ) and ( GGGGCC ) 48-MS2 . ( A ) Bottom left , a transfected neuron with nuclear ( GGGGCC ) 48-MS2 RNA foci . Top right , a transfected neuron that lacks ( GGGGCC ) 48-MS2 RNA foci in the nucleus . Image is representative of three biological replicates . Widefield epifluorescence micrograph . Bar: 10 μm . DAPI: blue . CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; DAPI , 4' , 6-diamidino-2-phenylindole; NES , nuclear export signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 00810 . 7554/eLife . 08881 . 009Table 1 . Behavior of repeat RNA particles in rat spinal cord neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 009Mean average velocity ( μm/s ) Average max velocity ( μm/s ) Velocity range ( μm/s ) Particles tracked in neurites Particles tracked in cell body Basal velocity* ( μm/s ) ( GGGGCC ) 48-MS2::GFP n = 11 cells 1 . 06 1 . 40 0 . 32–2 . 67 10 15 0 . 11 ( CAG ) 100-MS2::GFP n = 2 cells 1 . 30 1 . 85 0 . 30–4 . 73 5 28 0 . 13 Uninterrupted unidirectional anterograde and retrograde particle runs with an average run distance of 5 . 3 μm ( ( GGGGCC ) 48-MS2 ) , and 6 . 7 μm ( ( CAG ) 100-MS2 ) , were analyzed . ( * ) The basal velocity is given as a mean average and was estimated by analyzing five particles that underwent corralled movements with an average net displacement of <0 . 51 μm within 20s . Data are from four ( GGGGCC ) 48-MS2 and two ( CAG ) 100-MS2 independent live imaging sessions . ( GGGGCC ) 48-MS2 and ( CAG ) 100-MS2 were co-expressed with NLS-CP-GFP . CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; NLS , nuclear localization signal . 10 . 7554/eLife . 08881 . 010Video 1 . Movement of a distal ( GGGGCC ) 48-MS2 particle . Two identical videos ( 60 s real-time duration each ) are combined vertically; the bottom video displays a tracked particle in green , starting at 22 μm and reaching 40 μm from the cell body . Images were acquired at 1 frame/s and the video displays at 8 frames/s . The complete caption was 60 s . Selected images are shown in Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 01010 . 7554/eLife . 08881 . 011Video 2 . Movement of a distal ( CAG ) 100-MS2 particle . Two identical videos ( 40 s real-time duration each ) are combined vertically; the bottom video displays the tracked particle and its path in green , starting at 67 . 0 μm and reaching 111 . 4 μm from the cell body . Images were acquired at 1 frame/s and the video displays at 8 frames/s . The complete caption was 133 s . Selected images are shown in Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 01110 . 7554/eLife . 08881 . 012Video 3 . Movement of a proximal ( CAG ) 100-MS2 particle . Two identical videos ( 97 s real-time duration each ) are combined vertically; the bottom video displays the tracked particle and its path in green , starting in the cell body and reaching 6 . 5 μm into a neurite . Images were acquired at 1 frame/s and the video displays at 8 frames/s . The complete caption was 221 s . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 012 The presence of the GGGGCC repeat RNA in distal neuritic particles in C9orf72 hexanucleotide expansion carrier-derived neurons and in transfected rat spinal cord neurons raised the possibility that such expanded repeat RNA may confer local toxicity . We therefore analyzed neuritic arborization patterns in rat mixed spinal cord neurons with ( GGGGCC ) 48-MS2 RNA localized to the nucleus , the soma , or the neurites . While 37 . 5 ± 3 . 7% SD ( n=108 neurons total ) of neurons in the total population contained RNA in the soma ( Figure 1—figure supplement 1I ) , about half of these neurons also had neuritically localized ( GGGGCC ) 48-MS2 RNA ( 19 . 4 ± 7 . 5% SD of the total neuron population; n=108 neurons total , see also Figure 1M ) , and 10 . 0 ± 4 . 0% SD of the total neuron population ( n=108 neurons total ) contained nuclear ( GGGGCC ) 48-MS2 RNA foci ( Figure 3—figure supplement 1A ) . Neurons with neuritically localized ( GGGGCC ) 48-MS2 RNA had , on average , fewer primary branches ( 3 . 8 ± 1 . 6 SD; n=12 neurons ) than neurons with nuclear ( GGGGCC ) 48-MS2 RNA foci ( 6 . 2 ± 1 . 3 SD; n=12 neurons ) , or than neurons with somatic but without neuritic ( GGGGCC ) 48-MS2 RNA ( 6 . 1 ± 2 . 1 SD; n=20 neurons ) ( compare Figure 3F with 3D and E , see Figure 3H ) . They also had fewer primary branches than neurons expressing the ( GAA ) 100-MS2 control ( 8 . 4 ± 2 . 2 SD; n=21 neurons ) , which lacked neuritic RNA particles ( compare Figure 3F with 3C , see Figure 3H ) . In contrast , neurons with neuritically localized non-expanded ( GGGGCC ) 3-MS2 RNA did not show a dramatic primary branch loss Figure 3G ) . These neurons had a similar average number of primary branches ( 6 . 4 ± 1 . 3 SD; n=20 neurons ) compared with neurons with nuclear ( GGGGCC ) 48-MS2 RNA foci , or neurons with somatic but not neuritic ( GGGGCC ) 48-MS2 RNA ( Figure 3H , compare Figure 3G with 3D and E ) . We did not find a significant correlation between the presence of neuritic ( GGGGCC ) 48-MS2 RNA particles and expression level of the RNA in the soma ( Figure 3I , average normalized mean intensity 0 . 56 ± 0 . 23 SD and 0 . 42 ± 0 . 25 SD; n=18 and 24 neurons total with neuritic or somatic ( GGGGCC ) 48-MS2 RNA , respectively ) . Similarly , the expression level of the RNA in the soma did not affect primary branch number ( Figure 3J; n = 76 neurons total ) . These data show that when the ( GGGGCC ) 48-MS2 repeat RNA is present in neuritic particles , there is a dramatic reduction in primary neural branches—this is not the case when the RNA is nuclear or somatic . Furthermore , high expression is not required to drive ( GGGGCC ) 48-MS2 RNA association into neuritic particles or to induce branching defects . These data argue that the presence of the ( GGGGCC ) 48-MS2 RNA in neuritic particles is associated with deleterious effects on neuronal branching . Expanded repeat RNAs have also been reported to undergo translation into peptide repeat proteins . We looked for RAN translated peptide repeat proteins derived from the repeat RNAs ( Ash et al . , 2013; Mori et al . , 2013; Zu et al . , 2011 ) by immunostain utilizing FLAG , HA and Myc tags encoded 3’ of the repeat in the three different reading frames , but were unable to provide evidence for the presence of RAN peptides with these constructs in this system ( see Figure 1—figure supplement 2A and Materials and methods for construct details ) . Because we observe a dramatic reduction of primary branches only in neurons with neuritic RNA granules , branching defects in our system may be mediated by neuritically localized expanded repeat RNA . Moreover , our data suggest that the toxicity conferred by the expanded GGGGCC repeat RNA is not simply due to neuritic granule association , given that neuritically localized non-expanded ( GGGGCC ) 3-MS2 RNA did not confer dramatic primary branch loss . To further address the functional impact of the expanded microsatellite repeat on neuron morphology , we analyzed repeat RNA-induced dendritic degeneration in Drosophila . To visualize this , we used fly lines expressing ( GGGGCC ) 48 repeat RNA or DsRed control RNA ( Figure 1—figure supplement 2B ) in the highly branched class IV epidermal sensory dendritic arborization ( da ) neurons ( Grueber et al . , 2002 , 2003 ) . These neurons have a characteristic and elaborate dendritic branching pattern , allowing detailed analysis of branch complexity ( Figure 4A ) . Expression of UAS- ( GGGGCC ) 48 resulted in dramatic dendritic branching defects compared with the UAS-DsRed control at late third larval instar ( compare Figure 4B to 4A , see Figure 4—figure supplement 1A–B ) . To determine whether the defects resulted from compromised growth , degeneration of pre-established dendrites , or both , we scored da neuron morphology at two developmental time points: early and late third larval instar . As the animal body size increases during this time , the dendritic field undergoes expansion ( Figure 4I; also compare Figure 4C to 4A ) ; however , similar total number of intersections , branch segments per order , and number of endings indicated no overall major branch loss ( Figure 4—figure supplement 1C–D ) . At early third instar , neurons expressing UAS- ( GGGGCC ) 48 RNA appeared nearly normal ( compare Figure 4D with 4C ) , with a dendrite intersection distribution similar to UAS-DsRed control neurons ( compare yellow bars in Figure 4J and 4I , Figure 4—figure supplement 1A ) . By the late stage , however , dendrites in animals expressing UAS- ( GGGGCC ) 48 RNA had failed to extend far from the cell body ( compare pink bars in boxed areas in Figure 4I–J , Figure 4—figure supplement 1B ) , and there was a 42% decrease of distal intersections ( 140–360 μm from cell body ) compared to early stage neurons expressing UAS- ( GGGGCC ) 48 RNA , coinciding with a 53% loss of higher order branches ( orders 13–24 ) ( Figure 4—figure supplement 1E ) . These data indicated that neurons expressing UAS- ( GGGGCC ) 48 RNA were capable of establishing a complex dendritic arbor; however , they subsequently failed to extend and underwent late-stage degeneration of pre-established branches . 10 . 7554/eLife . 08881 . 013Figure 4 . ( GGGGCC ) 48-induced dendritic arborization defects are modulated by altered levels of transport granule components in Drosophila . ( A–H ) Tracings , from confocal Z-series projections , of the cell body and dendritic arbor of class IV da neurons located in the body wall of Drosophila early or late third instar larvae . Expression of ( GGGGCC ) 48 has a dramatic effect on the branching pattern compared with control ( a transgene expressing DsRed ) . The effect on branching is enhanced by upregulation and suppressed by downregulation of dFMR1 or orb2 . ( A–D ) GAL4477 ( Grueber et al . , 2003 ) driven expression of UAS-mCD8::GFP with ( A , C ) UAS-DsRed control ( Li et al . , 2008 ) , or with ( B , D ) UAS- ( GGGGCC ) 48 , in early and late third instar neurons . ( E–H ) GAL4477driven expression of UAS-mCD8::GFP with ( E ) UAS- ( GGGGCC ) 48 and UAS-dFMR1 , ( F ) UAS- ( GGGGCC ) 48 and UAS-orb2 , ( G ) UAS- ( GGGGCC ) 48 and UAS-dFMR1-RNAi , or with ( H ) UAS- ( GGGGCC ) 48 and UAS-orb2-RNAi , in late third instar neurons . ( I–L ) Sholl analysis of traced class IV da neurons shown in ( A–H ) , indicating the number of dendrite intersections with circles drawn at increasing radii from the cell body centroid . ( I–J ) Early ( yellow ) or late ( magenta ) third instar ( I ) UAS-DsRed control or ( J ) UAS- ( GGGGCC ) 48 expressing da neurons . Distal intersections ( 260–400 μm from the cell body centroid ) are boxed in red . ( K–L ) Late third instar neurons expressing UAS- ( GGGGCC ) 48 alone ( magenta ) , or ( K ) with UAS-dFMR1-RNAi ( cyan ) or UAS-dFMR1 ( black ) , or ( L ) with UAS-orb2-RNAi ( cyan ) or UAS-orb2 ( black ) . ( M–T ) Expression of the dFMR1 or orb2 modifier lines alone minimally alters the dendritic intersection distribution . Controls included comparison of the DsRed control ( see A ) to dFMR1 and orb2 lines in absence of UAS- ( GGGGCC ) 48 . ( M–P ) Tracings of class IV da neurons with GAL4477driven expression of UAS-mCD8::GFP with ( M ) UAS-dFMR1 , ( N ) UAS-orb2 , ( O ) UAS-dFMR1-RNAi , or ( P ) UAS-orb2-RNAi . ( Q–T ) Sholl analysis of traced class IV da neurons shown in ( M–P ) . One dorsal neuron from the third or fourth abdominal hemisegment was scored per larvae and three to five larvae were scored per genotype ( except for the late third instar control , A; n=2 ) . ( A–H , M–P ) Dorsal , up; anterior , right . The UAS constructs did not contain a translation reporter or MS2 tags , see Figure 1—figure supplement 2B . Standard deviations are shown . Data are representative of three biological replicates . Bar , 300 μm . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 01310 . 7554/eLife . 08881 . 014Figure 4—figure supplement 1 . Knock down of dFMR1 or orb2 restores ( GGGGCC ) 48-induced branching defects . ( A–G ) Tracings from confocal Z-series projections of GAL4477 driven UAS-mCD8::GFP expression ( Grueber et al . , 2003 ) in Drosophila class IV da neurons were used for the analyses . ( A–E ) The GAL4477 driver was used to express ( A–D ) UAS-DsRed ( control ) or ( A–B , E ) UAS- ( GGGGCC ) 48 . ( A ) Mild dendritic branching defects are observed in ( GGGGCC ) 48 da neurons at early third instar compared with the DsRed control . ( B ) Dramatic dendritic branching defects result from ( GGGGCC ) 48 expression at late third larval instar compared to the DsRed control . ( C–D ) Branch loss is not evident in control larvae during the early to late third instar transition . The total number of ( C ) dendrite intersections and endings , or ( D ) branch segments per order were scored at early ( yellow ) or late ( magenta ) third instar stages . ( E ) Reduction of high order branches at late third instar ( magenta ) compared to early third instar ( yellow ) in neurons expressing ( GGGGCC ) 48 . Branch segments at high orders are boxed in red ( orders 13–24 ) . ( F–G ) Sholl diagrams of traced late third instar larval neurons . The GAL4477 driver was used to express ( F–G ) UAS-DsRed ( control ) , ( F ) UAS- ( GGGGCC ) 48 and UAS-dFMR1-RNAi , or ( G ) UAS- ( GGGGCC ) 48 and UAS-orb2-RNAi . Distal intersections are restored upon ( F ) dFMR1 or ( G ) orb2 knockdown in larvae expressing UAS- ( GGGGCC ) 48 ( cyan ) . Comparisons with the UAS-DsRed control ( magenta ) are shown . Distal intersections at 140–400 μm from the cell body centroid are boxed in red . Standard deviations are shown . One dorsal neuron from the third or fourth abdominal hemisegment was scored per larvae and three to five larvae were scored per genotype ( except for the late third instar control; n=2 ) . The UAS constructs did not contain a translation reporter or MS2 tag , see Figure 1—figure supplement 2B . Data are representative of 3 biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 014 Transport mRNP function is critical for neural health and morphology ( Kiebler and Bassell , 2006; Holt and Schuman , 2013 ) . Our data suggested that the incorporation of the expanded microsatellite repeat into RNA-granules was conferring morphological abnormalities . We first asked whether the GGGGCC expansion might lead to dysregulated expression of RNA binding proteins ( Gerstberger et al . , 2014 ) in brain samples from C9orf72 patients ( Donnelly et al . , 2013 ) . Both RNA binding proteins as a general class and mRNA binding proteins , more specifically , were overrepresented among mRNAs in samples from the diseased brains ( Table 2 and Supplementary file 1 ) . To then assess whether the branching defects could be due to altered transport granule function , we reasoned that changing the levels of transport granule components might suppress or enhance the dendritic defects . We modulated the levels of fly fragile X mental retardation protein ( dFMRP ) , a component of mRNA transport granules and a local translational regulator ( Dictenberg et al . , 2008 ) , and assessed the effects . These studies showed that downregulation of dFMR1 dramatically mitigated the UAS- ( GGGGCC ) 48-induced dendritic branching defects ( compare Figure 4G with 4B ) with a near doubling ( 96% increase ) of distal intersections ( 140–360 μm from cell body; Figure 4K ) . In contrast , upregulation of dFMR1 in the context of UAS- ( GGGGCC ) 48 expression potentiated the branching defects , reducing intersections by 70% ( 120–360 μm from cell body; Figure 4K , compare Figure 4E with 4B ) . Studies on a second transport granule component that regulates the local translation of neuritic RNAs , Orb2 ( Cziko et al . , 2009; Mastushita-Sakai et al . , 2010; La Via et al . , 2013 ) , showed a similar dramatic modulation of the UAS- ( GGGGCC ) 48-induced branching defects: distal intersections were doubled upon downregulation ( 103% increase 140–380 μm from cell body ) , while upregulation resulted in a 34% overall loss ( Figure 4L , compare Figure 4F and H with 4B ) . Downregulation of either modifier restored the number of distal intersections ( 140–400 μm from cell body ) compared to the late stage control , to 91% ( dFMR1 RNAi ) , and to 94% ( orb2 RNAi ) ( compare Figure 4G and H to A; see Figure 4—figure supplement 1F–G ) . Our control experiments indicated that in the absence of UAS- ( GGGGCC ) 48 , up- or downregulation of dFMR1 and orb2 resulted in minimal changes in the dendrite intersection distribution . Upregulation of dFMR1 or orb2 alone did not reduce the number of intersections ( Figure 4M-N , Q-R ) . Knockdown of either dFMR1 or orb2 alone resulted in a 3 . 8 and 6 . 5% increase in distal dendrite intersections ( 140–400 μm from the cell body ) , respectively ( Figure 4O-P , S-T ) . The effects of dFMR1 modulation on da neuron morphology were milder than seen in previous studies , which used null animals , drove expression with a different Gal4 driver , and examined impacts on other specific neurons ( Lee et al . , 2003 ) . Taken together , our data show that expanded GGGGCC microsatellite repeat RNA is present and transported in neurites , and that modulation of levels of transport granule components impacts the neuritic defects induced by the expanded microsatellite repeat RNA in vivo in Drosophila . 10 . 7554/eLife . 08881 . 015Table 2 . Expression of transport-granule related transcripts in brains of C9orf72 patients . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 015Expression in C9orf72 CortexGene subsetExpected percentageObserved percentageChi-squared p-valueUpregulatedRNA binding proteins7 . 515 . 10 ( 167/1103 ) p<0 . 0001mRNA binding proteins3 . 47 . 80 ( 86/1103 ) p<0 . 0001FMRP targets4 . 24 . 90 ( 54/1103 ) p=0 . 2568STAT5B targets3 . 16 . 07 ( 67/1103 ) p<0 . 0001DownregulatedRNA binding proteins7 . 54 . 93 ( 129/2618 ) p<0 . 0001mRNA binding proteins3 . 41 . 72 ( 45/2618 ) p<0 . 0001FMRP targets4 . 23 . 40 ( 89/2618 ) p=0 . 0389STAT5B targets3 . 11 . 38 ( 36/2618 ) p<0 . 0001Comparison of uniquely-identified protein coding genes that were either up- or down-regulated in cortical samples from C9orf72 patients ( Donnelly et al . , 2013 ) with transcripts associated with the regulation of RNA . Supplementary file 1 lists the RNA binding proteins upregulated and downregulated , as noted above . The presence of expanded GGGGCC repeat RNA in transported granules and dramatic modulation of expanded GGGGCC repeat RNA toxicity by dFMR1 raised the possibility of a functional association between the repeat RNA in the RNA-granules and FMRP protein . In rat spinal cord neurons , both endogenous and exogenous FMRP colocalized in neuritic granules with ( GGGGCC ) 48-MS2 and ( CAG ) 100-MS2 repeat RNAs in neuronal processes ( Figure 5A–C , Figure 5—figure supplement 1A–F ) . Thus , association with FMRP was a property of multiple expanded repeat RNAs . Consistent with this observation , both FMRP as well as its interaction partners FXR1 and FXR2 , have been shown to interact with GGGGCC RNA repeats by assays that include pull down and proteome arrays ( Almeida et al . , 2013; Donnelly et al . , 2013; Haeusler et al . , 2014; Rossi et al . , 2015 ) . 10 . 7554/eLife . 08881 . 016Figure 5 . Misregulation of transport granule components in human iPSNs from carriers with a C9orf72 GGGGCC expansion . ( A–C ) Neuritic particles consisting of expanded GGGGCC repeat RNA co-label for FMRP . Rat primary spinal cord neurons were transfected with NLS-CP-GFP ( green ) , FMRP-RFP ( magenta ) , and ( GGGGCC ) 48-MS2 , and neuronal processes were defined as regions of interest . Colocalization coefficients M1 ( FMRP-RFP overlap with ( GGGGCC ) 48-MS2 RNA ) and M2 ( ( GGGGCC ) 48-MS2 RNA overlap with FMRP-RFP ) were 0 . 64 ± 0 . 15 SD and 0 . 68 ± 0 . 23 SD , respectively ( n=6 neurons ) . Colocalization coefficients for overlap between endogenous FMRP and ( GGGGCC ) 48-MS2 were M1=0 . 61 ± 0 . 06 SD and M2=0 . 56 ± 0 . 14 SD ( n=5 neurons; not shown ) . See Figure 1—figure supplement 2A and Materials and methods for construct details . Data are representative of three biological replicates . Confocal Z-series projections . ( D–K ) FMRP targets ( D–E ) PSD-95 and ( F–G ) FMRP , as well as ( H–I ) CPEB3 , a local translation regulator , are increased in human iPSNs from C9orf72 GGGGCC expansion carriers , with a concomitant increase in PSD-95 and CPEB3 foci . High magnification of cell bodies are shown as insets . Neurites were marked with α-b III Tubulin or are outlined ( dotted line ) , and ( D ) neuritic PSD-95 foci are indicated ( arrowheads ) . ( D–I ) Images for GGGGCC expansion carrier 2 are shown . ( J–L ) Key for carriers and controls is shown at top . ( J–K ) Quantitation of PSD-95 and CPEB3 ( J ) foci , and of ( K ) total protein levels by immunostain in human iPSNs from carriers with a C9orf72 GGGGCC expansion . Kruskal–Wallis analysis for carrier versus control for all conditions: p < 0 . 0001 . Post-hoc Dunn’s test , multiplicity adjusted p-values: ****p < 0 . 0005; ***p < 0 . 0015; **p < 0 . 018; N . S . p > 0 . 05 . ( J ) From left to right , PSD-95: control , n=812 foci in 29 neurons; carrier 1 , n=1590 foci in 30 neurons; control , n=851 foci in 49 neurons; carrier 2 , n=1455 foci in 38 neurons . CPEB3: control , n=980 foci in 35 neurons; carrier 1 , n=2067 foci in 39 neurons; control , n=735 foci in 21 neurons; carrier 2 , n=1980 foci in 44 neurons . ( K ) From left to right , PSD-95: control , n=29; carrier 1 , n=30; control , n=37; carrier 2 , n=48 neurons scored . FMRP: control , n=32; carrier 1 , n=30; control , n=25; carrier 2 , n=27 neurons scored . CPEB3: control , n=34; carrier 1 , n=37; control , n=17; carrier 2 , n=49 neurons scored . ( L ) FMRP is not sequestered in the nuclei of carrier iPSNs . Quantitation of the fraction nuclear to total FMRP in carrier vs . control iPSNs . Data are averages from carrier 1 , carrier 2 , and controls ± standard error of the mean . From left to right: control , n=5; carrier 1 , n=5; control , n=8; carrier 2 , n=6 neurons scored . All comparisons are non-significant by ANOVA and post-hoc Sidak’s t-test . Confocal Z-series projections are shown . Bars: ( E , G , I ) 10 μm; ( C ) 5 μm . DAPI: blue . See also Figure 5—figure supplement 1 . ANOVA , analysis of variance; CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; FMRP , fragile X mental retardation protein; NLS , nuclear localization signal; PSD , postsynaptic density protein; RFP , red fluorescent protein . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 01610 . 7554/eLife . 08881 . 017Figure 5—figure supplement 1 . FMRP colocalizes with neuritic ( GGGGCC ) 48-MS2 and ( CAG ) 100-MS2 RNA . Neuritic particles consisting of expanded GGGGCC or CAG repeat RNA co-label for FMRP . ( A–F ) Neurons were transfected with NLS-CP-GFP ( green ) , FMRP-RFP ( magenta ) , and ( A–C ) ( GGGGCC ) 48-MS2 , or ( D–F ) ( CAG ) 100-MS2 . Average ± standard deviation for colocalization coefficients M1 ( FMRP overlap with RNA ) and M2 ( RNA overlap with FMRP ) , are given . ( GGGGCC ) 48-MS2 ( n=6 neurons ) ; ( CAG ) 100-MS2 ( n=9 neurons ) . Neuronal processes were selected as regions of interest . See Figure 1—figure supplement 2A and Materials and methods for construct details . Data are representative of three biological replicates . Confocal Z-series projections are shown . CP-GFP , MS2 RNA-binding coat protein fused with green fluorescent protein; FMRP , fragile X mental retardation protein; NLS , nuclear localization signal . DOI: http://dx . doi . org/10 . 7554/eLife . 08881 . 017 To illuminate potential functional consequences of the association of the GGGGCC repeat with FMRP in cytoplasmic neuritic RNA granules , we examined whether FMRP-target genes were misregulated in samples derived from human C9orf72 expansion patients . However , we found no evidence of altered transcript levels of FMRP target genes ( Darnell et al . , 2011 ) in cortical samples from C9orf72 patients ( Donnelly et al . , 2013 ) ( Table 2 and Supplementary file 1 ) , which is consistent with a role for FMRP in post-transcriptional gene regulation . We therefore assayed the levels of an FMRP target protein , postsynaptic density protein ( PSD-95 ) ( Todd et al . , 2003; Muddashetty et al . , 2007; Zalfa et al . , 2007; Tsai et al . , 2012 ) , as a readout of FMRP translation regulation in iPSNs derived from two GGGGCC repeat expansion carriers . We found an 89–123% increase in the number , but not the size , of PSD-95 foci per neuron in iPSNs derived from GGGGCC repeat expansion carriers compared to controls ( Figure 5D-E , J ) . We also saw a 50–76% increase in total PSD-95 levels ( Figure 5D-E , K ) . These results contrast with those obtained when scoring exclusively neuritic PSD-95 , for which a change in PSD-95 neuritic puncta was not seen ( Almeida et al . , 2013 ) . We also examined the protein levels of FMRP ( which is subject to self-regulation at the mRNA level [Ashley et al . , 1993] ) , and found a 51–130% increase in FMRP in patient-derived iPSNs compared with controls ( Figure 5F-G , K ) . These results suggest that regulation of FMRP targets could be aberrant in iPSNs derived from C9orf72 GGGGCC repeat expansion carriers . We also analyzed a second transport granule component and local translation regulator , human CPEB3 ( Huang et al . , 2006; Darnell and Richter , 2012 ) . CPEB3 is a homolog of Drosophila Orb2 that modulates GGGGCC repeat toxicity in flies ( see Figure 4 ) , and is also present in FMRP granules ( Ferrari et al . , 2007 ) and postsynaptic densities ( Huang et al . , 2006 ) . Total CPEB3 levels were elevated 59–118% in iPSNs with a GGGGCC repeat expansion compared to controls ( Figure 5H-I , K ) . The upregulation correlated with a 60–89% increase in CPEB3 foci per neuron in carriers versus controls; foci size was not affected ( Figure 5H-I , J ) . The change in CPEB3 could be an independent effect of the toxic RNA or could be a consequence of FMRP-induced changes . Because we found no FMRP enrichment in the nuclei in carrier iPSNs ( Figure 5L ) , our data do not support a nuclear GGGGCC repeat RNA-mediated FMRP sequestration model . To investigate the functional significance of dysregulated CPEB3 levels , we asked whether its target genes might be misexpressed . CPEB3 can modulate expression of targets of the transcription factor STAT5B ( Peng et al . , 2010 ) . Consistent with disruptions in CPEB3/STAT5B-modulated transcription , cortical samples from C9orf72 patients ( Donnelly et al . , 2013 ) exhibit misregulation of STAT5B target genes ( Kanai et al . , 2014 ) ( Table 2 ) . These results indicate that expanded GGGGCC repeat RNA may interfere with the local translation machinery and indirectly modify transcriptional programs . Together , these data suggest that expanded microsatellite repeat RNAs like GGGGCC that are incorporated into granules within neurites may have local effects that contribute to neurodegeneration .
Here we have identified a novel function of expanded microsatellite RNA repeats in conferring neuritic RNA granule localization . Our data indicate that expanded repeat RNAs with specific structural context ( e . g . stem-loop for CAG , CUG , and CCUG , and G-quadruplex and stem-loop for GGGGCC repeat RNA ) can be recognized by the mRNA localization machinery , can become incorporated into neuritic RNA transport granules , and , at least for expanded GGGGCC hexanucleotide repeat RNA , may disrupt RNA granule function . The RNAs expressed are directed to the cytoplasm with a poly ( A ) tail , as are repeats that occur within the mRNA of the respective disease genes . In the case of the hexanucleotide expansion in C9orf72 , although the repeat is defined as intronic , we saw neuritic GGGGCC RNA granule localization in iPSNs , indicating the repeat can localize to the cytoplasm in disease . Notably among the large portion of mRNAs that are localized , RNAs with stem-loop structure commonly function as cis-acting localization signals ( Ferrandon et al . , 1994; Serano and Cohen , 1995; Cohen et al . , 2005; Snee et al . , 2005; Van De Bor et al . , 2005; Dienstbier et al . , 2009 ) . Indeed , in flies , 7 transcripts have been demonstrated to localize through minus-end directed transport along microtubules , and these mRNAs all contain one or more stem-loops within their localization signal . Although not similar to each other at the primary sequence level , all of these localization signals are recognized by the same localization machinery ( Dienstbier et al . , 2009 ) . In addition , G-quadruplex consensus RNA sequences have also been shown to be cis-acting elements that are both necessary and sufficient for neuritic localization of PSD-95 and CaMKIIα , two dendritically localized mRNAs ( Subramanian et al . , 2011 ) . Indeed , about one-third of the best characterized dendritic mRNAs contain a putative G-quadruplex in their 3’UTRs ( Subramanian et al . , 2011; Stefanovic et al . , 2015 ) . Hence , not only do G-quadruplex consensus sequences and disease-associated GGGGCC repeat RNA assume G-quadruplex structure , these RNAs also appear to have a similar common function as neuritic localization signals . These observations underscore the findings we report that , structured RNAs , like CAG and GGGGCC , are localizing to dynamic neuritic granules . We find that neuritic localization of the expanded GGGGCC hexanucleotide repeat RNA occurs in association with neuritic defects . Neurons with expanded GGGGCC RNA granules in neurites have a decrease in primary branches compared with controls . We do not see a comparable decrease when the expanded RNA is localized merely to the soma , or when it is present in nuclear foci , consistent with a recent report ( Tran et al . , 2015 ) . Importantly , the branching defects associated with the neuritically localized expanded repeat are not seen with a similarly localized non-expanded ( GGGGCC ) 3 repeat—branching in neurons with neuritic non-expanded repeat is not significantly different from branching in neurons with somatic or nuclear expanded repeat . These data indicate that the incorporation of expanded microsatellite repeat RNAs into granules within neurites induces dysfunction . The finding of branching defects in rat primary spinal cord neurons in culture was also extended to da neurons in Drosophila . In vivo , the dendritic arbors of da neurons are normal early , but later show a different pattern with fewer intersections and smaller field . This effect is distinct from defects in endosomal transport , as described for dynein loss-of-function mutations ( Satoh et al . , 2008 ) , indicating it is unlikely to be due to vesicular traffic transport defects . Furthermore , two transport granule components ( FMRP and Orb2 , the fly CPEB3 ortholog ) are novel modifiers of GGGGCC toxicity . Our studies also provide evidence that the expanded GGGGCC repeat RNA may compromise local translation regulation: the FMRP targets , PSD-95 and FMRP , appeared present at elevated levels in iPSNs from C9orf72 hexanucleotide expansion carriers . GGGGCC repeat RNA could disrupt FMRP-mediated translational repression or increase FMRP-mRNA target stability , the latter scenario being less likely because PSD-95 mRNA levels are similar in carrier vs . control iPSNs ( Almeida et al . , 2013 ) . In FMRP knockout mice , PSD-95 mRNA is destabilized and PSD-95 levels reduced ( Zalfa et al . , 2007; Zhu et al . , 2011 ) —similarly , knockdown of FMRP might lead to destabilization of its mRNA targets , thus counteracting translational derepression by toxic GGGGCC repeat RNA in disease . There are a large number of mRNAs regulated by FMRP and CPEB3 , many or all of which may factor into neurotoxicity . Consistent with our observation that CPEB3 protein levels are upregulated in iPSNs from GGGGCC expansion carriers , we find CPEB3/STAT5B-regulated genes are dysregulated in samples from C9orf72 patient cortex ( Donnelly et al . , 2013 ) , see Table 2 ) . The mechanisms of toxicity or pathogenesis of expanded microsatellite repeat RNA include protein translation and sequestration of binding proteins . A study in the fly showed that RAN translation products generated from the GGGGCC repeat RNA can be toxic . Moreover , they found that a GGGGCC repeat is toxic in vivo , but toxicity is minimal if the sequence is not a pure GGGGCC , but is interrupted by stop codons ( and thus could not code for peptides ) ( Mizielinska et al . , 2014 ) . However , alterations in the RNA sequence required to block RAN translation ( introduction of stop codons ) may well interfere with the intricacy of RNA-protein interactions , such as those required for subcellular RNA localization , and/or those that mediate toxicity . Mechanisms beyond RAN translation may well contribute to neurodegeneration conferred by expanded GGGGCC repeat RNA . Targeting of expanded microsatellite repeat RNA to the neuritic granules that we document may disrupt local mRNA translation , and might also interfere with proper trafficking of cellular RNAs . We speculate that the presence of microsatellite repeat RNA in neurites might also result in local RAN translation , and that RAN translation products in the neuritic subcellular compartment could contribute to neurite loss . Our data support a novel model in which neuritically localized expanded microsatellite repeat RNAs associate with neuritic RNP granule components and disrupt their function , resulting in neuritic defects . This mechanism may contribute to ALS/FTD disease in patients bearing the GGGGCC repeat expansion , as we have shown strong effects in iPSC-derived neurons from GGGGCC expansion carriers , in cultured rat spinal cord neurons , and in vivo in a Drosophila model . In culture , we have shown many different expanded microsatellite repeat RNAs are incorporated into neuritic granules , and at least several are actively transported . For the GGGGCC repeat , a number of proteins that bind the repeat ( hnRNP A3 ) or are modifiers of GGGGCC repeat toxicity ( Pur alpha and hnRNP A2/B1 ) are implicated in transport granule function ( Jin et al . , 2007; Sofola et al . , 2007; Xu et al . , 2013 ) . Interestingly , mutations in TDP-43 impair neuritic mRNA transport in primary and stem-cell derived neurons and are causative of ALS ( Alami et al . , 2014 ) ; TDP-43 pathology also characterizes many repeat expansion diseases ( Elden et al . , 2010; Toyoshima and Takahashi , 2014 ) . Thus , multiple lesions could converge at the functional level to result in disrupted mRNA transport granule function .
RFP-DCP1 , DsRed , and pGW were from Dr . Robert Kalb ( Department of Pediatrics , University of Pennsylvania School of Medicine ) , and FMRP-RFP was a kind gift from Dr . Ian Macara ( Department of Cell and Developmental Biology , Vanderbilt University ) . A backbone was designed to receive the repeat sequences ( CAG ) 40 , ( CAG ) 70 , ( CAG ) 100 , ( CUG ) 100 , ( CCUG ) 100 , ( GGGGCC ) 48 , and ( GAA ) 100 . The backbone , as well as the repeat sequences , were synthesized and ligated into pUC57 ( GenScript , Piscataway , NJ ) . The repeat sequences contained 5’ EcoRI and 3’ BamHI sites , and the first base of the first tandem repeat was omitted if it started with cytosine . The backbone contained the following in 5’ to 3’ order: a 6Stop sequence ( carrying six 5’ stop codons ( underlined ) in the leader sequence , two in each reading frame ) containing a 3’ EcoRI site ( TAGCTAGGTAACTAAGTAACTAGAATTC ( Renton et al . , 2011 ) ) , followed by a BamHI site ( GGATCC ) , then by sequences encoding FLAG- , HA- , and Myc-tags ( AGGATTACAAGGACGACGACGACAAGTAGCTACCCATACGACGTTCCAGATTAC CTTAACGAACAGAAACTCATCTCTGAAGAGGATCTGAACATGCATACGGGTCATC TCACCATCACCACTAATAGATAGTGAATAATGAATTTAAATTAATAGATAGTGAATA TGA ) , and then 12 MS2 stem-loops ( Haim-Vilmovsky and Gerst , 2009 ) ( of sequence ( CCTAGAAAACATGAGGATCACCCATGTCTGCAGGTCGACTCTAGAAAACATGAGGATCACCCATGTCTGCAG TATTCCCGGGTTCATTAGATCCTAAGGTACCTAATTG ) 5 CCTAGAAAACATGAGGATCACCCATGTCTGCAG GTCGACTCCAGAAAACATGAGGATCACCCATGTCTGCAG TATTCCCGGGTTCATT CTCGAG AGATCT ) . The backbone was then cloned into pGW using external restriction sites and the repeat sequences were then inserted between EcoRI and BamHI restriction sites of the backbone . ( CAG ) 20-MS2 and ( GGGGCC ) 3-MS2 were made by polymerase chain reaction ( PCR ) using complimentary oligos and ligated into pGW containing the backbone , as described above . LacZ was amplified by PCR and ligated into pGW-MS2 to generate LacZ-MS2 . ( CAG ) 100 , shown in Figure 1D–E and in Figure 1—figure supplement 1A–B , was inserted into 6Stop-FLAG-HA-Myc to generate 6Stop- ( CAG ) 100-FLAG-HA-Myc and cloned into pcDNA , and lacked an MS2 tag . CP- ( GFP ) 2 ( Haim-Vilmovsky and Gerst , 2009 ) with a 5’ NLS or nuclear export signal ( NES ) sequence was cloned into pGW . See Figure 1—figure supplement 2A for construct diagrams . Embryonic Sprague Dawley rat spinal cord neurons from embryonic day 14 were grown on previously established cortical postnatal d1-3 astrocyte monolayers ( Mojsilovic-Petrovic et al . , 2006 ) . Neurons were grown for 5 d before being transfected with Lipofectamine 2000 ( Invitrogen , Carlsbad , CA ) , according to the manufacturer , and using a 1:3:3 ratio of NES- or NLS-CP-GFP , MS2 tagged sequences , and DsRed or RFP plasmids , respectively . Neurons were fixed at 17–24 hr post transfection and processed according to standard procedures . Antibodies were added overnight at 4ºC and included chicken α-GFP ( 1:2000; A10262 , Invitrogen , Carlsbad , CA ) , mouse α-mRFP ( 1:2000; ab65856 , Abcam , Cambridge , MA ) . Mouse α-FMRP ( clone 2F5-1; Christie et al . , 2009 ) was added after steam antigen retrieval . Neurons on coverslips were mounted in Vectashield Mounting Medium with 4' , 6-diamidino-2-phenylindole ( DAPI; Vector Laboratories , Burlingame , CA ) . A minimum of three independent transfections with experimental samples along with controls were performed for all samples and yielded similar results across biological replicates . Fibroblast-derived iPSNs from GGGGCC hexanucleotide expansion carrier 1 ( line #5 ) and carrier 2 ( line #11 ) , and from control lines ( #17 and #20 ) ( Almeida et al . , 2013 ) were fixed , and stained using mouse α-FMRP as above , mouse α-PSD-95 ( 1:200; 6GG-IC9 , Pierce/Fisher , Rockford , IL ) , rabbit α-CPEB3 ( 1:200; ab10883 , Abcam , Cambridge , MA ) , or chicken α-b III Tubulin ( 1:1000; AB9354 , Millipore , Billerica , MA ) . Due to the extensive nature of required quantitation , each carrier was independently experimentally analyzed with a control . All iPSC lines were grown and differentiated to neurons in parallel . DIG labeled ( CUG ) 8 and ( CAG ) 8 sense and antisense oligonucleotide probes were generated ( IDT DNA , Coralville , IA ) , in situ hybridization was performed ( Wilk , 2010 ) , and the probe signal was amplified with the Tyramide Signal Amplification system ( Perkin Elmer , Whaltham , MA ) using a fluorescein kit according to the manufacturer . A Cy3-conjugated ( GGCCCC ) 4 oligonucleotide probe was used for in situ hybridization of iPSNs as described ( Almeida et al . , 2013 ) . A Leica confocal microscope equipped with a HyD detector was used for detection of GGGGCC RNA particles . Early and late third instar larvae were filleted in ice cold phosphate-buffered saline ( PBS ) , fixed in PBS/4% paraformaldehyde , and stained with chicken α-GFP as described above . Post-fix and post-stain washes included three rinses and 3 × 15 min in PBS/0 . 3% Triton X-100 . Secondary antibodies were conjugated to Alexa 488 ( Invitrogen , Carlsbad , CA ) . The Drosophila lines to knock down orb2 ( genotype y1 v1; P{TRiP . JF023076}attP2 ) , and dFMR1 ( genotype y1 sc* v1; P{TRiP . HMS00248}attP2 ) were from the Bloomington stock center . The UAS- ( GGGGCC ) 48 repeat sequence ( with the 6Stop sequence but without the translation or MS2 tags noted above; see Figure 1—figure supplement 2B ) was subcloned into pUAST to generate UAS- ( GGGGCC ) 48 and the construct was injected to generate transgenic strains ( Genetic Services , Inc . , Cambridge , MA ) . UAS-dFMR1 was from Dr . Thomas Jongens ( Department of Genetics , University of Pennsylvania School of Medicine ) . The UAS-DsRed strain was used as a control for UAS- ( GGGGCC ) 48 . UAS-DsRed ( Bilen and Bonini , 2007 ) and UAS-orb2 are described ( Dictenberg et al . , 2008 ) . Images of rat and da neurons were captured on a Leica TCS SP5 confocal microscope and processed with the Leica Application Suite ( LAS ) software ( Leica Microsystems , Wetzlar , Germany ) . Sequential acquisition was applied when capturing an image in multiple channels . Similar voltage settings were applied when capturing images of rat neurons transfected with different constructs , and a saturation threshold was applied . For Figure 1 , above-background fluorescence that was clearly discernable by eye as having a clear particle limit was scored as a particle . Images of spinal cord neurons at 17–24 hr post transfection , were collected with a Deltavision Core Deconvolution Microscope ( Applied Precision , Issaquah , WA ) , equipped with an Olympus IX70 microscope and a Photometrics CoolSNAP HQ camera , a 60X , 1 . 42 NA oil immersion PlanApo lens ( Olympus , Tokyo , Japan ) , and softWoRx ( Applied Precision , Issaquah , WA ) acquisition software . Environmental control was provided by a home-built plexiglass cage surrounding the entire microscope , kept at 37ºC and 5% CO2 . Individual frames were generated at 1 s intervals for single channel imaging . The percentage of neurons with distal ( GGGGCC ) 48-MS2 , and ( CAG ) 100-MS2 RNA particles detected by live imaging ( six sessions for ( GGGGCC ) 48-MS2 , and two sessions for ( CAG ) 100-MS2 was similar to that seen in fixed neurons from multiple biological replicates . Two sessions for ( GGGGCC ) 48-MS2 were excluded due to low transfection efficiency . Videos were generated and particles were tracked manually with Fiji software . Quantitative colocalization analysis was performed using Volocity software version 6 . 2 . 1 ( Perkin Elmer , Whaltham , MA ) . The colocalization coefficients ( M1 and M2 ) were computed ( Manders et al . , 1993 ) for regions of interest ( ROI ) . These ROIs were manually selected to only target neuronal processes . We analyzed >5 neurons for each condition to ensure that M1 and M2 were similar when comparing cells within the same sample and between distinct biological replicates ( >3 ) . The intensity thresholds for the colocalization coefficients were determined using an auto-threshold method ( Costes et al . , 2004 ) . Spinal cord neurons and Drosophila da neurons were traced , and the tracings were analyzed with Neurolucida and Neuroexplorer software , respectively ( MicroBrightField , Colchester , VT ) . For the rat dendritic arbor analysis , only neurons that had a cell body diameter of >20 μm , and had more than two primary arbors were included . A pre-established standard cell sample size ( n≥20; Drs . Lei Zhang and Robert Kalb , personal communication ) was used for this type of analysis , except for samples that had nuclear or neuritic ( GGGGCC ) 48-MS2 RNA ( n=12 ) , due to the limiting inclusion criteria used . For quantitation of PSD-95 and CPEB3 particles in iPSNs , z-stacks taken with a 63× objective acquired on a Leica confocal microscope equipped with a HyD detector were projected , and the cell body was outlined . The particle number and size were analyzed using the 'analyze particle' function of Image J ( NIH ) , using the Yen or Max Entropy auto-thresholding methods . Our analysis of PSD-95 in the entire cell body differs from previous quantitation solely in dendrites ( Almeida et al . , 2013 ) . Total protein levels of FMRP , PSD-95 , and CPEB3 in iPSNs was measured by outlining the entire neuron using Image J . For analysis of nuclear FMRP the nucleus ( based on DAPI stain ) , and the whole neuron were outlined , measured using Image J , and the nuclear intensity was divided by the total neuron intensity . For Figure 3I and J , expression levels were measured using Image J to calculate the mean intensity; the cell bodies of the neurons were selected as the ROI for these analyses . To determine the somatic expression levels , signal from the nucleus , defined by DAPI staining , was subtracted . All samples , including all animal experiments , were randomly assigned to processing order , and for cell transfections , the positions in the wells were random . Data was also collected randomly . Statistical tests were performed using R 3 . 1 . 2 ( Figure 1 ) or Prism 6 software from Graphpad , La Jolla , CA ( Figures 3 and 5 , and Table 2 ) . For Figure 1L–M , the data were expressed as a binomial , with cells categorized as having neuritic RNA or not . Within each condition ( for example , construct or carrier/control ) , the cells were grouped by experiment to account for potential variability between experiments . The data were fitted with a log-linear generalized linear model in R 3 . 1 . 2 ( Pumpkin Helmet ) using the glmer function of the lme4 package , with post-hoc analyses comparing each construct/condition to control . In the case of Figure 1M , where the LacZ-MS2 construct had no variance , the model could not converge . Therefore a single LacZ-MS2 data point was switched from nuclear to neuritic . The same operation was done for control iPSNs in Figure 1L . Both of these changes were conservative as they were in the opposite direction of the observed effect . For the analyses in Figure 5 , the Brown–Forsythe test indicated that the samples exhibited different variances . We therefore conducted nonparametric tests for these analyses . For the analysis of gene lists , uniquely identifiable , well-annotated protein coding transcripts ( i . e . , those with a refseq identifier beginning with 'NM' ) that were misregulated in C9orf72 patient samples were compared with several gene lists as indicated in the text: RNA binding proteins and mRBP classifications were from ( Gerstberger et al . , 2014 ) ; FMRP targets were GSE45148 from ( Darnell et al . , 2011 ) ; C9orf72 targets were from ( Donnelly et al . , 2013 ) ; Stat5b targets were from ( Kanai et al . , 2014 ) . The percent of C9orf72-regulated transcripts was compared with the percentage expected by chance given the prevalence of the RNAs in the ~20 , 000 protein-coding transcripts present in the human genome using a chi-squared analysis with a significance threshold of p=0 . 01 . | Genes contain instructions to build proteins , but these instructions are often interrupted by stretches of DNA that do not code for protein . Typically the entire length of a gene is copied to produce an intermediate molecule of RNA , which is processed to remove the non-coding regions before being translated to make a protein . The genes associated with various neurodegenerative diseases , including Huntington's disease and myotonic muscular dystrophies , often also carry short stretches of DNA sequence that are repeated one after the other . An increase in the number of the repeats within one of these genes can lead to a neurodegenerative disease . These disorders often have similar features , but are associated with different repeat sequences that can occur either in regions of the gene that code for protein or regions that do not . When the repeats lie in a non-coding region , it is thought that the RNA itself and not the protein causes the damage to nerve cells . While it is not known how this happens , it could be related to the shape of the RNA molecules , which in turn controls where the RNA molecules will go within a cell . Inside nerve cells , some RNAs ( but not all ) are directed to particles called 'transport granules' . These particles carry specific RNAs into the tips of the nerve fibers where they are then translated into proteins . Burguete et al . wanted to test whether disease-associated RNAs that contain repeats might interfere with this process in nerve cells . Microscopy showed that many RNAs with expanded stretches of repeats ended up in the transport granules by mistake , and were carried toward the tips of the nerve branches . When the repeat-containing RNAs localized to the transport granules , the nerve endings tended to form fewer branches . By analyzing the components of the granules , Burguete et al . could show that the incorrect localization of RNA molecules in the granules appeared to interfere with the production of proteins in the nerve branches . This disruption could contribute to the nerve cell defects seen in the many neurodegenerative diseases associated with these types of repeat expansions . These data suggest that preventing the disruption of transport granules’ activity could represent a potential therapeutic avenue against these diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | GGGGCC microsatellite RNA is neuritically localized, induces branching defects, and perturbs transport granule function |
LAMP1 and LAMP2 proteins are highly abundant , ubiquitous , mammalian proteins that line the lysosome limiting membrane , and protect it from lysosomal hydrolase action . LAMP2 deficiency causes Danon’s disease , an X-linked hypertrophic cardiomyopathy . LAMP2 is needed for chaperone-mediated autophagy , and its expression improves tissue function in models of aging . We show here that human LAMP1 and LAMP2 bind cholesterol in a manner that buries the cholesterol 3β-hydroxyl group; they also bind tightly to NPC1 and NPC2 proteins that export cholesterol from lysosomes . Quantitation of cellular LAMP2 and NPC1 protein levels suggest that LAMP proteins represent a significant cholesterol binding site at the lysosome limiting membrane , and may signal cholesterol availability . Functional rescue experiments show that the ability of human LAMP2 to facilitate cholesterol export from lysosomes relies on its ability to bind cholesterol directly .
Eukaryotic lysosomes are acidic , membrane-bound organelles that contain proteases , lipases and nucleases and degrade cellular components to regenerate catabolic precursors for cellular use ( Xu and Ren , 2015; Schwake et al . , 2013; Saftig and Klumperman , 2009 ) . Lysosomes are crucial for the degradation of substrates from the cytoplasm , as well as membrane bound compartments derived from the secretory , endocytic , autophagic and phagocytic pathways . The limiting membrane of lysosomes is lined with so-called lysosomal membrane glycoproteins ( LAMPs ) that are comprised of a short cytoplasmic domain , a single transmembrane span , and a highly , N- and O-glycosylated lumenal domain ( Wilke et al . , 2012; Kundra and Kornfeld , 1999; Granger et al . , 1990 ) . Because of their abundance and glycan content , LAMPs have been proposed to serve as a protective barrier to block hydrolase access to the limiting phospholipid bilayer . LAMP1 and LAMP2 are 37% identical and may overlap in function , but knockout of LAMP1 in mouse has a much milder phenotype than depletion of LAMP2 ( Tanaka et al . , 2000 ) : LAMP2-deficient mice have very short lifespans , and show massive accumulation of autophagic structures in most tissues . Indeed , LAMPs are required for fusion of lysosomes with phagosomes ( Huynh et al . , 2007 ) and LAMP2 has also been proposed to serve as a receptor for chaperone-mediated autophagy ( Cuervo and Dice , 1996 , 2000; Bandyopadhyay et al . , 2008 ) . Previous work has implicated LAMP2 in cholesterol export from lysosomes , as LAMP-deficient cells show cholesterol accumulation that can be rescued by LAMP2 expression ( Eskelinen et al . , 2004; Schneede et al . , 2011 ) . Proteome-wide analysis of cholesterol binding proteins included LAMP1 and LAMP2 among a long list of candidate proteins ( Hulce et al . , 2013 ) . Despite these hints , the precise function of LAMP proteins has remained unclear , and they are often presumed to be structural components . We show here that LAMP proteins bind cholesterol directly and this capacity contributes to their role in cholesterol export from lysosomes .
The structure of an individual LAMP domain from DC-LAMP protein is comprised of a novel , beta-prism fold that appears to contain a hydrophobic pocket ( Wilke et al . , 2012 ) ; we used this structure to model the structure of LAMP2 domain 1 ( Figure 6A ) . Site directed mutagenesis of hydrophobic residues predicted to line the walls of this cavity yielded purified LAMP2 proteins with impaired cholesterol binding activity . Thus , a soluble , LAMP2 domain 1-I111A/V114A construct yielded a secreted protein ( Figure 6B inset , right lane ) that bound significantly less cholesterol than its wild type counterpart ( Figure 6B inset , left lane and panel B ) . Because these proteins were obtained from cell secretions , they are likely to be properly folded , as they escaped the endoplasmic reticulum’s quality control machinery . These experiments show that residues facing the predicted , prism fold pocket are important for cholesterol binding and likely contribute to the cholesterol binding site . 10 . 7554/eLife . 21635 . 010Figure 6 . Cholesterol binding to LAMP2 domain 1 is required for its ability to rescue cholesterol export from LAMP-deficient lysosomes . ( A ) , predicted structure model of LAMP2 domain 1; residues I111 and V114 are highlighted in red . ( B ) Relative 3H-cholesterol binding to soluble LAMP 2 domain 1 or LAMP2 domain 1-I111A/V114A . Shown is combined data from 5 independent experiments carried out in duplicate in the presence of 50 nM 3H-cholesterol . Inset , SDS-PAGE analysis of wild type ( left ) and domain 1-I111A/V114A ( right ) proteins analyzed . P value was determined by two-tailed Student’s t-test . ( C ) flow cytometry analysis of mean fluorescence of GFP rescue constructs in lentivirus-tranduced cells ( >20 , 000 cells analyzed ) . ( D ) Cholesteryl oleate synthesis in MEF cells lacking LAMP1 and LAMP2 after rescue with either full length , membrane anchored LAMP2 , membrane anchored LAMP2 domain 1 , or membrane anchored LAMP2 domain 1-I111A/V114A . C-terminally GFP-tagged , rescue proteins were stably expressed using lentivirus transduction; shown is the combined result of 2 independent experiments , normalized for the amount of mature protein in each sample ( Figure 6—figure supplement 1 ) relative to the amount of rescue seen with full length LAMP2 protein . P-values are in relation to full length for domain 1 , or to domain 1 for the mutant protein , and were determined by one way ANOVA . E , confocal light microscopic analysis of GFP rescue construct localization ( green ) and endogenous LAMP1 protein ( red ) in transiently transfected HeLa cells; bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21635 . 01010 . 7554/eLife . 21635 . 011Figure 6—figure supplement 1 . Immunoblot analysis of LAMP2 constructs from lentivirus transduced LAMP1/LAMP2-knock out MEF cells . ( A ) total cell extract from cells expressing membrane anchored , GFP-LAMP2 domain 1 or GFP-LAMP2 domain 1-I111A/V114A; ( B ) migration of the constructs indicated , before or after 4 hr cycloheximide treatment ( 50 µg/ml ) . Panel A was developed as in Figure 4B; panel B was developed as in Figure 4A . DOI: http://dx . doi . org/10 . 7554/eLife . 21635 . 011 Finally , to verify the importance of cholesterol binding to LAMP2 protein as part of its physiological role , we tested the ability of wild type and mutant LAMP2 constructs to rescue the cholesterol accumulation seen in lysosomes from mouse embryonic fibroblasts missing LAMP1 and LAMP2 proteins ( Eskelinen et al . , 2004; Schneede et al . , 2011 ) . The ability of lysosomes to export cholesterol can be monitored by feeding cells cholesterol in the form of LDL , and using conversion of 14C-oleic acid to cholesteryl oleate that takes place after endocytosed cholesterol is transported to the endoplasmic reticulum ( Goldstein et al . , 1983 ) . Previous work showed that LAMP1/2 knockout MEF cells were impaired in cholesterol export using this assay ( Schneede et al . , 2011 ) . We used lentivirus transduction to test the ability of full length LAMP2 , a membrane anchored LAMP2 domain 1 ( LAMP2-GFP Δ194–368; Figure 1—figure supplement 1 ) , or a membrane anchored LAMP2 domain 1-I111A/V114A to rescue the ability of LAMP1/LAMP2 knockout MEF cells to export LDL-derived cholesterol from lysosomes . For these experiments , we used LAMP constructs containing a single LAMP domain , as full length LAMP2 constructs with mutations in both LAMP domains failed to fold properly or be transported efficiently to lysosomes . It was important to first verify the precise amounts of each construct in lysosomes , to evaluate any functional rescue findings . Flow cytometry analysis showed that the rescue constructs were expressed at comparable levels in each stably expressing cell population ( Figure 6C ) . Light microscopy confirmed that the constructs were capable of proper lysosome localization , as determined by their colocalization with endogenous LAMP1 protein ( Figure 6E ) in HeLa cells . ( Similar staining was observed in LAMP knockout MEF cells that lack LAMP protein markers ) . To fully confirm the folding of these artificial constructs , we analyzed their glycosylation status and stability after addition of cycloheximide to inhibit new protein synthesis ( Figure 6—figure supplement 1 ) . Full length GFP-LAMP2 protein migrated at ~140 kD and its abundance was not altered after 4 hr cycloheximide treatment , consistent with its long half life in cultured cells ( panel B ) . Similarly , the GFP-domain 1 construct was stable under these conditions and migrated at ~90 kD ( panels A , B ) . In contrast , the I111A/V114A mutant domain I protein displayed two distinct bands; the upper band was stable , while the lower band likely corresponded to an ER form that was largely degraded after 4 hr in cycloheximide ( panels A , B ) . From this we conclude that cells expressing membrane anchored LAMP2 domain 1 I111A/V114A are less efficient at folding the protein but some folded protein makes it to lysosomes , where it is stable . This difference was accounted for in subsequent functional rescue experiments ( Figure 6D ) . Figure 6D shows that as expected , full length , wild type LAMP2 rescued cholesterol export in LAMP1/2-deficient MEF cells; membrane anchored LAMP2 domain 1 showed a level of rescue consistent with its lower capacity for cholesterol binding ( cf . Figure 1F ) . Importantly , membrane anchored , LAMP2 domain 1 I111A/V114A failed to rescue cholesterol export from lysosomes ( Figure 6D ) , consistent with its inability to bind cholesterol; shown are the data corrected for the amount of mature proteins present in lysosomes in these cells . LAMP2 constructs mutated in both cholesterol-binding sites could not be tested , as they were only poorly delivered to lysosomes . These experiments demonstrate a direct role for LAMP2 in cholesterol export from lysosomes , and confirm that LAMP2’s ability to bind cholesterol correlates with its ability to support cholesterol export from LAMP-deficient MEF cells . In addition , LAMP proteins bind tightly to NPC proteins in vitro and in cells , and appear to facilitate cholesterol export from lysosomes . LAMP proteins are the most highly abundant membrane glycoproteins of the lysosome , and their lumenally oriented cholesterol binding sites represent a significant binding site for this important sterol . We measured ~7 × 106 molcules per HeLa cell , representing 0 . 3 mM binding sites in lysosomes . A recent cellular mass spectrometry analysis ( Itzhak et al . , 2016 ) estimated LAMP proteins to be present at 260 , 000 copies and NPC1 at 29 , 193 copies per HeLa cell . While the relative abundance of these proteins matches the values we report here , their total level was 25 fold lower in that study . It is possible that these transmembrane glycoproteins were under-represented in due to their unusual protease resistance as proteins of the lysosome membrane , differences in cell confluency and/or differences in HeLa cell lines employed . Lysosomes have recently been shown to sense and signal amino acid availability to influence lysosome biogenesis in relation to cellular need ( Settembre et al . , 2013 ) , and LAMP oligomerization has been reported to correlate with chaperone mediated autophagy ( Bandyopadhyay et al . , 2008 ) . Cholesterol levels may influence LAMP protein conformation or interaction with other partners to signal the availability of endocytosed cholesterol to influence autophagy and cellular metabolism . The ten fold higher abundance of LAMP proteins compared with NPC1 protein in HeLa cells suggests that LAMP proteins may do more than just facilitate NPC1 function in cholesterol export . Future experiments will be needed to fully understand the roles played by these highly abundant lysosomal membrane glycoproteins . We have shown that LAMP2 binds tightly to the N-terminal domain of NPC1 and also binds cholesterol with the same orientation as that domain . LAMP2 also aids in cholesterol export from lysosomes . How might LAMP2’s cholesterol binding site contribute to cholesterol export ? Current models suggest that the soluble NPC2 protein binds cholesterol from the internal membranes of lysosomes and delivers it to NPC1 at the limiting membrane of this compartment ( Kwon et al . , 2009 ) . One possibility is that NPC2 can deliver cholesterol to both NPC1 and to LAMP2 , which is more abundant . This would help drive the cholesterol export process by moving cholesterol from the accumulated , lumenal lipid stores to the lysosome’s limiting membrane . Because LAMP2 and NPC1 N-terminal domains bind cholesterol in the same orientation , it makes sense that NPC2 ( which binds in opposite orientation , [Xu et al . , 2007] ) could transfer the cholesterol between these two proteins . The recent crystal structure of NPC2 bound to the middle , lumenal domain of NPC1 ( Li et al . , 2016 ) supports a direct handoff between NPC1 and NPC2 ( Kwon et al . , 2009 ) . In future work , it will be important to elucidate precisely how LAMP2 interacts with both NPC2 and NPC1 to facilitate cholesterol export from lysosomes and how cholesterol binding contributes to LAMP2’s other cellular roles .
Buffer A: 50 mM ammonium acetate , pH4 . 5 , 150 mM NaCl , 0 . 004% NP-40; buffer B: 50 mM MES , pH5 . 5 , 150 mM NaCl , 0 . 004% NP40; buffer C: 50 mM MES , pH6 . 5 , 150 mM NaCl , 0 . 004% NP-40; buffer D: 50 mM HEPES , pH7 . 5 , 150 mM NaCl , 0 . 004% NP-40; buffer E: 25 mM Tris , pH7 . 4 , 150 mM NaCl; RIPA buffer: 50 mM Tris , pH7 . 4 , 150 mM NaCl , 1% NP-40 , 0 . 2% deoxycholic acid , 0 . 1% SDS . cDNAs encoding full length , soluble human LAMP1 ( 1–382 ) , human LAMP2 ( 1–375 ) and domain 1 of human LAMP2 ( 1–231 ) were PCR amplified from LAMP1-mGFP and pGEM-LAMP2 respectively . The PCR products were inserted into pEGFP-N3 vector . The constructs were assembled to have an unstructured GSTGSTGSTGA linker at the C terminus , followed by a His10 tag and a FLAG tag . For LAMP2 , another His10 tag was added downstream of the FLAG tag for improved purification . LAMP2 domain 2 was prepared by deleting residues 39–219 from the full length , soluble domain construct . FUGENE6 was used for transient transfection of HeLa cells . Membrane anchored rescue constructs were stably expressed in LAMP1/2 deficient MEF cells by lentivirus transduction and were comprised of full length LAMP2 bearing a C-terminal GFP ( LAMP2-GFP ) , or LAMP2-GFP Δ194–368 ( encoding membrane anchored domain 1 ) or the latter construct carrying point mutations . Authenticated HEK293F , HEK293T , and HeLa cells were from ATCC and used at low passage; Sf9 cells were purchased from Thermo Fisher Scientific ( Waltham , MA ) ; Mouse embryonic fibroblasts from LAMP1/LAMP2 double knockout mice ( Bandyopadhyay et al . , 2008; Eskelinen et al . , 2004 ) were the generous gift of Dr . Paul Saftig ( Christian-Albrechts-Universität Kiel , Germany ) . Mycoplasma contamination was monitored by DAPI staining . All cells were cultured at 37°C and under 5% CO2 in Dulbecco’s modified Eagle’s medium supplemented with 7 . 5% fetal bovine serum , 100 U/ml penicillin and 100 μg/ml streptomycin , unless indicated . HEK293F suspension cells were cultured at 37°C under 5% CO2 in Freestyle 293 medium . In some experiments , cells were cultured in lipoprotein deficient serum ( 5% ) . pFastBac NPC1-N-terminal domain plasmid was used to make virus for infection of Sf9 insect cells . 72 hr after infection , Sf9 cultures were spun down and ammonium sulfate added to achieve 60% saturation . The resulting precipitate was re-suspended buffer E and incubated with Ni-NTA resin overnight at 4°C . After washing with buffer E with 25 mM imidazole , the protein was eluted with buffer E plus 250 mM imidazole , and further purified using Q-Sepharose . HEK293F cells were transfected using 293fection according to the manufacturer . After 72 hr , supernatants were collected after spinning 3000 rpm for 5 min . To purify proteins for [3H] cholesterol binding , supernatants were subjected to 90% ammonium sulfate precipitation . After spinning at 13 , 000 rpm for 30 min , pellets were re-suspended in buffer E plus 25 mM imidazole and incubated with Ni-NTA resin overnight at 4°C , followed by washing with the same buffer . Bound proteins were eluted with buffer E plus 250 mM imidazole . Proteins were concentrated and buffer exchanged into buffer C with Pierce Protein Concentrators PES ( 10 kD cut-off ) . Proteins were either used immediately or stored at −80°C after snap freezing in liquid nitrogen . For cholesterol extraction and thin layer chromatography , supernatants were adjusted to pH 7 . 4 and incubated with Ni-NTA resin overnight at 4°C; after washing with buffer E plus 25 mM imidazole , proteins were eluted with buffer E plus 250 mM imidazole . Proteins were desalted into PBS using a PD-10 column . Each reaction was carried out in a final volume of 80–100 µl of buffer A , B , C or D containing 0 . 1–1 µg purified His-tagged protein , 1µg BSA and 10–400 nM 3H-cholesterol diluted with 0 . 1–50 µM cholesterol . For competition assays , reactions were in 80 µl buffer C ( 50 mM MES , 150 mM NaCl , 0 . 004% NP-40 , pH6 . 5 ) with 0 . 1 µg full length soluble LAMP2 protein and 50 nM 3H-cholesterol , competition was started by adding vehicle ( ethanol ) or different concentrations of competitors as indicated . After incubation overnight at 4°C , the mixture was loaded onto a column packed with 30 µl Ni-NTA agarose beads . After incubation for 10 min , each column was washed with 5 ml of buffer C plus 10 mM imidazole . The protein-bound 3H-cholesterol was eluted with 250 mM imidazole-containing buffer C and quantified by scintillation counting . Samples were analyzed by LC/MS on an Agilent 1260 HPLC and Bruker microTOF-Q II mass spectrometer . Full scan mass and product ion spectra were acquired in positive ion mode , using a Phenomenex Kinetex C18 2 . 6u 2 . 1 × 100 mm column , and an initial condition of 30% , 0 . 1% formic acid in water/70% methanol . Full length soluble LAMP2 , and domains 1 and 2 of LAMP2 were purified as described above . Extraction was performed by adding 3 sample volumes of chloroform/methanol ( 2:1 , v/v ) to the samples . After repeating once more , extracts were pooled and dried under nitrogen . The extracts were re-dissolved in 50–100µl chloroform/methanol ( 2:1 , v/v ) . Samples were spotted onto a Silica gel plate . The plate was developed with isopropanol until the front reached 1cm above the loading position; after drying under airflow , the plate was further developed using 2% methanol in chloroform until the front reached the top of the plate . The plate was sprayed with 10% CuSO4 in 4% or 8% phosphoric acid and heated at 180°C to visualize the samples . HEK293T cells expressing pEGFP-N1 , pEGFP-N1-mNPC1 or pEGFP-C3-MCOLN1 were harvested 24–48 hr post-transfection and lysed in lysis buffer ( 50 mM MES , pH 5 . 5 , 150 mM NaCl and 0 . 1% digitonin ) supplemented with protease inhibitors . After 30 min on ice , lysates were spun at 15 , 000 g for 15 min , and protein concentrations of the supernatants were measured . Equal amounts of extract protein were incubated with GFP-binding protein–conjugated agarose for 2 hr at 4°C . Immobilized proteins were washed 4 times with 1ml lysis buffer , eluted with 2× SDS loading buffer , and subjected to BioRad Mini-PROTEIN TGX 4–20% gradient gels . After transfer to nitrocellulose membrane and antibody incubation , blots were detected with ECL western blotting detection substrate or visualized using LI-COR Odyssey Imaging System . MST experiments were performed on a Monolith NT . 115Pico instrument ( Nanotemper Technologies ) . Briefly , His6-NPC1 N-terminal domain , RNase B or NPC2 were labeled using the RED-NHS ( Amine Reactive ) Protein Labeling Kit ( Nanotemper Technologies ) . A constant concentration of 6 nM labeled protein was mixed with binding partnerswith a final buffer condition of 50 mM MES , pH 5 . 5 , containing 150 mM NaCl , 0 . 004% NP-40 . Premium coated capillaries contained 16 sequential , 2 fold serial dilutions . Analysis was at 40% laser power for 30 s , followed by 5 s cooling . Data were normalized to fraction of bound ( 0 = unbound , 1 = bound ) . The dissociation constant KD was obtained by plotting the normalized fluorescence Fnorm against the logarithm of the different concentrations of the dilution series according to the law of mass action . HeLa and HEK293T cells were grown to sub-confluence in DMEM supplemented with 7 . 5% FBS . One 10cm dish of cells was washed 3 times with cold PBS , then lysed with 500µl RIPA buffer with protease inhibitor cocktail ( Sigma ) . After 30 min on ice , the lysate was centrifuged at 13 , 000 rpm for 15 min at 4°C . The resulting supernatant was transferred to a new tube and protein was measured by BCA assay . Lysates were resolved by SDS-PAGE , using different amounts of purified human LAMP2 or NPC1 protein as standards . After transfer to nitrocellulose , the blot was probed with anti-human LAMP2 or NPC1 antibody followed by IRDye 800CW labeled anti mouse ( for LAMP2 ) or rabbit ( for NPC1 ) secondary antibody , and visualized using a LI-COR Odyssey Imaging System and analyzed using ImageJ software . Calculations were based on molecular weights of 45 , 874 for LAMP2 and 142167 for NPC1 polypeptide chains , and neglected glycan contribution , which is not measured in the protein assay employed . Purified , full length NPC1 protein was the gift of Dr . Xiaochun Li ( Rockefeller University ) and was N-glycanase treated . Confocal immunofluorescence microscopy was carried out as described ( Li et al . , 2015 ) . Cells grown on coverslips were fixed with 3 . 7% ( vol/vol ) paraformaldehyde for 15 min at room temperature . LAMP1 staining was performed with sequential incubation of mouse anti-LAMP1 culture supernate and Alexa Fluor 594 goat anti-mouse antibody ( 1:1 , 000 , Invitrogen ) , each for 1 hr at room temperature . Coverslips were mounted using Mowiol and imaged using a Leica SP2 confocal microscope and Leica software with a 60 × 1 . 4 N . A . Plan Apochromat oil immersion lens and a charge-coupled device camera ( CoolSNAP HQ , Photometrics ) . Flow cytometry was carried out on a FACScan Analyzer on gently trypsinized cells fixed as described above ( Li et al . , 2015 ) . Structures were presented in drawings created using Chimera software ( Pettersen et al . , 2004 ) . LAMP1/LAMP2-deficient MEF cells were cultured in DMEM medium with 5% ( vol/vol ) LPDS for two days and assayed ( Goldstein et al . , 1983; Li et al . , 2015 ) with minor modification . After 48 hr , 100 µg/mL LDL , 50 µM lovastatin , and 50 µM sodium mevalonate were added for 5 hr . Cells were pulse labeled for 4 hr with 0 . 1 mM sodium [1-14C]oleate ( American Radiolabeled Chemicals ) –albumin complex . Cells were washed two times with 2 mL 50 mM Tris , 150 mM NaCl , 2 mg/mL BSA , pH 7 . 4 , followed by 2 mL 50 mM Tris , 150 mM NaCl , pH 7 . 4 . Cells were extracted and rinsed with hexane-isopropanol ( 3:2 ) , pooled , and evaporated . After resuspending each sample in 60 µl hexane , 4 µL of lipid standard containing 8 µg/mL triolein , 8 µg/mL oleic acid , and 8 µg/mL cholesteryl oleate was added . Samples were spotted onto a silica gel 60 plastic backed , thin layer chromatogram and developed in hexane . Cholesteryl oleate was identified with iodine vapor , scraped from chromatograms , and radioactivity determined by scintillation counting in 10 mL Biosafe II . Minimum sample sizes were determined assuming 5% standard error and >95% confidence level . p values were determined using Graphpad Prism software and are indicated in all figures according to convention: *p≤0 . 05; **p≤0 . 01; ***p≤0 . 001; ****p≤0 . 0001 . Error bars represent standard error of the mean . | Living cells contain many membrane-bound compartments surrounded by a gel-like substance called the cytoplasm . Lysosomes are compartments found in most animal cells , which contain enzymes that can break down virtually all kinds of biological molecules . Cell biologists around the world use two proteins called LAMP1 and LAMP2 to mark lysosomes to study them . The loss of LAMP2 causes a condition called Danon disease that is characterized by thickening of the heart muscle . However , relatively little is known about what these proteins actually do . Previous studies had hinted that these proteins might bind to the fatty molecule , cholesterol . Li and Pfeffer set out to test this directly and showed that LAMP1 and LAMP2 proteins do indeed bind to cholesterol . The two LAMP proteins also interact with another two proteins , called NPC1 and NPC2 , which export cholesterol out of lysosomes . Li and Pfeffer then showed that cells contain 5- to 10-times more LAMP proteins than they do NPC1-cholesterol exporters . This suggests that LAMP proteins have additional roles that need to be characterized and studied to see how important cholesterol binding is for these processes too . Future studies could also explore if LAMP proteins signal that free cholesterol is available for the cell’s needs . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology",
"short",
"report"
] | 2016 | Lysosomal membrane glycoproteins bind cholesterol and contribute to lysosomal cholesterol export |
Ribosomes can stall during translation due to defects in the mRNA template or translation machinery , leading to the production of incomplete proteins . The Ribosome-associated Quality control Complex ( RQC ) engages stalled ribosomes and targets nascent polypeptides for proteasomal degradation . However , how each RQC component contributes to this process remains unclear . Here we demonstrate that key RQC activities—Ltn1p-dependent ubiquitination and Rqc2p-mediated Carboxy-terminal Alanine and Threonine ( CAT ) tail elongation—can be recapitulated in vitro with a yeast cell-free system . Using this approach , we determined that CAT tailing is mechanistically distinct from canonical translation , that Ltn1p-mediated ubiquitination depends on the poorly characterized RQC component Rqc1p , and that the process of CAT tailing enables robust ubiquitination of the nascent polypeptide . These findings establish a novel system to study the RQC and provide a framework for understanding how RQC factors coordinate their activities to facilitate clearance of incompletely synthesized proteins .
Eukaryotic cells contain several cotranslational quality-control pathways that limit the production of aberrant proteins and thereby maintain protein homeostasis . One such pathway is activated when a ribosome fails to complete translation , leading to the recruitment of specialized factors that disassemble the stalled ribosome and facilitate degradation of the nascent protein ( Brandman and Hegde , 2016; Shoemaker and Green , 2012 ) . A key effector of this process is the highly conserved Ribosome-associated Quality control Complex ( RQC ) , which in budding yeast comprises the E3 ubiquitin ligase Ltn1p , the ATPase Cdc48p , and the poorly characterized proteins Rqc1p and Rqc2p ( Brandman et al . , 2012; Defenouillère et al . , 2013; Verma et al . , 2013 ) —the human homologs of which are Listerin , VCP/p97 , TCF25 , and NEMF , respectively . The stalled translation complex is first separated into subunits by ribosome splitting factors , allowing the small ribosomal subunit ( 40S ) and mRNA to be released . The RQC then recognizes and assembles on the large ribosomal subunit ( 60S ) that still contains a nascent polypeptide linked to a tRNA molecule ( 60S:peptidyl–tRNA ) . Ltn1p facilitates ubiquitination of the nascent chain while on the 60S subunit , marking the incompletely synthesized protein for proteasomal degradation ( Bengtson and Joazeiro , 2010; Shao et al . , 2013 ) . Additionally , Rqc2p recruits charged tRNAs to the 60S subunit to direct elongation of the nascent protein with a Carboxy-terminal Alanine and Threonine extension , or CAT tail ( Shen et al . , 2015 ) . Structural analysis of the yeast RQC , identification of the tRNA molecules that co-purify with the RQC , and biochemical characterization of failed nascent chains suggested that CAT tailing occurs on the 60S subunit by a unique mechanism that does not require an mRNA template or the 40S subunit ( Shen et al . , 2015 ) . However , many questions about the mechanism of CAT-tail synthesis and the consequences of elongating nascent polypeptides with CAT tails remain unanswered . Recent studies have suggested that one function of CAT tails is to facilitate aggregation of nascent polypeptides that fail to be ubiquitinated by Ltn1p ( due to either disruptions in LTN1 or the absence of a suitable ubiquitin acceptor ) . CAT tail-driven aggregation may limit the otherwise toxic effects of incomplete translation products accumulating in the cytoplasm ( Choe et al . , 2016; Defenouillère et al . , 2016; Yonashiro et al . , 2016 ) . However , our understanding of the functions of CAT tails in the context of an intact RQC or of the process of CAT tailing itself remains incomplete . Previous studies have analyzed the RQC in vitro by using cell-free translation systems based on rabbit reticulocyte lysates ( Shao et al . , 2013 ) or Neurospora crassa extracts ( Doamekpor et al . , 2016 ) . In the presence of a suitable mRNA substrate , both cell-free systems recapitulate Ltn1p-dependent ubiquitination and thereby provide valuable insight into the mechanism by which Ltn1p orthologs discriminate between elongating and stalled ribosomes ( Shao et al . , 2013 ) and the role of the N-terminal domain of Ltn1p in binding the 60S subunit ( Doamekpor et al . , 2016 ) . However , neither system recapitulates Rcq2p-dependent CAT tailing , leaving important unanswered questions about how CAT tails are synthesized and whether the two principal activities of the RQC—ubiquitination by Ltn1p and CAT tailing by Rqc2p—are functionally related . Although many studies have identified Rqc1p/TCF25 as a core component of the yeast and mammalian RQC required for nascent-chain degradation ( Brandman et al . , 2012; Defenouillère et al . , 2013; Shao and Hegde , 2014 ) , Rqc1p’s precise structural and functional roles in the complex remain unclear . Previous work in yeast suggested that Rqc1p acts after Ltn1p to promote nascent-chain degradation . This hypothesis emerged from two lines of evidence: The presence of polyubiquitinated proteins in purified RQC depends on Ltn1p ( and to a lesser extent on Rqc2p ) but not on Rqc1p or Cdc48p ( Brandman and Hegde , 2016; Brandman et al . , 2012 ) ; and recruitment of Cdc48p to the 60S subunit requires Rqc1p and nascent-chain ubiquitination ( Defenouillère et al . , 2013 ) . However , these studies did not determine whether Rqc1p is necessary for ubiquitination of the nascent chain itself or whether recruitment of Cdc48p requires a direct interaction with Rqc1p . Therefore , the mechanism by which Rqc1p promotes nascent-chain degradation in vivo has remained unclear . In this study , we provide an in vitro characterization of the RQC in a budding-yeast extract that uniquely recapitulates ubiquitination by Ltn1p and CAT tailing by Rqc2p , providing new insights into RQC action in promoting degradation of stalled translation products .
Because CAT tails have thus far only been observed in S . cerevisiae ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) , we used S . cerevisiae extracts to recapitulate Rqc2p-mediated elongation in vitro . Although S . cerevisiae has long been used for in vitro translation ( Hussain and Leibowitz , 1986; Iizuka et al . , 1994; Rojas-Duran and Gilbert , 2012; Tarun and Sachs , 1995 ) , these reactions are notoriously inefficient . Further exacerbating this problem , we aimed to program these reactions with truncated mRNA substrates that trigger quality control , which are translated less efficiently because they lack poly ( A ) tails that normally enhance translation . Thus , we found it necessary to first establish an optimized protocol that could reproducibly generate translation products that were detectable by immunoblotting ( see Materials and methods ) . Critical aspects of our protocol included: ( 1 ) lysing cells with a freezer mill under cryogenic conditions rather than by bead beating in the cold; ( 2 ) minimizing the number of lysis cycles; ( 3 ) removing small molecules by dialysis rather than by size-exclusion chromatography; and ( 4 ) programming translation reactions with an mRNA encoding a small protein ( i . e . , 23 kDa NanoLuc luciferase ) , which is translated more efficiently than an mRNA encoding a larger protein ( e . g . , 62 kDa firefly luciferase ) . To produce a substrate for the RQC , we used a truncated reporter mRNA that terminates with a sense codon ( i . e . , does not contain a stop codon , 3′–untranslated region ( 3′-UTR ) , or poly ( A ) tail ) . This type of mRNA substrate has been shown previously to generate a ribosome–nascent chain complex stalled at the 3′ end of the message ( Becker et al . , 2011; Shao et al . , 2013 ) . Our reporter mRNA encodes a NanoLuc luciferase ( NL ) protein in which the seven native lysine residues have been mutated to arginine to avoid potential confounding effects of lysine ubiquitination . The protein also includes an N-terminal 3xHA tag ( which is naturally devoid of lysines ) to allow detection of translation products by immunoblotting . Programming S . cerevisiae in vitro translation ( ScIVT ) reactions with a full-length control mRNA that contains a stop codon and 3′-UTR resulted in the time-dependent accumulation of a 23 kDa product corresponding to 3xHA-NL ( Figure 1A , left ) . In contrast , ScIVT of a truncated mRNA initially produced a ~43 kDa mass-shifted product not observed in control reactions ( Figure 1A , right ) , which we hypothesized corresponded to a peptidyl–tRNA intermediate . Remarkably , as the reaction proceeded we observed the disappearance of the initial ~43 kDa product and concomitant accumulation of smaller mass-shifted products ranging from 23 kDa to 43 kDa ( Figure 1A , right ) . These mass-shifted products were reminiscent of CAT-tailed species previously observed in vivo ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) . 10 . 7554/eLife . 27949 . 003Figure 1 . An S . cerevisiae in vitro translation system recapitulates synthesis of Rqc2p-dependent polypeptide extensions . ( A ) Time courses of S . cerevisiae in vitro translation ( ScIVT ) reactions . ScIVT reactions were prepared using wild-type ( WT ) extracts and 1 μg of either a full-length ( left; includes a stop codon and 3′-UTR ) or truncated ( right; encodes a terminal valine residue ) mRNA encoding lysine-free 3xHA-NanoLuc ( 3xHA-NL ) . At the indicated time points , aliquots of the reactions were quenched in 2X Laemmli Sample Buffer . Proteins were separated by SDS-PAGE , and HA-tagged translation products were visualized by immunoblotting . ( B ) Analyses of mass-shifted products . An ScIVT reaction was prepared using WT extracts and a lysine-free truncated mRNA substrate that also encodes a TEV cleavage site ( TCS ) . Translation was halted after 15 or 60 min by addition of 20 mM EDTA , after which reactions were treated without ( – ) or with ( + ) TEV and/or RNase A/T1 cocktail for 60 min . Translation products were analyzed by immunoblotting as in ( A ) . ‘Long’ and ‘Short’ refer to exposure times of the blots . ( C–E ) Genetic analysis of mass-shifted products . ScIVT reactions were prepared using extracts from strains of the indicated genotypes and a lysine-free truncated mRNA substrate . Reactions were performed and analyzed as in ( A ) but with less mRNA ( 480 ng ) . The species that migrate just below the peptidyl–tRNA in rqc2Δ extracts in ( D ) represent peptidyl–tRNA degradation products that arise due to prolonged incubation in the absence of Rqc2p . ( F ) Rescuing Rqc2p deficiency in vitro . ScIVT reactions were prepared using rqc2Δ extracts and a lysine-free truncated mRNA substrate . After 30 min of translation , reactions were supplemented with either protein storage buffer ( – ) or purified Rqc2p ( WT or CAT-tailing deficient D98A at 420 nM final concentration ) and indicated time points were analyzed by SDS-PAGE and immunoblotting . Dashed lines indicate where intervening lanes were removed for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 00310 . 7554/eLife . 27949 . 004Figure 1—figure supplement 1 . Purified wild-type and mutant Rqc2p . C-terminal polyhistidine-tagged Rqc2p ( WT and D98A mutant ) were purified as described in Materials and methods and analyzed by SDS-PAGE and Coomassie staining . 1/10th volume of the corresponding purified protein was added to ScIVT reactions . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 004 To further characterize the mass-shifted species , we added a sequence encoding a TEV protease cleavage site ( TCS ) at the 3′ end of our mRNA substrate ( Shen et al . , 2015 ) and determined the susceptibility of the mass-shifted species to TEV protease and RNase . We reasoned that a nascent polypeptide with its C-terminus covalently linked to a tRNA molecule would be liberated by either TEV protease or RNase . In contrast , a nascent polypeptide containing an untemplated C-terminal amino-acid extension ( e . g . , a CAT tail ) would be cleaved by TEV protease but not RNase , and a protein containing additional mass due to modifications anywhere except the C-terminus would be unaffected by either treatment . When treated with either RNase or TEV protease , the ~45 kDa intermediate observed at early time points was converted to a 25 kDa species , corresponding to the molecular weight of 3xHA-NL-TCS ( Figure 1B ) . Given that the average molecular weight of a tRNA is ~20 kDa , these results suggest that the ~45 kDa species contained a tRNA covalently linked to the C-terminus of 3xHA-NL-TCS ( ‘peptidyl–tRNA’ ) —which has previously been characterized as an intermediate of the quality-control pathway ( Shao et al . , 2013; Shoemaker et al . , 2010; Tsuboi et al . , 2012 ) . In contrast , the heterogeneous collection of ~25–45 kDa products that accumulated at later time points were only affected by TEV protease , converting them to a discrete 25 kDa species ( Figure 1B ) , as expected if the products originally contained additional mass downstream of the TCS ( i . e . , appended to the C-terminus of 3xHA-NL-TCS ) . Because the truncated mRNA substrate used for ScIVT contained no sequences downstream of the TCS , this additional mass was necessarily untemplated and therefore consistent with CAT tails . Notably , the prominent peptidyl–tRNA species that accumulated at early time points was largely absent after 60 min of translation ( Figure 1A and B ) , presumably due to peptidyl–tRNA hydrolysis that occurred after untemplated elongation of the nascent chain . In addition to being C-terminal and untemplated , another known feature of CAT tails is that their synthesis is strictly dependent on Rqc2p but not on Ltn1p or Rqc1p ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) . To determine if the ~23–43 kDa mass-shifted species share this property , we took advantage of the genetic tractability of S . cerevisiae and non-essential nature of the RQC by performing ScIVT using extracts prepared from ltn1Δ , rqc1Δ , and rqc2Δ strains . While reactions using ltn1Δ and rqc1Δ extracts yielded all of the mass-shifted products observed when using wild-type ( WT ) extracts ( Figure 1C ) , reactions using rqc2Δ extracts did not ( Figure 1D ) , consistent with those species corresponding to CAT-tailed protein . However , rather than producing the expected 23 kDa 3xHA-NL protein , reactions lacking Rqc2p generated a relatively stable 43 kDa peptidyl–tRNA species , indicating a defect in peptidyl–tRNA hydrolysis . This finding suggests that in addition to facilitating the incorporation of untemplated amino acids , Rqc2p may also be involved in promoting hydrolysis of the final peptidyl–tRNA bond and thereby liberating the nascent polypeptide . Previous biochemical and structural studies have suggested that Rqc2p engagement and subsequent CAT tailing must be preceded by ribosome splitting , which exposes the P-site tRNA and the surface of the Sarcin-Ricin loop ( SRL ) that are recognized by Rqc2p/NEMF ( Lyumkis et al . , 2014; Shao et al . , 2015; Shen et al . , 2015 ) . In the case of truncated mRNAs that generate a ribosome stalled at the mRNA 3′ end with an empty A site , ribosome splitting is effected by the release-factor mimics Hbs1p and Dom34p ( Shao et al . , 2013; Shoemaker et al . , 2010 ) . Accordingly , ScIVT of a truncated mRNA in hbs1Δ or dom34Δ extracts generated only the ~43 kDa product corresponding to an especially stable peptidyl–tRNA ( Figure 1E ) , which is presumably protected within a stalled but intact 80S ribosome . Collectively , these data demonstrate that ScIVT of a truncated mRNA generates polypeptides containing untemplated Rqc2p-dependent C-terminal extensions . Although we have not been able to confirm that these extensions are composed of alanine and threonine residues ( for technical reasons ) , we suspect that this is the case and therefore refer to the extensions as CAT tails for simplicity . The lack of CAT-tailing activity in rqc2Δ extracts ( Figure 1D ) is consistent with previous observations that CAT tails are absent from rqc2Δ strains in vivo ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) , which could reflect either a direct role for Rqc2p in CAT tailing ( as suggested by structural studies ) or indirect effects of RQC2 disruption on CAT tailing . To distinguish between these possibilities , we tested whether the absence of CAT-tailing activity in rqc2Δ extracts could be rescued by adding purified Rqc2p to ScIVT reactions already in progress . Remarkably , the addition of exogenous Rqc2p ( Figure 1—figure supplement 1 ) to rqc2Δ extracts restored both CAT-tail synthesis and peptidyl–tRNA hydrolysis , whereas the addition of a CAT-tailing-deficient Rqc2p mutant containing the D98A substitution ( Shen et al . , 2015; Yonashiro et al . , 2016 ) did not promote either CAT-tail synthesis or robust peptidyl–tRNA hydrolysis ( Figure 1F ) . These results provide direct evidence that Rqc2p is biochemically required for CAT tailing , consistent with its proposed role in recruiting alanine- and threonine-charged tRNAs to the 60S subunit ( Shen et al . , 2015 ) . Also , our ability to temporally separate CAT tailing from canonical translation in vitro ( by the addition of exogenous Rqc2p to rqc2Δ extracts ) provided an experimental strategy for specifically testing the requirements of CAT-tail elongation . Though previous structural studies of the RQC were instrumental in discovering CAT tailing and suggested a direct role for Rqc2p in the process ( Shen et al . , 2015 ) , the mechanism of CAT tailing has only been inferred from these data and otherwise remains poorly characterized . In particular , no published studies have investigated the extent to which CAT tailing by the 60S subunit is mechanistically similar to canonical elongation by the 80S ribosome . To address this question , we sought to examine the sensitivity of in vitro CAT tailing to a collection of well-characterized chemical inhibitors that target different sites of the ribosome or elongation factors ( Figure 2A ) and thereby interfere with canonical translation . Importantly , because canonical translation is required to generate the substrate for the RQC ( i . e . , a stalled ribosome at the post-splitting stage ) , it was necessary to temporally separate canonical translation from CAT tailing to isolate the effects of these inhibitors on the latter reaction . To do so , we first translated a truncated mRNA in rqc2Δ extracts in the absence of any inhibitors for 20 min to generate a pool of RQC substrate ( 60S:peptidyl–tRNA ) . We then supplemented the reactions with one of the inhibitors ( at a concentration that completely inhibited canonical translation in vitro; Figure 2—figure supplement 1A ) and purified Rqc2p ( Figure 2—figure supplement 1B ) ( or a buffer-only control ) to initiate CAT-tail synthesis . Finally , we allowed the reactions to proceed for an additional 40 min before analyzing the products by immunoblotting . 10 . 7554/eLife . 27949 . 005Figure 2 . CAT-tail synthesis is mechanistically distinct from canonical translation . ( A ) Schematics of small-molecule inhibitors that directly bind the ribosome ( top ) or that target the translation elongation factors eEF1a or eEF2 ( bottom ) . Inhibitors: ( A ) anisomycin; ( C ) cycloheximide; ( D2A ) didemnin 2A; ( DB ) didemnin B; ( E ) emetine; ( G ) G418; ( H ) hydrolyzable GTP; ( NH ) non-hydrolyzable GTP-analog GMP-PCP; ( P ) puromycin; ( S ) sordarin . ( * ) Denotes the peptidyl-transferase center of the 60S subunit . ( B–D ) Effects of small-molecule inhibitors on CAT tailing . ScIVT reactions were prepared using rqc2Δ extracts and a lysine-free truncated mRNA substrate . After 0 min ( t = 0 ) or 20 min ( t = 20 ) of translation , reactions were supplemented with either protein storage buffer ( – ) or purified Rqc2p at 670 nM final concentration ( + ) and the indicated inhibitor ( s ) . Indicated time points ( ‘Time ( min ) ' ) were analyzed by SDS-PAGE and immunoblotting . Additional t = 0 controls for the remaining inhibitors can be found in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 00510 . 7554/eLife . 27949 . 006Figure 2—figure supplement 1 . Inhibitors and purified Rqc2p used to dissect the mechanism of CAT-tail synthesis . ( A ) Effects of small-molecule inhibitors on translation . ScIVT reactions were prepared using WT extracts and a truncated mRNA substrate , supplemented without ( – ) or with ( + ) the indicated inhibitors after 0 min ( t = 0 ) of translation , and analyzed by SDS-PAGE and immunoblotting . Inhibitors: ( A ) anisomycin; ( C ) cycloheximide; ( D2A ) didemnin 2A; ( DB ) didemnin B; ( E ) emetine; ( G ) G418; ( S ) sordarin . Dashed lines indicate where intervening lanes were removed for clarity . ( B ) C-terminal polyhistidine-tagged Rqc2p was purified as described in Materials and methods and analyzed by SDS-PAGE and Coomassie staining . 1/10th volume of purified protein was added to ScIVT reactions . Note that this stock of protein was expressed and purified independently from the stock in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 006 As predicted by the structural analyses , treatment with drugs that target the peptidyl-transferase center ( PTC ) of the 60S subunit ( anisomycin and the chain terminator puromycin ) completely prevented CAT-tail synthesis , providing direct evidence that the catalytic activity of the ribosome is required for CAT tailing ( Figure 2B and C ) . Puromycin treatment also resulted in a complete collapse of the 43 kDa mass-shifted species to 23 kDa , confirming the identity of this larger species as peptidyl–tRNA ( Figure 2B ) . In contrast to the dramatic effects of PTC inhibitors , inhibitors that target the 40S subunit ( emetine and G418 ) did not inhibit CAT-tail synthesis ( Figure 2B ) , consistent with the proposed 40S subunit–independent mechanism of CAT tailing . Surprisingly , cycloheximide—which binds in the E site of the 60S subunit and sterically clashes with the 3′ end of the deacylated tRNA during canonical translation—did not inhibit CAT tailing ( Figure 2B ) . Cycloheximide insensitivity identifies an unanticipated feature of CAT-tail elongation that may reflect a distinct mechanism of deacylated-tRNA displacement in the absence of mRNA and the 40S subunit . It was previously proposed that specific Rqc2p–tRNA interactions impart selectivity for alanine- and threonine-tRNAs to CAT-tail elongation ( Shen et al . , 2015 ) . However , it is not known if the translation elongation factors eEF1a and eEF2—which deliver aminoacyl-tRNAs to the ribosome and promote translocation , respectively—collaborate with Rqc2p to facilitate CAT-tail synthesis . Strikingly , we found that drugs targeting either eEF1a or eEF2 ( didemnin variants ( Carelli et al . , 2015 ) or sordarin ( Justice et al . , 1998 ) , respectively ) had no effect on CAT tailing ( Figure 2C ) . Because many canonical translation factors are GTPases , including eEF1a and eEF2 , we also examined whether CAT tailing requires GTP hydrolysis . We tested for inhibition by the non-hydrolyzable GTP analog GMP-PCP , using a similar approach as before except that at the time of Rqc2p addition we stopped translation by adding emetine to prevent ongoing translation in the GTP control reaction . Consistent with the above differences between CAT-tail elongation and translation , treatment with GMP-PCP had no impact on CAT tailing ( Figure 2D ) , indicating that this mRNA-independent elongation mechanism does not require energy from GTP hydrolysis or the canonical activities of the translational GTPases eEF1a and eEF2 . Collectively , these findings provide direct evidence that CAT tailing is a 40S subunit–independent , PTC-catalyzed reaction and identify key differences from canonical translation that suggest an entirely different elongation cycle . The ability of ScIVT to recapitulate Rqc2p-dependent CAT tailing ( Figure 1 ) led us to explore whether this system also recapitulates the other key activity of the RQC , Ltn1p-dependent nascent-chain ubiquitination . Initial experiments comparing a lysine-containing truncated reporter mRNA to a lysine-free version revealed a faint smear of high-molecular-weight ( HMW ) products ( ~50–115 kDa ) unique to the lysine-containing reporter ( Figure 3A , compare lanes 1 and 4 ) . We reasoned that these HMW products were likely ubiquitinated proteins because lysine residues are the canonical ubiquitination sites ( Pickart , 2001 ) . However , because ScIVT extracts contain many ubiquitin ligase activities and a finite pool of endogenous ubiquitin , we suspected that Ltn1p-dependent ubiquitination of the reporter protein ( which occurs late in the reactions ) might have been limited by the amount of ubiquitin available to Ltn1p . Indeed , supplementing ScIVT reactions with exogenous ubiquitin resulted in enhanced accumulation of the lysine-dependent HMW products ( Figure 3A ) . Treatment of ubiquitin-supplemented reactions with TEV protease or RNase did not fully collapse the HMW species as it did for CAT-tailed species ( Figure 3B ) , consistent with the HMW species containing ubiquitin-modified residues rather than simply having exceptionally long CAT tails . To directly demonstrate that these HMW species contained ubiquitin , we translated truncated mRNAs ( with or without lysines ) in the presence of exogenous Myc-tagged ubiquitin and purified the reporter protein under denaturing conditions , followed by immunoblotting to detect Myc-tagged ubiquitin . In reactions conducted with WT extracts , we readily detected Myc-tagged ubiquitin within the purified HMW species ( Figure 3C , fourth lanes in left and right panels ) . 10 . 7554/eLife . 27949 . 007Figure 3 . S . cerevisiae in vitro translation recapitulates Ltn1p-mediated ubiquitination . ( A ) Effects of adding exogenous ubiquitin to ScIVT reactions . ScIVT reactions conducted in WT extracts with lysine-containing ( +Lys ) or lysine-free ( –Lys ) truncated mRNA were supplemented with the indicated concentrations of recombinant ubiquitin , incubated for 60 min , and then analyzed by SDS-PAGE and immunoblotting . Dashed line indicates where intervening lanes were removed for clarity . ( B ) Analysis of high-molecular-weight smears . RNase A/T1 and TEV protease treatment of ScIVT reactions programmed with lysine-containing truncated mRNA encoding a TEV cleavage site ( TCS ) in WT extracts supplemented with 100 μM recombinant ubiquitin . Translation was halted after 60 min by addition of 20 mM EDTA , after which reactions were treated without ( – ) or with ( + ) TEV and/or RNase A/T1 for 60 min , and then analyzed by SDS-PAGE and immunoblotting . Note that due to long incubations ( 120 mins ) , very little peptidyl–tRNA persists in these reactions . ( C ) Isolation and detection of ubiquitinated ScIVT products . ScIVT reactions were conducted with 1 . 2 μg of truncated mRNA ( 3xHA-10xHis-NanoLuc , with or without lysines or His tag as indicated ) in extracts prepared from strains of the indicated genotypes and supplemented with 10 μM recombinant Myc-ubiquitin . For input samples ( bottom panels ) , one-third of the ScIVT reaction was quenched with 2X Laemmli Sample Buffer . For Ni-NTA-purified samples ( top panels ) , two-thirds of the ScIVT reaction was quenched with 6 M guanidine-HCl . For SDS-PAGE , 30% of input samples and 100% of Ni-NTA-purified samples were separated on 12% NuPAGE gels and translation products were visualized by immunoblotting with antibodies indicated at left . Dashed lines indicate where intervening lanes were removed for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 007 As expected , we did not observe ubiquitination of the reporter protein in reactions performed with ltn1Δ extracts ( Figure 3C and Figure 4A ) , consistent with Ltn1p being the responsible E3 ubiquitin ligase . The addition of purified Ltn1p to ltn1Δ extracts restored ubiquitination , while the addition of purified Ltn1p containing the W1542E substitution—a RING domain mutant that does not support protein turnover in vivo ( Bengtson and Joazeiro , 2010 ) —did not ( Figure 4A and Figure 4—figure supplement 1 ) . These results demonstrate that the lack of ubiquitination in ltn1Δ extracts is a direct consequence of the absence of Ltn1p rather than an indirect effect . Collectively , these results demonstrate that in addition to Rqc2p-dependent CAT tailing the ScIVT system we established also recapitulates Ltn1p-dependent ubiquitination , as previously shown for lysates derived from N . crassa and rabbit reticulocytes ( Doamekpor et al . , 2016; Shao et al . , 2013 ) . 10 . 7554/eLife . 27949 . 008Figure 4 . Rqc1p and CAT tailing contribute to Ltn1p-dependent ubiquitination . ( A–C ) Genetic analysis of RQC-mediated ubiquitination in ScIVT . ScIVT reactions were prepared using extracts from strains of the indicated genotype , a lysine-containing truncated mRNA substrate , ubiquitin storage buffer ( – ) or 100 μM recombinant ubiquitin ( + ) , and either protein storage buffer ( – ) or the indicated purified proteins ( + ) : Ltn1p at 130 nM , Rqc1p at 70 nM , and Rqc2p at 420 nM final concentration . ( D ) ScIVT reactions were conducted using rqc2Δ extracts , a lysine-free or lysine-containing truncated mRNA substrate , and 100 μM exogenous ubiquitin . After 0 min ( t = 0 ) or 30 min ( t = 30 ) of translation , all reactions were supplemented with an equal volume of ‘mock ScIVT’ ( i . e . , without mRNA ) containing 1 . 34 μM purified Rqc2p , 100 μM exogenous ubiquitin , and the indicated inhibitor ( s ) . Indicated time points ( ‘Time ( min ) ' ) were analyzed by SDS-PAGE and immunoblotting . ‘Long’ and ‘Short’ refer to exposure times of the blots . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 00810 . 7554/eLife . 27949 . 009Figure 4—figure supplement 1 . Purified Ltn1p and Rqc1p . C-terminal polyhistidine-tagged Ltn1p ( WT and W1542E mutant ) and Rqc1p were purified as described in Materials and methods and analyzed by SDS-PAGE and Coomassie staining . 1/10th volume of the corresponding purified protein was added to ScIVT reactions . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 00910 . 7554/eLife . 27949 . 010Figure 4—figure supplement 2 . Impact of excess Ltn1p on ubiquitination in rqc1Δ extracts . ScIVT reactions were conducted as in Figure 4A–C except that 100 μM exogenous ubiquitin was present in all reactions . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 01010 . 7554/eLife . 27949 . 011Figure 4—figure supplement 3 . Impact of CAT-tailing inhibition on ubiquitination . ScIVT reactions were conducted as in Figure 4D , except all reactions were performed with lysine-containing mRNA and ‘mock ScIVT’ containing either protein storage buffer ( – ) or 1 . 34 μM purified Rqc2p ( + ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 011 Unexpectedly , we did not detect any Ltn1p-dependent ubiquitination of the reporter protein in rqc1Δ extracts ( Figure 3C ) . The addition of purified Rqc1p ( Figure 4—figure supplement 1 ) , however , fully rescued ubiquitination in rqc1Δ extracts ( Figure 3C and Figure 4B ) . Increasing the concentration of Ltn1p in the reaction did not bypass the requirement for Rqc1p in ubiquitination ( Figure 4—figure supplement 2 ) . These observations suggest that Rqc1p is directly involved in nascent-chain ubiquitination . Such a role is consistent with the fact that LTN1 deletion phenocopies RQC1 deletion in both the accumulation of stalling reporters in vivo and , more broadly , in their correlated set of genetic interactions ( Brandman et al . , 2012; Defenouillère et al . , 2013 ) . Although TCF25/Rqc1p was previously reported to be dispensable for Listerin/Ltn1p-mediated ubiquitination of purified 60S-bound stalled nascent chains ( Shao and Hegde , 2014 ) , the stringent purification of ribosome–nascent chain complexes in that study might have removed factors that otherwise impose a requirement for Rqc1p/TCF25 ( e . g . , chaperones that protect the nascent chain ) . Our results support a model in which Rqc1p directly promotes Ltn1p-mediated ubiquitination of the nascent chain via a mechanism that remains to be determined . Previous studies have shown that a CAT-tailing-deficient mutant of Rqc2p preserves degradation of stalled nascent chains ( Shen et al . , 2015 ) and that in vitro reconstitution of Listerin/Ltn1p-mediated ubiquitination does not strictly require NEMF/Rqc2p ( Shao and Hegde , 2014; Shao et al . , 2013 ) . Together , these studies suggested that CAT tailing is dispensable for degradation of the assayed reporter constructs . Based on these results and structural data , it was proposed that Rqc2p indirectly contributes to ubiquitination by recognizing the aberrant 60S:peptidyl–tRNA complex , stabilizing Ltn1p on the 60S subunit , and sterically preventing the 40S subunit from rejoining ( Shao et al . , 2015 ) . Together with the minimal impact of LTN1 disruption on CAT tailing in vivo ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) and in vitro ( Figure 1 ) , these studies led to a model in which ubiquitination by Ltn1p and CAT tailing by Rqc2p are independent activities of the RQC ( Inada , 2017 ) . The unique ability of our ScIVT system to recapitulate both activities of the RQC , combined with its genetic tractability , allowed us to directly test this model in vitro . Consistent with the proposed scaffolding function of Rqc2p in ubiquitination , disruption of RQC2 abrogated nascent-chain ubiquitination by Ltn1p ( Figure 3C and Figure 4C ) , and addition of wild-type Rqc2p rescued ubiquitination ( Figure 4C and Figure 1—figure supplement 1 ) . Unexpectedly , however , the addition of CAT-tailing-deficient ( D98A ) Rqc2p to rqc2Δ extracts only partially rescued ubiquitination ( Figure 4C ) . Similarly , reactions using extracts containing endogenously expressed Rqc2p ( D98A ) as the only RQC2 gene product yielded minimal ubiquitinated protein ( Figure 3C ) . To rule out the possibility that the effect of the D98A substitution on ubiquitination was due to disruption of the known scaffolding function of Rqc2p , we took an alternative approach to inhibit CAT tailing in the context of wild-type Rqc2p . As observed in the Rqc2p ( D98A ) experiments , preventing CAT tailing of stalled nascent chains—in this case with anisomycin treatment ( Figure 2C ) —substantially reduced ubiquitination ( Figure 4D and Figure 4—figure supplement 3 ) . These results suggest that Rqc2p not only provides structural support for Ltn1p but also that CAT tailing directly enhances Ltn1p-dependent ubiquitination of at least some substrates ( see Discussion ) . Collectively , our in vitro analyses reveal that all three components of the RQC—Ltn1p/Listerin , Rqc1p/TCF25 , and Rqc2p/NEMF—contribute to ubiquitination of the nascent chain .
We have shown that establishing a cell-free system that recapitulates both CAT tailing and ubiquitination opens new opportunities to explore how the fully functional RQC promotes clearance of aberrant translation products . Our analyses reveal that Rqc2p-mediated nascent-chain elongation is mechanistically distinct from canonical translation , that ubiquitination of the nascent polypeptide requires both Ltn1p and Rqc1p , and that the ubiquitination and CAT-tailing activities of the RQC are coupled through a mutual requirement for active Rqc2p ( Figure 5 ) . 10 . 7554/eLife . 27949 . 012Figure 5 . Model for CAT tailing and ubiquitination of stalled nascent chains . When an 80S ribosome stalls during translation , splitting factors recognize the stalled translation complex to facilitate dissociation of the 40S subunit and mRNA . Ltn1p , Rqc2p , and Rqc1p ( unknown location indicated by ‘ ? ' ) bind the resulting 60S:peptidyl–tRNA complex . Together with the peptidyl-transferase center of the 60S subunit , Rqc2p facilitates elongation of the stalled nascent chain with a CAT tail by recruiting alanine- and threonine-charged tRNAs to the A site . If the nascent chain contains a lysine residue ( red circle ) located within the vicinity of the Ltn1p RING domain ( or potentially hidden inside the ribosome exit tunnel ) , CAT tailing and Rqc1p enhance or facilitate Ltn1p-mediated ubiquitination of the nascent chain , respectively , for subsequent proteasomal degradation ( green box ) . If the nascent chain does not contain any lysine residues ( or contains lysine residues that are too distant from the Ltn1p RING domain ) , CAT tails may promote aggregation of incompletely synthesized proteins ( red box ) . In both instances , Rqc2p activity promotes hydrolysis of the peptidyl–tRNA linkage and liberation from the 60S subunit . DOI: http://dx . doi . org/10 . 7554/eLife . 27949 . 012 The key benefit of an in vitro system to study CAT tailing is the ability to perform experiments that would be intractable in vivo . Indeed , a major difficulty in studying CAT tailing is that it utilizes some of the same machinery as canonical translation ( i . e . , the 60S subunit ) and requires a substrate that is generated by canonical translation , making it difficult to perturb CAT tailing specifically . By temporally separating the translation- and CAT-tailing-phases of in vitro reactions , we overcame this obstacle and specifically tested the sensitivity of the CAT-tailing reaction to a wide range of well-characterized ribosome and translation inhibitors . These analyses revealed that aside from requiring the catalytic PTC of the ribosome , CAT tailing is otherwise fundamentally different from translation elongation in ways that could not have been fully anticipated from previous studies ( Figure 2 ) . In particular , our findings that CAT tailing does not require the canonical activities of the elongation factors or energy from GTP hydrolysis suggest a unique mechanism of elongation . Reexamining 60S:RQC structures ( Shao et al . , 2015; Shen et al . , 2015 ) , we noted that the network of interactions between Ltn1p/Rqc2p and the 60S subunit overlaps with the ribosome-binding sites of eEF1a ( Shao et al . , 2016 ) and eEF2 ( Taylor et al . , 2007 ) . The overlap in binding sites indicates that either eEF1a/eEF2 are dispensable for CAT tailing , or that eEF1a/eEF2 and the RQC interact transiently with the 60S subunit during elongation . Our findings favor the former model and suggest that Rqc2p directly recruits charged tRNAs with its selective tRNA-binding activity ( Shen et al . , 2015 ) , without the involvement of eEF1a . Following peptide-bond formation by the 60S subunit , the A/P and P/E tRNAs may spontaneously translocate in the absence of interactions with an mRNA template or 40S subunit; whereas during canonical translation such interactions impose an energy requirement for translocation that is fulfilled by the GTPase activity of eEF2 . Given that CAT tailing was proposed to occur on the 60S subunit ( Shen et al . , 2015 ) , we predicted that all translation inhibitors that bind the 60S subunit would inhibit CAT-tail synthesis . However , we found that CAT tailing was not inhibited by cycloheximide , suggesting that the deacylated tRNA may rapidly dissociate from the 60S subunit following peptidyl transfer . Together , these findings suggest that a minimal set of factors—a 60S:peptidyl–tRNA complex , charged alanine and threonine tRNAs , and Rqc2p—may be sufficient for CAT-tail synthesis . In 1969 , Monro demonstrated that in the presence of certain alcohols , isolated bacterial 50S ribosomal subunits could catalyze polymerization from aminoacyl-tRNAs in the absence of an mRNA template and 30S subunits ( Monro , 1969 ) . Thus , it is conceivable that Rqc2p may stimulate nascent-chain elongation much as in Monro’s minimal prokaryotic system . Many questions remain about Rqc2p dynamics during CAT tailing , including whether a single molecule of Rqc2p remains on the 60S subunit for successive cycles of peptide-bond formation or whether each cycle of elongation requires binding by a new Rqc2p–tRNA complex . CAT-tail synthesis eventually terminates by hydrolysis of the peptidyl–tRNA linkage , which is presumably needed to release the CAT-tailed nascent chain from the 60S subunit for its destruction by the proteasome . Indeed , one critical function of CAT tailing might be to provide a mechanism of termination in the absence of a stop codon . The wide range of CAT-tail lengths observed both in vivo ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) and in vitro suggests that termination is a stochastic process . Furthermore , our finding that peptidyl–tRNA intermediates are relatively stable in vitro in the absence of Rqc2p ( Figure 1D and F , Figure 2 , and Figure 4C ) indicates a potentially direct role of Rqc2p in the termination reaction . These observations lead to a model for CAT-tailing termination in which Rqc2p recruits a termination factor to the 60S subunit in a stochastic manner during the process of elongation . Because Rqc2p interacts with A-site tRNAs at a site distant from the acceptor stem ( Shen et al . , 2015 ) , it is possible that the ‘termination factor’ is an uncharged alanine- or threonine-tRNA ( Caskey et al . , 1971; Zavialov et al . , 2002 ) . Alternatively , Rqc2p might interact with the canonical termination factor eRF1 or another protein with peptidyl–tRNA hydrolase activity to facilitate termination . Taking advantage of the fact that our extracts also recapitulated Ltn1p-dependent ubiquitination , we found that Rqc1p plays a critical role in nascent-chain ubiquitination in vitro ( Figure 3C and Figure 4B ) . This direct role of Rqc1p in ubiquitination contrasts with its previously suggested role in recruiting Cdc48p downstream of ubiquitination ( Brandman et al . , 2012; Defenouillère et al . , 2013 ) . Nevertheless , our discovery that ubiquitination in vitro is as dependent on Rqc1p as it is on the E3 ligase Ltn1p is consistent with the fact that yeast strains lacking Rqc1p and Ltn1p have very similar phenotypes and genetic interaction profiles , suggesting a similar molecular defect in these strains ( Brandman et al . , 2012 ) . We speculate that Rqc1p may facilitate ubiquitination by positioning the nascent chain in proximity to the Ltn1p RING domain , by promoting binding of the E2 ubiquitin-conjugating enzyme , or by activating Ltn1p’s E3 ligase activity on the 60S subunit . With a system that uniquely recapitulates both nascent-chain elongation by Rqc2p and ubiquitination by Ltn1p , we discovered that Rqc2p can elongate the nascent chain to enhance ubiquitination , rather than just providing structural support for Ltn1p binding to the 60S subunit ( Figure 3C and Figure 4C ) . This finding was surprising given that previous studies have shown that a CAT-tailing-deficient mutant of Rqc2p preserves degradation of aberrant nascent chains in yeast cells ( Shen et al . , 2015 ) and that in vitro reconstitution of Listerin/Ltn1p-mediated ubiquitination does not strictly require NEMF/Rqc2p ( Shao and Hegde , 2014 ) . How can we reconcile these findings ? Structural studies of full-length Listerin/Ltn1p on the 60S subunit localized its RING domain ( which binds an E2 ubiquitin-conjugating enzyme ) near the ribosomal exit tunnel ( Shao et al . , 2015 ) , poised to facilitate ubiquitination of lysine residues close to or recently emerged from the tunnel . The physical tethering of the RING domain near the exit tunnel suggests that Ltn1p may not be able to access more distantly positioned lysines in the nascent polypeptide , nor can the most recently translated lysines—contained within the 30–60-amino-acid long exit tunnel ( Kramer et al . , 2009 ) —be accessed by Ltn1p until their emergence . Our observations lead to a model in which Ltn1p can only ubiquitinate a spatially restricted set of lysines , while CAT tailing enables access to other lysines—as previously proposed ( Brandman and Hegde , 2016; Simms et al . , 2017 ) and recently demonstrated in vivo ( Kostova KK et al . , 2017 ) . We reason that the few nascent chain RQC substrates previously studied in vivo could be degraded in a CAT-tailing-independent manner ( Choe et al . , 2016; Defenouillère et al . , 2016; Shen et al . , 2015; Yonashiro et al . , 2016 ) due to native lysines being fortuitously positioned proximal to the Ltn1p RING domain . Recent studies in budding yeast have demonstrated that CAT tails mediate formation of detergent-insoluble aggregates when the nascent chain cannot be degraded due to its limited ubiquitination potential or due to inactivation of Ltn1p ( Choe et al . , 2016; Defenouillère et al . , 2016; Yonashiro et al . , 2016 ) . In the context of a fully intact RQC , however , our findings suggest that CAT tailing and ubiquitination are interdependent activities . Elongation of the nascent chain with CAT tails can result in two outcomes: positioning lysine residues proximal to the Ltn1p RING domain for efficient ubiquitination; or distancing lysine residues from the Ltn1p RING domain , making ubiquitination less efficient . Thus , CAT tailing and ubiquitination must be tightly coordinated to promote nascent-chain degradation and to avoid , where possible , aggregate formation and the detrimental sequestration of cytosolic chaperones . These studies collectively suggest that rather than being the primary role of CAT tails , aggregation is more likely a backup pathway to mitigate the toxic effects of stalled polypeptides that cannot be efficiently ubiquitinated and degraded ( Figure 5 ) . A notable distinction between the aggregation- and ubiquitination-promoting functions of CAT tails is that the former depends on the alanine/threonine composition of the CAT tail ( Choe et al . , 2016 ) , while the latter depends on the process of CAT tailing itself . While CAT tailing has yet to be reported in metazoans , the conservation of Rqc1p/TCF25 , Rqc2p/NEMF , and Ltn1p/Listerin—including critical residues that we mutated in this study—suggest that both RQC activities are conserved and together provide a means of protecting cells against the accumulation of faulty translation products . Ltn1p-dependent ubiquitination has been detected in rabbit reticulocyte extracts , so a CAT-tailing-dependent mechanism to facilitate ubiquitination of inaccessible lysine residues likely operates in metazoans as well . Underscoring the importance of this quality-control mechanism in maintaining proteostasis , mutations in the mammalian homologs of HBS1 ( Ishimura et al . , 2014 ) and LTN1 ( Chu et al . , 2009 ) cause neurodegeneration in mice . With our newfound insights , the similarities between the bacterial tmRNA system and the eukaryotic RQC become even more striking than previously appreciated . In certain bacteria , a stalled ribosome is rescued by the recruitment of a hybrid tRNA/mRNA-like molecule ( tmRNA ) to the empty A-site of the ribosome , leading to translation of a tmRNA-encoded C-terminal degron that includes a dedicated stop codon ( Moore and Sauer , 2007 ) . In this way , stalled nascent chains are marked for degradation and translation can terminate even when the mRNA template lacks a stop codon . The RQC fulfills these same functions but in a different manner that is compatible with the ubiquitin-proteasome system: Stalled nascent chains are marked for degradation by ubiquitination ( in certain cases facilitated by CAT tailing ) , and translation can terminate without a stop codon with the help of Rqc2p . Thus , both mechanisms involve a ribosome-catalyzed peptidyl-transferase reaction that adds a C-terminal extension that is not templated by the parent mRNA molecule . The addition of the C-terminal extension , moreover , facilitates peptidyl–tRNA hydrolysis and nascent-chain release either directly ( tmRNA ) or indirectly ( RQC ) to promote degradation . These functional similarities between the tmRNA system and the RQC—despite not sharing any related factors other than the ribosome—provide a striking example of convergent evolution that emphasizes the physiological importance of discarding incompletely synthesized proteins and recycling the translation machinery .
Saccharomyces cerevisiae strains used in this study were derived from BY4741 ( MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ) and are listed in Supplementary file 1 . Yeast transformations were performed using the PEG–lithium acetate method ( Gietz and Woods , 2006 ) . To generate gene knockouts , the entire coding sequence of the gene of interest was replaced with the URA3 cassette of pRS416 . Strains containing point mutations at endogenous loci were generated from URA3-disrupted strains by transformation with PCR fragments encoding the mutant gene of interest and 5-FOA counterselection ( Boeke et al . , 1987 ) . Transformants were screened by PCR to identify integrants , which were subsequently verified by PCR and Sanger sequencing of the entire integrated cassette . pcDNA3 . 1 ( + ) ( Thermo Fisher Scientific , Waltham , MA ) was modified using Gibson Assembly Master Mix ( New England Biolabs , Ipswich , MA ) and appropriate DNA fragments according to the Gibson assembly method ( Gibson et al . , 2009 ) to generate pBAO1124 , which contains ( in order ) : T7 promoter , 46-nt 5′-UTR lacking any AUG or near-AUG codons ( i . e . , NUG , ANG , or AUN , where N is any nucleotide ) , 3xHA-NanoLuc luciferase ORF , 56-nt 3′-UTR , and 50-nt poly ( A ) sequence . RNAs were generated by run-off transcription with T7 RNA Polymerase using the MEGAscript T7 Transcription Kit ( Thermo Fisher Scientific ) according to the manufacturer’s instructions using PCR-amplified DNA templates derived from pBAO1124 or its variants ( DNA sequences are provided in Supplementary file 2 ) . Transcription reactions were terminated by addition of ammonium acetate stop solution . RNA was extracted with neutral phenol:chloroform:isoamyl alcohol ( 25:24:1 ) ( Sigma , St . Louis , MO ) , precipitated with ethanol , and resuspended in nuclease-free water . A 5′−7-methylguanosine cap was added to RNA post-transcriptionally using the Vaccinia Capping System ( New England Biolabs ) . Capping reactions were desalted using Micro Bio-Spin Columns with Bio-Gel P-30 ( Bio-Rad , Hercules , CA ) before RNA was extracted with phenol , precipitated with ethanol , and resuspended in nuclease-free water . S . cerevisiae strains were grown overnight to saturation in rich YPAD media , diluted the next morning to OD600 0 . 2 in a total volume of 1 L YPAD , and harvested at OD6001 . 4–1 . 8 by centrifugation at 3500 rpm for 6 min at 4°C . The cell pellet was washed with water and resuspended in lysis buffer A ( 30 mM HEPES-KOH pH 7 . 4 , 100 mM KOAc , 2 mM Mg ( OAc ) 2 , 2 mM DTT , and cOmplete mini EDTA-free protease inhibitor cocktail [Roche , Switzerland] ) using 1 ml per 6 g of wet cell pellet . The cell slurry was dripped into liquid nitrogen to produce frozen pellets , which were then pulverized using a 6970EFM Freezer/Mill ( SPEX SamplePrep , Metuchen , NJ ) by three cycles of 12 Hz agitation for 1 . 5 min with cooling for 2 min after each cycle . The resulting ‘grindate’ was combined with an equal volume of pre-chilled lysis buffer A ( i . e . , 1 ml per 1 g of grindate ) and allowed to thaw on ice . Cell debris was cleared by sequential centrifugation at 4°C at 1000 g for 5 min , 1350 g for 5 min , 14000 g for 30 min , and finally 14000 g for an additional 10 min . The clarified lysate was dialyzed twice for 2 hr against 250 ml lysis buffer A ( except without protease inhibitor cocktail ) using 3500 MWCO cassettes ( Thermo Fisher Scientific #87722 ) . After dialysis , lysates were flash frozen in 50 μl aliquots in liquid nitrogen and stored at –80°C . Endogenous mRNAs in thawed extracts were degraded by treatment with 0 . 3 U/μl micrococcal nuclease ( MNase ) and 480 μM CaCl2 for 10 min at room temperature , followed by addition of 2 mM EGTA and transfer to ice . ScIVT reactions were initiated by adding m7G-capped RNA ( 40 ng per μl of reaction volume ) to MNase-treated yeast extracts and incubating at 25°C for up to 90 min . Final concentrations of reaction components were 48 . 67% ( v/v ) MNase-treated yeast extract , 22 mM HEPES-KOH ( pH 7 . 4 ) , 120 mM potassium acetate , 1 . 5 mM magnesium acetate , 0 . 75 mM ATP , 0 . 1 mM GTP , 0 . 04 mM each amino acid , 1 . 7 mM DTT , 25 mM creatine phosphate , 0 . 34 μg/μl creatine kinase , 0 . 14 U/μl SUPERase•In RNase Inhibitor ( Thermo Fisher Scientific ) , and 0 . 16X cOmplete mini EDTA-free protease inhibitor cocktail ( Roche ) . Where indicated , reactions also included 10 or 100 μM recombinant human ubiquitin or Myc-ubiquitin ( Boston Biochem , Cambridge , MA ) . Reactions were halted by transferring to ice or by adding an equal volume of 2X Laemmli Sample Buffer ( Bio-Rad ) . The results shown for all ScIVT experiments are representative of at least two technical replicates ( i . e . , experiments conducted with independently prepared reagents ) . Concentrated stock solutions are diluted 1:10 into the ScIVT reactions . For example: 1 μl of 2 mM Anisomycin ( in 5% DMSO ) was added to a 10 μl ScIVT reaction . All stock solutions should be dissolved in water ( or diluted with water ) as ScIVT reactions cannot tolerate 100% DMSO or ethanol . Small molecules were used at the following final concentrations: 30 μl ScIVT reactions were assembled with 1 . 2 μg 3xHA-10xHis-NL mRNAs and 10 μM recombinant Myc-ubiquitin ( Boston Biochem ) and incubated at 25°C for 1 hr . For input samples , 10 μl was removed and quenched in an equal volume of 2X Laemmli Sample Buffer . For Ni-NTA-purified samples , the remaining 20 μl was quenched by addition of denaturing buffer ( 6 M guanidine-HCl , 50 mM Tris pH 7 . 8 , 300 mM KCl , 10 mM imidazole , 0 . 1% NP-40 , 5 mM β-mercaptoethanol [βME] ) , and then incubated with 10 μl of pre-washed Ni-NTA Magnetic Agarose Beads ( Qiagen , Germany ) at 4°C overnight with end-over-end rotation . Beads were washed three times with wash buffer I ( denaturing buffer except 500 mM KCl ) and three times with wash buffer II ( denaturing buffer except 50 mM KCl and no guanidine-HCl ) , each for 5 min at room temperature . Bound proteins were eluted from beads by adding 15 μl elution buffer ( wash buffer II except 200 mM imidazole ) and incubating at 22°C for 5 min with shaking at 1000 rpm . The elution step was repeated , and eluates were pooled and mixed with an equal volume of 2X Laemmli Sample Buffer . The results shown for all experiments that include a denaturing purification of ScIVT products are representative of at least three technical replicates ( i . e . , experiments conducted with independently prepared reagents ) . Protein samples were separated by SDS-PAGE using 12% Bolt Bis-Tris gels ( Thermo Fisher Scientific ) and transferred in 1X CAPS Buffer onto 0 . 22 micron PVDF membrane ( Bio-Rad ) . Blots were probed with the following antibodies diluted 1:5000 in 1X TBS-T containing 5% nonfat dry milk: mouse anti-HA ( RRID:AB_627809 , Santa Cruz Biotechnology [Dallas , TX] sc-7392 ) , rat anti-HA high sensitivity ( RRID:AB_390918 , Roche #11867423001 ) , mouse anti-Myc ( RRID:AB_331783 , Cell Signaling Technology [Danvers , MA] #2276 ) , HRP-conjugated goat anti-mouse IgG ( RRID:AB_631736 , Santa Cruz Biotechnology sc-2005 ) , and HRP-conjugated goat anti-rat IgG ( RRID:AB_631755 , Santa Cruz Biotechnology sc-2032 ) . Blots were developed using Clarity ECL Western Blotting Substrate ( Bio-Rad ) , and chemiluminescence was detected on a ChemiDoc Imaging System ( Bio-Rad ) . S . cerevisiae strain yRH101 ( a gift from Stephen Bell , MIT ) derived from ySC7 ( Chen et al . , 2007 ) containing a 2 μm PGAL1-[protein]−10xHis plasmid ( a gift from Bob Stroud , UCSF ) was grown overnight in SC–His media containing 2% raffinose , diluted the next day , and grown for an additional night to early saturation . Protein expression was induced by adding an equal volume of Yeast-Peptone media containing 2% galactose , and cells were grown for 5 hr at 30°C . Cells were harvested by centrifugation at 3500 rpm for 6 min at 4°C , and the cell pellet was washed with water before resuspending in Lysis Buffer ( 20 mM HEPES-KOH pH 7 . 4 , 500 mM KCl , 20 mM imidazole; 2 mM βME and cOmplete EDTA-free protease inhibitor cocktail [Roche] added just prior to use ) at a ratio of 1 ml per gram of cell pellet . The resulting cell slurry was dripped into liquid nitrogen to produce frozen pellets , which were pulverized using a 6970EFM Freezer/Mill ( SPEX SamplePrep ) by three cycles of 12 Hz agitation for 1 . 67 min with 2 min cooling after each cycle . The resulting powder was briefly thawed before adding Lysis Buffer ( 1 mL per 1 g of powder ) supplemented with additional protease inhibitors ( 292 μM Pepstatin , 8 . 4 mM Leupeptin , 1 . 23 mM Aprotinin , 1 mM Phenylmethylsulfonyl fluoride [PMSF] ) . Cell debris was cleared by sequential centrifugation at 14000 rpm at 4°C for 10 min and then 30 min , followed by sequential filtration through 2 . 7 and 1 . 6 μm Whatman GD/X filters ( GE Healthcare Life Sciences , Marlborough , MA ) . His-tagged proteins were purified from lysate using Ni-NTA Sepharose beads ( Qiagen ) as follows . Beads ( ~1 ml 50% slurry per 1 L yeast culture ) were washed with water and equilibrated in Lysis Buffer . Lysate was added to semi-dry beads and rotated at 4°C for 2 hr . Beads were collected by centrifugation at 800 rpm for 3 min , resuspended in an equal volume of Lysis Buffer , loaded over a disposable Bio-Spin column ( Bio-Rad ) , and washed once with 10 ml Lysis Buffer . The column was then washed as follows: once with 10 ml Wash Buffer ( 20 mM HEPES-KOH pH 7 . 4 , 500 mM KCl , 10% glycerol , 2 mM βME ) containing 20 mM imidazole and 1 mM PMSF; once with 10 ml Wash Buffer containing 20 mM imidazole; and twice with 10 ml Wash Buffer containing 50 mM imidazole . Proteins were sequentially eluted from the beads by gravity rinses as follows: once with 250 μl Wash Buffer containing 250 mM imidazole; twice with 500 μl Wash Buffer containing 250 mM imidazole; and twice with 500 μl Wash Buffer containing 500 mM imidazole . Elution fractions were analyzed by SDS-PAGE and staining with Coomassie Brilliant Blue R-250 to identify those containing the protein of interest , which were then pooled and concentrated to ~500 μl before overnight dialysis into Rqc2p/Ltn1p Storage Buffer ( 50 mM HEPES-KOH pH 7 . 4 , 300 mM KOAc , 5% glycerol , 2 mM DTT ) or Rqc1p Storage Buffer ( 20 mM HEPES-KOH pH 7 . 4 , 500 mM KOAc , 10% glycerol , 2 mM DTT ) . Dialyzed protein was passed through a 0 . 1 μm centrifugal filter ( EMD Millipore [Hayward , CA] #UFC30VV00 ) before flash freezing in liquid nitrogen . Protein concentration was determined by both spectrophotometry using a Nanodrop 2000 ( Thermo Fisher Scientific ) and Coomassie staining against BSA standards . | Cells make proteins by reading instructions encoded in molecules called messenger RNAs . Structures called ribosomes move along the messenger RNAs and translate the coded instructions to build new proteins from building blocks known as amino acids . Normally , a ribosome will encounter a stop signal on the messenger RNA , which ends translation and allows the newly built protein to be released . Sometimes , however , ribosomes stall before they reach the genuine stop signal , which can happen due to defects in the messenger RNAs or ribosomes . To prevent incomplete proteins from accumulating and causing damage , cells contain a group of other proteins called the Ribosome-associated Quality-control Complex ( or RQC for short ) . This quality-control complex is composed of three components that assemble on stalled ribosomes and attach two different tags to the incomplete protein . One component adds a degradation tag called ubiquitin . A second component works with the ribosome to tag the incomplete protein with a ‘tail’ that contains the amino acids alanine and threonine . These amino acids are abbreviated to A and T , and are added to the end of the protein known as the ‘C-terminus’ , so this tag is named a ‘CAT tail’ . Although all three RQC components are needed to degrade incomplete proteins , little was known about why the CAT tails are added , or what the third component – a protein called Rqc1p – actually does . Osuna et al . have now investigated how the apparently unrelated activities of RQC components are coordinated to destroy incomplete proteins . Ribosomes and RQC components from yeast cells were extracted and mixed in the laboratory with a messenger RNA that stalls ribosomes . In this cell-free system , the RQC components could still tag incomplete proteins with both ubiquitin and CAT tails . Osuna et al . then used this system to show that the way the ribosome added amino acids to form a CAT tail was different from how it normally builds proteins . The experiments also showed that in order for the ubiquitin tags to be added efficiently , Rqc1p must be present , and in some cases , the incomplete proteins also need to be ‘CAT tailed’ . When either were missing , very few ubiquitin tags could be added to the incomplete proteins . The results show that the three core RQC components need to work together to degrade incomplete proteins . This quality-control complex is also found in mice and humans , and mice with mutations in the genes that encode RQC components often have damaged nervous systems . In the future , researchers building upon these findings and other studies of the RQC may eventually understand the relationship between the RQC and neurodegenerative diseases in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2017 | In vitro analysis of RQC activities provides insights into the mechanism and function of CAT tailing |
Immunotherapy , represented by immune checkpoint inhibitors ( ICI ) , is transforming the treatment of cancer . However , only a small percentage of patients show response to ICI , and there is an unmet need for biomarkers that will identify patients who are more likely to respond to immunotherapy . The fundamental basis for ICI response is the immunogenicity of a tumor , which is primarily determined by tumor antigenicity and antigen presentation efficiency . Here , we propose a method to measure tumor immunogenicity score ( TIGS ) , which combines tumor mutational burden ( TMB ) and an expression signature of the antigen processing and presenting machinery ( APM ) . In both correlation with pan-cancer ICI objective response rates ( ORR ) and ICI clinical response prediction for individual patients , TIGS consistently showed improved performance compared to TMB and other known prediction biomarkers for ICI response . This study suggests that TIGS is an effective tumor-inherent biomarker for ICI-response prediction .
Immunotherapy , represented by immune checkpoint inhibitors ( ICI ) , including anti-PD-1 antibodies , anti-PD-L1 antibodies , anti CTLA-4 antibodies or their combinations , is transforming the treatment of cancer . Compared to conventional therapies , ICI can induce significantly improved clinical responses in patients with various types of late-stage metastatic cancers . However , the majority of unselected patients will not respond to ICI . Most tumor types show response rates below 40% to PD-1 inhibition , and the response rates of each tumor type are reported to be correlated with the tumor mutational burden ( TMB ) of that tumor type ( Yarchoan et al . , 2017 ) . Multiple factors are reported to affect ICI effectiveness , including: PD-L1 expression ( Herbst et al . , 2014; Shukuya and Carbone , 2016 ) , TMB ( Rizvi et al . , 2015; Snyder et al . , 2014 ) , DNA mismatch repair deficiency ( Le et al . , 2015 ) , the degree of cytotoxic T cell infiltration ( Tang et al . , 2016 ) , mutational signature ( Miao et al . , 2018; Wang et al . , 2018 ) , antigen presentation defects ( Chowell et al . , 2018; Zaretsky et al . , 2016 ) , interferon signaling ( Ayers et al . , 2017 ) , tumor aneuploidy ( Davoli et al . , 2017 ) and T-cell signatures ( Jiang et al . , 2018 ) . These biomarkers have various rates of accuracy and utility , and the identification of a robust ICI-response biomarker is still a critical challenge in the field ( Nishino et al . , 2017 ) . ICI help a patient’s immune system to recognize and attack cancer cells . The immunogenicity of cancer cells is the fundamental determinant of ICI response . Theoretically , tumors of very low or no immunogenicity will not respond to therapeutic strategies that enhance the immune response . Hence , ICI can only be used to treat tumors that have sufficient immunogenicity . Furthermore , enhancing tumor immunogenicity can potentially transform an immunotherapy-non-responsive tumor into an immunotherapy-responsive tumor . The actual immunogenicity of a tumor is not easy to measure . In theory , tumor immunogenicity is determined by the tumor cell itself , and is also influenced by factors related to the tumor microenvironment , such as the functioning of professional antigen-presenting cells like dendritic cells ( DCs ) ( Mellman and Steinman , 2001 ) . Fundamental determinants of tumor immunogenicity include tumor antigenicity , and antigen processing and presenting efficiency ( Blankenstein et al . , 2012 ) . Antigen presentation defects have already been shown to contribute to ICI-response failure ( Chowell et al . , 2018; Zaretsky et al . , 2016 ) . To measure antigen processing and presenting efficiency systematically , we applied a gene set variation analysis ( GSVA ) method to generate an antigen processing and presenting machinery ( APM ) score ( APS ) ( Hänzelmann et al . , 2013 ) , which was calculated from the mRNA expression status of APM genes . Tumor immunogenicity score ( TIGS ) was then calculated by combining the APM score and the TMB . The antigen-presentation gene expression signature and tumor immunogenicity landscape of 32 cancer types from The Cancer Genome Atlas ( TCGA ) project are provided . TIGS exhibits improved performance in both pan-cancer ICI objective response rate ( ORR ) correlation and accuracy of ICI clinical response prediction when compared with TMB . Our results suggest that TIGS represents a novel and effective tumor-inherent biomarker for the prediction of immunotherapy response .
Cell surface presentation of peptides by major histocompatibility complex ( MHC ) class I molecules is critical to CD8+ T-cell mediated adaptive immune responses , including those against tumors . The generation and loading of peptides onto MHC class I molecules require the functioning of the APM . Several steps are involved in this process , including: 1 ) peptide generation and trimming in the proteasome; 2 ) peptide transport; 3 ) assembly of the MHC class loading complex in the endoplasmic reticulum ( ER ) ; and 4 ) antigen presentation on cell surface ( Leone et al . , 2013 ) . The efficiency of antigen processing and presentation is one determinant of tumor immunogenicity . Here , we used the mRNA expression status of genes involved in the APM process as an indicator of the efficiency of these antigen-processing and -presenting steps . A GSVA approach was applied to measure the overall expression enrichment of APM genes ( Hänzelmann et al . , 2013 ) . On the basis of a review paper about APM ( Leone et al . , 2013 ) , the following genes were selected for quantification: PSMB5 , PSMB6 , PSMB7 , PSMB8 , PSMB9 , PSMB10 , TAP1 , TAP2 , ERAP1 , ERAP2 , CANX , CALR , PDIA3 , TAPBP , B2M , HLA-A , HLA-B and HLA-C ( Figure 1—source data 1 ) . GSVA calculates the per sample overexpression level of a particular gene list by comparing the ranks of the genes in that list with those of all other genes . The resulting GSVA enrichment score is defined as the APS . To explore the pan-cancer distribution pattern of APS , we analyzed about 10 , 000 tumors of 32 cancer types from TCGA ( Figure 1 ) . The boxplot in Figure 1A shows large variance in APS across TCGA cancer types , which uncovers significant distinction in antigen-processing and -presenting efficiency among different cancer types . This analysis is similar to a previous study of seven APM genes ( Şenbabaoğlu et al . , 2016 ) whose expression signature is highly correlated with the APS quantified in this study ( Figure 1—figure supplement 1 ) . Patient Harmonic Best Rank ( PHBR ) I and II scores have recently been proposed to quantify a patient’s antigen presentation ability on the basis of the genotypes of their MHC class I or class II genes , respectively ( Marty Pyke et al . , 2018; Marty et al . , 2017 ) . However , no significant correlations can be observed between APS and PHBR scores ( Figure 1—figure supplement 1 ) , probably because these two methods capture different information about antigen presentation: PHBR are based on MHC genotype information , whereas APS are based on information about the expression of antigen-presentation genes . Univariate Cox regression analyses suggest that APS is associated with cancer patients' survival , and some are statistically significant ( Figure 1B ) . Meta-analysis with pan-cancer hazard ratio values suggests that APS do not associate with prognosis ( Figure 1B ) . To identify the specific gene signatures that determine patients’ APS status , we initially ran differential gene expression analysis for each TCGA cancer type on the basis of APS status . Patients with APS above the median were defined as ‘APS-High’ , patients with APS below the median were defined as ‘APS-Low’ . Differential expression genes ( p-value < 0 . 01 , FDR < 0 . 05 ) were ranked by logFC from high to low and then selected for gene set enrichment analysis ( GSEA ) with gene sets from MSigDB ( Subramanian et al . , 2005 ) . In results from hallmark gene sets , several gene signatures ( especially interferon alpha/gamma response ) were found to be enriched in most TCGA cancer types with high APS , suggesting that high APS is strongly associated with the interferon alpha/gamma signaling pathway ( Figure 2A ) . GSEA using Reactome gene sets further validated this result ( Figure 2—figure supplement 1 ) . Interestingly , interferon gamma was reported to regulate APM gene expression ( Beatty and Paterson , 2001; Ikeda et al . , 2002 ) , which is consistent with this observation . Immune infiltration score ( IIS ) was calculated with GSVA using a list of marker genes for immune cell types and has been validated by the CIBERSORT method ( Şenbabaoğlu et al . , 2016 ) ( Figure 2—source data 1 ) . TIMER ( Li et al . , 2016 ) is another method that can accurately resolve the relative fractions of diverse cell types on the basis of gene expression profiles from complex tissues . To further validate the calculated IIS , we performed TIMER analysis ( Li et al . , 2016 ) and found that the TIMER results were highly correlated with the calculated IIS ( Figure 2—figure supplement 2 ) . Significant associations between APS and IIS at both the level of cancer types and the level of individual patients were observed ( Figure 2B and C ) . The gene list for APS calculation did not overlap with the gene list for IIS calculation . Pan-cancer distribution of TMB was also analyzed with the TCGA dataset ( Figure 2—figure supplement 3 ) . Different cancer types show different prognosis in relation to high TMB ( Figure 2—figure supplement 3 ) . Meta-analysis including all TCGA cancer types suggests that patients with high TMB tend to have poor prognosis ( Figure 2—figure supplement 3 ) . TMB reflects tumor antigenicity and predicted improved survival after immunotherapy . However , in cancer patients not treated with immunotherapy , high TMB tends to be associated with poor prognosis , probably because tumors accumulate mutations during progression as a result of genome instability , and consequently , high TMB is usually associated with late-stage cancer . The immune cell subsets were assessed with both IIS and CIBERSORT ( Newman et al . , 2015 ) methods , and the associations between immune cell subsets with APS were analyzed further ( Figure 2—figure supplement 4 ) . Several types of immune cells , including cytotoxic cells , show strong correlation with APS values ( Figure 2—figure supplement 4 ) . TMB and IIS show relatively weak intercorrelation ( Figure 2D and E ) . The significant correlation between APS and IIS could be due to the following reasons: first , the immune response coordinated by interferon signaling could regulate both APS and IIS; and second , the immunogenicity contributed by APS could stimulate immune response . Tumor immunogenicity is determined by two factors: the antigenicity of tumor cells and the processing and presentation of tumor antigens . These two factors are independent , and are both required for tumor immunogenicity determination . Theoretically , tumor immunogenicity score ( TIGS ) can be represented as [“Tumor antigenicity”] x [“Antigen processing and presenting status”] . Non-synonymous tumor mutation and , consequently , the production of neoantigens can elicit immune response ( Schumacher and Schreiber , 2015 ) . Pan-cancer TMB distribution was analyzed , and log-based TMB values were found to show a Gaussian distribution ( Figure 4—figure supplement 1 ) . In addition , a previous study had already indicated that log ( TMB ) shows linear correlation with pan-cancer immunotherapy ORR ( Yarchoan et al . , 2017 ) . Thus , we used log ( TMB ) as a simple representation of ‘Tumor antigenicity’ . APS calculated on the basis of GSVA range from −1 to 1 . To multiply with tumor antigenicity , we used normalized APS values , which range from 0 to 1 , as a representation of ‘Antigen processing and presenting status’ . APSnormalized= APS-APSpancan_minAPSpancan_max- APSpancan_min We calculated tumor immunogenicity score ( TIGS ) by using the following formula: ( TMB ) TIGS= APSnormalized × log ( TMB ) TIGS were calculated for TCGA samples for which both TMB and RNA-seq gene expression data are available ( 32 cancer types , 8413 samples ) ( Figure 3A ) . Cancer types with high TIGS include: skin cutaneous melanoma ( SKCM ) , diffuse large B-cell lymphoma ( DLBC ) , colon adenocarcinoma ( COAD ) , head and neck squamous cell carcinoma ( HNSC ) ( Figure 3A ) . Univariate Cox regression analysis suggests that TIGS is associated with cancer patients' survival , and this association is statistically significant for some cancer types ( Figure 3B ) . Meta-analysis involving all TCGA cancer types suggested that high TIGS tends to be associated with a poor prognosis in patients not treated with immunotherapy ( Figure 3B ) , which may be due to a mechanism that is the same as that which leads to high TMB . Previous studies have shown that TMB can predict pan-cancer ICI ORR ( Yarchoan et al . , 2017 ) . Here , we evaluated and compared the performance of APS , TIGS with TMB in pan-cancer ICI ORR correlation . The ORR for anti–PD-1 or anti–PD-L1 therapy were plotted against the corresponding median APS , TIGS , TMB across multiple cancer types . In an extensive literature search , we identified 25 tumor types or subtypes for which ORR data are available . For each tumor type , we pooled the response data from the largest published studies that evaluated ORR . We included only studies of anti–PD-1 or anti–PD-L1 monotherapy that enrolled at least 10 patients who were not selected for PD-L1 tumor expression . ( Identified individual studies and references are available in Figure 4—source data 1 and Figure 4—source data 2 . ) To calculate TIGS , two different approaches can be applied . In the first approach , the APS and TMB information are obtained from different studies . This approach can include a greater number of different cancer datasets . In a second approach , all APS and TMB information is obtained from the same TCGA datasets , and in this case , fewer cancer types are available for investigation . When using the first approach , in order to calculate TIGS , the median TMB for each tumor type was obtained from a validated comprehensive genomic profiling assay that was performed and provided by Foundation Medicine ( Chalmers et al . , 2017 ) . The APS information for 23 tumor types was calculated on the basis of TCGA datasets , whereas the APS for Merkel cell carcinoma , cutaneous squamous cell carcinoma and small-cell lung cancer were calculated on the basis of GEO microarray datasets . Significant correlations between APS , TMB , TIGS and the ORR were observed ( Figure 4 ) . The correlation coefficients between APS and ORR and between TMB and ORR were 0 . 42 ( p=0 . 038 ) and 0 . 71 ( p=6 . 8e-5 ) , respectively ( Figure 4 ) , suggesting that 18% and 50% of the difference in the ORR across cancer types could be explained by APS and TMB , respectively . The correlation coefficient between TIGS and ORR is 0 . 78 ( p=5 . 4e-6 ) ( Figure 4C ) , indicating that 60% of the difference in ORR could be explained by TIGS . These pan-cancer ORR analyses imply that TIGS performs better than TMB or APS in correlations with immunotherapy ORR . When using the second approach for TIGS calculation , TIGS still outperformed both TMB and APS in pan-cancer ORR correlation ( Figure 4—figure supplement 1 ) . Compared with TMB and APS , TIGS showed improved correlation with immunotherapy ORR in various types of cancer . Here , we further evaluate the performance of TIGS in predicting ICI clinical response for individual cancer patients . Recently , several prediction biomarkers for immunotherapy response that are based on gene-expression profiling have been reported ( Ayers et al . , 2017; Jiang et al . , 2018 ) . Ayers et al . ( 2017 ) reported an IFN-γ-related mRNA expression signature that predicts clinical response to PD-1 blockade . Benci et al . ( 2019 ) recently described two distinct interferon-related gene expression signatures: ISG . RS , which is associated with resistance to ICI , and by contrast , IFNG . GS , which is derived from an IFNG hallmark geneset and associated with response to ICI . Jiang et al . ( 2018 ) reported a T-cell dysfunction and exclusion gene expression signature ( named ‘TIDE’ in the original paper ) as a biomarker for cancer immunotherapy response . TIDE outperforms known immunotherapy biomarkers — TMB , PD-L1 expression , and interferon gamma gene expression signature — in predicting the response to immunotherapy in melanoma and lung cancer ( Jiang et al . , 2018 ) . The predictive power of TIGS in ICI clinical response was evaluated and compared with those of TMB and biomarkers based on gene expression profiling using ICI datasets , which contain both TMB and transcriptome data for individual patients . In total , two melanoma datasets ( Hugo et al . , 2016; Van Allen et al . , 2015 ) and one urothelial cancer ( Snyder et al . , 2017 ) dataset were available for this analysis . To evaluate performance in predicting clinical response to ICI , we used the receiver operating characteristic ( ROC ) curve to measure the true-positive rates against the false-positive rates at various thresholds of TMB , TIDE or TIGS values ( Figure 5A–C ) . When compared to the widely used ICI-response biomarker TMB , TIGS consistently achieved better performance in all three ICI datasets ( Figure 5A–C ) . The predictive power of TIGS was comparable to that of TIDE in the two melanoma datasets . However , TIDE failed to predict response to immunotherapy in urothelial cancer , so TIGS showed better performance in the urothelial cancer dataset ( Figure 5C ) . TIGS also outperforms other immunotherapy biomarkers that are based on gene expression profiling , including IIS , IFNG , ISG . RS , IFNG . GS and CD8 , in all three datasets ( Figure 5D–F and Figure 5—figure supplement 1 ) . The list of genes used to calculate IFNG , ISG . RS , IFNG . GS and CD8 signatures are available in Figure 5—source data 1 . Interestingly , APS itself also shows improved or similar prediction power when compared to other gene-expression-profiling-based biomarkers ( Figure 5D–F and Figure 5—figure supplement 2 ) . The expression profiles of randomly selected genes ( named ‘APSr’ in Figure 5D–F ) , which were used as a negative control , failed to predict immunotherapy response in all three datasets . In all three available datasets , Kaplan–Meier overall survival curves were further compared in patients with high vs low TIDE , TMB or TIGS level ( Figure 5G–O ) . Patients with TIGS above the median were defined as ‘TIGS-High’ while the remaining patients were defined as ‘TIGS-Low’ . ‘TMB-High’ , ‘TMB-Low’ , ‘TIGS-High’ and ‘TIGS-Low’ were similarly defined . Comparison of survival curves showed better survival for TMB-High patients than for TMB-Low patients in all three ICI datasets , even though the difference did not reach significance in any of the three datasets , probably because of the limited sample size ( Figure 5G–I ) . As defined in the original paper ( Jiang et al . , 2018 ) , TIDE-Low indicates low tumor immune dysfunction and low immune escape , and consequently high immunotherapy response . In the Van Allen et al . ( 2015 ) melanoma dataset , significantly improved survival was observed in TIDE-Low patients when compared to TIDE-High patients ( Figure 5M ) . In the urothelial cancer dataset ( Snyder et al . , 2017 ) , TIDE-Low patients did not have the expected immunotherapy response ( Figure 5O ) . However , TIGS-High patients showed significantly better survival curves than TIGS-Low patients in all three ICI datasets ( Figure 5J–L ) . These analyses suggest that in all three available datasets , TIGS outperforms TMB and other biomarkers that are based on gene-expression profiling ( TIDE , IFNG etc . ) in accurately predicting clinical response to immunotherapy and in pan-cancer applicability .
Immunogenicity is an important inherent feature of tumor cells . This feature is determined by the tumor cell itself , and is also influenced by the tumor microenvironment . Two key determinants of tumor immunogenicity are tumor antigenicity and the ability to present such antigenicity . Here , we proposed an initial method to measure the immunogenicity of a tumor . This measured tumor immunogenicity score ( TIGS ) shows consistently improved correlations with immunotherapy ORR in various types of cancer when compared to TMB . TIGS also shows improved performance in ICI clinical response prediction when compared with TMB and other biomarkers that are based on gene expression profiling ( TIDE , interferon gamma signature and so on ) in both prediction accuracy and pan-cancer applicability . Furthermore , our tumor-immunogenicity-based biomarker could guide the treatment to transform some ICI-non-responsive tumors into ICI-responsive tumors . Stimulating the APM pathway could enhance tumor immunogenicity , and possibly ICI responsiveness . Our study demonstrates that TIGS is an effective biomarker for ICI-response prediction . TIGS capture two key aspects of tumor immunogenicity , antigen presentation and tumor antigenicity , which could be the reason for its improved performance in ICI-response prediction when compared to known biomarkers . Furthermore , our formula for TIGS calculation can point to a new way to transform some ICI-non-responsive tumors into responsive tumors by enhancing the tumor immunogenicity . One approach is to enhance the efficiency of antigen presentation . Our GSEA indicates that interferon signaling is the top gene signature associated with APS-High , and interferon signaling has been reported to influence APM gene expression ( Beatty and Paterson , 2001; Ikeda et al . , 2002 ) . We may enhance antigen presentation by stimulating interferon signaling in patients who are initially not responsive to ICI , especially in cancer types that have low APS , such as prostate cancer and breast cancer . Our study identified several cancer types in which antigen presentation status makes a significant contribution in ICI response . Breast cancer and prostate cancer have usual TMB but fairly low ICI-response rates , probably because of low APS; renal clear cell carcinoma has good ICI response rate , possibly as a result of high APS . Furthermore , our linear correlation formula — ORR = 21 . 4 × TIGS – 2 . 7 ( this formula is based on the data in Figure 4C ) — can be used to make hypotheses with respect to the ORR in tumor types for which anti–PD-1 therapy has not been explored . For example , we anticipate a clinically meaningful ORR of 12 . 3% ( 95% confidence interval [CI] , 8 . 8% to 15 . 8% ) for uterine corpus endometrial carcinoma ( UCEC ) on the basis of a median TIGS of 0 . 7 . This study reports the first quantification of tumor immunogenicity . Several situations need to be considered for future improvement of this quantification . First , other factors including tumor germline antigen , copy number variation status , tumor purity and intra-tumor heterogeneity should also be considered to enable more accurate measurement of the antigenicity of tumor cells . Second , for quantifying antigen presentation efficiency , APM protein expression and function assessment will be more accurate than APM mRNA expression measurement . Third , other factors that influence TIGS should also be considered , including the function of professional antigen presentation cells ( dendritic cells for example ) in the immune microenvironment . This manuscript primarily focused on the cytosolic or endogenous neoantigen presentation pathway mediated by MHC class I . This does not mean that the potential neoantigen presentation by MHC class II is not important , and further studies are needed to improve the methods for the quantification of antigen presentation in cancer patients . In addition , a sex difference in the predictive power of TMB has been reported recently in lung cancer ( Wang et al . , 2019b; Wang et al . , 2019c ) . To explore the potential sex difference in TIGS’s predictive power , we need larger datasets with more patients . TIGS is an extension and enhancement of the immunotherapy biomarker TMB . TIGS is tumor cell-based , and is distinct from the recent immunotherapy biomarkers immunophenoscore ( Charoentong et al . , 2017 ) or T-cell dysfunction and exclusion signature ( Jiang et al . , 2018 ) . Both of these ICI biomarkers are based on tumor immune microenvironment . As a tumor inherent biomarker , TIGS can not only be used for predicting immunotherapy response , but also point ways to manipulate the immunogenicity of tumors , and consequently the response to immunotherapy .
The pancan normalized gene-level RNA-Seq data and clinical information for 33 TCGA cohorts were downloaded from UCSC Xena ( https://xenabrowser . net/ ) with R package UCSCXenaTools ( Wang and Liu , 2019a ) . Samples with ‘pathologic stage’ 0 or X were filtered out and only ‘sample type’ is ‘Primary Tumor’ ( 32 cancer types , N = 9109 ) were saved for further analysis . Pre-compiled , curated somatic mutations ( MC3 version ) for TCGA cohorts were downloaded by the R package TCGAmutations ( Ellrott et al . , 2018 ) . Microarray gene expression datasets for Merkel cell carcinoma , cutaneous squamous carcinoma and small cell lung cancer were downloaded from the GEO database via R package GEOquery ( Davis and Meltzer , 2007 ) . Specifically , GSE39612 ( Harms et al . , 2013 ) , GSE22396 ( Paulson et al . , 2011 ) , GSE36150 ( Masterson et al . , 2014 ) , GSE50451 ( Daily et al . , 2015 ) , GSE99316 ( Sato et al . , 2013 ) were identified and downloaded . APM gene expression status and infiltration levels for immune cell types were quantified using the GSVA method implemented in the R package GSVA ( Hänzelmann et al . , 2013 ) . RNA-Seq or microarray datasets were provided as input and output is a near-Gaussian list of decimals that can be used in visualization or downstream statistical analysis . Lists of genes for quantifying immune cell types were as previously described ( Şenbabaoğlu et al . , 2016 ) . Gene lists for APM score and quantification of immune cell type are provided in Figure 1—source data 1 and Figure 2—source data 1 . The immune infiltration score ( IIS ) for a sample was defined as the mean of standardized values for macrophages , DC subsets ( total , plasmacytoid , immature , activated ) , B cells , cytotoxic cells , eosinophils , mast cells , neutrophils , NK cell subsets ( total , CD56 bright , CD56 dim ) , and all T-cell subsets ( CD8 T , T helper , T central and effector memory , Th1 , Th2 , Th17 , and Treg cells ) . In vitro validation with multiplex immunofluorescence , in silico validation using simulated mixing proportions and comparison between CIBERSORT ( Newman et al . , 2015 ) and IIS have been described previously ( Şenbabaoğlu et al . , 2016 ) . TIMER ( Li et al . , 2016 ) is another method that can accurately resolve the relative fractions of diverse cell types on the basis of gene expression profiles from complex tissues . To further validate the calculated IIS , we performed TIMER analysis and found that the result of TIMER was highly correlated with the calculated IIS ( Figure 2—figure supplement 1 ) . Original APM scores ( APS ) from GSVA are in the range of −1 to 1 . To calculate TIGS , original APM score from GSVA implementation was rescaled by the minimal and maximal APM score from TCGA Pan-cancer analysis . The formula isAPSnormalized= APS−APSpancan_minAPSpancan_max− APSpancan_minwhere APSpancan_min is the minimal APM score among TCGA pan-cancer samples; and APSpancan_max is the maximal APM score among TCGA pan-cancer samples . The normalized APM scores are in the range of 0 to 1 . The normalized APS is set to 0 if a loss of function mutation exists in the B2M gene . TMB was defined as the number of non-synonymous alterations per megabase ( Mb ) of genome examined . As reported previously ( Chalmers et al . , 2017 ) , we used 38 Mb as the estimate of the exome size . For studies reporting mutation number from whole exome sequencing , the normalized TMB = ( whole exome non-synonymous mutations ) / ( 38 Mb ) . We calculated TIGS as following:TIGS= APSnormalized × log ( TMB ) The natural logarithm was used here . Notably , some tumors have a TMB level below one mutation/Mb , so to avoid a negative number in quantifying ‘tumor antigenicity’ , we added a pseudo count of one to normalized TMB . So the TIGS formula is:TIGS= APSnormalized × ln ( TMB +1 ) orTIGS= APSnormalized × ln ( whole exome mutation number38+1 ) The dataset search strategy for assessment of cancer immunotherapy ORR ) assessment has been described previously ( Yarchoan et al . , 2017 ) . We searched MEDLINE ( from January 1 , 2012 to September 1 , 2018 ) , as well as abstracts in the American Society of Clinical Oncology ( ASCO ) , the European Society for Medical Oncology ( ESMO ) , and the American Association for Cancer Research ( AACR ) , to identify clinical studies for anti-PD1 or anti-PDL1 therapy in various tumor types or subtypes . We searched for clinical trials using the following keywords: nivolumab , BMS-936558 , pembrolizumab , MK-3475 , atezolizumab , MPDL3280A , durvalumab , MEDI4736 , avelumab , MSB0010718C , BMS-936559 , cemiplimab , and REGN2810 . We excluded studies that enrolled fewer than 10 participants , studies that investigated anti-PD- ( L ) one therapies only in combination with other agents , and studies that selected patients on the basis of PD-L1 expression or other immune-related biomarkers . Of the remaining studies , only the largest published study for each anti-PD- ( L ) one therapy was included in the final assessment of pooled ORR for each tumor type or subtype . The final identified individual studies are summarized and presented in Figure 4—source data 1 . The TMB information for major solid tumor types or subtypes has been described previously ( Chalmers et al . , 2017 ) . The APS of most tumor types or subtypes are based on TCGA RNA-seq data , except those for Merkel cell carcinoma , cutaneous squamous carcinoma and small cell lung cancer , which do not have available TCGA RNA-seq data . For these cancer types , the GEO datasets GSE39612 , GSE22396 , GSE36150 , GSE50451 , GSE99316 were used to generate APS . In total , 28 cancer types have both TMB and ORR values , and 25 of them also have transcriptome data that can be used for calculating APS . Therefore , TIGS were calculated for these 25 cancer types which have both TMB and APS information available ( Figure 4—source data 2 ) . Linear regression models were constructed to correlate ORR with APS , TMB and TIGS for each of the cancer types or subtypes . To evaluate the power of TIGS to predict clinical response to ICIs , we searched PubMed for ICI clinical studies for which TMB and gene transcriptome information was available for individual patients . In total , three datasets were identified after this search . The Van Allen et al . ( 2015 ) dataset was downloaded from the supplementary files of reference ( Van Allen et al . , 2015 ) . This dataset related to CTLA-4 blockade in metastatic melanoma , and defined ‘clinical benefit’ using a composite end point of complete response or partial response to CTLA-4 blockade as assessed by RECIST criteria or stable disease by RECIST criteria with overall survival greater than 1 year , ‘no clinical benefit’ was defined as progressive disease by RECIST criteria or stable disease with overall survival less than 1 year ( Van Allen et al . , 2015 ) . The Hugo et al . ( 2016 ) dataset was downloaded from the supplementary files of reference ( Hugo et al . , 2016 ) . This dataset related to anti-PD-1therapy in metastatic melanoma: responding tumors were derived from patients who have complete or partial responses or stable disease in response to anti-PD-1 therapy; non-responding tumors were derived from patients who had progressive disease ( Hugo et al . , 2016 ) . The Snyder et al . ( 2017 ) dataset ( Snyder et al . , 2017 ) was downloaded from https://github . com/hammerlab/multi-omic-urothelial-anti-pdl1 . This dataset related to PD-L1 blockade in urothelial cancer: durable clinical benefit was defined as progression-free survival >6 months ( Snyder et al . , 2017 ) . RNA-Seq data were used to calculate the APS for each patient . Only patients for whom both APS and TMB value were available were used to calculate the TIGS . The median of TMB or TIGS was used as the threshold to separate the TMB-High and TMB-Low groups or the TIGS-High and TIGS-Low group in Kaplan-Meier overall survival curve analysis . The immunotherapy clinical response prediction performance of TIGS and APS have been compared with those of the following biomarkers: TMB , TIDE , IFNG , IFNG . GS , ISG . RS , PDL1 , IIS , and CD8 . The TIDE score was calculated using online software that is available on the website http://tide . dfci . harvard . edu . We followed the instructions on the website to generate input data for TIDE score calculation and exported the results to CSV files . The TIDE scores in the result files were used to predict response . The calculation of scores for the gene-expression-profiling-based biomarkers ( i . e . IFNG , CD8 , and PDL1 ) has been described by Jiang et al . ( 2018 ) . The average expression values among all members defined by the original publications were used to quantify each biomarker . The interferon gamma gene expression signature ( Ayers et al . , 2017 ) ( IFNG ) used genes IFNG , STAT1 , IDO1 , CXCL10 , CXCL9 , and HLA-DRA . The calculation of IFNG . GS and ISG . RS scores were previously described in Benci et al . ( 2019 ) . CD8 used genes CD8A and CD8B . PDL1 used gene CD274 . As a negative control , we performed GSVA with 18 randomly selected genes , and the resulting score was named ‘APSr’ here . This GSVA with random genes was repeated for 100 times , and APSr were used to predict immunotherapy response . The average AUC of these 100 APSr is shown . Univariate cox analysis was performed by R package survival . P values were adjusted using the FDR method , and FDR < 0 . 1 is considered statistically significant . Hazard ratios and their 95% confidence intervals for TCGA cancer types were collected and used for meta-analysis with the random effect model in the R package metafor ( Viechtbauer , 2010 ) . The receiver operator characteristic ( ROC ) curve was generated by plotting the rate of response at various threshold settings of TMB , TIDE or TIGS within the R package pROC ( Robin et al . , 2011 ) . The area under the curve ( AUC ) was reported for each analysis . On the basis of the median of TMB , TIDE or TIGS , we separated patients into High and Low group in the survival analysis . Keplan-Meier curves of overall survival were thus plotted with log-rank test p-value in the R package ggpubr . For GSEA enrichment analysis , we compared samples that had APS above the median with those that had APS below the median across TCGA tumor types using the limma package ( Ritchie et al . , 2015 ) . Genes with p-value < 0 . 01 and FDR < 0 . 05 were ranked by logFC from top to bottom and then inputted into the GSEA function of the R package clusterProfiler ( Yu et al . , 2012 ) with custom gene sets downloaded from Molecular Signature Database v6 . 2 ( Liberzon et al . , 2015; Subramanian et al . , 2005 ) . Normalized enrichment score ( NES ) was used to rank the differentially enriched gene sets . Correlation analysis was performed using the spearman method . All reported p-values are two-tailed , and for all analyses , p<=0 . 05 is considered statistically significant , unless otherwise specified . Statistical analyses were performed using R ( version 3 . 6 . 0 ) . All of the code and data used to generate the figures are freely available at https://github . com/XSLiuLab/tumor-immunogenicity-score ( Wang , 2019; copy archived at https://github . com/elifesciences-publications/tumor-immunogenicity-score ) . Analyses can be read online at https://xsliulab . github . io/tumor-immunogenicity-score/ . Source data files have been provided for Figures 1 , 2 , 4 and 5 . | In the last decade a new kind of cancer therapy , called immunotherapy , has changed how doctors treat cancer patients . These therapies mean that previously incurable cancers , including some skin and lung cancers , can now sometimes be cured . Immunotherapy does this by activating the patient’s own immune system so that it will attack the cancer cells . But for this to work , the cancer cells , much like invading bacteria or viruses , need to be recognized as foreign . Cancer cells contain many DNA mutations that cause the cell to make mutated proteins it would not normally make . These proteins betray the cancer cells as foreign to the immune system . The extent to which cancer cells make mutated proteins – also called the ‘tumor mutational burden’ – can sometimes predict whether a patient will respond to immunotherapy . In general , patients with a high mutational burden respond well to immunotherapy , but overall fewer than one in five cancer patients are cured by this treatment . An important question is whether there are better ways of predicting if a cancer patient will respond to immunotherapy . Wang et al . have addressed this problem by adding a second variable to the prediction . Not only do cancer cells have to make mutated proteins , but these proteins also have to be ‘seen’ by immune cells . Cancer cells , like normal cells , have mechanisms to present protein fragments to immune cells . Wang et al . hypothesized that patients with a high mutational burden would not respond to immunotherapy if they were lacking the machinery required for presenting protein fragments . The experiments revealed that measuring both tumor mutational burden and the levels of the machinery that presents protein fragments resulted in better predictions of patients’ responses to immunotherapy than measuring tumor mutational burden alone . Additionally , this new way of predicting responses to immunotherapy was successful across many different cancer types . The combined measurement of these two variables could be applied in clinical practice as a way to predict cancer patients’ response to immunotherapy . This should allow doctors to determine which course of treatment will work best for a specific patient . The results also suggest that inducing tumor cells to produce more of the machinery that presents protein fragments to the immune system could increase their responsiveness to immunotherapy . In the future , predicting how well a patient will respond to immunotherapy could become even more accurate by incorporating additional variables . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics",
"cancer",
"biology"
] | 2019 | Antigen presentation and tumor immunogenicity in cancer immunotherapy response prediction |
Regulation of cytoplasmic dynein's motor activity is essential for diverse eukaryotic functions , including cell division , intracellular transport , and brain development . The dynein regulator Lis1 is known to keep dynein bound to microtubules; however , how this is accomplished mechanistically remains unknown . We have used three-dimensional electron microscopy , single-molecule imaging , biochemistry , and in vivo assays to help establish this mechanism . The three-dimensional structure of the dynein–Lis1 complex shows that binding of Lis1 to dynein's AAA+ ring sterically prevents dynein's main mechanical element , the ‘linker’ , from completing its normal conformational cycle . Single-molecule experiments show that eliminating this block by shortening the linker to a point where it can physically bypass Lis1 renders single dynein motors insensitive to regulation by Lis1 . Our data reveal that Lis1 keeps dynein in a persistent microtubule-bound state by directly blocking the progression of its mechanochemical cycle .
Cytoplasmic dynein ( ‘dynein’ here ) , the largest and least understood of the cytoskeletal motors , uses the energy from ATP hydrolysis to move towards the minus ends of microtubules ( Vale , 2003; Carter , 2013 ) . As the major minus-end-directed motor in most eukaryotic cells , dynein's many roles include transporting a range of macromolecular cargo ( Blocker et al . , 1997; Jordens et al . , 2001; Kural et al . , 2005; Pilling et al . , 2006; Driskell et al . , 2007 ) , constructing and positioning the mitotic spindle ( Heald et al . , 1996; Merdes et al . , 1996; Kiyomitsu and Cheeseman , 2013 ) , and polarizing and anchoring mRNAs during development ( Wilkie and Davis , 2001 ) . To perform its diverse biological functions , dynein partners with a range of regulatory co-factors , an important subset of which can alter dynein motility directly . Despite progress in understanding the architecture and mechanism of dynein's large motor domain , how this structure is acted upon by regulatory factors is not yet known . Dynein is a homodimer of force generating units ( ∼500 kDa each ) ( Figure 1A , B ) . The N-terminal region of each monomer forms the ‘tail’ domain , which mediates dimerization and cargo attachment via adaptor proteins . Removal of the tail yields the ‘motor’ , the minimal portion of dynein that can exert force . At the core of the motor are six AAA+ modules ( AAA1–6 ) that fold into a ring . AAA1 is the main site of ATP hydrolysis for motility but AAA2 , 3 , and 4 can also bind ATP , and AAA3 and 4 can hydrolyze it ( Gibbons et al . , 1991; Kon et al . , 2004 , 2012; Cho et al . , 2008; Schmidt et al . , 2012 ) . AAA5 and AAA6 have lost the ability to bind nucleotide ( Kon et al . , 2012; Schmidt et al . , 2012 ) . Two appendages to the ring are essential for dynein function; the ‘stalk’ , an intramolecular anti-parallel coiled-coil at the end of which lies the microtubule-binding domain ( Gee et al . , 1997; Carter et al . , 2008 ) and the ‘linker’ , which is dynein's key mechanical element and is an elongated structure N-terminal to AAA1 . The linker spans the ring and moves in a nucleotide dependent manner that is thought to transmit force to dynein's cargo ( Burgess et al . , 2003; Kon et al . , 2005; Shima et al . , 2006; Roberts et al . , 2009 , 2012 ) . In order for dynein to move along microtubules , ATP binding/hydrolysis at AAA1 must be coupled with linker motion and microtubule binding and release at the tip of the stalk , located 250 Å away ( Gibbons et al . , 2005; Imamula et al . , 2007; Kon et al . , 2009; Redwine et al . , 2012 ) . 10 . 7554/eLife . 03372 . 003Figure 1 . The binding of Lis1 to dynein changes the position of dynein's linker domain . ( A ) Domain organization of dynein and Lis1 constructs used in this study . Dynein's AAA+ domains are labeled AAA1–6 . MTBD: microtubule binding domain; CC: coiled coil; LisH: Lis-homology ( dimerization ) motif . ( B ) Schematic representation of dynein and Lis1 , color-coded as in ( A ) and throughout the paper . NT: N terminus; CT: C terminus . ( C ) Cryo-NS EM reconstruction of the dynein motor domain in complex with Lis1 and ( D ) of the motor domain alone . AAA4 and AAA5 are labeled . A density present only in the dynein–Lis1 map is highlighted in ( C ) ( brown arrowhead ) . The linker occupies different positions in the two maps ( compare labeled densities and gray arrows ) , and its position in the dynein alone map is sterically incompatible with Lis1 , as indicated by a semi-transparent Lis1 density . ( E ) Structural model of dynein's motor domain docked into the EM maps of dynein–Lis1 and ( F ) dynein alone . The model was built from crystal structures of the S . cerevisiae dynein ring ( PDB ID: 4AKG [Schmidt et al . , 2012] ) and D . discoideum linker aligned to the yeast linker position ( PDB ID: 3VKG [Kon et al . , 2012] ) , the D . discoideum linker being closer in length to that in our EM construct . In ( E ) , a homology model of the S . cerevisiae Lis1 β-propeller ( brown ) has been docked into the new density highlighted in ( C ) . The linker domain in the EM map ( gray arrow ) is shifted away from its position in the crystal structure ( purple arrow ) , which protrudes from the EM density and clashes with the Lis1 density . In contrast , the linker is within the EM density in the dynein alone map ( F ) . Green circle: location of known interactions between the linker and AAA5 module in dynein ( Schmidt et al . , 2012 ) . ( G ) Close-up view of the Lis1 density with homology model docked in , viewed along the axis indicated by the arrowhead in ( C ) . ( H ) A rotated , smaller view of ( E ) , showing the interface between Lis1 ( brown arrowhead ) and dynein . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 00310 . 7554/eLife . 03372 . 004Figure 1—figure supplement 1 . Three-dimensional ( 3D ) classification and refinement of the dynein and dynein–Lis1 reconstructions . ( A ) SDS-PAGE of dynein and Lis1 , affinity purified from S . cerevisiae . ( B ) Comparison between re-projections of the dynein and dynein–Lis1 reconstructions and the best-matching reference-free class averages ( no nucleotide conditions ) . ( C ) Fourier Shell Correlation plots for all EM maps presented . The plots are shown as a function of resolution ( 1/frequency ) . The thresholds for the 0 . 5 FSC and 0 . 143 FSC criteria are shown . ( D and E ) Class averages of monomeric dynein in complex with dimeric ( D ) or monomeric ( E ) Lis1 , with a schematic representation alongside each . ( F–H ) 3D classification of linker positions in the dynein–Lis1 ( no nucleotide ) data set . ( F ) A reconstruction using the entire data set without sorting out linker conformations . The N-terminus of the linker is mostly averaged out in this map . ( G ) 3D class average of a subset of dynein motors whose linkers are located on the AAA5 proximal side of Lis1 . The purple arrow points to the N-terminus of the linker . ( H ) The linker density from ( G ) was overlaid on the dynein–Lis1 structure ( Figure 1C ) to highlight the different positions adopted by the linker in the presence of Lis1 . The purple and gray arrows point to the N-terminus of the linker domain in the two conformations . The Lis1 density is indicated in all three maps . ( I ) A side view of the dynein motor domain in surface representation with the linker domain colored by conservation ( 100% , purple; 11% , white ) . The alignment was carried out using cytoplasmic dynein from the following species: M . musculus , H . sapiens , S . cerevisiae , C . albicans , A . fumigatus , A . nidulans , D . discoideum , D . melanogaster , and C . elegans using Muscle ( Edgar , 2004 ) . For orientation , a cartoon representation of the view is shown bottom right . The truncation site for the short linker dynein is marked with a dashed line and scissors ( see Figure 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 00410 . 7554/eLife . 03372 . 005Figure 1—figure supplement 2 . The linker's displaced position in the presence of Lis1 does not appear to involve a specific interaction with AAA4 . ( A ) Zoomed out view of dynein–Lis1; only the portion of the crystal structure corresponding to AAA4 is displayed , in yellow ( PDB ID: 4AKG [Schmidt et al . , 2012] ) . ( B ) Close-up of the N-terminal portion of the linker ( left ) and a view rotated by 60° ( right ) . The latter shows a density connecting the linker and AAA4 . Residues in an AAA4 helix that are located in the density connecting AAA4 to the linker are shown in atomic representation , colored by element , and labeled . ( C ) Kymographs of in vitro motility experiments with TMR-labeled wild-type GST-dynein331kDa or mutant GST-dynein331kDa with the five residues labeled in ( C ) changed to alanine ( AAA4 mut ) . Assays were performed with dynein alone or in the presence of 200 nM Lis1 . Horizontal scale bar = 2 μm , vertical = 30 s . ( D ) Histogram of mean velocities for each experiment ± S . D . , N = 228–612 , ***p < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 005 Across a wide range of species , dynein interacts with a conserved regulator called Lis1 ( also known as Pac1 in Saccharomyces cerevisiae ) that is necessary for many dynein driven processes . Mutations in the Lis1 gene cause the neurodevelopmental disorder lissencephaly ( Reiner et al . , 1993 ) . Lis1 is the only dynein regulator known to interact directly with its motor domain ( McKenney et al . , 2010; Huang et al . , 2012 ) . Like dynein , Lis1 acts as a dimer , with each monomer comprising an N-terminal dimerization domain ( LisH ) followed by a coiled-coil , a flexible loop and a C-terminal β-propeller domain of 7 WD motifs ( Kim et al . , 2004; Tarricone et al . , 2004 ) ( Figure 1A , B ) . We previously showed that the propeller domain alone can regulate dynein in vitro and used negative stain electron microscopy ( EM ) and two-dimensional ( 2D ) image processing to show that Lis1 binds to dynein's motor domain near AAA3/4 ( Huang et al . , 2012 ) . We and others have shown that Lis1 induces a slow-moving microtubule-attached state in dynein ( Yamada et al . , 2008; McKenney et al . , 2010; Torisawa et al . , 2011; Huang et al . , 2012 ) . Interestingly , Lis1 can accomplish this without substantially affecting dynein's overall ATP hydrolysis rate ( Yamada et al . , 2008; McKenney et al . , 2010; Huang et al . , 2012 ) . This led us to propose that Lis1 acts as a ‘clutch’ , uncoupling dynein's engine ( AAA+ ring ) from its track-binding region . Our previous 2D EM data , which indicated that Lis1 binds in the vicinity of AAA3/4 , raised at least two possibilities for how Lis1 can affect dynein's mechanochemistry . On the one hand , Lis1 may regulate dynein allosterically , influencing the structure or motions of AAA3 or AAA4 in the ring , and thus preventing the propagation of a signal for microtubule detachment to the stalk . Alternatively , because the linker domain lies close to AAA4 in certain nucleotide-states , Lis1 may regulate dynein sterically , affecting the linker's movement directly , and thus dynein's mechanochemistry . Establishing Lis1's mode of regulation is not possible without three-dimensional ( 3D ) data , as it is not known if Lis1 binds on the same face of the AAA+ ring as the linker , as would be required for direct Lis1–linker interactions to occur . Moreover , it is not clear to what extent the Lis1 binding site encompasses AAA3 , AAA4 , or both of these modules . Thus , 3D structural information is critical to understanding the mechanistic basis of Lis1's regulation of dynein . We set out to establish how Lis1 induces a persistent microtubule-bound state in dynein . We obtained the 3D structure of S . cerevisiae dynein bound to Lis1 to determine which elements of the motor Lis1 directly affects . Our structure revealed that Lis1 sterically prevents the linker from reaching its normal post-powerstroke locations on the ATP hydrolyzing ring that are involved in its conformational cycle . Structure-based mutagenesis also allowed us to identify residues in Lis1 responsible for binding to dynein . Single molecule analysis of a dynein motor with a shortened linker that can physically bypass Lis1 indicated that removing the steric block renders dynein insensitive to Lis1 . Our combined data show that Lis1 directly blocks dynein's mechanochemical cycle , inducing a persistent microtubule-bound state , by acting on its linker domain .
In order to visualize the spatial relationship between Lis1 and dynein's multiple structural elements and to better understand the mechanism by which Lis1 regulates dynein , we used cryo-negative stain ( cryo-NS ) EM and single-particle image processing to obtain the 3D structure of the dynein–Lis1 complex ( Figure 1C ) . Cryo-NS combines the structural preservation of vitrification with the high contrast provided by the negative stain ( De Carlo and Stark , 2010 ) . We found this increased contrast to be instrumental to our ability to computationally sort the different conformations that co-existed in most of our samples . We also determined a 3D map of dynein alone , as a reference , to establish whether Lis1 alters dynein's structure ( Figure 1D ) . We used a well-characterized monomeric dynein construct ( Reck-Peterson et al . , 2006 ) but chose to use dimeric rather than monomeric Lis1 for our reconstructions . We previously showed that while a Lis1 monomer is sufficient to slow down dynein , a much higher concentration of it is required ( Huang et al . , 2012 ) , presumably due to the high local concentration of the β-propeller in the context of a Lis1 dimer . Using a Lis1 dimer at much lower concentrations allowed us to minimize the background in our images . Both dynein and Lis1 were expressed from S . cerevisiae at their genomic loci ( Table 1 ) . We imaged dynein–Lis1 and dynein alone in the absence of nucleotide and obtained structures at resolutions of 21 Å for the complex ( Figure 1C , E , Figure 1—figure supplement 1C and Video 1 ) and of 15 Å for dynein alone ( Figure 1D , F , Figure 1—figure supplement 1C and Video 2 ) . The dynein alone map accommodates the crystal structures of the dynein motor domain well ( Kon et al . , 2012; Schmidt et al . , 2012 ) ( Figure 1F ) , with a Fourier Shell Correlation between the EM map and yeast motor domain structure ( Schmidt et al . , 2012 ) of 0 . 143 at a resolution of 18 . 8 Å . 10 . 7554/eLife . 03372 . 006Table 1 . Yeast strainsDOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 006StrainGenotypeFigure ( s ) ReferenceRPY753MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-GFP-3xHA-GST-DYN1331kDa-gs-DHA , pac1Ä::URA3 , ndl1Ä::cgLEU2Figure 2 , Figure 2—figure supplement 1 , 2 , Figure 5—figure supplement 1Huang et al . , 2012RPY816MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1 , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigures 1–5 , Figure 2—figure supplement 1 , 2 , Figure 1—figure supplement 1 , Figure 4—figure supplement 1 , Figure 5—figure supplement 1Julie Huang , Harvard Medical SchoolRPY842MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1-g-1xFLAG-ga-SNAP-KanR , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigures 3 , 5 , Figure 3—figure supplement 1 , Figure 5—figure supplement 1Huang et al . , 2012RPY844MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDa , pac1Ä::HygroRFigures 1 , 4 , Figure 1—figure supplement 1 , Figure 3—figure supplement 1Huang et al . , 2012RPY1198MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDa-gs-DHA-KanR , pac1Ä::HygroRFigure 5 , Figure 5—figure supplement 1Huang et al . , 2012RPY1245MATa , ura3-52 , lys2-801 , leu2-Ä1 , his3-Ä200 , trp1-Ä63 , SPC110-GFP::TRP1 , HXT1-tdTomato::HIS3Figure 2Jeff Moore , University of ColoradoRPY1248MATa , ura3-52 , lys2-801 , leu2-Ä1 , his3-Ä200 , trp1-Ä63 , SPC110-GFP::TRP1 , HXT1-tdTomato::HIS3 , dyn1Ä::URA3Figure 2This workRPY1302MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-DYN1331kDa , pac1Ä::HygroRFigures 1 , 3This workRPY1400MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDa-L2441ybbR , pac1Ä::HygroRFigure 3 , Figure 3—figure supplement 1This workRPY1422MATa , his3-11 , 15 , ura3-52 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-GFP-3xHA-DYN1314kDa-gs-DHA , pac1Ä::HygroRFigures 4 , 5 , Figure 4—figure supplement 1 , Figure 5—figure supplement 1This workRPY1436MATa , his3-11 , 15 , ura3-52 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev- DYN1314kDa , pac1Ä::HygroRFigure 5This workRPY1439MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-GFP-3xHA-GST-DYN1314 kDa-gs-DHA-KanR , pac1Ä:URA3 , ndl1Ä::cgLEU2Figure 5—figure supplement 1This workRPY1509MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-DYN1331kDa-gs-DHA-KanR , pac1Ä::HygroRFigure 5—figure supplement 1This workRPY1510MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-DYN1314kDa-gs-DHA-KanR , pac1Ä::HygroRFigure 5—figure supplement 1This workRPY1523MATa , ura3-52 , lys2-801 , leu2-Ä1 , his3-Ä200 , trp1-Ä3 , SPC110-GFP::TRP1 , HXT1-tdTomato::HIS3 , pac1Ä::URA3Figure 2This workRPY1524MATa , ura3-52 , lys2-801 , leu2-Ä1 , his3-Ä200 , trp1-Ä63 , SPC110-GFP::TRP1 , HXT1-tdTomato::HIS3 , PAC1R378AFigure 2This workRPY1525MATa , ura3-52 , lys2-801 , leu2-Ä1 , his3-Ä200 , trp1-Ä63 , SPC110-GFP::TRP1 , HXT1-tdTomato::HIS3 , PAC1R275A , R301A , R378A , W419A , K437AFigure 2This workRPY1543MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1R275A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2—figure supplement 1This workRPY1544MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1R378A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2 , Figure 2—figure supplement 1 , 2This workRPY1545MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1W419A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2—figure supplement 1This workRPY1546MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1K437A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2—figure supplement 1This workRPY1547MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1 R275A , R301A , R378A , W419A , K437A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2 , Figure 2—figure supplement 1 , 2This workRPY1548MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-PAC1R301A , dyn1Ä::cgLEU2 , ndl1Ä::HygroRFigure 2—figure supplement 1This workRPY1553MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDaE1849Q , pac1Ä::HygroRFigure 4 , Figure 4—figure supplement 1This workRPY1554MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDaE2819Q , pac1Ä::HygroRFigure 4 , Figure 4—figure supplement 1This workRPY1555MATa , his3-11 , 15 , ura3-52 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-GFP-3xHA-DYN1314kDaK3438E , R3445E , F3446D-gs-DHA , pac1Ä::HygroRFigure 4—figure supplement 1 , Figure 5—figure supplement 1This workRPY1557MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PAC11-13xMYC-TRP1 , PGAL1-ZZ-Tev-GFP-3xHA-DYN1331kDaK3438E , R3445E , F3446D-gs-DHA-KanR , pac1Ä::HygroRFigure 4 , Figure 4—figure supplement 1This workRPY1623MATa , his3-11 , 15 , ura3-1 , leu2-3 , 112 , ade2-1 , trp1-1 , pep4Ä::HIS5 , prb1Ä , PGAL1-ZZ-Tev-GFP-3xHA-GST- DYN1331kDaR2857A , N2858A , K2859A , R2861A , S2862A-gs-DHA , pac1Ä::URA3 , ndl1Ä::cgLEU2Figure 1—figure supplement 2This workDYN1 , PAC11 , PAC1 , and NDL1 encode the dynein heavy chain , dynein intermediate chain , Lis1 and Nudel orthologs , respectively . DHA , SNAP , and ybbR refer to the HaloTag ( Promega ) , SNAP-tag ( NEB ) , and ybbR tag ( Yin et al . , 2005 ) , respectively . TEV indicates a Tev protease cleavage site . PGAL1 denotes the galactose promoter , which was used for inducing strong expression of Lis1 and dynein motor domain constructs . Genes encoding proteases Pep4 and Prb1 were deleted as noted . Amino acid spacers are indicated by g ( glycine ) , ga ( glycine-alanine ) , and gs ( glycine-serine ) . 10 . 7554/eLife . 03372 . 007Video 1 . The three-dimensional structure of dynein–Lis1 . The movie shows the 3D reconstruction of dynein in complex with Lis1 with 360° rotation about the Y-axis . After this rotation , the EM density is made transparent to display the docked dynein crystal structure model and Lis1 homology model and is again rotated by 360° about the Y-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 00710 . 7554/eLife . 03372 . 008Video 2 . The three-dimensional structure of dynein . The movie shows the 3D reconstruction of dynein alone with 360° rotation about the Y-axis . After this rotation , the EM density is made transparent to display the docked dynein crystal structure model and is again rotated by 360° about the Y-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 008 The dynein–Lis1 map shows two major differences relative to the dynein alone reconstruction . First , a prominent donut-shaped density is resolved in contact with the dynein ring , adjacent to the stalk ( Figure 1C , brown arrowhead ) . This extra density matches the dimensions of a β-propeller , including the hole at its center ( Figure 1E , G , H ) . We thus conclude that the density corresponds to Lis1 . The second , and striking difference between the two maps is in dynein itself: Lis1 binds on the same face of the ring where dynein's linker domain is located and the linker is displaced by ∼44 Å in the dynein–Lis1 map relative to the dynein alone reconstruction ( Figure 1C–F ) . Our previous 2D image analysis of the dynein–Lis1 complex did not allow us to determine whether their interaction involved one or both of Lis1's β-propellers or whether Lis1's N-terminal LisH dimerization domain was part of the interaction as well . The extra density in the dynein–Lis1 3D map fits well a single homology model of the S . cerevisiae Lis1 β-propeller built from the crystal structure of the mouse protein ( Tarricone et al . , 2004 ) ( Figure 1E , G , H ) . Because our map resolved the hole at the center of the propeller , the homology model could be unambiguously docked within the density in terms of its translation ( Figure 1G ) . In Lis1 , the β-propeller is connected to the N-terminal LisH dimerization domain by a loop , predicted to be flexible , and a coiled coil ( Kim et al . , 2004; Tarricone et al . , 2004 ) . Consequently , the rest of Lis1 would be expected to adopt a wide range of positions relative to the dynein-bound propeller domain . In agreement with this , we did not resolve density beyond that of the single β-propeller in our dynein–Lis map . 2D image analysis of dynein–Lis1 complexes with either monomeric or dimeric Lis1 showed the same density and location for Lis1 ( Figure 1—figure supplement 1D , E ) , further supporting a stoichiometry of one Lis1 propeller to one dynein motor domain . Our structure of the dynein–Lis1 complex shows that the Lis1 β-propeller contacts dynein primarily at a surface-exposed helix at the junction of AAA3 and AAA4 ( Figure 2A and Video 3 ) , explaining why mutagenesis of four conserved , charged residues ( KDEE ) on this helix virtually abolished Lis1 binding and dynein regulation ( Huang et al . , 2012 ) . Since the resolution of the reconstruction does not allow us to determine unambiguously the rotational orientation of the Lis1 homology model within the corresponding density , we used mutagenesis to probe the dynein–Lis1 interface and further constrain our model of the complex . 10 . 7554/eLife . 03372 . 009Figure 2 . Disrupting the putative dynein–Lis1 interface impairs Lis1's ability to bind to and regulate dynein . ( A ) The Lis1 β-propeller engages dynein primarily at a surface helix connecting AAA3 and AAA4 ( yellow arrowhead , see Video 3 ) . Inset: a zoomed out view . ( B ) ( Left ) View along the axis highlighted in ( A ) by the eye/arrow; ( right ) rotated view . Except for the helix ( yellow ) , the dynein density was removed for clarity . Five conserved residues on Lis1 that were mutated to alanine , either in combination ( Lis15A ) or individually , are labeled and shown in atomic representation . Also displayed are residues ( KDEE ) in dynein known to be involved in the interaction with Lis1 ( Huang et al . , 2012 ) . Basic and acidic residues are labeled in blue and red , respectively . ( C ) No co-migration of dynein and Lis1 was detected by size-exclusion chromatography with the Lis15A and Lis1R378A mutants . Traces show elution profiles of GST-dynein331kDa ( ‘dynein’ ) with wild-type Lis1 ( black ) , Lis15A ( purple ) and Lis1R378A ( green ) . SDS-PAGE for collected fractions are shown below . ( D ) Kymographs of in vitro motility experiments with TMR-labeled GST-dynein331kDa alone or in the presence of 200 nM wild-type or mutant Lis1 . Horizontal scale bar = 2 μm , vertical = 20 s , N = 274–542 . ( E ) In vivo spindle oscillation assays comparing S . cerevisiae strains carrying either wild-type or mutant Lis1 or full deletions of dynein or Lis1 . Inset is a Z-projection of a dividing cell with markers for the membrane ( purple ) and spindle pole bodies ( SPBs ) ( cyan ) . BN = bud neck . Bud neck crossings by the SPBs were counted over 20 min . WT N = 53 , DyneinΔ N = 32 , Lis1Δ N = 55 , Lis1R378A N = 58 , Lis15A N = 47 . For each strain the mean and SE are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 00910 . 7554/eLife . 03372 . 010Figure 2—figure supplement 1 . Probing of the proposed dynein–Lis1 interface by mutagenesis . ( A ) Sequence identity ( 100% , purple; 0% , white ) mapped onto the Lis1 homology model . The alignment was carried out with the following species: M . musculus , H . sapiens , S . cerevisiae , A . nidulans , D . discoideum , D . melanogaster , X . laevis , C . elegans , and D . rerio using Muscle ( Edgar , 2004 ) . ( B ) The Lis1 homology model was docked into the dynein–Lis1 map in two possible orientations , with either the top ( middle panel ) or bottom ( right panel ) face interacting with dynein . The rotations relating the two orientations are indicated . A cross-correlation coefficient calculated for each fit is shown below the structures . These coefficients were calculated between the Lis1 EM density and the homology model filtered to the same resolution ( 21 Å ) ( as implemented in UCSF Chimera [Pettersen et al . , 2004] ) . ( C ) Size-exclusion chromatography traces for wild-type and Lis1 mutants . ( D ) SDS-PAGE of size-exclusion chromatography fractions for GST-dynein331kDa ( shortened to ‘dynein’ in the figure ) mixed with wild-type of mutant Lis1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01010 . 7554/eLife . 03372 . 011Figure 2—figure supplement 2 . Velocity distributions for dynein alone or in the presence of wild-type or mutant Lis1 . Histogram showing the velocity distribution of single TMR-labeled GST-dynein331kDa molecules in the absence of Lis1 ( black ) and with 200 nM wild-type Lis1 ( light gray ) , Lis1R378A ( medium gray ) and Lis15A ( dark gray ) . Velocity distributions were unimodal and could be well fit by a single Gaussian ( R2 values between 0 . 8221 and 0 . 9937 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01110 . 7554/eLife . 03372 . 012Video 3 . The dynein–Lis1 interface . The movie shows the 3D reconstruction of dynein–Lis1 , with the crystal structure of the dynein motor domain and the Lis1 homology model docked in . After a few frames , the EM density disappears to show only the atomic structures and the view changes to show the interaction between dynein and Lis1 in closer detail , finishing with an open-book view of Lis1 . The conserved residues that were mutated in Lis1 are annotated as well as the conserved residues in the AAA4 helix in dynein that have been shown to be necessary for Lis1 binding . Note: the rotational fit of the Lis1 propeller within the Lis1 EM density is uncertain at the current resolution of the dynein–Lis1 map . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 012 Within Lis1 , sequence conservation is much greater on one face of the β-propeller ( ‘top’ ) compared to the other ( Figure 2—figure supplement 1A ) , suggesting that this top face may interact with dynein . Consistent with this idea , the docked β-propeller showed a better qualitative fit and a slightly higher cross-correlation coefficient with our density map when the top face is placed at the dynein interface ( Figure 2—figure supplement 1B ) . To test this docking orientation , we mutated highly conserved residues on the top face of the propeller ( Figure 2B ) . Our previous finding that the KDEE residues in dynein are critical for Lis1 binding ( Huang et al . , 2012 ) suggested that interactions between Lis1 and dynein have an electrostatic component . We therefore targeted four positively charged residues on the top propeller face , as well as a surface tryptophan , all of which are conserved ( Figure 2B ) . We mutated these residues to alanine , both singly and in combination . We first used size-exclusion chromatography to test the ability of the Lis1 mutants to interact with a functional dimerized dynein construct ( GST-dynein331kDa ) ( Reck-Peterson et al . , 2006 ) . When all five residues are mutated to alanine ( Lis15A ) , no binding could be detected by size-exclusion chromatography ( Figure 2C and Table 2 ) . We also did not detect an interaction with two of the single point mutants , Lis1R378A and Lis1W419A ( Figure 2C and Figure 2—figure supplement 1D ) . The remaining single point mutants showed decreased but detectable binding to dynein ( Table 2 and Figure 2—figure supplement 1D ) . Thus , we conclude that highly conserved amino acids on the top face of Lis1's β-propeller are critical for dynein binding , in support of our structural model for the dynein–Lis1 complex ( Figure 2A ) . 10 . 7554/eLife . 03372 . 013Table 2 . Dynein:Lis1 ratios in complexes purified by size-exclusion chromatographyDOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 013GST-dynein331kDaLis1Lis1 ( normalized to WT ratio ) WT Lis10 . 82 ± 0 . 010 . 18 ± 0 . 011 . 00Lis1R275A0 . 85 ± 0 . 010 . 15 ± 0 . 010 . 80Lis1R301A0 . 88 ± 0 . 010 . 12 ± 0 . 010 . 62Lis1R378A1 . 00 ± 0 . 000 . 00 ± 0 . 000 . 00Lis1W419A1 . 00 ± 0 . 000 . 00 ± 0 . 000 . 00Lis1K437A0 . 85 ± 0 . 010 . 15 ± 0 . 010 . 80Lis15A1 . 00 ± 0 . 000 . 00 ± 0 . 000 . 00In relation to Figure 2 and Figure 2—figure supplement 1 . Fractions were run on SDS-PAGE , stained with SYPRO red , and the bands corresponding to GST-dynein331kDa and wild-type/mutant Lis1 were quantified using ImageJ . The quantification was done using three adjacent lanes corresponding to the peak from size-exclusion . Values are averages of the three lanes ± SD . The ratio for each mutant normalized against that of wild-type Lis1 is also shown . We next examined if the binding-deficient Lis1 mutants Lis15A and Lis1R378A were able to regulate dynein in vitro . Wild-type Lis1 decreases dynein velocity in vitro in a concentration-dependent manner ( Huang et al . , 2012 ) . These assays , where the motion of single , fluorescently labeled dynein molecules along microtubules is monitored over time , are more sensitive than size-exclusion chromatography for detecting dynein–Lis1 interactions . Therefore , we expected that some of the Lis1 mutants that did not co-migrate with dynein might still exhibit weak but measurable regulation of the motor . The Lis15A mutant showed no reduction in dynein velocity , consistent with an inability to bind dynein ( Figure 2D and Figure 2—figure supplement 2 ) . The Lis1R378A mutant , on the other hand , showed a slight reduction in dynein velocity compared to dynein alone , suggesting that its binding to the motor is compromised ( Figure 2D and Figure 2—figure supplement 2 ) . Thus , the effect of the Lis1 mutations on dynein binding correlates with the ability of the Lis1 mutants to regulate dynein at the single-molecule level . Lastly , we tested our model for the dynein–Lis1 complex by measuring the effect of disrupting the dynein–Lis1 interface in vivo . In yeast , spindle pole bodies ( SPB ) span the nuclear envelope and coordinate microtubule minus ends that emanate from its nuclear and cytoplasmic faces ( Jaspersen and Winey , 2004 ) . Lis1 assists in concentrating dynein at the plus ends of cytoplasmic microtubules , from where dynein is offloaded to the cell cortex ( Lee et al . , 2003; Sheeman et al . , 2003; Roberts et al . , 2014 ) . Cortically anchored dynein exerts a pulling force that results in displacements of the entire mitotic spindle ( Moore et al . , 2009 ) , giving rise to a brief series of oscillations across the bud neck . Deletion of dynein eliminates these oscillations ( Yeh et al . , 1995 ) . We quantified the effect of the Lis15A and Lis1R378A mutants on spindle movement in cells treated with hydroxyurea , which prolongs the period of oscillations , by tracking fluorescently labeled SPBs over the course of 20 min . We found that disruption of the dynein–Lis1 interface resulted in a decrease in the number of bud neck crossings to a level similar to that caused by the deletion of Lis1 ( Figure 2E ) . These results indicate that the dynein–Lis1 interface identified in our structural model is crucial for dynein's biological function . The end of the linker domain is displaced ∼44 Å in the dynein–Lis1 structure relative to the dynein alone map , mainly along the plane of the ring ( Figure 1C , D ) . The structure suggests that this displacement may be a direct result of Lis1's binding to dynein: the linker position in the dynein alone structure is sterically incompatible with the presence of Lis1 ( Figure 1D ) . This is consistent with a model where Lis1 regulates dynein motility through a steric mechanism , by physically blocking the linker's normal position in the no nucleotide state . We next sought to test if Lis1 sterically blocks the linker in other nucleotide states . As the main mechanical element of dynein , the linker is thought to adopt at least two additional conformations during the ATPase cycle . In the presence of ATP ( or ATP plus Vi , which leads to the formation of the transition state analog ADP . Vi ) , the linker is displaced across the ring to a position near AAA2 ( Kon et al . , 2005; Roberts et al . , 2009 , 2012 ) ( the ‘pre-powerstroke’ position ) . In the presence of ADP the linker lies over AAA4 in the crystal structure of the Dictyostelium discoideum dynein ( Kon et al . , 2012 ) , a ‘post-powerstroke’ position slightly different from that seen in the S . cerevisiae dynein crystal structure in the absence of nucleotide , where the linker is docked onto AAA5 ( Schmidt et al . , 2012 ) . However , since dynein from the same species had not been visualized in both the no nucleotide and ADP states , and because different constructs were used in the studies cited above , it was uncertain whether the AAA4 and AAA5 linker positions corresponded to distinct mechanochemical states or were due to differences between dynein species and/or constructs . To address this , we first obtained the structure of S . cerevisiae dynein alone in the presence of ADP . Conformational sorting revealed that the linker adopts two positions in this condition ( Figure 3A ) . One is over AAA4 , coinciding with that observed in the D . discoideum crystal structure . The other is the AAA5-docked position seen in S . cerevisiae dynein with no nucleotide . The AAA4 position was seen only in the presence of ADP and was not detectable in no nucleotide conditions . These results suggest that the linker docks at AAA5 in the absence of nucleotide but can coexist in the AAA4- and AAA5-interacting states in the presence of ADP . Importantly , both the AAA4 linker position in the ADP state and the AAA5 position in the no nucleotide state are sterically incompatible with the presence of Lis1 ( Figure 3B ) . Thus , we conclude that binding of Lis1 to the dynein ring results in a displaced linker , away from its normal docking sites under both no nucleotide and ADP conditions . 10 . 7554/eLife . 03372 . 014Figure 3 . Lis1 sterically blocks the linker domain's normal position on dynein's ring in ADP and no nucleotide conditions but does not prevent it from reaching the pre-powerstroke position at AAA2 . ( A ) Cryo-NS maps of S . cerevisiae dynein in 100 μM ADP displaying the linker next to either AAA5 ( left ) or AAA4 ( right ) . The S . cerevisiae linker domain ( lacking nucleotide at AAA1 , PDB ID: 4AKG [Schmidt et al . , 2012] ) and the D . discoideum linker domain ( with ADP at AAA1 , PDB ID: 3VKG [Kon et al . , 2012] ) are displayed in purple ribbon representation and have been docked into the linker-AAA5 and linker-AAA4 maps , respectively . To enable unambiguous comparison of linker positions between the EM density and crystal structure , we aligned each EM map to the corresponding dynein motor domain crystal structure after computationally removing the linker . ( B ) The dynein maps in no nucleotide ( blue ) and ADP ( purple ) conditions ( the latter with the linker at the AAA4 location ) are overlaid to compare linker positions . The location of Lis1 in the dynein–Lis1 map is shown as a transparent brown density . Both linker positions are sterically incompatible with the presence of Lis1 . Note: since the ADP AAA5 linker position is the same as that seen under no nucleotide conditions , we only show the ADP map with the linker at AAA4 . ( C ) Schematic representation of the dynein FRET construct used to test dynein's linker swing in the presence of Lis1 . eGFP ( green sphere–donor ) was fused to the N-terminus of the linker domain , and TMR ( red sphere–acceptor ) was inserted into the AAA2 domain in the ring . A pre-powerstroke linker position , where the linker moves close to AAA2 in ATP plus vanadate ( Vi ) conditions , would display an increased FRET efficiency between the two fluorophores ( bottom ) relative to the no nucleotide state , where the linker is docked at AAA5 ( top ) . ( D ) FRET efficiency between the eGFP and TMR fluorophores in the absence or presence of 200 μM ATP + Vi and 840 nM Lis1 , ***p < 0 . 001 . The order of addition for the reactions containing ATP + Vi and Lis1 is indicated by arrows . Averages of three experiments ± SD are shown . ( E ) Cryo-NS reconstruction of dynein–Lis1 in ATP + Vi conditions with the crystal structure of the motor domain docked in ( PDB ID: 4AKG [Schmidt et al . , 2012] ) . The Lis1 density is indicated . ( F ) At lower contour levels , the N-terminal portion of the linker can be resolved ( purple arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01410 . 7554/eLife . 03372 . 015Figure 3—figure supplement 1 . FRET analysis of linker movement towards the pre-powerstroke position in the presence of Lis1 . ( A ) Diagram of a microtubule-gliding assay . Monomeric GFP-dynein molecules are immobilized on the coverslip via anti-GFP antibodies ( Y shape ) . Dynein-driven gliding of fluorescently labeled ( purple asterisks ) microtubules is visualized using TIRF microscopy . ( B ) A dynein FRET construct with CoA-TMR inserted into its AAA2 domain ( GFP-dyneinFRET/A2 ) has a microtubule gliding activity similar to that of a control construct lacking it ( GFP-dynein331kDa ) . ( C ) Fitted FRET emission spectra for dynein in different nucleotide conditions ( no nucleotide or 200 μM ATP + Vi ) and with 0 nM or 840 nM Lis1 . Arrows indicate the order of addition for the last two plots . Emission spectra for eGFP and TMR used for the fit are shown as green and red traces , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 015 We then used two approaches to determine if Lis1 also influences the position of the linker at its pre-powerstroke position , near AAA2 . First , we designed a monomeric dynein construct to use fluorescence resonance energy transfer ( FRET ) to measure linker movement to AAA2 . We based our design , which used S . cerevisiae dynein , on a linker sensor developed for D . discoideum dynein ( Kon et al . , 2005 ) . In our construct , we fused an eGFP donor to the N-terminus of the linker and coupled a tetramethylrhodamine ( TMR ) acceptor to AAA2 via a small acetyl-CoA-binding tag ( ybbR [Yin et al . , 2005] ) ( Figure 3C ) . This dynein construct ( GFP-dyneinFRET/A2 ) slides microtubules robustly , with gliding rates ∼90% of wild-type dynein ( GFP-dynein331kDa ) ( Figure 3—figure supplement 1A , B ) , showing that the tags are compatible with motor function . Under no nucleotide conditions , and in the absence of Lis1 , this construct showed a low FRET efficiency ( ∼2% ) , as expected when the linker is docked at AAA5 ( post-powerstroke position ) and the FRET donor and acceptor are far apart ( Figure 3D ) . In the presence of ATP and vanadate ( Vi ) , which trap dynein as an ADP . Vi-bound complex after hydrolysis , FRET increased to ∼26% ( Figure 3D ) . Under these conditions , the linker is biased towards the pre-powerstroke position at AAA2 and the fluorophores lie closer together . When Lis1 was added to ADP . Vi–dynein , there was no significant change in the FRET efficiency relative to ADP . Vi–dynein alone ( Figure 3D ) . When dynein was incubated with Lis1 before adding ATP and Vi , the FRET efficiency decreased somewhat but remained close to that observed for ADP . Vi–dynein alone ( Figure 3D ) . These results suggest that Lis1 does not affect the linker pre-powerstroke AAA2 position ( ATP + Vi added before Lis1 ) and has only a minor effect on linkers undergoing the AAA5 to AAA2 transition ( Lis1 added before ATP + Vi ) . As a second method to determine if Lis1 affects the linker's ability to reach the AAA2 position under ADP . Vi conditions , we determined the EM structure of the ADP . Vi–dynein–Lis1 complex ( Lis1 added before ATP + Vi ) . We could resolve the linker in the expected pre-powerstroke position towards AAA2 and the density for the Lis1 β-propeller at the AAA3/4 junction ( Figure 3E , F ) . In summary , these results indicate that the presence of Lis1 does not interfere with linker movement towards its pre-powerstroke position at AAA2 . In contrast , Lis1 occludes the linker binding sites at its two post-powerstroke positions ( AAA4 and AAA5 ) . We and others have shown that Lis1 reduces dynein's velocity without significantly affecting the motor's overall ATPase rate ( Yamada et al . , 2008; McKenney et al . , 2010; Huang et al . , 2012 ) . However , which of dynein's AAA+ modules is responsible for this continued ATP hydrolysis was not previously addressed . To determine this , we measured microtubule-stimulated ATPase rates , with and without Lis1 , in different monomeric dynein constructs . As expected , dynein monomers continued to hydrolyze ATP in the presence of Lis1 at levels similar to those of dynein alone ( Figure 4A ) . This hydrolysis , however , was virtually abolished , both in the presence and absence of Lis1 , in a construct where AAA1 , the main site of ATP hydrolysis in dynein ( Gibbons et al . , 1987 ) , was rendered hydrolysis-deficient with an E to Q mutation in its Walker B motif ( Kon et al . , 2004 ) ( Figure 4B ) . This result suggests that an intact ATP hydrolysis site at AAA1 is required for ATPase activity in the presence of Lis1 . 10 . 7554/eLife . 03372 . 016Figure 4 . ATP turnover in the presence of Lis1 requires a hydrolysis-competent AAA1 and a functional AAA5 linker-docking site . Microtubule-stimulated ATPase activity of dynein monomers carrying ( A ) wild-type AAA+ modules , ( B ) a hydrolysis deficient E1849Q mutation in AAA1 ( Kon et al . , 2004 ) , ( C ) a hydrolysis deficient E2819Q mutation in AAA4 ( Cho et al . , 2008 ) , ( D ) AAA5 mutations ( K3438E , R3445E , F3446D ) that prevent linker docking ( Schmidt et al . , 2012 ) . ATPase traces are of dynein alone ( light green ) or in the presence of 140 nM Lis1 ( brown ) . Measurements were done in triplicate ( A and C ) or duplicate ( B and D ) from one preparation . Diagrams of the dynein constructs used to generate the plots are shown next to them . See Table 3 for fit equation and rate quantifications . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01610 . 7554/eLife . 03372 . 017Figure 4—figure supplement 1 . Lis1 binds to dynein ATPase mutants . ( A–D ) SDS-PAGE of fractions eluted from size-exclusion chromatography runs of Lis1 mixed with each of the dynein constructs used in the ATPase assays . Lis1 co-elutes with all of the constructs . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01710 . 7554/eLife . 03372 . 018Table 3 . ATPase assay rate measurementsDOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 018SampleKm ( MT ) ( ìM ) kbasal ( Motor domain−1 . s−1 ) kcat ( Motor domain−1 . s−1 ) Full-length linker1 . 06 ± 0 . 163 . 51 ± 0 . 3116 . 75 ± 0 . 49+Lis11 . 09 ± 0 . 204 . 36 ± 0 . 3015 . 06 ± 0 . 49Short linker0 . 92 ± 0 . 104 . 45 ± 0 . 2216 . 98 ± 0 . 32+Lis12 . 05 ± 0 . 447 . 14 ± 0 . 2116 . 12 ± 0 . 61Full-length linker , AAA4 ATPase mutant ( E2819Q ) 1 . 55 ± 0 . 144 . 53 ± 0 . 1718 . 80 ± 0 . 38+Lis11 . 10 ± 0 . 154 . 60 ± 0 . 1913 . 93 ± 0 . 31Data were fit to the following equation: kobs = ( kcat − kbasal ) − [MT]/ ( Km ( MT ) + [MT] ) + kbasal . Km ( MT ) is the microtubule concentration that gives half-maximal activation . Values are the averages of triplicate readings ± SE of the fit . Given that Lis1 binds at AAA4 , one of the hydrolysis-competent AAA+ modules in dynein , it was possible that Lis1 might be stimulating ATP hydrolysis at that site , with AAA1 playing only an indirect role . However , dynein carrying an E to Q mutation in the Walker B motif of AAA4 ( Cho et al . , 2008 ) showed a near wild-type ATPase rate with or without Lis1 ( Figure 4C ) . Therefore , a hydrolysis-competent AAA4 is not required for the ATPase activity observed in the presence of Lis1 . Mutations in AAA5 ( an AAA+ module that cannot bind ATP ) that prevent linker docking have been shown to severely reduce dynein's ATPase activity ( Schmidt et al . , 2012 ) . We wondered whether Lis1 binding might rescue this mutation and restore ATPase activity to dynein . This was not the case; dynein constructs carrying the AAA5 mutation did not hydrolyze ATP even in the presence of Lis1 ( Figure 4D ) . Taken together , these results indicate that sustained ATP hydrolysis in a Lis1-regulated dynein requires a hydrolysis-competent AAA1 and a functional linker-docking site at AAA5 . The experiments discussed above showed that Lis1 does not regulate dynein by affecting the linker's ability to reach its pre-powerstroke position at AAA2 . Our dynein–Lis1 structure shows that Lis1 , however , does affect post-powerstroke linker positions as Lis1 and the linker are sterically incompatible in no nucleotide and ADP conditions ( Figure 1D , E , Figure 3B ) . We wondered whether motility regulation was a result of this steric blocking by Lis1 . Specifically , we wanted to test the hypothesis that steric blocking of the linker is necessary for inducing dynein's Lis1-dependent state of persistent microtubule attachment . To test this hypothesis , we used a dynein construct with a truncated linker that is long enough to form a functional motor but is too short to be sterically blocked by Lis1 . This construct is generated by deleting 145 amino acids at the N-terminus of the dynein motor ( Figure 5A ) . 10 . 7554/eLife . 03372 . 019Figure 5 . A shortened linker that can physically bypass Lis1 renders dynein Lis1 insensitive . ( A ) A short linker construct was designed by docking the crystal structure of the D . discoideum linker ( purple ribbon ) ( PDB ID: 3VKG [Kon et al . , 2012] ) into our EM map of dynein alone and overlaying the position of Lis1 ( brown mesh ) . Truncating the linker at residue 1365 ( dashed line ) yields a linker that is functional ( see Figure 5—figure supplement 1 ) but that can no longer contact Lis1 . ( B ) Cryo-NS reconstruction of the short linker dynein–Lis1 complex; the linker assumes the same conformation with Lis1 bound as in the absence of Lis1 . ( C ) Diagram of the single-molecule microtubule release assay we used to test Lis1 regulation of dynein . Release from microtubules of TMR-labeled ( red asterisk ) dynein monomers on addition of ATP is monitored by TIRF microscopy . ( D ) Diagrams of predicted outcomes . Dynein's linker domain in purple , microtubule in gray , Lis1 in brown . ( i ) Dynein monomers release from microtubules in ATP conditions in the absence of Lis1 . ( ii ) Our model proposes that Lis1 sterically blocks a full-length linker from assuming the normal conformation on dynein's ring , keeping dynein bound to the microtubule . ( iii ) In the absence of Lis1 , shortening the linker would have no effect on dynein's mechanochemical cycle . ( iv ) Our model predicts that a shortened linker that can bypass the Lis1 steric block should render dynein insensitive to Lis1 . ( E ) Kymographs of TMR-labeled full-length ( left ) or short linker ( right ) dynein molecules . After pre-binding to microtubules , release of dynein molecules is monitored after addition of 5 mM ATP , with and without 300 nM Lis1 . Kymographs correspond to the dynein constructs shown in ( D ) . Scale bar = 5 s . ( F ) Quantification of the kymographs in ( D ) , showing the duration of microtubule attachment after addition of ATP , in the absence ( gray ) or presence ( brown ) of Lis1 . Data were binned into 1 s intervals and the histograms show alternating no Lis1 and +Lis1 bars . Rare attachments longer than 10 s were excluded from the analysis and plot , N = 179–183 . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 01910 . 7554/eLife . 03372 . 020Figure 5—figure supplement 1 . The short linker dynein construct shows robust motility , hydrolyzes ATP , and binds Lis1 . ( A ) Single-molecule motility assays . Kymographs of GST-dimerized full-length and short linker dyneins . Horizontal scale bar = 2 μm , vertical = 30 s . ( B ) Velocity and run length for short linker and full-length linker dyneins . Similar values are seen for the two constructs , N = 265–333 . ( C ) SDS-PAGE of elution fractions from size-exclusion chromatography of monomeric short linker dynein mixed with Lis1 . Lis1 co-migrates with short linker dynein . ( D ) Microtubule-stimulated ATPase activity of short linker dynein with either wild-type AAA+ modules ( left ) or with mutations ( K3438E , R3445E , F3446D ) that prevent linker docking at AAA5 ( Schmidt et al . , 2012 ) ( right ) . ATPase traces are of dynein alone ( light green ) or in the presence of 140 nM Lis1 ( brown ) . Measurements were done in triplicate ( wild type ) or duplicate ( AAA5 mutant ) from one preparation . Diagrams of the dynein constructs used to generate the plots are shown next to them . ( E ) Addition of buffer lacking ATP to flow chambers containing TMR-labeled full-length or short linker dynein does not cause microtubule detachment . Scale bar = 5 s . ( F ) The short linker construct used in Figure 5 contains an N-terminal GFP connected to the short linker via 3 HA tags . We expected the GFP to be flexible in its location relative to dynein based on previous EM studies ( Roberts et al . , 2009 ) and the fact that it is averaged out in our 3D reconstruction of dynein ( Figure 1D ) . However , to rule out interference from GFP , we also carried out the microtubule release assays with a GFP-less construct . We observed the same results: addition of ATP to GFP-less dyneins in the presence of 300 nM Lis1 caused short-linker dyneins to release from the microtubule while full-length dyneins remain attached . Scale bar = 5 s . ( G ) Quantification of the kymographs in ( F ) , showing the duration of microtubule attachment of full-length ( light brown ) or short linker ( dark brown ) dynein molecules after addition of ATP , in the presence of 300 nM Lis1 . Data were binned into 1 s intervals and the histograms show alternating full-length and short linker dynein bars . Rare attachments longer than 10 s were excluded from the analysis and plot , N = 141–197 . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 020 We first verified this construct functionally and structurally . A dimeric dynein motor containing this shortened linker shows robust motility properties in in vitro motility assays ( Figure 5—figure supplement 1A , B ) ( Reck-Peterson et al . , 2006 ) . We also tested whether shortening the linker affects the microtubule-stimulated ATPase activity of dynein monomers . Monomers with a short linker showed ATPase levels comparable to those seen with a full-length linker , both in the context of a wild-type set of AAA+ modules and in the linker docking-deficient AAA5 mutant ( Figure 5—figure supplement 1D ) . As a monomer , the short linker construct can bind Lis1 as shown both by their co-migration in size-exclusion chromatography ( Figure 5—figure supplement 1C ) and by our ability to obtain a 3D reconstruction of the short linker dynein–Lis1 complex ( Figure 5B ) . Central to our testing the steric block hypothesis , our 3D structure of the short linker dynein–Lis1 complex shows the same conformation for the linker in the presence of Lis1 as we had observed for the full-length linker in the absence of Lis1 ( Figure 1 , Figure 5A , B ) . Therefore , the short linker is functional and able to physically bypass Lis1 . To directly test whether Lis1 was capable of regulating dynein with a short linker , we used a single-molecule microtubule release assay ( Figure 5C ) . In this study , the duration of single monomeric dynein's attachments to microtubules can be measured by kymograph analysis in a flow chamber by TIRF microscopy ( Huang et al . , 2012 ) . Addition of ATP triggers a low-affinity state in dynein ( Kon et al . , 2005; Imamula et al . , 2007; Huang et al . , 2012 ) and the dynein monomers release from microtubules , resulting in a loss of fluorescence signal . Microtubule rebinding events are short lived , likely corresponding to single turnovers of ATP . In the presence of a full-length linker , Lis1 converted dynein to a state of persistent microtubule attachment and dynein monomers stayed bound in the presence of ATP for extended periods as previously shown ( Figure 5D–F ) ( Huang et al . , 2012 ) . Strikingly , shortening of dynein's linker eliminated Lis1's ability to induce this persistent microtubule-bound state . We quantified the durations of microtubule attachments after the addition of ATP and found the same short-lived attachments seen with dynein in the absence of Lis1 ( Figure 5D–F ) . Thus , Lis1 is not capable of regulating microtubule attachment in the short linker construct . These data support a steric mode of dynein regulation where Lis1 physically blocks the linker .
We previously described Lis1 as a ‘clutch’ for dynein , based on its ability to uncouple the cycles of ATP hydrolysis , which take place in the motor domain , from the cycles of microtubule binding and release at the microtubule binding domain ( Huang et al . , 2012 ) . One of the functional consequences of the dynein–Lis1 interaction is that Lis1 keeps dynein in a persistent microtubule-bound state . In this study , we have determined six 3D EM structures of dynein and dynein–Lis1 in different nucleotide states . By combining these structures with single molecule motility experiments , we have established that Lis1 regulates dynein's microtubule attachment by sterically blocking its linker domain . Together , our data suggest the following model of dynein regulation by Lis1 ( Figure 6 ) . In the current view of dynein's mechanochemical cycle , the motor domain encounters the microtubule with ADP . Pi bound at AAA1 , with the linker in a pre-powerstroke position at AAA2 ( Kon et al . , 2005; Roberts et al . , 2009 , 2012 ) . Strong microtubule binding stimulates Pi release , inducing the linker to swing to AAA4 ( Kon et al . , 2012 ) . Finally , linker docking at AAA5 is thought to promote the release of ADP from AAA1 , resetting the mechanochemical cycle ( Schmidt et al . , 2012 ) . Our data suggest that when Lis1 is present , the linker retains its ability to adopt the pre-powerstroke AAA2 position but is prevented from reaching its normal post-powerstroke positions at AAA4 and AAA5 on dynein's ring ( Figure 6 ) . This blocking of the linker by Lis1 is critical for motility regulation; its removal by shortening dynein's linker renders the motor Lis1 insensitive . 10 . 7554/eLife . 03372 . 021Figure 6 . Model for the regulation of dynein by Lis1 . ( A–G ) Current view of dynein's mechanochemical cycle . ( A ) ATP binding to AAA1 induces the low-affinity conformation in dynein's microtubule-binding domain and ( B ) release from the microtubule . ( C ) The linker domain changes its position from AAA5 towards AAA2 , the ‘pre-powerstroke’ and ATP is hydrolyzed . ( D ) Binding of dynein to a new site on the microtubule triggers a change in the microtubule-binding domain to its high affinity conformation ( E ) . ( F ) Release of Pi results in the ‘powerstroke’ , a movement of the linker back towards AAA5 . ( G ) Docking of the linker at AAA5 is thought to promote nucleotide exchange at AAA1 , resetting the motor for a new cycle . ( H–J ) Model for the Lis1-regulated cycle . Lis1 prevents the linker from completing its normal conformational cycle , keeping dynein in a persistent microtubule-attached state , despite continuing ATP hydrolysis . ( H ) Binding of Lis1 to dynein blocks the linker from docking onto the ring at AAA5 , preventing the conformational changes in the stalk and microtubule binding domain that ultimately result in dynein's release from the microtubule . ( I ) The linker is still capable of moving to the pre-powerstroke position at AAA2 in the presence of Lis1 , and ATP is hydrolyzed . ( J ) Presumably , by analogy to the dynein alone cycle , Pi release triggers the power-stroke , but Lis1 sterically blocks the linker's normal position on dynein's ring in the ADP state . Our current understanding of Lis1 regulation does not yet explain the mechanism of nucleotide exchange at AAA1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03372 . 021 Why does Lis1's blocking the linker from adopting its normal post-powerstroke positions prevent dynein's microtubule detachment ? One possibility suggested by our structures is that Lis1 disrupts the interaction between the linker and AAA5 , preventing normal progression through the mechanochemical cycle . Consistent with this notion , when linker docking at AAA5 is abolished by mutagenesis , dynein displays reduced velocity and prolonged microtubule attachments ( Schmidt et al . , 2012 ) , reminiscent of Lis1's effects . However , while Lis1 has little effect on dynein's ATPase , AAA5 linker docking mutants display severely reduced ATPase rates , both in the absence ( Schmidt et al . , 2012 ) and the presence of Lis1 ( Figure 4D , Figure 5—figure supplement 1D ) . Given these results , it is not clear at this point what the mechanistic basis is for dynein's continuing ATPase in the presence of Lis1 . On the one hand , it is possible that the AAA5 mutations may , in addition to preventing linker docking , disrupt dynein's mechanochemical cycle and thus also prevent ATP hydrolysis . A method to reversibly block linker docking at AAA5 ( e . g . , via a small molecule ) would be required to determine if AAA5 docking is truly required for dynein ATPase activity . On the other hand , Lis1 may uncouple ATP hydrolysis from linker docking at AAA5 through an allosteric effect on the ring . In this scenario , the linker–AAA5 interaction , which is blocked by Lis1 , would be required for the conformational changes that ultimately shift dynein's microtubule-binding domain to its low-affinity state , but not for dynein's continuing ATPase activity . Higher resolution structures of the dynein–Lis1 complex will be required to establish whether Lis1 has an effect on the structure of dynein's ring . It is conceivable that blocking of the normal linker-docking sites by Lis1 might induce a new interaction between the linker and the AAA+ ring . Similarly , Lis1 may interact specifically with the linker itself . Either ( or both ) of these scenarios could in turn be responsible for preventing microtubule release . However , current evidence does not favor these possibilities . Low sequence conservation in the portion of the linker facing Lis1 argues against a specific Lis1–linker interaction ( Figure 1—figure supplement 1I ) . Likewise , a specific interaction between a Lis1-displaced linker and dynein's ring is not supported by the apparent conformational heterogeneity of the N-terminus of the linker in the presence of Lis1 , where 3D sorting is required to resolve linker positions ( Figure 1—figure supplement 1F–H ) . Also mutating five amino acids on AAA4 , proximal to the linker's displaced position ( the most likely candidates to interact with the displaced linker ) , had minimal effect on Lis1-mediated motility regulation ( Figure 1—figure supplement 2 ) . A direct test of whether specific interactions exist among these different elements will also require a higher resolution structure , where the rotational orientation of the Lis1 homology model within its density in the EM map is unequivocal and specific interactions between the linker and Lis1 as well as the linker and the dynein ring can be distinguished from physical proximity . In conclusion , our data show that Lis1 , a conserved dynein regulator , directly disrupts dynein's mechanochemical cycle by physically blocking conformations that are required to couple the cycles of ATP hydrolysis taking place in the motor domain from those of track binding and release happening at the microtubule binding domain . This allows Lis1 to keep dynein in a persistent microtubule-bound state . This modulation of dynein's interaction with its microtubule track likely contributes to dynein's ability to carry out the variety of cellular functions it performs in different organisms , given the conservation of the amino acids at the dynein–Lis1 interface . For example , Lis1 is involved in initiation of cargo transport ( Lenz et al . , 2006; Egan et al . , 2012; Moughamian et al . , 2013 ) , in transport of high load cargo ( McKenney et al . , 2010 ) , and in targeting dynein molecules to the cell cortex via the microtubule plus end ( Lee et al . , 2003; Sheeman et al . , 2003; Roberts et al . , 2014 ) . The displaced linker observed in the presence of Lis1 in our 3D dynein–Lis1 reconstruction may contribute to this latter task , generating an ‘unmasked’ tail domain that has been shown necessary for cortical dynein localization ( Markus and Lee , 2011 ) . In the case of the mammalian proteins , dynein and Lis1 were previously shown to form a stable complex only in ATP and Vi conditions ( McKenney et al . , 2010 ) . Our 3D reconstruction of dynein–Lis1 under those conditions suggests that this might be a consequence of the linker's moving to its pre-powerstroke site at AAA2 , where the linker and Lis1 are no longer sterically incompatible . The work presented here has helped dissect the molecular mechanism by which Lis1 regulates a single dynein motor domain . The next challenge will be to understand the interactions between Lis1 and dynein dimers and of those with other regulatory factors . Future structural studies with full-length dimeric dynein–Lis1–Nudel complexes , free and bound to microtubules , will be required to answer these exciting questions .
The S . cerevisiae strains used in this study are listed in Table 1 . Deletions or modifications of endogenous genomic copies of the dynein heavy chain ( DYN1 ) and Lis1 ( PAC1 ) were done using PCR-based methods as previously described ( Longtine et al . , 1998 ) , using the URA3/5FOA ‘pop-in/pop-out’ method ( Guthrie and Fink , 1991 ) . Transformations were performed using the standard lithium acetate method ( Gietz and Woods , 2002 ) . Point mutants were generated using the PCR stitching method and verified by DNA sequencing . Cultures of S . cerevisiae for protein purification were grown , harvested , and frozen as described previously ( Reck-Peterson et al . , 2006 ) . Dynein and Lis1 constructs were purified and labeled as described previously ( Reck-Peterson et al . , 2006; Huang et al . , 2012 ) , except that a modified TEV buffer for Lis1 purification was used; 50 mM Tris–HCl ( pH 8 . 0 ) , 150 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 5% glycerol , 1 mM DTT , and 1 mM PMSF . We chose to use cryo-NS EM , where a carbon support , combined with a heavy metal stain , resulted in highly reproducible grids with high contrast . Prior attempts at getting dynein reproducibly in open holes for standard cryo-EM were unsuccessful . Furthermore , cryo-EM on continuous carbon gave micrographs where individual dynein particles were difficult to see above the noise . The high reproducibility we were able to achieve with cryo-NS allowed us to sample a much greater range of constructs/nucleotide conditions in the same time frame than we would otherwise have been able to do in unstained , unsupported conditions . Most importantly , the improved contrast was instrumental in allowing us to sort the different dynein conformations that were present in most of our data sets . 4 μl of monomeric dynein ( 80–120 nM ) , or monomeric dynein pre-incubated for 10 min with Lis1 dimer at a 1 . 5-fold excess ( 120–180 nM ) , was applied to a glow discharged , continuous carbon coated , C-flat EM grids ( Protochips , Raleigh , NC ) . Dynein samples stated to be prepared in no nucleotide conditions were treated with apyrase ( 0 . 14 U/ml ) for 15 min prior to grid application to hydrolyze residual ADP left over from the dynein purification procedure . Dynein samples stated to be prepared in ADP and ATP + Vi conditions contained 100 μM ADP and 500 μM Mg-ATP/NaVO4 , respectively . For the latter , nucleotide was added after the dynein–Lis1 pre-incubation step . Once applied to the grid , the samples were stained with 2% uranyl formate by floating the grid sample face down on a pool of stain . Samples were then sandwiched with a thin layer of freshly evaporated carbon , and grids were lightly blotted from the non-sample containing side and plunged into liquid nitrogen . Grids were then stored at liquid nitrogen temperatures . Samples were imaged at liquid nitrogen temperatures using a Gatan 626 cryo holder ( Gatan , Inc . , Pleasanton , CA ) on a Tecnai F20 TEM microscope ( FEI , Hillsboro , OR ) , operating at 120 kV , equipped with a US4000 4k × 4k CCD camera ( Gatan ) . Data were collected either manually or automatically using Leginon ( Carragher et al . , 2000 ) . Dynein alone samples ( no nucleotide [strain RPY844] and ADP conditions [strain RPY844] ) and dynein–Lis1 ( ATP + Vi conditions [strains RPY1302 and RPY816] ) were imaged at 62 , 000× nominal magnification ( 1 . 73 Å/pixel ) . Dynein–Lis1 ( no nucleotide [strains RPY1302 and RPY816] ) was imaged at 50 , 000× nominal magnification ( 2 . 14 Å/pixel ) . Short linker dynein–Lis1 ( no nucleotide [strains RPY1436 and RPY816] ) was imaged at 80 , 000× nominal magnification ( 1 . 34 Å/pixel ) . Low-dose conditions during imaging ( dose ∼25 e−/Å2 ) were used for all data sets , and micrographs were collected using a defocus range of −0 . 6 to −1 . 5 μm . For all data sets , ∼1 , 000 particles were initially selected manually in Boxer ( EMAN1 ) ( Ludtke et al . , 1999 ) and reference-free 2D classified in IMAGIC ( van Heel et al . , 1996 ) to give class averages that were then used as templates for automated particle picking in Appion ( Lander et al . , 2009 ) . Reference-free 2D classification in IMAGIC was subsequently used on the data sets to remove averages with blurred appearance or incorrect size . CTF determination and correction of image phases were carried out in Appion using Ace2 ( NRAMM ) . Particles were band-pass filtered ( high-pass = 250 Å , low-pass = 3 × sampling ) in Imagic and normalized in Xmipp ( Sorzano et al . , 2004 ) . For 3D classification and initial 3D refinement particles were binned by two; final 3D refinements were carried out using unbinned data . EM maps have been deposited with the EMDataBank . Accession codes as follows; dynein–Lis1 ( no nucleotide conditions ) EMDB-6008; dynein alone ( no nucleotide conditions ) EMDB-6013; dynein alone ( ADP conditions ) with position 1 and 2 linker domains , EMDB-6015 and EMDB-6014 respectively; dynein–Lis1 ( ATP + Vi conditions ) EMDB-6016; short linker dynein–Lis1 ( no nucleotide conditions ) EMDB-6017 . For each entry , in addition to the final masked and filtered maps , raw half maps for each reconstruction have been deposited . We also deposited an XML file of the FSC plot between the dynein alone map and the fitted crystal structure of the motor domain ( PDB ID: 4AKG [Schmidt et al . , 2012] ) , as a supplementary file to the dynein alone submission ( EMDB-6013 ) . Dynein and Lis1 were tested for complex formation , and Lis1 mutants were tested for structural integrity ( Figure 2—figure supplement 1C ) by size-exclusion chromatography . 400–800 nM dynein and 475–800 nM Lis1 were loaded separately or after being mixed for 10 min at 4°C . Samples were fractionated on a Superose 6 PC 3 . 2/30 column using an ÄKTAmicro system ( GE Healthcare ) that had been equilibrated with degassed gel filtration buffer ( 50 mM Tris–HCl pH 8 . 0 , 150 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 1 mM DTT ) . Fractions ( 50 μl or 90 μl ) were analyzed by SDS-PAGE on 4–12% Tris-Bis gels ( Invitrogen , Grand Island , NY ) with SYPRO Red staining ( Invitrogen ) and imaged using an ImageQuant 300 ( BioRad , Hercules , CA ) or Typhoon ( Amersham , UK ) gel imaging system . Single-molecule motility assays were performed using flow chambers as previously described ( Case et al . , 1997 ) . Dynein was labeled with TMR ( Promega , Madison , WI ) , and microtubules contained ∼10% biotin-tubulin for surface attachment and ∼10% HyLite488-tubulin ( Cytoskeleton Inc . , Denver , CO ) for visualization . For assays that included Lis1 , dynein was incubated with 200 nM Lis1 for 10 min at 4°C prior to addition to the flow chamber . The imaging buffer consisted of 30 mM HEPES ( pH 7 . 2 ) , 50 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 10% glycerol , 1 mM DTT , 20 mM taxol , 1 . 25 mg/ml casein , 1 mM Mg-ATP , and an oxygen scavenger system . Images were recorded every 2 s for 5 or 10 min , and dynein velocities and run-lengths were calculated from kymographs generated in ImageJ ( National Institutes of Health ) . In vitro motility assays were visualized on either a Zeiss Elyra PS . 1 microscope with a 100× 1 . 46 N . A . oil immersion TIRF objective ( Carl Zeiss GmbH , Germany ) with an Andor EM-CCD camera or an Olympus IX-81 TIRF microscope with a 100× 1 . 45 N . A . oil immersion TIRF objective ( Olympus , Japan ) with a Hamamatsu EM-CCD camera . TMR-labeled dynein and HyLite488-microtubules were excited with 561 nm and 488 nm solid state laser lines , respectively . Images were recorded with a 100 ms exposure using Zen Black ( Zeiss ) or Metamorph software . Microtubule gliding assays and microtubule binding and release assays were performed as described ( Huang et al . , 2012 ) . Control experiments for the microtubule release assays examined dynein release in buffer lacking ATP ( Figure 5—figure supplement 1E ) , where dynein remained bound to microtubules as expected ( Huang et al . , 2012 ) and with dynein lacking N-terminal tags ( Figure 5—figure supplement 1F , G ) , where untagged dynein behaved similar to tagged dynein ( Figure 5E , F ) . To track the dynein-dependent movement of spindle pole bodies ( SPBs ) , we used a strain containing a GFP-labeled SPB marker , SPC110 , and a tdTomato-labeled cell membrane marker , HXT1 ( kindly provided by Jeff Moore , University of Colorado ) . Mutations were introduced into the PAC1 ( Lis1 ) locus in this strain . For control experiments , strains containing deletions of the dynein heavy chain ( DYN1 ) and PAC1 loci were constructed . All strains were PCR verified , and mutations were additionally verified by DNA sequencing . For image analysis , saturated overnight cultures for each strain were diluted to an OD600 of 0 . 1 in a total volume of 5 ml YPD media . The dilution of cultures was staggered such that the data could be collected for all strains during a single imaging session . Following dilution , each culture was incubated with rotation at 30°C for 3 hr . Hydroxyurea ( HU ) was then added to a final concentration of 200 mM , and the culture was incubated for an additional 2 hr with rotation at 30°C . The cells were collected by centrifugation , the media was discarded , and the cells were resuspended in 250 μl of fresh YPD + 200 mM HU . The cells were loaded into an Y04C microfluidic yeast plate ( CellASIC , EMD-Millipore , Germany ) and introduced into the viewing chamber with the ONIX controller ( CellASIC ) . Imaging was performed at the Nikon Imaging Center at Harvard Medical School . All images were collected with a Yokagawa CSU-X1 spinning disk confocal with Borealis modification , on a Nikon Ti inverted microscope equipped with a Plan 60 × 1 . 4 N . A . objective and the Perfect Focus System ( Nikon Corp . , Japan ) . GFP-labeled SPB and tdTomato-labeled cell membrane fluorescence were excited with the 488 nm and 561 nm lines , respectively , from a LMM-5 solid state laser merge module controlled with an ATOF ( Spectral Applied Research Inc . , Canada ) . Images were acquired with a Hamamatsu ORCA-AG CCD controlled with MetaMorph 7 . 0 software . Images were collected for SPBs as 100 ms exposures , spanning 9 × 500 nm Z-sections ( 4 . 5 μm total Z stack ) every 30 s for a total of 20 min . Cell membranes were imaged as single central Z-sections at the first and last time point . Membrane image pairs were digitally merged to allow for drift analysis; those cells with visible drift were excluded from analysis . Maximum intensity projections were calculated for Z-series at each time point for GFP-labeled SPBs . At each time point , SPBs were independently detected in the Z-projection using a wavelet detection algorithm ( Aguet et al . , 2013 ) , and the two spindles were tracked throughout the course of the movie using a nearest neighbor tracking method ( unpublished Matlab [Mathworks , Natick , MA] scripts ) . The location of the bud neck and the mother–daughter orientation were determined using the first tdTomato-labeled cell membrane exposure . The locations of tracked SPBs were used to calculate the number of bud neck crossings . To generate the dynein FRET construct , eGFP ( the FRET donor ) was inserted at the dynein N-terminus and the acceptor site was inserted after L2241 in AAA2 . We used the ybbR tag ( GGGTVLDSLEFIASKLAGGG [Yin et al . , 2005] ) labeled with TMR-CoA ( NEB , Ipswich , MA ) as the FRET acceptor . Dynein was incubated with or without Lis1 for one hour on ice , followed by apyrase ( 6 . 6 U/ml ) or 200 μM ATP . Vi for 2 min at room temperature ( RT ) . For some experiments , dynein was first incubated with 200 μM ATP . Vi for 2 min at RT , followed by Lis1 for 1 hr on ice . Assays were performed in 30 mM HEPES ( pH 7 . 2 ) , 50 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 1 mM DTT and the final concentrations of dynein and Lis1 were 84 nM and 840 nM , respectively . The sample was excited with 485 nm ( eGFP ) light , and the emitted light was detected from 505 nm to 650 nm in a SpectraMax M5 fluorimeter ( Molecular Devices , Sunnyvale , CA ) at RT . In order to normalize across experiments , the samples were also excited with 535 nm ( TMR ) light and the emitted light was detected from 555 nm to 700 nm . To analyze the FRET data , we first subtracted the fluorescence background from the buffer alone . We then used the emission spectra of dynein-labeled with eGFP and free TMR dye alone to decompose each channel in the experimental spectra . FRET efficiencies ( E ) were calculated using the method of Clegg ( Clegg , 1995 ) : E = {FaFRET/FaDIR − εa ( 485 ) /εa ( 535 ) }εa ( 535 ) /εd ( 485 ) , where the superscripts ‘d’ and ‘a’ refer to the donor ( eGFP ) and the acceptor ( TMR ) , respectively . FaFRET is the fluorescence intensity of the acceptor excited at 485 nm and FaDIR is the fluorescence intensity of the acceptor excited at 535 nm . εd ( 485 ) εa ( 485 ) and εa ( 535 ) are the molar extinction coefficients at the designated wavelengths . In our experiments εa ( 535 ) /εd ( 485 ) = 37 , 900 M−1 cm−1/40 , 000 M−1 cm−1 and εa ( 485 ) /εa ( 535 ) = 0 . 2 . Dynein constructs used in ATPase assays were tested for complex formation with Lis1 by size-exclusion chromatography ( Figure 4—figure supplement 1 ) . ATPase assays were performed using an EnzChek phosphatase kit ( Molecular Probes , Thermo Fisher Scientific Inc . , Cambridge , MA ) as previously described ( Reck-Peterson et al . , 2006; Cho et al . , 2008 ) . The final reaction consisted of 10–20 nM dynein ( monomeric constructs , see Figure 4 and Figure 5—figure supplement 1D ) , 0 or 140 nM Lis1 , 0–7 . 5 μM taxol-stabilized microtubules , 2 mM Mg-ATP , 200 mM MESG ( 2-amino-6-mercapto-7-methyl-purine riboside ) , 1 U/ml purine nucleoside phosphorylase , and assay buffer ( 30 mM HEPES ( pH 7 . 2 ) , 50 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 1 mM DTT , and 10 mM taxol ) . A SpectraMax384 plate reader ( Molecular Devices ) was used to monitor the coupled reaction at OD360 every 12 s for 10 min . Data were fit according to Nishiura et al . ( 2004 ) . | Cells use motor proteins to move ‘cargo’ from one location to another inside the cell . This cargo can range in size from a single macromolecule to something as large as the nucleus of the cell . A motor protein called dynein is the largest and least understood of the motor proteins found in cells . Dynein molecules work in pairs to take ‘steps’ along tracks called microtubules . Dynein contains two domains: a motor domain , which is responsible for generating movement , and a ‘tail’ domain to which the cargo is attached . The motor domain is composed of a ring-like shape and two appendages—the stalk and the linker . The linker undergoes large-scale movements relative to the ring that transmits force to the tail domain . Dynein also interacts with various accessory proteins to do its job inside the cell . One of these is a protein called Lis1 that is found across a wide range of species from yeast to humans . Defects in the gene for Lis1 result in brain developmental disorders in humans . However , it is not clear how the Lis1 protein influences the activity of dynein . Now Toropova , Zou et al . have visualized the structure of dynein bound to Lis1 and compared it with the structure of dynein on its own in order to work out if dynein changes its shape as a result of binding to Lis1 . These experiments show that when Lis1 binds to dynein , it physically blocks the linker , preventing it from making contacts with the ring-like shape that are important for the normal function of the motor . To test the idea that this physical block is responsible for dynein molecules spending a relatively long time attached to their microtubules , Toropova , Zou et al . shortened the linker to a point where the Lis1 protein could no longer block it: this resulted in a dynein motor that was no longer sensitive to Lis1 . A challenge for the future is to understand , at a molecular level , how the Lis1-mediated slowing down of dynein affects the multiple functions the motor carries out in a cell . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Lis1 regulates dynein by sterically blocking its mechanochemical cycle |
To avoid mutations in the genome , DNA replication is generally followed by DNA mismatch repair ( MMR ) . MMR starts when a MutS homolog recognizes a mismatch and undergoes an ATP-dependent transformation to an elusive sliding clamp state . How this transient state promotes MutL homolog recruitment and activation of repair is unclear . Here we present a crystal structure of the MutS/MutL complex using a site-specifically crosslinked complex and examine how large conformational changes lead to activation of MutL . The structure captures MutS in the sliding clamp conformation , where tilting of the MutS subunits across each other pushes DNA into a new channel , and reorientation of the connector domain creates an interface for MutL with both MutS subunits . Our work explains how the sliding clamp promotes loading of MutL onto DNA , to activate downstream effectors . We thus elucidate a crucial mechanism that ensures that MMR is initiated only after detection of a DNA mismatch .
To enable the correct and complete transfer of genetic information during cell division , DNA polymerases efficiently replicate the genome by pairing nucleotide bases opposite their complementary template base . However , despite the polymerase proofreading ability , incorrect nucleotides are occasionally incorporated into the new DNA strand , resulting in mutations when left uncorrected . To reduce the number of such mismatches and maintain genomic stability , replication is followed by DNA mismatch repair ( MMR ) in almost all cellular organisms ( Kunkel and Erie , 2005; Jiricny , 2013 ) . The initiation of this MMR system is evolutionarily conserved , although in eukaryotes heterodimeric homologs replace the bacterial homodimeric components . Defects in MMR result in a mutator phenotype and in humans in predisposition for cancer , known as Lynch syndrome or HNPCC ( Lynch and de la Chapelle , 1999 ) . MMR is initiated when a MutS homolog binds to a mismatch . In this mismatch recognition step , the MutS dimer kinks the DNA at the site of the mismatch and stacks a phenylalanine onto the mispaired base ( Lamers et al . , 2000; Obmolova et al . , 2000; Warren et al . , 2007 ) . Upon ATP binding MutS releases the mismatch ( Allen et al . , 1997; Gradia et al . , 1997 ) and travels as a ‘sliding clamp’ along the DNA helix ( Gradia et al . , 1999; Acharya et al , 2003; Jeong et al . , 2011 ) , and this specific state of MutS is recognized by MutL or its homologs ( Grilley et al . , 1989; Prolla et al . , 1994; Drotschmann et al . , 1998; Acharya et al . , 2003 ) . MutL proteins are constitutive dimers through their C-terminal domains , while the N-terminal ATPase domains reorganize and dimerize upon ATP binding ( Grilley et al . , 1989; Ban and Yang , 1998; Ban et al . , 1999; Guarné et al . , 2004 ) . Once recruited by the MutS sliding clamp , the MutL homologs activate downstream repair . This includes the nicking of the newly replicated strand by a nuclease , which is either part of the MutL C-terminal domain ( Kadyrov et al . , 2006 ) , or a separate protein such as MutH in Escherichia coli ( Hall and Matson , 1999 ) . MutL also activates UvrD in bacteria to unwind the DNA ( Yamaguchi et al . , 1998 ) , after which the new DNA strand can be removed and re-replicated ( Kunkel and Erie , 2005 ) . As loss of MutS homologs ( MSH2 , MSH3 and MSH6 in humans ) or MutL homologs ( MLH1 and PMS2 in humans ) leads to mutator and/or cancer phenotypes , these proteins evidently have critical roles in mismatch repair and it is therefore important to understand their exact mechanism . Despite extensive studies ( Gradia et al . , 1999; Mendillo et al . , 2005; Cho et al . , 2012; Qiu et al . , 2012 ) , it is unclear how MutS achieves the sliding-clamp state , how this promotes MutL recognition and why this results in activation of the MutL protein . Here , we trap the transient complex between MutS and MutL to resolve a crystal structure of the MutS sliding clamp bound to MutL . This is , to our knowledge , the first time that not only this MutS conformation but also the complex between MutS and MutL could be observed . We show how rearrangements in MutS promote interactions from both MutS subunits with a single MutL N-terminal domain , and how this domain is then positioned to load onto DNA running through a novel channel in the MutS dimer . We use biophysical methods to analyze the transient states and mechanistically understand the specificity and effect of MutL binding to MutS , and functional assays to address how this affects MMR initiation .
To trap the E . coli MutS/MutL complex we used site-specific chemical crosslinking of single-cysteine variants of MutS and MutL , with a flexible BM ( PEO ) 3 crosslinker . First all cysteines in MutS and MutL were replaced and functionality of the resulting protein was confirmed ( Giron-Monzon et al . , 2004; Manelyte et al . , 2006; Winkler et al . , 2011 ) . Then single cysteines were introduced to find positions where crosslinking was dependent on sliding clamp formation . MutS D246C crosslinks specifically to MutL N131C only when a DNA mismatch and a nucleotide are present ( Winkler et al . , 2011; Figure 1A , Figure 1—figure supplement 1A ) , indicating that a complex relevant for MMR is trapped . 10 . 7554/eLife . 06744 . 003Figure 1 . Crystal structure of the crosslinked MutSΔC800/MutLLN40 complex . ( A ) DNA and ATP-dependent crosslinking of MutSΔC800 D246C ( S ) and MutLLN40 N131C ( L ) and large-scale purification . Constructs and domain definitions are shown . ( B ) Crystal structure of the trapped transient complex of MutSΔC800 dimer ( blue/cyan ) with MutLLN40 ( green ) . ( C ) Comparison between MutSΔC800 in mismatch-recognition state ( 1E3M . pdb ) and the MutSΔC800/MutLLN40 complex , with MutS subunit B colored as in ( A ) . ( D ) The dimer subunits ( blue/cyan ) tilt across each other ( connector and mismatch-binding domains not shown for clarity ) compared to the mismatch-bound state ( red/pink ) . ( E ) The connector domain ( blue/cyan ) rotates around residues 265–266 compared to the mismatch-bound state ( red/pink ) relative to other domains . Reorientation of residues 128 and 246 indicated . ( F ) Each MutLLN40 subunit ( green ) interacts via two interfaces ( orange/yellow ) with the MutSΔC800 dimer ( blue/cyan ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 00310 . 7554/eLife . 06744 . 004Figure 1—figure supplement 1 . Crosslinking , purification and crystal structure of the 856 MutSΔC800/MutLLN40 complex . ( A ) Single-cysteine MutS D246C and single-cysteine MutL 857 N131C constructs with replaced and introduced cysteine positions are shown , and colored 858 according to domain definitions in main text Figure 1 . ( B ) Final size-exclusion 859 chromatography profile and corresponding SDS-PAGE gel for the purification of the 860 MutSΔC800/MutLLN40 complex ( SL ) . Pooled fractions are indicated . ( C ) Two rounds of 861 crosslinking and purification of MutSΔC800 D246C and MutLLN40 N131C result in almost all 35 MutSΔC800 subunits crosslinked to MutLLN40 , as shown on SDS-PAGE ( 862 elutions from Talon 863 beads and size-exclusion chromatography [SEC] are indicated ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 00410 . 7554/eLife . 06744 . 005Figure 1—figure supplement 2 . Electron density for different crystal forms of the MutSΔC800/MutLLN40 complex . Electron density shown in region around the domain indicated at contour level 1 . 0 rmsd and 3 . 50 rmsd for the difference density map . ( A ) Crystal form 1 , ( B ) Crystal form 2 , ( C ) Crystal form 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 005 For structural studies , we scaled up the reaction and removed C-terminal domains from MutS and MutL ( Figure 1A ) , to capture the complex between MutSΔC800 D246C ( which we will refer to as MutSΔC800 ) and the 40 kDa N-terminal LN40 domain ( Ban and Yang , 1998 ) of MutL N131C ( which we will refer to as MutLLN40 ) . The proteins were crosslinked in the presence of mismatched DNA and ATP , followed by purification to obtain the protein , and then this cycle was repeated in order to obtain fully crosslinked material . This generated a complex where each MutSΔC800 subunit in the dimer binds to a MutLLN40 monomer ( Figure 1A , Figure 1—figure supplement 1B , C ) , which was sufficiently homogeneous and stable to allow crystallization . We crystallized the MutSΔC800/MutLLN40 complex in the presence of DNA containing a G:T mismatch and the non-hydrolyzable ATP analog AMP-PNP ( adenylyl-imidodiphosphate ) . The complex crystallized in several different space groups , diffracting to resolutions from 7 . 6 to 4 . 7 Å . In all crystal forms , we could elucidate the same structure of the protein complex ( Figure 1B , Figure 1—figure supplements 2 , Table 1 ) , using parts of higher-resolution MutSΔC800 and MutLLN40 structures for molecular replacement . 10 . 7554/eLife . 06744 . 019Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 019Crystal form 1 27-bp DNACrystal form 2 27-bp DNACrystal form 3 100-bp DNAData collection Space groupC2C2P21 Cell dimensions a , b , c ( Å ) 165 . 9 , 188 . 5 , 200 . 4380 . 6 , 126 . 5 , 243 . 3192 . 6 , 109 . 4 , 277 . 5 α , β , γ ( ° ) 90 . 0 , 94 . 8 , 90 . 090 . 0 , 91 . 4 , 90 . 090 . 0 , 90 . 0 , 90 . 0 Resolution ( Å ) *82 . 7–4 . 71 ( 4 . 96–4 . 71 ) 49 . 94–6 . 6 ( 7 . 13–6 . 6 ) 49 . 3–7 . 6 ( 8 . 5–7 . 6 ) Rmerge19 . 4 ( 79 . 7 ) 21 . 3 ( 80 . 1 ) 16 . 8 ( 91 . 9 ) I/σI2 . 5 ( 1 . 0 ) 3 . 4 ( 1 . 1 ) 4 . 3 ( 1 . 0 ) Completeness ( % ) 97 . 3 ( 98 . 0 ) 96 . 8 ( 97 . 7 ) 81 . 3 ( 82 . 5 ) Redundancy2 . 4 ( 2 . 4 ) 2 . 9 ( 3 . 0 ) 2 . 3 ( 2 . 2 ) Refinement Resolution ( Å ) 4 . 76 . 67 . 6 No . reflections31 , 05221 , 30511 , 763 Rwork/ Rfree31 . 8/35 . 025 . 6/28 . 726 . 2/30 . 5 No . atoms21 , 90645 , 05445 , 054 Protein21 , 81344 , 86844 , 868 Ligand/ion93186186 Water000 B-factors Protein212255221 Ligand/ion220212171 Watern/an/an/a R . m . s deviations Bond lengths ( Å ) 0 . 0090 . 01030 . 0113 r . m . s . Z ( bonds ) 0 . 450 . 510 . 55 Bond angles ( ° ) 1 . 321 . 351 . 31 r . m . s . Z ( angles ) 0 . 590 . 700 . 68*Highest resolution shell is shown in parenthesis . The crystal structure shows a novel conformation of MutS , in which the subunits in the dimer are tilted across each other by ∼30° , compared to the mismatch recognition state ( Figure 1C , D ) . The subunits are tilted as a rigid body , but the C-terminal HTH domains hinging around residues 765–766 , move with the opposite subunit , maintaining their role in stabilizing MutS dimers ( Biswas et al . , 2001 ) . Meanwhile , the connector domains are rotated by ∼160° around a hinge at residues 265–266 , which moves these domains out of the center of the molecule and packs them against the ATPase domains ( Figure 1C , E ) . The mismatch-binding domain could not be resolved in the density , probably because it is flexible in this state . While the mismatch recognition state of MutS is asymmetric ( Lamers et al . , 2000 ) , this MutLLN40-bound conformation shows a more symmetrical MutSΔC800 dimer . The MutLLN40 interaction with MutSΔC800 involves two interfaces ( Figure 1F ) . The first interface is formed by the largest β-sheet of the ATPase domain of MutLLN40 , and the ATPase and core domains of one subunit of MutSΔC800 . The second interface involves the side of this same β-sheet and a looped-out helix of MutLLN40 , and the newly positioned connector domain of the other MutSΔC800 subunit . Each MutLLN40 monomer is therefore interacting with both subunits in the MutSΔC800 dimer . The novel conformation of MutS in our crystal structure reveals a rearrangement of the subunits in the MutSΔC800 dimer , tilting around the interface formed by the two ATPase sites ( Figure 1D , Video 1 ) . The tilting creates a new MutS dimer interface of ∼500 Å2 where the clamp domains cross over , partially from interactions between the helices themselves ( 200 Å2 ) , the rest from the ends of the clamp domains with the helices . 10 . 7554/eLife . 06744 . 009Video 1 . Interpolation between two MutS conformations . Interpolation between the mismatch-bound conformation of MutS and the conformation as observed in complex with MutLLN40 shows tilting of the MutS subunits across each other . The connector domain rotates outward , although the exact trajectory may be different than in this visualization . Mismatch-recognition domains are not shown since they are not visible in the MutSΔC800/MutLLN40 structure . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 009 We observe nucleotide density in the ATP binding sites of both subunits in the MutSΔC800 dimer ( Figure 2—figure supplement 1A ) , and since we crystallized the protein with AMP-PNP we modeled these nucleotides in the density . This type of ATP-induced tilting and increased packing of ATPase domains is more often observed upon ATP binding in ABC ATPases , such as ATP transporters , SMCs and RAD50 ( Hopfner and Tainer , 2003 ) . Based on comparison to RAD50 ( Hopfner et al . , 2000 ) we previously predicted a tilting motion ( Lamers et al . , 2004 ) , and an open-to-closed transition has been supported by deuterium exchange mass spectrometry ( Mendillo et al . , 2010 ) , but the crossing of the clamp domains of MutS and the effect that this has on DNA binding were unexpected . The type of rearrangement of the MutS N-terminal region was similarly unexpected . In this movement the connector domains have rotated onto the so-called ‘signature helix’ ( residues 670–684 ) ( Hopfner and Tainer , 2003 ) , whose amino terminus interacts with the γ-phosphate of the ATP in the opposite subunit in ABC ATPases . Therefore the observed connector domain movement could be the result of binding of ATP in the opposing subunit . In RAD50 this tilting or ‘closing’ motion is transmitted through a ‘signature coupling helix’ via charged interactions with the signature helix ( Williams et al . , 2011; Deshpande et al . , 2014 ) . This ‘coupling helix’ is found at the beginning of a long stretch ( 144–767 ) in RAD50 that includes the coiled coil region and ends in the signature helix . The equivalent region in MutS is only 10 residues long ( 660–669 ) and it is disordered in all structures . It is feasible that this 10-residue loop is critical for transmission of the ATP signal , but the details must be different , since the basic residues in the signature helix of RAD50 are not conserved in MutS . To validate the relevance of the observed conformational changes for the MMR process , MutS proteins with a single cysteine at position 449 were site-specifically labeled with two different Alexa fluorophores and combined into heterodimers by random subunit exchange ( Figure 2A , Figure 2—figure supplement 2 ) . When labeled protein was bound to end-blocked DNA containing a G:T mismatch , FRET increased upon ATP addition . This indicates that ATP-induced sliding clamp formation moves these residues toward each other , in line with the shorter distance in the new conformation ( from 50 Å in the mismatch-recognition state to 43 Å in the MutLLN40-bound structure ) . 10 . 7554/eLife . 06744 . 006Figure 2 . The structure of the MutSΔC800/MutLLN40 complex reveals the MutS sliding clamp conformation . ( A ) FRET within MutS dimers ( normalized for unbound protein ) reveals residues 449 coming closer together upon ATP addition . Error bars depict mean ± SD , n = 3 . ( B ) FRET assay agrees with residue 246 on the connector domain of MutS moving towards residue 798 upon ATP addition after mismatch recognition . ( C ) Mismatch and ATP-induced conformational changes open a channel lined by positively charged residues ( left: arginines and lysines as red sticks , middle: electrostatic surface ) , which would fit a DNA helix ( right ) . ( D ) FRET assay agrees with movement of DNA away from residues 449 in MutS , while approaching residues 336 upon ATP addition as schematically depicted ( DNA mismatch represented by pink star ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 00610 . 7554/eLife . 06744 . 007Figure 2—figure supplement 1 . ATP-analog and DNA in the crystal structure . ( A ) Difference density map for AMP-PNP ( mFo-DFc at 2 . 8σ ) after refinement without the nucleotide is shown for MutS subunit A in the 4 . 7 Å crystal structure . ( B ) The asymmetric unit in the P21 crystal form that diffracted to 7 . 6 Å , which was crystallized with 100-bp DNA aligns the channels in the three MutSΔC800 dimers ( green , violet and pink cartoon representations ) such that a DNA strand would fit through all complexes simultaneously . ( C ) Sequence conservation of the positively charged residues ( indicated by asterisks ) in the DNA channel of the MutS sliding clamp . Red: positive charge fully conserved; orange: conserved in >50% of the species shown; yellow: conserved in <50% of the species shown . ( D ) No change in FRET between labeled residue 246 on MutS and labeled DNA is observed upon addition of ATP , in agreement with reorientation of the connector domain concomitant with repositioning of the DNA to the new channel . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 00710 . 7554/eLife . 06744 . 008Figure 2—figure supplement 2 . FRET assay – controls and raw data . ( A ) FRET ( normalized for unbound protein ) within MutS D835R heterodimers ( 449-AF488/449-AF594 ) reveals residues 449 coming closer together upon ATP addition when bound to mismatch DNA . Bars depict mean +/− SD , n = 3 . ( B ) FRET assay within MutS D835R heterodimers ( 246-AF594/798-AF488 ) shows that residue 246 on the connector domain of MutS moving towards residue 798 upon ATP addition when bound to mismatch DNA . Bars depict mean +/− SD , n = 3 ( C ) Example of emission spectra with excitation at 485 nm ( solid lines ) or 590 nm ( dashed lines ) of MutS D835R heterodimers ( 449-AF488/449-AF594 ( left ) or 246-AF594/798-AF488 ) ( right ) ( in the presence of the indicated DNA ( GT-59 or homoduplex DNA λ-DNA ) in the absence ( black ) or presence of ATP ( red ) . ( D ) Example of emission spectra with excitation at 435 nm ( solid lines ) or 590 nm ( dashed lines ) of MutS D835R variants 246 ( top ) that corresponds to bar graph in Figure 2—supplement 1D , 336 ( middle ) or 449 ( bottom ) that correspond to bar graph in Figure 2D . The proteins labeled with Alexa Fluor 594 in the presence of the indicated DNA ( GT30 or AT30 ) stained with Sytox Blue in the absence ( black ) or presence of ATP ( red ) . ( see Materials and Methods for details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 008 The new position of the connector domain brings it closer to the ATPase domain ( Figure 1E , Video 1 ) . To analyze this movement we combined two single-cysteine variants of MutS , labeled in the connector domain ( residue 246 ) and the ATPase domain ( residue 798 ) respectively , into heterodimers , and measured the FRET signal between these sites upon sliding clamp formation ( Figure 2B , Figure 2—figure supplement 2 ) . Indeed , after ATP addition the FRET increased , indicating that these residues come closer together . As this is measured in the absence of MutL it suggests that after mismatch binding , ATP is sufficient to induce movement of the connector domain away from the mismatch-recognition position . Although the complex was crystallized in the presence of DNA containing a mismatch , the DNA is not visible in the structure . This could be due to smearing out of the electron density over multiple positions or the DNA may not be present in the crystal , both indicating that the mismatch has been released , as expected for the ATP-bound state of MutS . The subunit tilting has occluded the original DNA binding site , but because the connector and mismatch-binding domains have moved , a large channel ( ∼35 Å wide ) in MutS has become accessible , which could easily accommodate a DNA duplex ( 20 Å diameter ) . The new channel is lined by conserved lysines and arginines ( Figure 2C , Figure 2—figure supplement 1C ) , which can govern nonspecific contacts with the negative backbone of DNA , as expected for the MutS sliding clamp state ( Cho et al . , 2012 ) . Moreover , in our crystal forms these channels are aligned between symmetry mates or even within the asymmetric unit ( Figure 2—figure supplement 1B ) . This packing of MutS/MutL complexes is most likely a crystallographic artefact , as it could not occur in the presence of MutL dimers , but the alignment could reflect the path of the DNA present during crystallization . We hypothesize that the DNA is pushed down to this channel during the ATP-induced conformational changes of MutS after mismatch recognition . To test whether DNA moves down into the new channel in solution , we analyzed FRET signals between fluorescently labeled DNA ( end-blocked ) and specific sites in single-cysteine MutS variants ( Figure 2D ) . After addition of ATP , DNA moves away from residues 449 at the DNA-clamp position ( FRET/acceptor ratio reduction ∼1 . 5 fold ) , while an increase in FRET/acceptor ratio ( >3 . 6 fold ) was observed when MutS was labeled at position 336 . Since the connector domain moves down itself , no substantial change in FRET/acceptor ratio is observed between residue 246 and DNA ( Figure 2—figure supplement 1D , Figure 2—figure supplement 2 ) . Combined , these FRET data are in agreement with repositioning of the DNA towards the channel created by the new conformation . Based on these validations , we conclude that the observed MutS conformation in our crystal structure is induced by ATP after mismatch recognition . Since the new position of the DNA would allow MutS to fit as a loose ring around the DNA duplex ( with a channel size similar to that of PCNA ( Krishna et al . , 1994 ) ) , consistent with free movement over DNA ( Cho et al . , 2012 ) , we propose that this is the MutS sliding clamp conformation . In the structure MutLLN40 makes two interfaces with MutSΔC800 . Interface 1 orients MutLLN40 on the ATPase and core domains of MutS . Recently , a loop in Bacillus subtilis MutS was found to be essential for MutL interaction ( Lenhart et al . , 2013 ) . Although the equivalent loop is shorter in E . coli MutS and the explicit residues ( F319/F320 ) are missing , the corresponding region is located within the ∼590 Å2 interface ( interface 1 ) with MutLLN40 . We validated the observed interaction at interface 1 by a crosslinking experiment with a short crosslinker . We created single-cysteine mutants MutSΔC800 A336C and MutLLN40 T218C ( Figure 3A ) , which are located ∼7 . 4 Å apart in the structure , and then showed that we could crosslink them efficiently with a short cysteine-specific crosslinker ( 8 Å , BMOE ) , dependent on the presence of both mismatched DNA and ATP ( Winkler et al . , 2011 ) . Only background crosslinking occurred when using MutSΔC800 D246C ( connector domain ) with MutLLN40 T218C ( interface 1 ) under these conditions ( Figure 3—figure supplement 1A ) , indicating that the crosslinking between MutSΔC800 A336C and MutLLN40 T218C is specific . 10 . 7554/eLife . 06744 . 010Figure 3 . Interaction of the MutSΔC800 sliding clamp with MutLLN40 . ( A ) Crosslinking occurs between MutSΔC800 A336C and MutLLN40 T218C using BMOE ( right panel ) , as suggested by the structure ( left panel ) . ( B ) Spontaneous mutation rates after complementing MutS or MutL-deficient cells with the indicated mutants . Error bars represent 95% confidence intervals . ( C ) SPR assay to measure MutL binding to pre-formed MutS sliding clamps on end-blocked DNA . MutL contribution ( green dotted line ) is approached by subtracting MutS-only contribution ( blue line ) from the total signal ( solid line ) . Data normalized to maximum MutS response . ( D ) MutL and MutS mutants with deficiency in MMR show reduced MutS/MutL complex formation in SPR . Error bars represent SD for averages between two experiments . ( E ) The yellow patch of MutSΔC800 interacts with MutLLN40 in the new conformation after rearrangement of the connector domain . ( F ) Residues in MutSΔC800/MutLLN40 interfaces . Red: full MMR deficiency upon mutation; orange: deficiency upon combination; white: mild effect . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 01010 . 7554/eLife . 06744 . 011Figure 3—figure supplement 1 . MutS–MutL interaction . ( A ) MutSΔC800 D246C and MutLLN40 T218C do not crosslink efficiently with either a short ( BMOE , 8 Å ) or a long ( BM[PEO]3 , 18 Å ) crosslinker , as e . g . seen by lack of MutS and MutL depletion . ( B ) MutS P595A/I597A/M759D shows ATPase activity with similar Km ( 9 . 5 ± 1 μM ) as WT MutS ( 8 . 1 ± 0 . 7 μM ) but differs in Kcat ( mutant: 3 . 8 ± 0 . 1 min−1; WT 8 . 3 ± 0 . 2 min−1 ) . Data points are averages between two measurements and error bars indicate standard deviations . ( C ) MutS P595A/I597A/M759D shows similar sliding clamp formation as WT MutS ( assay on end-blocked 21-bp DNA with dT20 linker; Groothuizen et al . , 2013 ) . ( D ) Without preformed MutS sliding clamps , there is only little binding of MutL to DNA . ( E ) The region in the ATPase domain that is solvent exposed during mismatch recognition but showed reduced deuterium exchange upon MutL binding ( red , residues 673–686 ) ( Mendillo et al . , 2009 ) is buried by the connector domain ( light blue surface; orange patch is close to residues 673–686 ) . ( F ) The association with the MutSΔC800 sliding clamp ( blue/cyan ) does not sterically hinder potential dimerization by the MutLLN40 domains ( dimer modeled in green and grey as present in pdb entry 1NHJ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 01110 . 7554/eLife . 06744 . 012Figure 3—figure supplement 2 . ( A ) MutLLN40 ( L ) coelutes with crosslinked MutSΔC800/MutLLN40 complex ( SL ) from size-exclusion chromatography ( right ) , after incubation with 100-bp DNA with a G:T mismatch and AMP-PNP , indicating that MutL can still dimerize in this complex . ( B ) Model for MSH2/MSH6 interaction with MLH1/PMS2 , in which the N-terminus of MLH1 simultaneously binds to the connector domain of MSH2 and the APTase and core domains of MSH6 . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 012 To further verify interface 1 between MutS and MutL , we tested whether mutations in the interface affected MMR activity in vivo , in a complementation assay with MutS or MutL deficient cells ( Figure 3B , F , Table 2 ) . We found several mutants of MutL ( A138E , A138E/H139A , R55D/R57D , or combinations ) and a triple mutant in MutS ( P595A/I597A/M759D ) that could not complement loss of wild type ( WT ) protein . We purified the mutants that impaired MMR and characterized their defects . The MutS triple mutant has a slight defect in ATPase activity but this does not impair its sliding clamp formation ( Figure 3—figure supplement 1B , C ) , and other mutants with similar ATPase effects ( e . g . MutS F596A ) can almost fully reconstitute MMR ( Junop et al . , 2003 ) , suggesting that the in vivo effect we observe is due to the perturbed interface with MutL . 10 . 7554/eLife . 06744 . 013Table 2 . Mutation rates for MutS and MutL mutants as determined using in vivo complementation assaysDOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 013ProteinMutations per 107 ( 95% confidence interval ) MutS variant ( MutL interface ) Empty vector0 . 601 ( 0 . 446–0 . 772 ) WT MutS0 . 0686 ( 0 . 0408–0 . 101 ) MutS P595A/I597A0 . 0545 ( 0 . 0310–0 . 0826 ) MutS M759D0 . 0819 ( 0 . 0490–0 . 121 ) MutS Y563A0 . 0488 ( 0 . 0272–0 . 0749 ) MutS P595A/I597A/M759D0 . 704 ( 0 . 556–0 . 864 ) MutS Y563A/P595A/I597A0 . 317 ( 0 . 233–0 . 411 ) MutS Y563A/P595A/I597A/M759D0 . 773 ( 0 . 618–0 . 941 ) MutL variant ( MutS interface ) Empty vector5 . 43 ( 4 . 00–7 . 00 ) WT His-MutL0 . 121 ( 0 . 0542–0 . 206 ) His-MutL A138E2 . 76 ( 2 . 12–3 . 46 ) His-MutL H139A0 . 103 ( 0 . 0439–0 . 179 ) His-MutL A138E/H139A4 . 87 ( 3 . 55–6 . 33 ) His-MutL R55D/R57D6 . 41 ( 4 . 99–7 . 95 ) His-MutL R200D0 . 663 ( 0 . 432–0 . 932 ) His-MutL R55D/R57D/H139A5 . 33 ( 3 . 93–6 . 89 ) His-MutL R55D/R57D/A138E/H139A6 . 13 ( 4 . 58–7 . 84 ) His-MutL R55D/R57D/H139A/R200D5 . 22 ( 3 . 84–6 . 76 ) His-MutL R55D/R57D/A138E/H139A/R200D5 . 48 ( 4 . 04–7 . 06 ) MutL variant ( DNA binding ) His-MutL R266E5 . 87 ( 4 . 78–7 . 04 ) His-MutL R162E/R266E/R316E5 . 39 ( 4 . 37–6 . 49 ) Mutation rates and 95% confidence intervals were determined using the Fluctuation AnaLysis CalculatOR ( http://www . mitochondria . org/protocols/FALCOR . html ) using the MSS-MLE method . For MutS , at least 24 independent colonies were picked; for MutL at least 12 independent colonies were picked . To assess the effect of these mutations on binding of MutL to the transient MutS sliding clamp we designed a two-stage assay using Surface Plasmon Resonance ( SPR ) . We first formed and trapped MutS sliding clamps on 100-bp end-blocked DNA in the presence of ATP ( Groothuizen et al . , 2013 ) . Next , MutL was injected , which could then bind to these MutS clamps . By subtraction of the MutS signal , the contribution of MutL could be evaluated for the different mutants ( Figure 3C ) , since MutL alone shows no DNA binding under these conditions ( Figure 3—figure supplement 1D ) . Indeed the interface 1 mutants that were deficient for MMR conferred a deficiency in MutS/MutL complex formation ( Figure 3D ) . The rearrangement of the connector domain creates a second interface with MutLLN40 ( interface 2 , Figure 3E ) . Previous deuterium exchange experiments ( Mendillo et al . , 2009 ) indicated that the connector domain interacts with MutL , particularly via MutS glutamines 211 and 212 . Indeed in our structure these residues are buried within this ∼670 Å2 interface with MutLLN40 ( Figure 3F ) . Interestingly , the deuterium exchange experiments identified a second region on the MutS surface that was protected upon MutL interaction in the ATPase domain ( residues 673–686 ) . These residues are located in the ‘signature helix’ of MutS ( Hopfner and Tainer , 2003 ) and in the complex structure this region is masked by the MutSΔC800 connector domain in its new position ( Figure 3—figure supplement 1E ) . ATP binding is sufficient to displace the connector domain ( Figure 2B ) , and MutLLN40 interaction may stabilize the position of the connector domain that we see in the crystal structure . At the resolution of our structure , there is no clear electron density for the connecting crosslinker that we used to stabilize the complex , and the crosslinked residue 131C on MutLLN40 could not be modeled . However , the distance between Cα atoms of crosslinked residue 246C in MutSΔC800 and residue 132 in MutLLN40 is shorter ( ∼15 . 5 Å ) than the 18 Å crosslinker ( further spaced by cysteine side-chains ) , showing that the crosslinker can not enforce the observed position of the connector domain . On the MutL side of interface 2 , residue K52 of MutLLN40 is involved in the interaction with the connector domain of MutSΔC800 ( Figure 3F ) . This explains the previously reported unexpected mutator phenotype of MutL K52C ( Giron-Monzon et al . , 2004 ) . To confirm its role in the interface we measured the binding of MutL K52C to the MutS sliding clamp in our SPR assay ( Figure 3D ) . Indeed , the binding of this mutant is reduced compared to WT MutL . The ATP-induced tilting of the subunits within MutS and the accompanying connector domain movement positions interfaces 1 and 2 such that they become simultaneously available for binding to the N-terminal domain of MutL ( Figure 1F ) . Perturbing either interface 1 or interface 2 impairs MutL binding and MMR ( Figure 3F ) . This explains the specificity of MutL for the MutS sliding clamp , which has never been understood before . MutL proteins dimerize through the C-terminal LC20 domains . The LN40 domains are monomeric in isolation , but can form unstable dimers after ADP or ATP binding or stable dimers when incubated with AMP-PNP ( Ban and Yang , 1998; Ban et al . , 1999 ) . Our crosslinked protein crystallizes as MutSΔC800 dimers bound to MutLLN40 monomers , and does not show the MutLLN40 dimer arrangement through crystal contacts . Accordingly the MutL monomers have the apo-conformation of residues 80–103 ( Ban and Yang , 1998 ) and no density for a nucleotide is visible . However , the interfaces with MutS sterically allow MutL dimerization ( Figure 3—figure supplement 1F ) , and in analytical gel filtration , MutLLN40 coelutes with the S2/L2 complex after incubation with DNA and AMP-PNP ( Figure 3—figure supplement 2A ) . The stoichiometry of the MutS/MutL complex in vivo is a topic of interest ( Hombauer et al . , 2011; Elez et al . , 2012 ) . To obtain crystallizable complexes , MutLLN40 was bound to each MutSΔC800 subunit in our experiments , but during MMR a symmetric complex may not be necessary . Indeed the asymmetry of the eukaryotic MMR proteins suggest that this is not required and that a single heterodimeric MutLα will bind to one MSH2/MSH6 or MSH2/MSH3 heterodimer . Literature suggests that interface 2 will be made with MSH2 ( Mendillo et al . , 2009 ) , implying that interface 1 will be with MSH6 . The observed MutLLN40 protein would then correlate with the MLH1 subunit ( Plotz et al . , 2003 ) ( Figure 3—figure supplement 2B ) . MutL and homologs have weak DNA binding ability ( Bende and Grafström , 1991; Ban et al . , 1999; Hall et al . , 2001; Plotz et al . , 2003 ) which is only clearly observed in low salt conditions , and retention of MutS on DNA upon MutL interaction has been observed ( Drotschmann et al . , 1998; Schofield et al . , 2001 ) . Although different from the proposed DNA orientation in the crystal ( Figure 2—figure supplement 1B ) , a model can be constructed in which the DNA running through the channel in the MutS sliding clamp is simultaneously bound by the proposed DNA binding grooves of the MutLLN40 subunits ( Schorzman et al . , 2011 ) ( Figure 4A , Figure 4—figure supplement 1A ) . While such DNA binding may require additional conformational changes of MutL , it suggests a mechanism where MutS loads MutL onto DNA . 10 . 7554/eLife . 06744 . 014Figure 4 . The MutS sliding clamp positions MutL onto DNA . ( A ) Model of DNA binding by the MutSΔC800/MutLLN40 complex . Three arginines in the MutLLN40 DNA-binding groove are shown as red spheres . ( B ) In the presence of ATP , MutSΔC800 has a fast off-rate from 100-bp DNA and MutLLN40 alone does not bind DNA under physiological salt ( 150 mM KCl ) , while the crosslinked MutSΔC800/MutLLN40 complex releases slowly from DNA . ( C , D ) Mutations in the DNA-binding groove of MutL reduce its DNA-binding ability ( observed in low salt , 50 mM KCl ) ( C ) and affect release rates of the MutSΔC800/MutLLN40 complex in physiological salt conditions ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 01410 . 7554/eLife . 06744 . 015Figure 4—figure supplement 1 . DNA binding by the MutSΔC800/MutLLN40 complex . ( A ) The model for MutSΔC800/MutLLN40 complex on DNA ( orange ) sterically allows for LN40 dimerization ( dimer modeled in green and grey as present in pdb entry 1NHJ ) . ( B ) Analysis as in Figure 4B , but using shorter , 41-bp mismatched DNA ( ATAGGACGCTGACACTGGTGCTTGGCAGCTTCTAATTCGAT annealed with ATCGAATTAGAAGCTGCCAGGCACCAGTGTCAGCGTCCTAT ) shows that crosslinked LN40 slows down the MutS sliding clamp on with a G:T mismatch , but point mutations in the MutLLN40 DNA-binding groove can abolish this . ( C ) Absolute response from the assay on 41-bp DNA fitted ( single-site binding mode ) to estimate maximum signal . Comparison shows that maximum response for MutSΔC800/MutLLN40 binding ( 150 RU ) is 45% higher than for MutSΔC800 alone ( 104 RU ) . Since the MutSΔC800/MutLLN40 complex is 43% larger than MutSΔC800 dimer alone , this indicates that in this assay a single MutSΔC800/MutLLN40 complex binds the DNA duplex . ( D ) Co-injection of WT MutL results in slower release from DNA than MutSΔC800 alone , while uncrosslinked MutLLN40 only has a minor effect ( traces normalized to maximum response ) . ( E ) While end-blocks on the DNA slow down MutS release , there is no effect on the already slow MutSΔC800/MutLLN40 release . This is more pronounced for WT MutS , which forms more stable dimers than MutSΔC800 . Crosslinked MutLLN40 with mutations in the DNA-binding groove is deficient in the ability to retain MutSΔC800 on DNA . ( F ) Normalized overlay of the 1280 nM traces from Figure 4B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 015 We tested for MutLLN40 loading onto DNA in the context of the MutS/MutL complex in an SPR assay , comparing MutSΔC800 alone with MutSΔC800 crosslinked to MutLLN40 when it is flowed over 100-bp DNA with a G:T mismatch in the presence of ATP ( Figure 4B ) . MutSΔC800 alone displays fast release from the DNA due to ATP-dependent sliding-clamp formation ( Groothuizen et al . , 2013 ) , as shown by the effect of blocking the end of the DNA ( Figure 4—figure supplement 1E ) . The presence of crosslinked MutLLN40 greatly reduces the rate of release , suggesting additional DNA binding . The magnitude of the signal in response units on a 41-bp oligomer shows that a single MutSΔC800/MutLLN40 complex is sufficient for this effect ( Figure 4—figure supplement 1B , C ) . This delay in release from DNA is also observed when using a mixture of WT MutL and MutSΔC800 , although to a lesser extent ( Figure 4—figure supplement 1D ) . The remaining slow release of the crosslinked complex is not affected by blocking of the free DNA end by antibody ( Figure 4—figure supplement 1E ) indicating that the constitutive interaction with crosslinked MutLLN40 completely stops MutSΔC800 dissociation from DNA ends . To validate that the slower release from DNA is indeed due to MutLLN40 binding to DNA , we made point mutants of the MutLLN40 protein and crosslinked them to MutS . Mutation R266E reduces DNA binding by MutL ( Junop et al . , 2003; Robertson et al . , 2006 ) ( Figure 4B ) , most pronounced in full-length context . This mutation also reduces the ability of crosslinked MutLLN40 to retain the MutSΔC800 sliding clamp on DNA ( Figure 4D , Figure 4—figure supplement 1B , F ) . When introducing two additional mutations ( R162E and R316E ) in the MutLLN40 DNA binding site as suggested by the crystal structure ( Figure 4A ) , DNA binding is completely abolished ( Figure 4C ) and the MutSΔC800/MutLLN40 complex releases as fast as MutSΔC800 alone ( Figure 4D , Figure 4—figure supplement 1B , F ) . This indicates that MutL binds DNA when interacting with the MutS sliding clamp . To assess whether the loading of MutLLN40 onto DNA is kinetically distinct from MutS mismatch recognition , we set up an assay to separate events . We read out mismatch recognition ( Lamers et al . , 2000; Obmolova et al . , 2000; Warren et al . , 2007 ) by the kinking of DNA , which can be assessed using 45-bp heteroduplex DNA labeled with Alexa fluorophores on each side of the mismatch ( Cristóvão et al . , 2012 ) , Figure 5A , Figure 5—figure supplement 1A ) , in a stopped-flow set up . In parallel we follow DNA interaction using fluorescence polarization ( FP ) of TAMRA-labeled DNA with the same sequence . This shows that the kinking is concurrent with DNA binding by MutSΔC800 , while kinking is not observed when homoduplex is used ( Figure 5A , Figure 5—figure supplement 1A ) . When the assay is performed in the presence of ATP , MutSΔC800 binds and kinks the DNA but subsequently releases due to sliding clamp formation , after which an equilibrium is reached between rebinding and release ( Figure 5B , Figure 5—figure supplement 1B ) . 10 . 7554/eLife . 06744 . 016Figure 5 . Implications for DNA mismatch repair initiation . ( A ) Stopped-flow FRET and FP assay shows kinking of 45-bp DNA by MutSΔC800 binding only if there is a mismatch . Magnitude of FRET events are indicated by stars in the cartoon . ( B ) While MutSΔC800 initially kinks the DNA and subsequently releases in the presence of ATP , the MutSΔC800/MutLLN40 shows a secondary FP event without kinking the DNA . ( C ) Nicking assay of mismatch containing closed circular DNA ( ccDNA ) shows that WT or single-cysteine MutL can activate MutH , while mutations in the DNA-binding groove of MutL strongly impair the activation . ( D ) Spontaneous mutation rates after complementing MutL-deficient cells shows that the DNA-binding ability of MutL is essential for MMR in vivo . Error bars represent 95% confidence intervals . ( E ) Model for MMR initiation . After MutS undergoes an ATP-induced conformational change to allow binding of both subunits to one MutL molecule , MutL N-termini can interact and possibly dimerize , to be loaded onto DNA where MutL can activate downstream effectors . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 01610 . 7554/eLife . 06744 . 017Figure 5—figure supplement 1 . DNA kinking by MutSΔC800 and MutSΔC800/MutLLN40 . ( A ) Stopped-flow FRET and FP assay shows kinking of 45-bp DNA by MutS binding only if there is a mismatch . Separate traces for the fluorophores are shown ( orange: acceptor; green: donor fluorophore; grey: FP ) . Size of FRET events are indicated by stars . ( B ) While MutSΔC800 initially kinks the DNA and subsequently releases in the presence of ATP , the MutSΔC800/MutLLN40 complex remains bound to unkinked DNA . ( C ) Using WT MutL ( 400 nM ) mixed with the MutSΔC800 in this assay results in more binding at equilibrium than MutSΔC800 alone , while there is less FRET than for MutSΔC800 alone . Under the graph with the FRET ratio ( black: FRET; grey: FP ) , separate traces for the fluorophores are shown as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 017 In the presence of the crosslinked complex we observed a two-step sequence of events ( Figure 5B ) . The first increase in FP is consistent with mismatch recognition by MutSΔC800 , simultaneous with an increase in FRET due to kinking of the DNA . A second event increases FP even more but reduces the FRET signal to below starting value ( Figure 5—figure supplement 1B ) . This can be explained by release of the mismatch ( unkinking ) and sliding clamp formation . Now , however , the complex does not slide off the DNA but instead the MutLLN40 is docked onto the DNA to keep the complex bound , as observed in the SPR assays ( Figure 4B ) and by the increase in FP ( Figure 5B ) . At this time , since DNA has been pushed to the new channel , it is not kinked any more but kept relatively rigid by the MutLLN40 binding . This , and interaction of the fluorophore itself with bound protein , can explain the lowered FRET . A similar straightening of DNA relative to the unbound DNA was previously observed upon ATP-dependent MutS release in SAXS experiments using DNA labelled with gold-clusters ( Hura et al . , 2013a ) . The effect is also present to lesser extent when using a mixture of MutSΔC800 with WT MutL in this setup ( Figure 5—figure supplement 1C ) . The result indicates that MutLLN40 loading occurs after mismatch recognition and sliding clamp formation by MutSΔC800 . Since we observed that upon sliding clamp formation , MutS loads MutL onto DNA , we wondered whether this DNA loading step is essential for MMR . Indeed we observed a correlation with the DNA binding ability of MutL for MutH activation ( Figure 5C ) . Moreover , the DNA-binding mutants of MutL impair in vivo MMR ( Robertson et al . , 2006 ) ( Figure 5D , Table 2 ) , indicating that loading of MutL onto DNA after mismatch recognition is essential for MMR .
Taken together , our data reveal how the large conformational changes within MutS after mismatch recognition promote MMR activation . In the mismatch and ATP activated state MutS pushes DNA into a new channel , which allows sliding of the protein over DNA . The new state with the clamps crossed over the DNA explains the stability of the MutS sliding clamp on DNA ( Schofield et al . , 2001; Lebbink et al . , 2010; Jeong et al . , 2011 ) , as electrostatic interactions between DNA and the positive charges lining the new channel may stabilize the new clamp conformation . The conformational change pushes the connector domain away from the center and on top of the ATPase domains , to provide a second interface for the MutL protein that binds to the opposing MutS subunit , while DNA in the new MutS channel can also contribute to MutL binding . This loads the N-terminal domains of MutL onto the DNA and the MutL binding delays the sliding of MutS ( Figure 5E , Video 2 ) . The loading step of MutL onto DNA is required for MutH activation and nicking ( Figure 5C ) ( Junop et al . , 2003; Robertson et al . , 2006 ) , while UvrD loading and activation at this nick ( Yamaguchi et al . , 1998 ) would follow similar validation . In this way , the requirement of the MutS conformational change for full MutL interaction is a sophisticated validation mechanism , which presumably is conserved in the eukaryotic homologs . It ensures that repair is only initiated when necessary , and due to the MMR system DNA replication can be completed with few errors incorporated in the genome . 10 . 7554/eLife . 06744 . 018Video 2 . Model for initiation of DNA mismatch repair . After MutS ( cyan/blue ) has recognized a mismatch in DNA ( in orange; mismatch shown as pink spheres ) , it will bind ATP which triggers a conformational change in which the subunits tilt across each other and the connector domains move outward . This pushes the DNA to a new channel , where MutS fits as a loose ring around the DNA duplex and can behave as a sliding clamp . The N-terminal domain of MutL ( green ) can specifically recognize this state by binding two interfaces simultaneously . This loads MutL onto the DNA , where the N-terminal domains could dimerize and downstream effectors can be activated . DOI: http://dx . doi . org/10 . 7554/eLife . 06744 . 018 The complete transition from mismatch binding to sliding clamp state is likely to take multiple steps ( Qiu et al . , 2012 ) . First a single ATP will bind , leading to a stabilized asymmetric nucleotide state of MutS on the mismatch ( Antony and Hingorani , 2004; Antony et al . , 2006; Monti et al . , 2011 ) , followed by binding of the second ATP ( Mazur et al . , 2006; Hargreaves et al . , 2010 ) . Meanwhile MutS will undergo two separate ATP-induced events , the tilting of the subunits that push DNA into a new channel and the rearrangement of the connector domain ( and the associated mismatch binding domain ) that together generate a new MutL interface . These two movements could potentially be uncoupled . MutS binding to a non-hydrolysable ATP analog can already cause a closed clamp-like state , ( i . e . perform the tilting movement ) as supported by SAXS analysis ( Hura et al . , 2013b ) , but may possibly not change the conformation of the mismatched binding domain ( Qiu et al . , 2012 ) , as consequence of the connector domain movement . This would explain how MutS with ATPγS ( or with ATP for a mutant that cannot hydrolyse nucleotides [E694A] [Jacobs-Palmer and Hingorani , 2007] ) could form a closed clamp state that can no longer be loaded onto DNA ( Gradia et al . , 1999; Jacobs-Palmer and Hingorani , 2007; Cristóvão et al . , 2012 ) , but nevertheless is not sufficient to bind MutL . Our data do not address the order of the two events , tilting and connector movement , or how they relate to the two ATP binding events . Observed conformational changes resulting in ternary complex and sliding clamp formation have previously been suggested to be independent ( Mendillo et al . , 2010 ) . Indeed our structure does suggest that rearrangement of a single connector domain ( in the subunit equivalent to the ‘MSH2’ subunit; ( Mendillo et al . , 2009 ) is sufficient for the complex formation with MutLα ( Hess et al . , 2006; Hargreaves et al . , 2010 , 2012 ) . This might allow MSH6 to initially remain bound to the mismatch , consistent with models that consider transient asymmetric nucleotide states involved in mismatch verification and possibly ternary complex formation ( Antony and Hingorani , 2004; Hess et al . , 2006; Lebbink et al . , 2006; Mazur et al . , 2006; Hargreaves et al . , 2010; Monti et al . , 2011; Qiu et al . , 2012 ) . Another question that is unclear is where the loading of MutL onto DNA takes place . It could occur on or close to the mismatch itself , but it is also possible that MutS first slides before loading MutL on DNA . Once the sliding clamp conformation is reached , the complex no longer interacts with the mismatch ( Gorman et al . , 2012 ) . The clamp state loads MutL onto DNA , stabilizes a straight form of the DNA ( Figure 5B ) ( Hura et al . , 2013a ) and triggers the conformational changes of MutL . These involve movements in the C-terminal domains ( Guarné et al . , 2004 ) to form a ring around the DNA and ATP binding by the N-terminal domains of MutL to generate the state that activates MutH and UvrD ( Prolla et al . , 1994; Drotschmann et al . , 1998; Ban et al . , 1999; Acharya et al . , 2003 ) . In conclusion , we have used single-cysteine mutants and chemical crosslinking to trap and analyze a relevant MMR intermediate state that has long been elusive . This sliding clamp state of MutS bound to a MutL domain is highly informative . It corresponds to a reaction intermediate that occurs during a series of conformational changes triggered by mismatch recognition , and explains why specifically this conformation of MutS is able to recruit MutL . The presented combination of structural and biophysical methods provides a powerful approach to resolve conformational changes within large and transient protein complexes that form and act during biologically relevant processes .
MutS mutants were created in the mutS gene in vector pET-3D ( Lamers et al . , 2000; Giron-Monzon et al . , 2004; Manelyte et al . , 2006; Winkler et al . , 2011 ) or vector pET15b ( Manelyte et al . , 2006; Winkler et al . , 2011 ) ( for His-tagged MutS constructs in FRET assays ) . MutL mutants were generated in the mutL gene in plasmid pTX418 ( Feng and Winkler , 1995; Ban and Yang , 1998 ) . Single-cysteine MutS and MutL constructs were obtained as described ( Giron-Monzon et al . , 2004; Groothuizen et al . , 2013 ) . Mutant and WT MutS and MutL proteins were purified as described previously ( Lamers et al . , 2000; Manelyte et al . , 2006 ) , except that in the buffers KCl was used instead of NaCl ( final gel filtration buffer for MutS: 25 mM Hepes pH 7 . 5 , 150 mM KCl , 1 mM DTT; for MutL: 20 mM Tris pH 8 . 0 , 0 . 5 M KCl , 10% glycerol , 1 mM DTT ) . MutH was purified as follows: E . coli BL21 ( DE3 ) cells were transformed with MutH expression plasmid pTX417 ( Feng and Winkler , 1995 ) and plated onto LB agar with 50 μg/ml carbenicillin . A colony was picked and cells were grown in LB with 50 μg/ml carbenicillin at 37°C to OD600 ∼0 . 6 and induced with 1 mM isopropyl 1-thio-β-D-galactopyranoside for 4 hr . Cells were harvested and resuspended in binding buffer ( 25 mM Tris pH 8 . 0 , 300 mM KCl , 10 mM imidazole , 0 . 2 mM DTT ) with 1 mM PMSF and protease inhibitors ( Roche Diagnostics , F . Hoffmann-La Roche Ltd , Switzerland ) and lysed by sonication . The cleared supernatant was incubated with Talon resin ( Clonetech Laboratories , Takara holdings inc , Japan ) for 30 min on ice . Beads were washed using binding buffer with 1 M KCl , and MutH was eluted with 250 mM imidazole in binding buffer . The His-tag was removed by cleavage with Thrombin protease ( ∼5 units thrombin/mg MutH; GE Healthcare , Fairfield , California ) while dialyzing against 20 mM Tris pH 8 . 0 , 100 mM KCl , 0 . 2 mM DTT for 2 hr at 22°C followed by overnight incubation at 4°C . The mixture was brought to 20 mM imidazole , incubated with Talon beads to remove uncleaved protein , and loaded onto a heparin column equilibrated in buffer A ( 25 mM Tris pH 8 . 0 , 0 . 1 M KCl , 1 mM DTT ) . MutH was eluted using a gradient of 0 . 1–1 . 0 M KCl in buffer A , pooled and diluted twofold with buffer A and loaded onto a MonoQ column equilibrated with buffer A . MutH was eluted using the same gradient , pooled and dialyzed overnight against 25 mM MES pH 5 . 5 , 150 mM KCl , 1 mM DTT . MutH was loaded onto a MonoS column equilibrated with 25 mM MES pH 5 . 5 , 0 . 1 M KCl , 1 mM DTT and eluted using a 0 . 1–1 . 0 M KCl gradient . Peak fractions were pooled , concentrated using Centriprep 10 and loaded onto a Superdex 75 column equilibrated with 25 mM Tris pH 8 . 0 , 250 mM KCl , 1 mM DTT . Peak fractions were pooled , concentrated , flash frozen in 25 mM Tris pH 8 . 0 , 250 mM KCl , 1 mM DTT , 50% glycerol and stored at −80°C . Single cysteine MutSΔC800 and His-tagged MutLLN40 proteins were reduced with 10 mM DTT for 20 min and O/N dialyzed into buffer B ( 25 mM Hepes pH 7 . 5 , 400 mM KCl , 5 mM MgCl2 , 10% glycerol ) at 4°C , to remove DTT . MutSΔC800 ( 0 . 57 μM ) was incubated with 100-bp DNA containing a G:T mismatch ( AAACAGGCTTAGGCTGGAGCTGAAGCTTAGCTTAGGATCATCGAGGATCGAGCTCGGTGCAATTCAGCGGTACCCAATTCGCCCTATAGGCATCCAGGTT annealed with AACCTGGATGCCTATAGGGCGAATTGGGTACCGCTGAATTGCACCGAGCTTGATCCTCGATGATCCTAAGCTAAGCTTCAGCTCCAGCCTAAGCCTGTTT , 0 . 29 μM ) for 25 min on ice in buffer C ( 25 mM Hepes pH 7 . 5 , 125 mM KCl , 5 mM MgCl2 ) . MutLLN40 ( 4 μM ) was incubated with 5 mM ATP for 25 min on ice . MutSΔC800/DNA and MutLLN40/ATP samples were then combined to final protein concentrations 0 . 4 μM ( DNA concentration 0 . 2 μM ) and additional ATP was added to a final concentration of 1 mM . Samples were then incubated for 10 min at RT , after which they were adjusted to 37°C for 2 min . Crosslinker ( BMOE or BM[PEO]3 , Pierce , Thermo Fisher scientific , Waltham , MA , dissolved to 0 . 5 mM in DMSO ) was added to a final concentration of 50 μM and samples were incubated for exactly 2 min at 37°C . Reactions were stopped by adding protein loading buffer with DTT and crosslinking was assessed on SDS-PAGE gels stained with coomassie . To obtain crystallizable amounts of crosslinked MutSΔC800/MutLLN40 complex , equimolar amounts of MutSΔC800 D246C and His-tagged MutLLN40 N131C ( or with additional arginine mutations ) were reduced and dialyzed separately , as described above . MutLLN40 was diluted to 2 μM in buffer D ( 25 mM Hepes pH 7 . 5 , 400 mM KCl , 10% glycerol ) and incubated with a 5-fold molar excess of BM ( PEO ) 3 ( from 50 mM solution in DMSO ) for 10 min at 4°C . The low MutLLN40 concentration prevented the formation of MutLLN40-MutLLN40 crosslinks , while the excess crosslinker ensured each MutLLN40 to react with one maleimid group so that the other reactive side of the crosslinker remained available . The MutLLN40 was then bound to Talon beads and the beads were subsequently washed with 20 column volumes of buffer D and 20 column volumes of buffer E ( 25 mM Hepes pH 7 . 5 , 150 mM KCl , 10% glycerol , 5 mM imidazole ) to remove excess crosslinker . MutSΔC800 was incubated for 10 min with equimolar amounts of 30-bp DNA with a G:T mismatch at position 9 ( AGCTGCCAGGCACCAGTGTCAGCGTCCTAT annealed with ATAGGACGCTGACACTGGTGCTTGGCAGCT ) in buffer C . The DNA-bound MutSΔC800 was then added to the Talon-bound MutLLN40 , and 30-fold excess ATP was immediately added after which everything was incubated to crosslink for 1 hr at 4°C . The beads were then washed with 10 column volumes buffer E to remove MutSΔC800-MutSΔC800 crosslinks , after which the protein was eluted in buffer E with 300 mM imidazole and DTT was added to quench excess crosslinker . The protein was bound to a 5 ml heparin column ( GE Healthcare , Fairfield , California ) and eluted with a 0 . 1–1 M KCl gradient , which removed the DNA from the protein . The elution was subsequently concentrated and purified with size-exclusion chromatography in buffer B containing 1 mM DTT , for which two S200 16/60 columns were coupled resulting in one long column . The MutSΔC800/MutLLN40 protein peak was then concentrated , after which the MutSΔC800 concentration was estimated using ε = 95 , 238 and the whole process ( including DTT incubation and dialysis ) was repeated to obtain S2L2 complexes . The resulting protein ( 5–10% final yield ) was concentrated to 80–90 μM ( expressed in MutS monomer concentrations; ε = 94 , 660 ) and flash-frozen until further use . For crystallization , 50 μM MutSΔC800/MutLLN40 complex was incubated with 25 μM DNA containing a G:T mismatch ( 27-bp: TGCCAGGCACCAGTGTCAGCGTCCTAT annealed with ATAGGACGCTGACACTGGTGCTTGGCA or 100-bp , same sequence as above ) for 25 min on ice . AMP-PNP was subsequently added to a concentration of 1 mM and the protein was crystallized at 4°C using vapor diffusion in 9–12% PEG-8000 , 100 mM Tris pH 7 . 0 , 200 mM MgCl2 , and 80–450 mM sodium malonate . Microseeding was used to increase crystal nucleation . Crystals were cryoprotected in mother liquor supplemented with 25% ethylene glycol and 100 mM KCl before flash-cooling in liquid nitrogen . Diffraction data were collected at 100 K at beamline ID-29 at the ESRF or beamline PX-III at the SLS . Crystallographic data were processed with XDS ( Kabsch , 2010 ) or iMOSFLM ( Powell et al . , 2013 ) and scaled using Aimless from the CCP4 suite ( Winn et al . , 2011 ) . Crystal structures were solved in consecutive steps of finding domains using Phaser ( McCoy et al . , 2007 ) . Several search models were used , but best results were obtained with domains from chain A of PDB entry 1W7A as search models for MutSΔC800 and chain A from PDB entry 1BKN for MutLLN40 , while clear density for residues 150–164 of MutLLN40 allowed building as in PDB entry 1NHJ . The search process was improved by going back and forth between the different datasets to find missing domains . Initial structure solution was performed starting from crystal form 1 as follows: first , a search model consisting of residues 267–800 of chain A of PDB entry 1W7A ( MutS ) was searched twice using Phaser , which resulted in a solution with these chains forming a tilted MutS dimer . Next , this solution was used together with a search model consisting of chain A of 1BKN ( MutLLN40 ) , which placed this protein against the ATPase domain of one MutS subunit . Then , the second MutLLN40 was found with Phaser using a brute rotation search of 15° around the angle that would orient this MutLLN40 on the other side of the MutS dimer in a similar manner as the first , and automated translation , packing and refinement steps by Phaser indeed placed the MutLLN40 in the symmetrically equivalent position . One connector domain ( residues 128–266 of chain A of 1W7A ) was then found with Phaser , and the second connector domain was placed using similar steps as for the second MutLLN40 search . Thus the search identified the equivalent dimeric counterpart three times for separate parts of the complex ( the main MutS chain , MutLLN40 and connector domains ) . The resulting MutS-MutLLN40 complex structure could then be used as a search model in all crystal forms and easily identified equivalent complexes in each of those ( present three times in the asymmetric units in the 6 . 6 Å and 7 . 6 Å datasets ) . Finally , for crystal form 1 , an additional ‘half complex’ was found with Phaser using one MutS chain and one MutLLN40 chain of the existing complex structure . This second complex forms a symmetry-generated dimer over a twofold axis , with similar MutS-MutLLN40 interfaces , but the MutS clamp domains in this crystallographic dimer could not be modeled . This second conformer forms a more compact MutS dimer , probably due to crystal packing , but since it has identical interfaces with MutLLN40 we focussed on the main conformation throughout this paper . Excellent quality of the structure solutions after molecular replacement with the complete but unrefined models is evident from the Phaser statistics: TFZ = 9 . 0/LLG = 996 for 4 . 7 Å; TFZ = 14 . 2/LLG = 899 for 6 . 6 Å; and TFZ = 13 . 0/LLG = 795 for the 7 . 6 Å dataset . Refinement was first performed using rigid body refinement in REFMAC5 ( Murshudov et al . , 1997 , 2011 ) , for which the following domains of MutS were defined: residues 128–266 , 267–765 , 766–800; and for MutL: residues 20–204 , 205–331 . Next , limited restrained refinements were performed , first using ProSMART-generated external restraints ( Nicholls et al . , 2012 ) to the PDB_REDO-optimized ( Joosten et al . , 2012 ) entries of chain A of 1W7A and chain A of 1BKN in order to ensure consistency with prior observations , followed by TLS and jelly-body refinement in latter stages . PDB_REDO-optimized homologues were used for external restraint generation in order to maximize reliability of the prior structural information . All refinements were performed using REFMAC5 ( Murshudov et al . , 1997 , 2011 ) . During refinement , clear density became visible for missing residues 150–164 of the MutLLN40 subunits , which followed the conformation of PDB entry 1NHJ . Interestingly , this conformation was different from that in the MutL search model state , indicating this to be real signal , and not due to bias from the search model . Also , AMP-PNP could be placed in density in the nucleotide binding sites of MutS . During intermediate stages , PDB_REDO and MolProbity ( Chen et al . , 2010 ) were used to correct geometry and perform side-chain flips . After refinement , all structures were in the 97th–100th Clashscore and 98th–100th MolProbity score percentiles . Refinement and data collection statistics can be found in Table 1 . Figures and videos were generated with MacPyMOL ( http://www . pymol . org ) , interpolations between conformations were created with LSQMAN ( Kleywegt and Jones , 1994 ) and electrostatic surface with CCP4mg ( Winn et al . , 2011 ) . Protein interface areas were calculated using PISA ( Winn et al . , 2011 ) for which the missing loop of residues 126–131 of MutLLN40 in interface 2 was modeled as in PDB entry 1NHJ . To look at changes within MutS dimers , we used MutS D835R dimer ( Manelyte et al . , 2006; Groothuizen et al . , 2013 ) variants that do not form tetramers , with single cysteines in positions R449C ( His-tagged ) , D246C , S798C , or A336C . The proteins were labeled with Alexa Fluor 488 or Alexa Fluor 594 maleimide ( Invitrogen , Thermo Fisher scientific , Waltham , MA ) according to the manufacturers instruction . Excessive dye was removed using Zeba Spin Desalting columns ( Thermo Fisher scientific , Waltham , MA ) and the degree of labeling determined from the absorbance spectra recorded from 220–700 nm ( nanodrop ) according to the manufactures instructions . Clamp-domain crossover movement and connector domain movement within MutS dimers were measured using FRET in which fluorescence emission spectra were recorded with excitation at either 485 nm ( 5 nm slit width ) for FRET or 590 nm ( 5 nm slit width ) for direct acceptor measurements . FRET was determined by the ratio between signal at 485 and 615 nm while direct acceptor was determined by the ratio between signal at 590 and 615 nm and after correction for spectral crosstalk the ratio FRET/acceptor was calculated , and normalized for unbound protein . Heterodimers of single-cysteine MutS variants labeled with Alexa Fluor 488 and Alexa Fluor 594 , respectively , were allowed to form by mixing 200 nM of each protein and incubation at 22°C for at least 30 min in the absence of ADP in buffer F ( 25 mM Hepes pH 7 . 2 , 150 mM KCl and 5 mM MgCl2 ) supplemented with 0 . 05% TWEEN-20 . Next , 200 nM of 59-bp DNA with a G:T mismatch ( TGAAGCTTAGCTTAGGATCATCGAGGATCGAGCTCGGTGCAATTCAGCGGTACCCAATT annealed with AATTGGGTACCGCTGAATTGCACCGAGCTTGATCCTCGATGATCCTAAGCTAAGCTTCA , with blocked ends as described above ) was added , followed by addition of 1 mM ATP . As a homoduplex control 240 pM λ-DNA ( corresponding to 200 nM of the 59 bp blocked Heteroduplex-DNA ) was used . MutS-DNA FRET was measured in a Hitachi Fluorescence spectrofluorimeter F-4500 ( Hitachi Ltd , Japan ) ( Program FL Solutions 2 . 0 ) . Fluorescence emission spectra ( 600–700 nm ) were recorded with excitation at either 435 nm ( 5 nm slit width ) for FRET or 590 nm ( 5 nm slit width ) for direct acceptor measurements . FRET was determined by the ratio between signal at 435 and 615 nm while direct acceptor was determined by the ratio between signal at 590 and 615 nm and after correction for spectral crosstalk the ratio FRET/acceptor was calculated . We used 30-bp DNA with or without a G:T mismatch ( AATTGCACCGAGCTTGATCCTCGATGATCC annealed with complementary strand or GGATCATCGAGGATCGAGCTCGGTGCAATT ) , where the T-containing strand had 5' and 3' digoxigenin labels so that both DNA ends were blocked with anti-digoxigenin Fab fragments ( Roche Diagnostics , F . Hoffmann-La Roche Ltd , Switzerland ) . 100 nM of the DNA with 6 µM SYTOX Blue ( Invitrogen , Thermo Fisher scientific , Waltham , MA ) was mixed with 200 nM MutS variants labeled with Alexa Fluor 594 in buffer F , after which ATP was added to 1 mM to induce the conformational change in MutS . Spontaneous mutation rates were assessed using acquired rifampicin resistance . Strains KR1517 ( mutS , as in [Lamers et al . , 2004] ) or GM4250 ( gift from M Marinus , [Aronshtam and Marinus , 1996] ) ( mutL ) were transformed with empty vector or plasmid containing WT or mutant MutS or His-MutL genes , and plated on LB/agar plates with 50 μg/ml carbenicillin and 30 μg/ml kanamycin . After overnight incubation at 37°C , single colonies were picked and grown in 10 ml LB with antibiotics to OD600 ∼1 . 0 . Next , 0 . 33 × 108 or 1 × 108 cells were plated on LB plates with antibiotics and 0 . 1 mg/ml rifampicin . Plates were O/N incubated at 37°C and rifampicin resistant colonies were counted . Mutation rates and 95% confidence intervals were determined using Fluctuation AnaLysis CalculatOR with the MSS maximum-likelihood method ( http://www . mitochondria . org/protocols/FALCOR . html ) . SPR experiments for binding MutSΔC800 D246C or crosslinked MutSΔC800/MutLLN40 complex to DNA were performed in a Biacore T200 system ( GE Healthcare , Fairfield , CA ) as described ( Groothuizen et al . , 2013 ) . The experiments were performed in SPR buffer containing 25 mM Hepes pH 7 . 5 , 150 mM KCl , 5 mM MgCl2 , 1 mM DTT , 0 . 05% TWEEN-20 and 1 mM ATP , using immobilized biotinylated 100-bp DNA ( same sequence as above ) with a fluorescein moiety at the other end . Full-length His6-MutL binding to the full-length MutS sliding clamp was assessed using a two-step SPR assay . The resulting graphs are not strictly affinity curves , as changes in MutS stability on DNA contribute to the observed response , but serve to assess the effect of mutations . The SPR buffer was supplemented with 20% glycerol to ensure MutL stability . Before every measurement , anti-fluorescein Fab fragment ( Invitrogen , Thermo Fisher scientific , Waltham , MA ) was injected to block the fluorescein-labeled DNA ( 100 bp , see above ) ends . MutS sliding clamps were captured on the end-blocked DNA by injecting 200 nM WT or mutant MutS protein ( in buffer with 1 mM ATP ) for 120 s . Then WT or mutant MutL protein ( in buffer with 1 mM ATP ) was injected for 120 s , followed by dissociation with buffer only . This was repeated with varying concentrations of MutL . Fluorescence polarization measurements to assess DNA-binding of MutLLN40 mutants were performed in low-salt FP buffer with 25 mM Hepes pH 7 . 5 , 50 mM KCl , 5 mM MgCl2 , 1 mM DTT and 0 . 05% TWEEN-20 . For full length MutL , the buffer was supplemented with 10% glycerol . A concentration of 0 . 5 nM of 5′ labeled TAMRA-41-bp DNA ( ATAGGACGCTGACACTGGTGCTTGGCAGCTTCTAATTCGAT annealed with complementary strand ) was used . MutL proteins were serial diluted in black 96-well microplates ( PerkinElmer Inc , Waltham , MA ) in 100 μl volumes . Polarization of the TAMRA label was read out in a PHERAstar FS machine ( BMG Labtech GmbH , Germany ) with an 540/590 ( excitation/emission ) FP module . Stopped-flow assays to assess DNA binding and kinking were performed in buffer containing 25 mM Hepes pH 7 . 5 , 150 mM KCl , 5 mM MgCl2 , 1 mM DTT , 0 . 05% TWEEN-20 and 10 μM ADP , with or without 1 mM ATP . One syringe contained 100 nM of 45-bp DNA with or without a G:T mismatch ( GTCATCCTCG[T*]CTCAAGCTGCCAGGCACCAGTGTCAGCGTCCTAT annealed with complementary strand or ATAGGACGC[T*]GACACTGGTGCTTGGCAGCTTGAGACGAGGATGAC ) which was either labeled with Alexa Fluor 594 at position 11 in the top strand and Alexa Fluor 488 at position 10 in the bottom strand ( indicated by T* ) , or with 5′-labeled with TAMRA in the top strand . The other syringe contained 400 nM MutSΔC800 D246C or crosslinked MutSΔC800/MutLLN40 complex . For assays with double-labeled DNA , donor fluorophores were excited at 473 nm and measured using filters 540IB + 540IK , while acceptor fluorophores were measured at the same time using an OG590 filter . For experiments with TAMRA-labeled DNA , the fluorophore was excited at 545 nm and OG540 filters were used for read-out . Samples were co-injected in a KinetAsyst SF-61DX2 stopped-flow machine ( TgK Scientific , UK ) fitted with R10699 photomultiplier tubes ( Hamamatsu Photonics K . K . , Japan ) at 15°C and measured for 100 s , which was repeated 5–10 times and averages were calculated . Circular DNA containing a single G:T mismatch and 12 hemi-methylated GATC sites was prepared via primer extension on single stranded DNA from a derivative of pGEM-13Zf ( gift from J Jiricny ) as described ( Baerenfaller et al . , 2006 ) with the exception that closed circular DNA was purified from gel using a Wizard gel purification kit ( Promega Corporation , Madison , WI ) . To enable quantification , an Alexa Fluor 647 labeled oligo ( IBA GmbH , Germany ) was used: CCAGACGTCTGTCGACGTTGGGAAGCT[T*]GAGTATTCTATAGTGTCACCT , where the G is nucleotide forming a G:T mismatch and the T* is the Alexa Fluor 647 labeled nucleotide . Nicking reactions contained 25 mM Hepes KOH pH 7 . 5 , 150 mM KCl , 0 . 1 mg/ml BSA , 5 mM MgCl2 , 1 mM DTT , 1 mM ATP , 0 . 5 nM circular DNA , 200 nM MutS , 200 nM WT MutL , single-cysteine MutL N131C , MutL N131C R266E or MutL N131C R162E/R266E/R316E and 100 nM MutH as well as twofold dilutions thereof . Control reactions contained either no protein or 200 nM MutS and 100 nM MutH . Reactions were incubated for 5 min at 37°C and stopped with an equal volume of 20% glycerol , 1% SDS and 50 mM EDTA . Samples were analyzed on 0 . 8% agarose gels supplemented with 1 µg/ml ethidium bromide , run in 1x TAE . Conversion of covalently closed circles into nicked product was visualized using the fluorescence of the Alexa Fluor 647 label using a Typhoon Trio Imager ( GE Healthcare , Fairfield , CA ) with excitation at 633 nm and emission filter 670BP30 . ATPase activity of WT MutS and MutS P595A/I597A/M759D was measured by coupling ATP hydrolysis to oxidation of NADH as in ( Lamers et al . , 2004 ) . MutS protein ( 5 μM ) was mixed with 3 . 125–500 μM ATP and hydrolysis was measured in a spectrophotometer during 5 min . Crosslinked MutSΔC800/MutLLN40 complex ( 1 mg/ml ) was incubated for 5 min on ice with equimolar amounts of 100-bp DNA containing a G:T mismatch ( sequence as in main text ) . MutLLN40 ( 2 mg/ml ) was incubated with the MutSΔC800/MutLLN40/DNA complex or with DNA only , and 1 mM AMP-PNP as described ( Ban and Yang , 1998 ) . Samples were injected onto a S200 5/150 column in buffer containing 20 mM Tris pH 8 . 0 , 150 mM KCl , 0 . 1 mM EDTA , 5 mM MgCl2 , 1 mM DTT and 5% glycerol . Eluted fractions were analyzed on SDS-PAGE stained with coomassie . | The genetic code of DNA is written using four letters: “A” , “C” , “T” , and “G” . Molecules of DNA form a double helix in which the letters in the two opposing strands pair up in a specific manner—“A” pairs with “T” , and “C” pairs with “G” . A cell must replicate its DNA before it divides , and sometimes the wrong DNA letter can get added into the new DNA strand . If left uncorrected , these mistakes accumulate over time and can eventually harm the cell . As a result , cells have evolved several ways to identify these mistakes and correct them , including one known as “mismatch repair” . Mismatch repair occurs via several stages . The process starts when a protein called MutS comes across a site in the DNA where the letters are mismatched ( for example , where an “A” is paired with a “C” , instead of a “T” ) . MutS can recognize such a mismatch , bind it , and then bind to another molecule called ATP . MutS then changes shape and encircles the DNA like a clamp that can slide along the DNA . Only when it forms this “sliding clamp” state can MutS recruit another protein called MutL . This activity in turn triggers a series of further events that ultimately correct the mismatch . However , it remains poorly understood how MutS forms a clamp around DNA and how and why this state recruits MutL in order to start the repair . To visualize this short-lived intermediate , Groothuizen et al . trapped the relevant complex in the presence of DNA containing a mismatch and then used a technique called X-ray crystallography to determine the three-dimensional structure of MutS bound to MutL . The structure reveals that two copies of MutS tilt across each other and open up a channel , which is large enough to accommodate the DNA . In this manner , MutS is able to form a loose ring around the DNA . The changes in the structure and the movement of the DNA to the new channel were confirmed using another technique , commonly referred to as FRET . Groothuizen et al . observed that the movements in the MutS protein also serve to make the interfaces available that can recognize MutL . If these interfaces were disturbed , MutS and MutL were unable to associate with each other , which resulted in a failure to trigger mismatch repair . Further analysis revealed that that MutL binds to DNA only after MutS has recognised the mismatch and formed a clamp around it . This is the first time that the MutS clamp and the MutS/MutL complex have been visualized , and further work is now needed to understand how MutL triggers other events that ultimately repair the mismatched DNA . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | MutS/MutL crystal structure reveals that the MutS sliding clamp loads MutL onto DNA |
The in vivo roles for even the most intensely studied microRNAs remain poorly defined . Here , analysis of mouse models revealed that let-7 , a large and ancient microRNA family , performs tumor suppressive roles at the expense of regeneration . Too little or too much let-7 resulted in compromised protection against cancer or tissue damage , respectively . Modest let-7 overexpression abrogated MYC-driven liver cancer by antagonizing multiple let-7 sensitive oncogenes . However , the same level of overexpression blocked liver regeneration , while let-7 deletion enhanced it , demonstrating that distinct let-7 levels can mediate desirable phenotypes . let-7 dependent regeneration phenotypes resulted from influences on the insulin-PI3K-mTOR pathway . We found that chronic high-dose let-7 overexpression caused liver damage and degeneration , paradoxically leading to tumorigenesis . These dose-dependent roles for let-7 in tissue repair and tumorigenesis rationalize the tight regulation of this microRNA in development , and have important implications for let-7 based therapeutics .
MicroRNAs are thought to control cellular responses to stresses such as tissue damage and transformation ( Leung and Sharp , 2010; Chivukula et al . , 2014 ) , but the impact of this idea is unclear because microRNAs have been understudied in vivo . let-7 is one of the most ancient and omnipresent microRNAs , yet relatively little is known about its functional roles in mammalian development and physiology . let-7 was first identified as a gene that regulates the timing of developmental milestones in a C . elegans screen ( Reinhart et al . , 2000 ) . In mammals , mature let-7 is undetectable in early embryos and embryonic stem cells , but becomes highly expressed in most adult tissues ( Schulman et al . , 2005; Thomson et al . , 2006 ) . A handful of previous studies have implicated let-7 in body size regulation , metabolism , stem cell self-renewal , and colon carcinogenesis ( Zhu et al . , 2011; Frost and Olson , 2011; Shyh-Chang , et al . , 2013; Nishino et al . , 2013; Madison , et al . , 2013 , but the core functions of let-7 in regeneration and disease remain incompletely understood . In addition to questions about what let-7 does , it is unknown why so many let-7s are expressed at such high levels . In mice and humans , the let-7 family is comprised of 10 to 12 members who are thought to share a common set of mRNA targets . It has been thought that deep redundancy might make it difficult to discern any phenotypes that individual let-7s might have . Essential unanswered questions regarding let-7 biology include whether let-7 members are redundant , have unique functions , or are regulated to maintain a specific total dose . Our previous study of Lin28a , which inhibits the biogenesis of each let-7 member similarly ( Heo et al . , 2008; Nam et al . , 2011 ) , suggests that total let-7 dose alterations , rather than regulation of specific members , is important . In transgenic mice , modest increase in Lin28a and consequent 40% suppression of total let-7 levels promote increased glucose uptake and an overgrowth syndrome ( Zhu et al . , 2010 ) . In this study we examined the consequences of let-7 dose disruption in cancer and organ regeneration in genetic mouse models . While let-7s have been implicated as a tumor suppressor , this has predominantly been shown in cell lines and xenograft assays ( Guo et al . , 2006; Chang et al . , 2009; Iliopoulos et al . , 2009; Viswanathan et al . , 2009; Wang et al . , 2010; Lan et al . , 2011 ) , as well as using exogenous let-7 delivery to mouse cancer models ( Esquela-Kerscher et al . , 2008; Trang et al . , 2010; Trang et al . , 2011 ) . Here , we confirmed the tumor suppressor activity of an endogenous transgenic let-7 in a MYC-driven hepatoblastoma model . However , we found that this same level of let-7 overexpression impaired liver regeneration after partial hepatectomy ( PHx ) . Furthermore , chronic high-dose let-7 resulted in severe liver damage and paradoxical liver cancer development . Overall , we provide in vivo evidence that let-7 expression levels have been developmentally constrained to balance the need for regenerative proliferation against the need to antagonize malignant proliferation , findings with implications for let-7 based therapies .
To study the effect of let-7 on carcinogenesis , we employed an inducible MYC-driven hepatoblastoma model ( Shachaf et al . , 2004 ) . In this model , most let-7s are suppressed by more than 60% ( Nguyen et al . , 2014 ) . However , MYC affects the expression of many other microRNAs ( Chang et al . , 2009; Kota et al . , 2009 ) . To test if let-7 suppression is specifically required for MYC’s oncogenic program , we simultaneously overexpressed let-7g and MYC using a triple transgenic , liver-specific , tet-off model ( Figure 1A: LAP-tTA; TRE-MYC; TRE-let-7S21L transgenic mice ) . This transgenic form of let-7g is an engineered let-7 species called let-7S21L ( let-7g Stem + miR-21 Loop ) ( Zhu et al . , 2011 ) , in which the precursor microRNA loop derives from mir-21 and cannot be bound and inhibited by Lin28b ( Figure 1B ) , which is highly expressed in MYC-driven tumors ( Chang et al . , 2009; Nguyen et al . , 2014 ) . 10 . 7554/eLife . 09431 . 003Figure 1 . let-7g inhibits the development of MYC-driven hepatoblastoma . ( A ) Schema of the liver-specific inducible LAP-MYC +/- let-7S21L cancer model . ( B ) let-7S21L is a chimeric construct containing the let-7g stem , miR-21 loop , and let-7g flanking sequences . ( C ) Schema showing that LAP-MYC +/- let-7S21L mice were induced at 14 days of age , tissues were collected at 90 days of age , and survival was followed . ( D ) Ninety-day old mice bearing tumors in the LAP-MYC ( 87 . 5% , 7/8 ) and LAP-MYC + let-7S21L ( 27 . 3% , 3/11 ) mouse models . ( E ) Livers showing tumors from the above mice . ( F ) Liver surface area occupied by tumor . ( G ) Kaplan-Meier curve of LAP-MYC alone and LAP-MYC + let-7S21L mice . ( H ) Mature let-7 expression levels in as determined by RT-qPCR . ( I ) Human c-MYC mRNA expressionin tumors as determined by RT-qPCR . ( J ) Heat map of let-7 target gene expression in WT normal livers , MYC tumors , and MYC + let-7S21L tumors as measured by RT-qPCR . Red is higher and blue is lower relative mRNA expression . ( K ) Gene expression plotted as bar graphs to show relative changes . ( L ) Evolutionarily conserved let-7 target sites within 3’UTRs ( Targetscan . org ) . All data in this figure are represented as mean ± SEM . *p < 0 . 05 , **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 00310 . 7554/eLife . 09431 . 004Figure 1—figure supplement 1 . H&E staining of LAP-MYC and LAP-MYC + let-7S21L tumor-adjacent normal tissues and tumor tissues . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 004 Induction of MYC with or without let-7S21L was initiated at 14 days of age ( Figure 1C ) . By 90 days of age , large multifocal tumors had formed in 88% of the MYC alone group , whereas single small tumors appeared in only 27% the MYC + let-7S21L group ( Figure 1D–F ) and overall survival was dramatically improved ( Figure 1G ) . The level of let-7g was increased more than eightfold in both non-tumor and tumor tissues ( Figure 1H ) . Tumors from both groups showed similar histology ( Figure 1—figure supplement 1 ) and MYC expression ( Figure 1I ) . Gene-expression within tumors showed that previously validated let-7 targets involved in proliferation and growth including Cdc25a ( Johnson et al . , 2007 ) , Cdc34 ( Legesse-Miller et al . , 2009 ) , E2f2 ( Dong et al . , 2010 ) , E2f5 ( Kropp et al . , 2015 ) , and Map4k4 ( Tan et al . , 2015 ) were upregulated in MYC-tumors , but suppressed back down to normal levels in the context of let-7 overexpression ( Figure 1J–L ) , suggesting that the repression of these targets restrains MYC-dependent tumorigenesis . These data indicated that let-7g has potent tumor suppressor activity in the context of MYC-driven hepatoblastoma . Since increasing let-7g was extremely effective at suppressing hepatoblastoma without compromising overall health , we asked if this increase in levels would impact tissue homestasis . We examined let-7g overexpression in the setting of liver injuries that drive rapid proliferation and growth . After PHx , let-7s in regenerating tissues fell , but returned to normal after forty hours ( Figure 2A ) , findings consistent with a previous report ( Chen et al . , 2011 ) . Similarly , let-7s also declined acutely after chemical injury with the xenobiotic TCPOBOP ( 1 , 4-bis-[2-[3 , 5-dichloropyridyloxy]] benzene ) ( Figure 2—figure supplement 1A ) . This shows that while let-7 increases in a temporally defined fashion during development ( Figure 2—figure supplement 1B ) , it can transiently fluctuate after environmental perturbations . To test if the observed let-7 suppression is necessary for regeneration , we induced let-7g in LAP-let-7S21L mice and performed PHx ( Figure 2B–D ) . The body weight ( Figure 2—figure supplement 1C ) , liver function ( Figure 2—figure supplement 1D ) , resected liver mass ( Figure 2E ) and histology ( Figure 2—figure supplement 1E ) were unaffected in LAP-let-7S21L mice compared to control mice . Forty hours after PHx , there was reduced liver mass and decreased Ki-67 in LAP-let-7S21L mice ( Figure 2F–H ) . Liver mass was no different at four and fourteen days , indicating a kinetic delay but not a permanent impairment ( Figure 2—figure supplement 1F , G ) . 10 . 7554/eLife . 09431 . 005Figure 2 . let-7g overexpression inhibits liver regeneration after partial hepatectomy . ( A ) Mature endogenous let-7 expression levels in WT C57Bl/6 mice at different time points after PHx as determined by RT-qPCR ( n=4 and 4 for each time point ) . ( B ) Schema of the LAP-let-7S21L dox-inducible model . LAP-tTA single transgenic mice served as the controls . ( C ) Schema showing that let-7S21L control and LAP-let-7S21L mice were induced at 42 days of age , PHx was performed after 14 days of induction , and tissues were collected 40 hr post PHx . ( D ) Mature let-7 expression levels in let-7S21L and LAP-let-7S21L livers after 14 days of induction ( n=4 and 4 ) . ( E ) Resected liver/body weight ratios of LAP-tTA Control and LAP-let-7S21L mice at the time of PHx ( n=4 and 4 ) . ( F ) Liver/body weight ratios 40 hr after PHx ( n=4 and 4 ) . ( G ) Ki-67 staining on resected and post-PHx liver tissues . ( H ) Quantification of Ki-67-positive cells ( n=2 and 2 mice; ten 20x fields for each mouse were quantified ) . ( I ) Resected liver/body weight ratios 2 days after intravenous injection of 0 . 5 mg/kg negative control or let-7g microRNA mimics packaged in C12-200 LNPs ( n=5 and 5 ) . ( J ) Liver/body weight ratios 40 hr after PHx ( n=4 and 4 ) . ( K ) Mature let-7g expression levels in mimic treated livers ( n=5 and 5 ) . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 00510 . 7554/eLife . 09431 . 006Figure 2—figure supplement 1 . Data associated with Figure 2 . ( A ) Sum of the absolute sequencing reads for mature let-7 microRNA family members after TCPOBOP treatment , as measured by small RNA sequencing . ( B ) Mature let-7 microRNA family expression in WT mouse livers at different ages as determined by RT-qPCR . Numbers over bars indicate the fold change normalized to that of 1 day old mice . ( C ) Body weights of let-7S21L alone and LAP-let-7S21L mice pre-PHx ( n=4 and 4 ) . ( D ) Liver function tests: ALT ( U/L ) and AST ( U/L ) of let-7S21L alone and LAP-let-7S21L mice pre-PHx ( n=5 and 5 ) . ( E ) H&E staining of let-7S21L alone and LAP-let-7S21L mice pre-PHx . ( F ) Schema showing that let-7S21L control and LAP-let-7S21L mice were induced at 42 days of age , PHx was performed after 14 days of induction , and tissues were collected 4 and 14 days after PHx . ( G ) Liver to body weight ratio of let-7S21L alone and LAP-let-7S21L mice 4 and 14 days after PHx ( n=5 and 5 ) . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 006 To rule out increasing demands on microRNA biogenesis machinery as a mechanism of proliferative suppression , we delivered mature control or let-7g microRNA mimics ( 0 . 5 mg/kg ) into wild-type mice two days prior to hepatectomy using C12-200 lipidoid nanoparticles ( LNPs ) ( Love et al . , 2010 ) . let-7g , but not control mimics , inhibited regeneration ( Figure 2I–K ) . While let-7 overexpression blocked MYC-induced tumorigenesis , these data show that a similar increase in let-7 levels inhibited post-injury organ growth and regeneration . To assess the physiological relevance of our gain-of-function experiments , we examined knockout mice to determine if let-7 is a bona fide regeneration suppressor . Both let-7b and let-7c2 were conditionally deleted from the liver by crossing Albumin-Cre into a let-7b/c2 floxed mouse ( "let-7b/c2 LKO" mice , Figure 3A ) . Small RNA-sequencing data from Xie et al . showed that let-7 is one of the most highly expressed microRNA families in the liver and that let-7b and let-7c2 together comprise approximately 18% of the let-7 total ( Figure 3B ) ( Xie et al . , 2012 ) . Thus , let-7b/c2 LKO mice have substantial , but far from a complete reduction of total let-7 levels . 10 . 7554/eLife . 09431 . 007Figure 3 . Loss of let-7b and let-7c2 is sufficient to enhance liver regeneration . ( A ) Schema of liver-specific let-7b and let-7c2 knockout mice ( let-7b/c2 LKO ) . Albumin-Cre excises loxPs in the embryonic liver of let-7b/c2Fl/Fl mice . Mice without Cre serve as the controls . ( B ) Small RNA sequencing showing the distribution of 10 let-7s in WT mice ( n=2 ) ( Data obtained from Xie et al . 2012 ) . ( C ) Schema showing that PHx was performed on let-7b/c2Fl/Fl and let-7b/c2 LKO mice at 56 days of age and tissues were collected 40 hr post PHx . ( D ) Resected liver/body weight ratios at the time of PHx , and ( E ) Liver to body ratios of let-7b/c2Fl/Fl ( n=11 ) and let-7b/c2 LKO mice ( n=10 ) 40 hr after PHx . ( F ) Ki-67 staining and ( G ) Quantification of Ki-67-positive cells on resected and 40 hr post-PHx liver tissues ( n=3 and 3 mice; total of five 40x fields/mouse were used for quantification ) . ( H ) RT-qPCR on let-7 family members from let-7b/c2Fl/Fl and let-7b/c2 LKO mice pre- and 40 hr post-PHx . ( I ) Viability of H2 . 35 immortalized human hepatocytes treated with either scrambled , let-7a , or let-7b antiMiRs , measured at two and three days after transfection ( n=10 each ) . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 00710 . 7554/eLife . 09431 . 008Figure 3—figure supplement 1 . Characterization of let-7b/c2 LKO mice . ( A ) Body weight , liver weight , and liver to body weight percent of let-7b/c2Fl/Fl ( n = 7 ) and let-7b/c2 LKO mice ( n=5 ) at 56 days of age . ( B ) H&E staining of let-7b/c2Fl/Fl ( n=3 ) and let-7b/c2 LKO livers ( n=3 ) . ( C ) Schema showing that PHx was performed on let-7b/c2Fl/Fl and let-7b/c2 LKO mice at 56 days of age and tissues were collected at 4 , 7 , and 14 days post PHx . ( D ) Liver to body weight percent of let-7b/c2Fl/Fl ( n=11 ) and let-7b/c2 LKO mice ( n=10 ) 4 , 7 , and 14 days after PHx . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 00810 . 7554/eLife . 09431 . 009Figure 3—figure supplement 2 . Post-natal deletion of let-7b/c2 also enhances liver regeneration . ( A ) Schema showing that let-7b/c2 +/+ and let-7b/c2Fl/Fl mice were injected with AAV-Cre at 49 days of age , PHx was performed 7 days after viral injection , and tissues were collected 40 hr post PHx . ( B ) DNA gel showing excised let-7b/c2 band in let-7b/c2Fl/Fl + AAV-Cre mice ( n=5 ) but not in let-7b/c2 +/+ + AAV-Cre mice . ( C ) Percentage of resected liver/body weight ratios of let-7b/c2 +/+ + AAV-Cre ( n=5 ) and let-7b/c2Fl/Fl + AAV-Cre mice ( n=5 ) at the time of PHx . ( D ) Liver/body weight ratios of the above mice 40 hr after PHx . ( E ) Ki-67 staining on resected and 40 hr post-PHx livers from the above mice ( n=3 and 3 ) . ( F ) Quantification of Ki-67-positive cells ( n=2 and 2 mice; total of ten 40x fields/mouse were used for quantification ) . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 009 These LKO mice were healthy and showed normal liver/body weight ratios and histology at baseline ( Figure 3—figure supplement 1A , B ) . An identical amount of liver mass was resected from let-7b/c2Fl/Fl control and let-7b/c2 LKO mice ( Figure 3C , D ) , but LKO mice exhibited significant increases in liver mass and proliferation 40 hr after surgery ( Figure 3E–G ) . Four and seven days after PHx , there were no differences in liver weights , indicating that other phases of regeneration were unaffected ( Figure 3—figure supplement 1C , D ) . At fourteen days , the liver weight precisely achieved pre-surgery levels in control and LKO mice , indicating accelerated but not excessive regeneration ( Figure 3—figure supplement 1C , D ) . There was no compensatory upregulation of other let-7s in pre- or post-PHx tissues ( Figure 3H ) , supporting the concept that let-7 is a dose-dependent regeneration suppressor . Cre under the Albumin promoter is expressed in embryonic hepatoblasts that give rise to both hepatocyte and bile duct compartments ( Postic and Magnuson , 1999 , 2000; Xu et al . , 2006; Weisend et al . , 2009; Malato et al . , 2011 ) , so developmental influences of let-7 loss could have led to adult phenotypes . To define cell- and temporal-specific roles of let-7b/c2 , we used adeno-associated virus expressing Cre ( AAV8 . TBG . PI . Cre . rBG , hereafter called "AAV-Cre" ) , known to mediate efficient gene excision in hepatocytes but never in biliary epithelial cells ( Yanger et al . , 2013 ) ( Figure 3—figure supplement 2A , B ) . These adult and hepatocyte-specific conditional knockout mice also exhibited significantly enhanced regenerative capacity ( Figure 3—figure supplement 2C–F ) . To test if proliferative effects are specific to particular let-7 species , we knocked-down either let-7a or let-7b in SV40 immortalized hepatocytes ( H2 . 35 cells ) and found that both led to increased proliferation ( Figure 3I ) . Collectively , our data shows that physiological let-7 levels regulate the kinetics of adult liver regeneration by hepatocytes . let-7 was previously demonstrated to regulate the insulin-PI3K-mTOR pathway ( Zhu et al . , 2011; Frost and Olson , 2011 ) , which is also important in liver regeneration ( Okano et al . , 2003; Chen et al . , 2009; Haga et al . , 2009; Espeillac et al . , 2011 ) . To avoid auto-regulatory feedback and compensation as confounding factors , we focused on liver tissues exposed to acute let-7 gain or loss . In regenerating livers treated with let-7g mimic ( Figure 2I–K ) , we found significant protein suppression of insulin receptor β , Igf1rβ , and Irs2 , previously validated let-7 targets at the top of the insulin pathway ( Figure 4A , B ) ( Zhu et al . , 2011 ) . In addition to insulin signaling components , the expression of cell cycle genes ( Ccnb1 , Cdc34 , and Cdk8 ) and Map4k4 were also downregulated ( Figure 4C ) . In mice with acute let-7b/c2 deletion by AAV-Cre ( Figure 3—figure supplement 2 ) , there was a small increase in insulin receptor β protein levels ( Figure 4D , E ) . Increased mTOR signaling was also evident in the increased phospho-S6K/Total S6K and phospho-S6/Total S6 ratios ( Figure 5E ) . 10 . 7554/eLife . 09431 . 010Figure 4 . let-7g suppresses liver regeneration through insulin-PI3K-mTOR . ( A ) Western blots of insulin receptor β , Igf1rβ , Irs2 , and β-Actin in negative control or let-7g microRNA mimic treated liver tissues 40 hr after PHx . ( B ) Quantification of intensity of insulin receptor β , Igf1rβ , Irs2 ( Image J ) . ( C ) Cell cycle gene expression in let-7S21L alone ( n=4 ) and LAP-let-7S21L ( n=4 ) livers before and 40 hr after PHx as determined by RT-qPCR . ( D ) Western blots of insulin receptor β , Igf1rβ , p-S6K , total S6K , β-Actin , p-S6 ( Ser235/236 ) , and total S6 in AAV-Cre treated let-7b/c2 +/+ and let-7b/c2Fl/Fl livers ( n=5 and 5 ) . ( E ) Quantification of intensity of insulin receptor β/β-Actin , p-S6K/total S6K , and p-S6/total S6 , 40 hr after PHx ( Image J ) . ( F ) Rapamycin treatment during and after PHx in let-7b/c2Fl/Fl control and let-7b/c2 LKO mice . Shown are liver weights 40 hr post PHx . ( G ) INK128 treatment during and after PHx in let-7b/c2Fl/Fl control and let-7b/c2 LKO mice . Shown are liver weights 40 hr post PHx . ( H ) Western blots of p-S6K , total S6K , and β-Actin in let-7b/c2Fl/Fl control and let-7b/c2 LKO livers treated with either vehicle or INK128 at 40 hr post PHx . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 01010 . 7554/eLife . 09431 . 011Figure 5 . Chronic high-dose let-7g causes hepatotoxicity and liver carcinogenesis . ( A ) Schema showing that let-7S21L control and Rosa-rtTA; let-7S21L mice were induced at 42 days of age and collected at 84 days . ( B ) Images showing the whole body , extremities , and livers of Rosa-rtTA ( n=4 ) and Rosa-let-7S21L mice ( n=3 ) given 1 mg/mL dox between 42 and 84 days of age . ( C ) Liver function tests: AST ( U/L ) , ALT ( U/L ) , and total bilirubin ( mg/L ) in these mice . ( D ) H&E staining of livers . ( E ) RT-qPCR of mature let-7s and other microRNAs in let-7g overexpressing mice ( n=4 and 3 ) . ( F ) Body weight 3 days after injection of 2 . 0 mg/kg negative control or let-7g microRNA mimics packaged in C12-200 LNPs relative to pre-injection weight ( n=5 and 4 ) . ( G ) Liver function tests: AST ( U/L ) and ALT ( U/L ) in WT C57Bl/6 mice before and 3 days after mimic injection ( n=5 and 4 ) . ( H ) Mature let-7 levels in wild-type C57Bl/6 mice treated with mimics as determined by RT-qPCR ( n=5 and 4 ) . ( I ) Kaplan-Meier curve for Rosa-let-7S21L induced with 1 . 0 g/L dox at 6 weeks old ( n=15 and 17 ) . ( J ) Gross images of the liver of Rosa-let-7S21L mice induced for 18 months . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 01110 . 7554/eLife . 09431 . 012Figure 5—figure supplement 1 . Data associated with Figure 5 . ( A ) Body , liver , and liver/body weight ratios of control ( n=4 ) and Rosa-let-7S21L ( n=3 ) mice after 42 days of induction . ( B ) Gross image of Rosa-rtTA ( non-induced ) and Rosa-miR26a-2 ( induced ) mice under 1 . 0 g/L dox for 42 days . ( C ) Liver function tests: AST ( U/L ) , ALT ( U/L ) , and total bilirubin ( mg/L ) ( n=2 and 2 ) after induction . ( D ) H&E staining of Rosa-rtTA ( non-induced ) and Rosa-miR26a-2 ( induced ) mice after induction . All data in this figure are represented as mean ± SEM . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09431 . 012 To determine if mTOR signaling is functionally relevant in LKO mice , we treated mice with rapamycin two hours prior to and immediately after PHx . Rapamycin abrogated differences in regenerating liver weights between control and LKO mice ( Figure 4F ) , demonstrating that let-7b/c2 loss promotes additional mTOR activation to enhance regeneration . Rapamycin’s allosteric inhibition of mTOR can lead to pleiotropic and unpredictable effects due to cell-type specific and feedback related phenomena ( Thoreen et al . , 2009 ) . INK128 is a second generation mTOR inhibitor that directly competes with ATP at the catalytic domains of mTORC1/2 , leading to more complete abrogation of 4EBP and S6K1 ( Hsieh et al . , 2012 ) . INK128 , similar to rapamycin , completely abrogated the regenerative enhancement associated with let-7b/c2 loss ( Figure 4G ) . Analysis of p-S6K confirmed that mTOR is hyperactivated in LKO livers and that INK128 extinguishes the mTOR dependent phosphorylation of this substrate ( Figure 4H ) . Similar results after rapamycin and INK128 indicated that mTOR and its substrates play an essential role in driving increased regeneration in the context of let-7 suppression . Since acute let-7g induction interferes with hepatocyte proliferation , we asked what the effects of chronic high-dose let-7g induction might be . To answer this question , we induced let-7g using rtTA under the control of the Rosa promoter , which drives higher expression than the LAP promoter ( Figure 5A ) . These mice lost significant body weight and became jaundiced ( Figure 5—figure supplement 1A and Figure 5 ) . Liver function tests indicated severe liver injury ( Figure 5C ) and histology showed prominent microvesicular steatosis , characterized by intra-cytoplasmic lipid droplets ( Figure 5D ) , a finding associated with drug-induced liver injury , acute fatty liver of pregnancy , or Reye’s syndrome in humans . Using this system , mature let-7g was overexpressed by more than twentyfold , as compared to ∼eightfold induction in the LAP-let-7S21L system ( Figure 5E ) . Liver dysfunction was not seen after low-dose let-7 overexpression in LAP-let-7S21L mice ( Figure 2—figure supplement 1 ) or after high-dose miR-26a-2 overexpression in Rosa-rtTA; TRE-miR-26a-2 mice ( Figure 5—figure supplement 1B–D ) , suggesting a dose and let-7 microRNA specific effect . Another possibility was that the miR-21 loop of the let-7S21L construct saturated the microRNA biogenesis machinery , thus causing non-specific toxicity independent of the let-7 seed sequence . To address this we again delivered a higher dose ( 2 . 0 mg/kg ) of mature let-7g and control microRNA mimics , which do not harbor loops or tails , into wild-type mice . Mice receiving let-7g mimic lost significantly more weight ( Figure 5F ) and suffered hepatocyte destruction leading to increased AST/ALT levels ( Figure 5G ) , while control mice remained healthy . Mimic delivery achieved a twelvefold increase of let-7g ( Figure 5H ) . These results suggest that above certain doses let-7 is incompatible with hepatocyte survival , and that let-7’s anti-proliferative activities would interfere with normal tissue homeostasis . When the Rosa-let-7S21L mice were induced chronically , approximately 50% of the mice survived the acute liver injury seen after dox induction ( Figure 5I ) . Over the course of 18 months , 5 of 10 ( 50% ) of these surviving mice developed large liver tumors , whereas only 1 of 12 ( 7 . 7% ) of the non-induced mice had any tumors ( Figure 5J ) . Chronic let-7 overexpression likely caused hepatocyte toxicity and selected for pre-malignant hepatocytes that eventually become cancer . Our long-term experiments revealed the potential dangers of chronic let-7 treatment , and the consequent disruption of the balance between tissue regeneration , degeneration , and cancer risk .
The role of let-7s in adult animal physiology is unclear in part because the redundancy of this large microRNA family has made loss of function studies challenging . Deep redundancy of multiple highly conserved genes raises the possibility that dose regulation is important . Despite this , overexpression has been helpful in uncovering physiological functions of let-7 ( Zhu et al . , 2011; Frost and Olson , 2011; Shyh-Chang et al . , 2013; Nishino et al . , 2013; Madison et al . , 2013 ) . Using overexpression tools , we have shown that let-7 suppression is a fundamental requirement for MYC-mediated liver transformation , and that let-7 is capable of counteracting strong oncogenic drivers in vivo . However , one negative consequence of raising the level of let-7 expression is a limitation in the ability to regenerate after major tissue loss . More surprisingly , our knockout mouse model showed that the loss of two out of the ten let-7 members in the liver resulted in improved liver proliferation and regeneration . These data suggest a lack of complete redundancy between let-7 microRNA species , but rather a precisely regulated cumulative dose that when increased or decreased , leads to significant alterations in regenerative capacity . The knowledge that let-7 suppresses both normal and malignant growth will have particular relevance to malignancies that arise from chronically injured tissues . In these tissues , winners of the competition between cancer and host cells might ultimately dictate whether organ failure or tumor progression ensues . It has been thought that one key advantage of using microRNAs therapeutically is that they are already expressed at high levels in normal tissues , thus making increased dosing likely to be safe and tolerable . Surprisingly , we showed that chronic let-7 overexpression caused hepatotoxicity , disrupted tissue homeostasis , ultimately leading to carcinogenesis . This is likely due to the high levels of overexpression achieved in the Rosa-rtTA transgenic system , as opposed to the lower dose in the LAP-tTA system . These high doses are likely to be toxic to hepatocytes , a phenomena compounded by the fact that excess let-7 impairs proliferation in surviving cells that might serve to replenish lost tissues . Eventually , certain clones must epigenetically or genetically evolve to evade let-7 growth inhibition in order to transform . It is also interesting that let-7 overexpression led to dramatically different outcomes in distinct cancer contexts . While dose is most likely the critical variable between the Rosa-rtTA and LAP-tTA systems , Rosa-rtTA does induce expression in cells other than hepatocytes and bile duct epithelia , leaving the possibility that non-cell autonomous influences of let-7 overexpression play a role in liver injury and cancer development . A more interesting possibility would be if distinct genetic subtypes of cancer respond differently to let-7 overexpression . Since let-7 has been conceptualized as a general tumor suppressor , it is surprising that it can cause opposing phenotypes in distinct cancer models . MYC liver cancers show a dramatic suppression of let-7 , rendering it especially sensitive to let-7 replacement . Tumors or tissues with more normal levels of let-7 might not respond to increases in let-7 . Alternatively , the growth of other cancer models may not depend on the overproduction of let-7 target genes/proteins . let-7 overexpression in these contexts would probably not elicit growth suppression , but may instead exacerbate tissue injury . It would be interesting to evaluate the effects of let-7 overexpression in hepatocellular carcinomas caused by different driver mutations . Together , our data suggest that let-7 therapy directed at hepatocellular carcinomas could be risky , given that most of these cancers occur in severely compromised , cirrhotic livers ( Yang et al . , 2011 ) . We speculate that the total dose of let-7 is evolutionarily determined via regulation of the expression levels of individual let-7 members , and is postnatally maintained at a level that can suppress cancer , but which also allows for adequate levels of mammalian regenerative capacity . Clearly , let-7 levels are not static throughout life , since let-7 levels are dynamic after environmental perturbations . However , when baseline let-7 levels are altered permanently by genetic means , compromises in tumor suppression or tissue regeneration were revealed . Our study underscores the importance of regulating appropriate levels of this small RNA to maintain health during times of regenerative stress .
All mice were handled in accordance with the guidelines of the Institutional Animal Care and Use Committee at UTSW . MYC tumor models and the LAP-let-7S21L inducible mice were carried on a 1:1 FVB/C57Bl/6 strain background . Please see ( Nishino et al . , 2013 ) for more details about the let-7b/c2 floxed mice , which are on a C57Bl/6 background . The chronically injured let-7 inducible mice were on a mixed B6/129 background . All experiments were done in an age and sex controlled fashion unless otherwise noted in the figure legends . Two-thirds of the liver was surgically excised as previously described ( Mitchell and Willenbring , 2008 ) . Total RNA was isolated using Trizol reagent ( Invitrogen ) . For RT-qPCR of mRNAs , cDNA synthesis was performed with 1 ug of total RNA using miScript II Reverse Transcription Kit ( Cat . #218161 , Qiagen ) . Gene expression levels were measured using the ΔΔCt method as described previously ( Zhu et al . , 2010 ) . Mouse liver tissues were ground with a pestle and lysed in T-PER Tissue Protein Extraction Reagent ( Thermo Scientific Pierce ) . Western blots were performed in the standard fashion . The following antibodies were used: Anti-Insulin receptor β ( Cell Signaling #3025 ) , Anti-Igf1rβ ( Cell Signaling #9750 ) , Anti-Irs2 ( Cell Signaling #3089 ) , Anti-total S6K ( Cell Signaling #9202 ) , Anti-p-S6K ( Cell Signaling #9205 ) , Anti-total S6 ( Cell Signaling #2217 ) , Anti-p-S6 Ser235/236 ( Cell Signaling #2211 ) , Anti-mouse β-Actin ( Cell Signaling #4970 ) , Anti-rabbit IgG , HRP-linked Antibody ( Cell Signaling #7074 ) and Anti-mouse IgG , HRP-linked Antibody ( Cell Signaling #7076 ) . Tissue samples were fixed in 10% neutral buffered formalin or 4% paraformaldehyde ( PFA ) and embedded in paraffin . In some cases , frozen sections were made . Immunohistochemistry was performed as previously described ( Zhu et al . , 2010 ) . Primary antibodies used: Ki-67 ( Abcam #ab15580 ) . Detection was performed with the Elite ABC Kit and DAB Substrate ( Vector Laboratories ) , followed by Hematoxylin Solution counterstaining ( Sigma ) . Blood samples ( ∼50 ul ) were taken retro-orbitally in heparinized tubes . Liver function tests were analyzed by the UTSW Molecular Genetics core . 100 μL of AAV8 . TBG . PI . Cre . rBG ( University of Pennsylvania Vector Core ) was retro-orbitally injected at a titer of 5 x× 1010 genomic particles to mediate 90%-–100% Cre excision . The H2 . 35 cell line was directly obtained from ATCC and has been cultured for less than 6 months . The cells were authenticated by ATCC using Short Tandem Repeat ( STR ) DNA profiling . Cells were cultured in DMEM with 4% ( vol/vol ) FBS , 1x Pen/Strep ( Thermo Scientific ) and 200 nM Dexamethasone ( Sigma ) . Cells were transfected with control ( Life Technologies Cat . AM17010 ) , let-7a ( Life Technologies Cat . #4464084-Assay ID MH10050 ) , or let-7g ( Life Technologies Cat . #4464084-Assay ID MH11050 ) miRVana antiMiRs . AntiMiRs were packaged by RNAiMAX ( Invitrogen ) and transfected into H2 . 35 cells cultured in 96-well plates at a concentration of 50 nM . The number of viable cells in each well was measured at 2 and 3 days after transfection using CellTiter-Glo Luminescent Cell Viability Assay ( Promega Cat . #G7570 ) . For in vivo experiments , formulated C12-200 lipidoid nanoparticles ( LNPs ) were used to package either Control ( Life Technologies Cat . #4464061 ) or let-7g ( Life Technologies Cat . 364 #4464070-Assay ID MC11758 ) miRVana mimic at either 0 . 5 or 2 mg/kg and delivered intravenously through the tail vein . LNPs were formulated following the previously reported component ratios ( Love et al . , 2010 ) with the aid of a microfluidic rapid mixing instrument ( Precision Nanosystems NanoAssemblr ) and purified by dialysis in sterile PBS before injection . Rapamycin ( LC Biochem ) was dissolved in 25% ethanol/PBS and then injected at 1 . 5 mg/kg 2 hr prior to and 20 hr after PHx . INK128 ( LC Biochem ) was formulated in 5% polyvinylpropyline , 15% NMP , 80% water and administered by oral gavage at 1 mg/kg 2 hr prior to and 20 hr after PHx . Female CD1 mice were treated with 3 mg/kg TCPOBOP in DMSO-corn oil by gavage ( Tian et al . , 2011 ) , sacrificed at 3 , 6 , 9 , 12 , and 18 hr after treatment , and compared to untreated controls . Replicate libraries were made from two individual mice for each condition . RNA was purified with the Qiagen miRNeasy Mini kit . Small RNA libraries were constructed using an Illumina Truseq Small RNA Sample Prep Kit . 12 indexed libraries were multiplexed in a single flow cell lane and received 50 base single-end sequencing on an Illumina HiSeq 2500 sequencer . Sequence reads were aligned to mm9 using Tophat and quantified with Cufflinks by the FPKM method ( Trapnell et al . , 2012 ) . Data for each experimental condition represent the average values from two libraries . The data in most figure panels reflect multiple experiments performed on different days using mice derived from different litters . Variation is always indicated using standard error presented as mean ± SEM . Two-tailed Student's t-tests ( two-sample equal variance ) were used to test the significance of differences between two groups . Statistical significance is displayed as p<0 . 05 ( * ) or p<0 . 01 ( ** ) unless specified otherwise . In all experiments , no mice were excluded form analysis after the experiment was initiated . Image analysis for the quantification of cell proliferation , cell death , and fibrosis were blinded . | The development of animals is guided by the expression of certain genes at critical moments . Many different mechanisms control development; in one of them , the expression of genes can be decreased by molecules called microRNAs . In particular , the group of microRNAs called let-7 has been intensively studied in roundworms and fruit flies . Although mammals have extremely similar let-7 microRNAs they seem to be more important during adulthood . Previous studies using cells grown in the laboratory have shown that mammalian let-7 microRNAs decrease cell proliferation and cell growth . Furthermore , in mouse models of various cancers , let-7 microRNAs often reduce tumour growth when they are supplied to adult mice . Therefore , overall the let-7 group has been classified as genes that act to suppress tumors , and thus protect mice ( and most likely humans too ) from cancers . However , in-depth analysis of let-7 microRNAs was still missing . Wu and Nguyen et al . have now studied mice with liver cancer using strains where they were able to regulate the levels of let-7 . These mice overproduce a strong cancer-inducing gene in the liver; half were used as controls and the other half were further engineered to have moderately elevated levels of let-7 expression . Most of the control mice got large cancerous tumors , but only a few mice in the other group developed cancers and the tumors were smaller . This confirmed that let-7 hinders tumor formation . Wu and Nguyen et al . also observed that the protected mice were less able to regenerate their liver tissues . Further experiments showed that deleting just two out of ten let-7 microRNAs enhanced the mice’s ability to regenerate liver tissue after injury . These findings indicate that let-7 microRNAs slow down the growth of both cancerous and normal cells . Lastly , when let-7 levels were raised to very high levels for a prolonged amount of time this actually led to liver damage and subsequent tumor formation . This last observation may have important consequences for possible cancer therapies . Some scientists have shown that providing extra let-7 can slow or even reverse tumour growth , but the findings here clearly point out that too much let-7 could actually worsen the situation . Since the let-7 family comprises a handful of microRNAs in mammals , in the future it will also be important to find out to what extent these molecules play overlapping roles and how much they differ . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cell",
"biology"
] | 2015 | Precise let-7 expression levels balance organ regeneration against tumor suppression |
Kinesin-1 is a dimeric motor that transports cargo along microtubules , taking 8 . 2-nm steps in a hand-over-hand fashion . The ATP hydrolysis cycles of its two heads are maintained out of phase by a series of gating mechanisms , which lead to processive runs averaging ∼1 μm . A key structural element for inter-head coordination is the neck linker ( NL ) , which connects the heads to the stalk . To examine the role of the NL in regulating stepping , we investigated NL mutants of various lengths using single-molecule optical trapping and bulk fluorescence approaches in the context of a general framework for gating . Our results show that , although inter-head tension enhances motor velocity , it is crucial neither for inter-head coordination nor for rapid rear-head release . Furthermore , cysteine-light mutants do not produce wild-type motility under load . We conclude that kinesin-1 is primarily front-head gated , and that NL length is tuned to enhance unidirectional processivity and velocity .
Kinesin-1 , hereafter referred to simply as kinesin , is an ATP-driven , dimeric motor protein that facilitates the unidirectional transport of intracellular cargo along microtubule ( MT ) filaments ( Vale et al . , 1985; Howard et al . , 1989; Block et al . , 1990; Hackney , 1995 ) . Each kinesin dimer is composed of a pair of identical catalytic motor domains , or heads , which are connected to a common , coiled-coil stalk by a ∼14-amino-acid-long sequence known as the neck linker ( NL ) ( Kozielski et al . , 1997 ) . Kinesin translocates towards the plus-ends of MTs ( Svoboda et al . , 1993 ) via an asymmetric hand-over-hand mechanism ( Asbury et al . , 2003; Yildiz et al . , 2004 ) , hydrolyzing one molecule of ATP ( Hua et al . , 1997; Schnitzer and Block , 1997; Coy et al . , 1999 ) for each 8 . 2-nm step ( Svoboda et al . , 1993 ) the motor takes . The biochemical states associated with each kinesin head during stepping are coupled to mechanical transitions in an overall mechanochemical cycle: biochemical events modulate the affinities of heads to the MT and influence mechanical transitions , and mechanical states , in turn , influence the rates of biochemical processes ( Block , 2007 ) . During the mechanochemical cycle , each kinesin head transitions between one or more states that are strongly bound to the MT ( the ATP-containing state , and also the no-nucleotide state ) , and one or more states that are weakly bound to the MT ( the ADP-containing state ) ( Block , 2007 ) . A simplified version of this cycle ( Figure 1A ) may arbitrarily be taken to begin with the one-head-bound ( 1-HB ) , ATP-waiting state [α] , where the nucleotide-free front head is strongly bound to the MT while the rear , ADP-bound tethered head remains unbound ( Hackney , 1994; Asenjo and Sosa , 2009; Guydosh and Block , 2009; Toprak et al . , 2009 ) . Following ATP binding to the MT-bound head , the NL of its motor domain undergoes a structural reconfiguration and forms a β-sheet with the head , in a process termed NL docking ( Rice et al . , 1999; Schnitzer et al . , 2000; Rosenfeld et al . , 2001; Asenjo et al . , 2006; Tomishige et al . , 2006; Khalil et al . , 2008; Sindelar and Downing , 2010; Clancy et al . , 2011 ) . NL docking shifts the position of the tethered head towards the MT plus-end , beyond the position of the bound head [β1 , β2] . Recent work has suggested that NL docking may occur in two stages: first , ATP binding induces a load-dependent mechanical transition , leading to partial NL docking [β1] , whereas subsequent ATP hydrolysis completes the docking and enables the tethered head to bind the MT [β2] ( Milic et al . , 2014 ) . Once the bound head hydrolyzes ATP , fully docks its NL , and the tethered head binds the MT , the motor enters a mechanically strained , two-heads-bound ( 2-HB ) state [γ] ( Rice et al . , 1999; Rosenfeld et al . , 2003; Block , 2007; Yildiz et al . , 2008; Gennerich and Vale , 2009; Clancy et al . , 2011 ) . Finally , rear-head release ( Klumpp et al . , 2004 ) completes the step and returns the motor to its initial 1-HB waiting state [α] , primed to begin the cycle anew , after having translocated forward by one step ( 8 . 2 nm ) along the MT . The processivity of kinesin—evinced by the ability of a single dimer to undergo >100 consecutive stepping cycles before dissociation—relies upon a tight coordination of the biochemical and mechanical events , collectively known as gating mechanisms ( Block , 2007 ) . 10 . 7554/eLife . 07403 . 003Figure 1 . A general gating framework based on mechanical states of dimeric motors . ( A ) The kinesin mechanochemical cycle . Kinesin starts from the one-head-bound ( 1-HB ) ATP-waiting state [α] , characterized by a strongly bound , nucleotide-free ( Ø ) front head ( red ) and an unbound , ADP-containing tethered head ( blue ) . ATP binding induces a force-dependent transition involving partial NL docking , shifting the tethered head past the bound head [β1] . ATP hydrolysis completes NL docking and facilitates tethered-head binding [β2] . At this point , kinesin may access a dissociated state [Off] , induced by premature phosphate release from the bound head , leading to dimer detachment . However , if the tethered head reaches the forward MT binding site and completes the step before the bound head can dissociate , kinesin enters the two-heads-bound ( 2-HB ) state [γ] . Rear-head release returns the dimer to the ATP-waiting state [α] , having moved forward by 8 . 2 nm . ( B ) A simplified general gating framework , based on the cycle in ( A ) . Stepping , binding , and unbinding gates are shown with associated rate constants between each of the three gated states , [A] , [B] , and [C] . The cycle begins at the 1-HB ATP-waiting state [A] , where the stepping gate promotes processivity by inhibiting rear-head ( blue ) rebinding and premature bound-head ( red ) release . Following a force-dependent step that shifts the tethered head past the bound head [B] , the binding gate promotes binding of the tethered head at the forward MT binding site while inhibiting release of the bound head . Also shown is a competing dissociated state [Off] , arising from premature release of the bound head from the 1-HB state , accessible from either [A] or [B] . Tethered-head binding leads to the 2-HB state [C] , where the unbinding gate promotes rear-head release while inhibiting front-head release , returning the motor to the start of the cycle [A] . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 003 We note that the term ‘gating’ encompasses any mechanism where the state of one head influences its partner in such a way as to ensure that the mechanochemical cycles of the heads are maintained out of phase , leading to alternate-head stepping . This term includes , but is not limited to , mechanisms that modulate detachment rates , alter nucleotide affinities , and affect ATP hydrolysis ( Block et al . , 1990; Hackney , 1994; Vale et al . , 1996; Hua et al . , 1997; Schnitzer and Block , 1997; Hancock and Howard , 1999; Rosenfeld et al . , 2003; Block , 2007 ) . To establish a general framework for gating in the kinesin cycle , we begin by recognizing that the kinesin dimer transitions through a fixed series of 1-HB [A , B] and 2-HB [C] states during each cycle , where the unbound ( tethered ) head of a 1-HB motor may be positioned either predominantly behind [A] or in front of [B] , the MT-bound head ( Figure 1B ) . Since dissociation from the 2-HB state [Off] necessarily requires passage through a 1-HB intermediate , processive cycling can be distilled into the following set of three gating properties which , taken all together , promote forward stepping while suppressing dissociation:i . The MT-bound head must remain attached in any 1-HB state . ii . When the motor is in either a 2-HB state or in a 1-HB state where the tethered head is behind the MT-bound head , the unbinding of the rear head should be promoted ( or maintained , if already unbound ) . iii . When the motor is in either a 2-HB state or in a 1-HB state where the tethered head is in front of the MT-bound head , the binding of the front head should be promoted ( or maintained , if already bound ) . Taken together , these principles lead to the mechanical gating framework presented in Figure 1B . Starting from a 1-HB state , with the tethered head positioned behind the MT-bound head [A] , unidirectional processivity necessitates a stepping gate that stabilizes binding by the front head while inhibiting any rebinding by the rear head . Following a structural transition that shifts the tethered head beyond the bound head [B] , a binding gate acts to promote tethered-head binding at the forward site , while preventing premature unbinding of the rear head , which would otherwise lead to dissociation [Off] . Once the tethered head binds successfully , and the motor achieves the 2-HB state [C] , an unbinding gate is required to retain the front head on the MT , while promoting the release of the rear head . Finally , rear-head release completes the cycle , returning the motor to its initial state [A] , but advanced by one step . We note that this abstraction is agnostic with respect to the biochemical state associated with each mechanical transition . However , precisely because these biochemical states are unspecified , this general framework ( Figure 1B ) can be applied not only to kinesin but also to other processive , two-headed motors that may couple mechanical and biochemical states differently ( Block , 2007; Gennerich and Vale , 2009; Kull and Endow , 2013; Cleary et al . , 2014 ) . Here , we used the gating framework to guide an investigation of the role of the NL domain in determining kinesin processivity , which remains incompletely understood , despite considerable research ( Hackney et al . , 2003; Block , 2007; Yildiz et al . , 2008; Shastry and Hancock , 2010 , 2011; Clancy et al . , 2011 ) . Because intramolecular forces within the kinesin dimer are thought to be transmitted through the NL of each head , the NL domain is well situated to play a role in gating . Consistent with this , extending the NL by mutation has previously been shown to affect kinesin processivity , velocity , and stepping behavior ( Hackney et al . , 2003; Block , 2007; Yildiz et al . , 2008; Shastry and Hancock , 2010 , 2011; Clancy et al . , 2011 ) . In principle , perturbing the properties of the NL by inserting additional amino acids ( AA ) at the junction of the NL domain and the coiled-coil stalk could affect some , or all , of the gates in Figure 1B . When kinesin is in its 1-HB state [A , B] , the NL might serve to suppress rear-head rebinding in the ATP-waiting state [A] , as part of the stepping gate , or to promote tethered-head binding ( following NL docking ) in the bound head [B] , as part of the binding gate . When kinesin is in a 2-HB state [C] , the NL can transmit inter-head tension , affecting the unbinding gate ( Rosenfeld et al . , 2003; Guydosh and Block , 2006; Block , 2007; Yildiz et al . , 2008; Hariharan and Hancock , 2009; Shastry and Hancock , 2010 , 2011; Clancy et al . , 2011 ) . The unbinding gate may act through ( i ) front-head gating , where biochemical events on the front head are suppressed until the rear head has a chance to detach ( candidate mechanisms include the suppression of ATP hydrolysis in the front head , reduced ATP binding arising from inter-head tension , or reduced ATP binding caused by the rearward-pointing configuration of the NL ) , or through ( ii ) rear-head gating , where the rear-head release rate is accelerated by the inter-head tension , or ( iii ) some combination of the two ( Block , 2007 ) . Because inter-head tension is directly controlled by the length of the NL domain , an understanding of the relationship between NL length and the unbinding gate is central to evaluating tension-based gating models . The predominant mechanism for gating used by kinesin has been the subject of some controversy . Published models have invoked both versions of rear-head gating ( Hancock and Howard , 1999; Crevel et al . , 2004; Schief et al . , 2004; Yildiz et al . , 2008 ) and front-head gating ( Rosenfeld et al . , 2003; Klumpp et al . , 2004; Guydosh and Block , 2006 , 2009; Toprak et al . , 2009; Clancy et al . , 2011 ) . The conversation about kinesin gating has heretofore focused on how inter-head coordination might be achieved from the 2-HB state , that is , from the unbinding gate . Comparatively little attention has been paid to gating at other points of the kinesin cycle , specifically at the stepping and binding gates ( Figure 1B ) , and quantitative measures of the competing rates responsible for gating have been notably lacking . Here , we examine the quantitative contribution of each of the three gates to maintaining unidirectional processivity by assessing the effects of NL length on kinesin motility , and attempt to reframe the discussion of gating from a debate about front- vs rear-head gating at the 2-HB state to a more unified view that admits to gating at multiple states of the mechanochemical cycle , both 1-HB and 2-HB . Towards that end , we examined a series of truncated Drosophila kinesin constructs ( DmK ) containing NLs that were extended incrementally , from 1 to 6 AA ( DmK-1AA to DmK-6AA ) , using a combination of single-molecule optical trapping and bulk fluorescence approaches .
Our first experiments were designed to probe the influence of NL length on the stepping gate , which might contribute to unidirectional processivity by suppressing the rebinding of the ADP-bound tethered head while the motor is in the ATP-waiting state [A] ( Figure 1B ) . To investigate the influence of NL length on this rebinding , we performed a series of half-site ADP release experiments using mantADP , a fluorescent ADP analog , on a set of truncated constructs with extended NL domains . Because free kinesin heads in solution have a high affinity for ADP , whereas MT-bound heads exhibit a substantially lower affinity , the release of mantADP from kinesin heads serves as a proxy for the binding of free heads to MTs ( Hackney , 1988 , 1994 , 2002; Gilbert et al . , 1995; Hackney et al . , 2003; Clancy et al . , 2011 ) . Binding was assayed by pre-incubating motors in mantADP and subsequently monitoring the MT-stimulated decay of fluorescence upon the introduction of MTs via stopped-flow . Wild-type ( WT ) Drosophila kinesin ( DmK-WT ) , as well as constructs with NLs extended by up to 3 AA , displayed a 50% reduction in fluorescence following the addition of MTs ( Figure 2A ) , consistent with being in a 1-HB state , where only one of the two heads is bound to the MT and releases mantADP . By contrast , constructs with NL inserts consisting of 4 , 5 , or 6 AA exhibited progressively greater drops in fluorescence upon the addition of MTs ( Figure 2A ) . These decreases correspond to 16% of DmK-4AA , 38% of DmK-5AA , and 65% of DmK-6AA motors releasing mantADP from both heads , respectively . In principle , the increased mantADP release could arise from ( i ) the additional NL length facilitating transient rear-head rebinding events , ( ii ) more motors adopting a stable 2-HB ATP-waiting state , or ( iii ) some combination of both . These findings are consistent with previous work , which has also reported a ∼50% fluorescence decrease for WT kinesin , and a nearly 100% fluorescence decrease for constructs carrying a 6-AA NL insert ( Hackney , 1994 , 2002; Hackney et al . , 2003; Clancy et al . , 2011 ) . Additional mantADP exchange experiments—where the frequency of transient rear-head interactions with the MT was assessed based on the effective rate of mantADP exchange at the rear head—revealed an order-of-magnitude increase relative to WT kinesin once the NL domain was extended by 4 or more AA ( Figure 2B ) . The low mantADP exchange rates observed for our series of constructs are consistent with previous reports ( Hackney , 2002; Hackney et al . , 2003 ) . Taken all together , these results indicate that kinesin can maintain a stable , 1-HB ATP-waiting state with up to 3 additional AA in its NL , but that extending the NL by 4 AA or more leads to an abrupt increase in rebinding of the rear-head to the MT . 10 . 7554/eLife . 07403 . 004Figure 2 . Kinesin with as many as 3 AA inserted in the NL maintains a 1-HB ATP-waiting state . ( A ) Half-site mantADP release measurements as a function of NL insert length ( mean ± SE; N = 3 ) . Upon MT binding , both DmK-WT and DmK-3AA , pre-incubated with mantADP ( mantD ) , lose ∼50% of their initial fluorescence . The fluorescence loss exceeds 50% for constructs containing NL inserts longer than 3 AA . Inset , A 50% loss of fluorescence corresponds to dimers binding to MTs in a 1-HB state , whereas a 100% fluorescence decrease is consistent with the release of all bound mantADP ( mantD ) upon MT binding . ( B ) MantADP exchange by the tethered head as a function of NL insert length ( mean ± SE; N = 3 ) , measured by rapidly diluting mantADP·kinesin·MT complexes into nucleotide-free buffer via stopped flow . The cartoon depicts the measured reaction . The exchange rate increased significantly for constructs with NL inserts of 4 AA or more . In the insets ( A and B ) , white shading indicates non-fluorescent , nucleotide-free heads ( Ø ) ; yellow indicates fluorescent , mantADP-bound heads . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 004 After kinesin undertakes its force-producing mechanical step [B] , the binding gate may promote front-head binding to the forward MT binding site while keeping the rear head bound to the MT ( Figure 1B ) . The overall distance kinesin travels along the MT before dissociating—the run length—is governed by the probability of dissociation during each stepping cycle , which is itself determined , to a first approximation , by a competition between the binding of the tethered front head and the premature release of the MT-bound rear head ( Milic et al . , 2014 ) . To determine whether the rate of front-head binding at the binding gate depends upon NL length , we performed run-length measurements as a function of load for constructs with extended NLs . Extending the NL by just a single AA decreased the unloaded run length by a factor of ∼3 relative to the WT run length . Further increments in the NL length ( up to 6 AA ) yielded only minor decreases in run length relative to DmK-1AA ( Figure 3 ) . Under unloaded conditions , DmK-6AA was capable of taking several dozen consecutive forward steps , on average , before dissociating , indicating that kinesin retains significant processivity with as many as 6 AA introduced into its NL . In the presence of forces applied either against ( hindering loads ) or along ( assisting loads ) the direction of kinesin motility ( Figure 3 , inset ) , the differences in processivity among constructs with differing NL lengths were substantially more pronounced under hindering forces than under assisting forces . Although the unloaded run lengths for DmK-WT and DmK-1AA differed by a factor of 3 , the differences in run length between these constructs under assisting loads were comparatively minor . A gradual , exponential decrease in run length was observed for all constructs under increasing hindering loads , but the application of even a small assisting force ( +2 pN ) abruptly reduced the run length for both the WT and NL-mutant constructs . Based on exponential fits to the mean run lengths as a function of applied load , the characteristic distance parameter for the force dependence of run length , δL , was nearly an order of magnitude greater under hindering-load conditions than under assisting-load conditions for all constructs ( Table 1 ) . These results show that the highly asymmetric force dependence of DmK run lengths previously reported for the WT motor ( Milic et al . , 2014 ) is also exhibited by kinesin constructs with NLs extended up to 6 AA . Interestingly , although processivity was found to depend upon the NL length , particularly under hindering-load and unloaded conditions , the distance parameters for both hindering and assisting loads were nearly independent of NL length ( Table 1 ) . 10 . 7554/eLife . 07403 . 005Figure 3 . Extending the NL by a single AA compromises processivity . Mean run lengths as a function of applied force ( mean ± SE; N = 49–818 ) for the constructs studied , acquired with an optical force clamp in the presence of 2 mM ATP ( solid circles; color-coded according to the legend ) . DmK-WT ( Milic et al . , 2014 ) and hindering load data sets for DmK-3AA ( Andreasson et al . , 2015 ) are reproduced from our previous work . For all constructs , mean run lengths exhibited significant asymmetry , depending upon the direction of load . To obtain the unloaded run length ( L0 ) and characteristic distance parameter ( δL ) for each construct , run length ( L ) data for hindering ( −6 to 0 pN ) and assisting loads ( +2 to +20 pN ) were separately fit to exponentials ( solid lines; color-coded according to the legend ) of the form L ( F ) =L0exp[−|F|δL/kBT] , where F is the force applied by the optical trap and kBT is Boltzmann's constant times the absolute temperature; parameter values are in Table 1 . Inset cartoon , a graphical representation of the experimental geometry of the single-molecule assay ( not to scale ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 00510 . 7554/eLife . 07403 . 006Table 1 . Parameters from exponential fits to the run length data of Figure 3DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 006ConstructL0− ( nm ) *δL− ( nm ) *L0+ ( nm ) *δL+ ( nm ) *DmK-WT†1120 ± 602 . 0 ± 0 . 187 ± 60 . 27 ± 0 . 03DmK-1AA360 ± 301 . 6 ± 0 . 1120 ± 70 . 48 ± 0 . 02DmK-2AA410 ± 601 . 8 ± 0 . 2115 ± 170 . 42 ± 0 . 05DmK-3AA320 ± 201 . 6 ± 0 . 176 ± 40 . 31 ± 0 . 02DmK-5AA440 ± 602 . 4 ± 0 . 248 ± 140 . 35 ± 0 . 14DmK-6AA270 ± 301 . 9 ± 0 . 159 ± 80 . 27 ± 0 . 10L0− , unloaded run length for hindering loads ( −6 to 0 pN ) ; δL− , distance parameter for hindering loads ( −6 to 0 pN ) ; L0+ , unloaded run length for assisting loads ( +2 to +20 pN ) ; δL+ , distance parameter for assisting loads ( +2 to +20 pN ) . *Parameter values correspond to mean ± SE . †Values from Milic et al . ( 2014 ) . Previous work has shown that under assisting loads , added phosphate ( Pi ) nearly doubles the run length of DmK-WT ( Figure 4A ) , indicating that the probability of dissociation at the binding gate—and therefore run length—is determined by a competition between the rate of Pi release from the bound head and the rate of productive tethered-head binding ( Milic et al . , 2014 ) . To determine the extent to which the NL length affects this competition , we examined the influence of 100 mM potassium phosphate on the processivity of DmK-6AA , under moderate assisting loads ( +4 pN ) . Under saturating levels of ATP , the addition of Pi increased the DmK-6AA run length relative to that in its absence ( Figure 4A ) . However , no statistically significant changes in processivity were detected at low ATP concentrations , nor under conditions where ATP was replaced with the slowly hydrolyzable analog , ATPγS , nor in the presence of 100 mM potassium chloride ( used as a control for ionic strength effects ) ( Figure 4 ) . Although run lengths under otherwise identical assay conditions were systematically lower for DmK-6AA than for DmK-WT , those differences disappeared when the data were normalized relative to the baseline run lengths ( no added Pi ) for each motor ( Figure 4B ) . 10 . 7554/eLife . 07403 . 007Figure 4 . Added phosphate enhances the processivity of DmK-WT and DmK-6AA . ( A ) Mean run lengths ( mean ± SE; N = 84–210 ) under moderate assisting load ( +4 pN ) , in the presence of nucleotide analogs ( green ) , 100 mM potassium chloride ( KCl; purple ) , or 100 mM potassium phosphate ( KPi; orange ) . Run lengths for DmK-WT ( shaded bars , data from Milic et al . , 2014 ) are shown paired with DmK-6AA data ( unshaded bars ) . Decreasing the ATP concentration , replacing ATP by ATPγS , or adding KCl elicited no statistically significant change in run length relative to the baseline run length for saturating ATP ( 2 mM ) in the absence of added salt . The mean run length increased significantly in the presence of phosphate for both DmK-WT ( p < 10−4; t-test ) and DmK-6AA ( p < 10−4; t-test ) . ( B ) Run-length data from ( A ) , normalized to the baseline run length value for each construct . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 007 Once the tethered head binds the MT to generate the 2-HB state [C] , the unbinding gate may ensure unidirectional processivity by inhibiting front-head unbinding while promoting rear-head release ( Figure 1B ) . As discussed previously , the unbinding gate might consist of a rear-head gating mechanism , where the inter-head tension enhances the rate of rear-head release . If detachment of the trailing head is accelerated by inter-head tension , then extending the NL would be expected to relieve the tension and decrease the rear-head release rate , thereby reducing the overall rate at which kinesin proceeds around the reaction cycle . To explore the effect of inter-head tension on rear-head gating , we performed single-molecule measurements of the velocity of our kinesin constructs as a function of load . Both WT and mutant constructs exhibited a force–velocity relationship that is highly asymmetric with respect to the direction of the applied load ( Figure 5 ) , consistent with previous findings ( Block et al . , 2003; Block , 2007 ) . Whereas the insertion of a single AA in the NL elicited a drop in velocity across all loads , additional NL insertions produced no further changes in velocity: indeed , Drosophila kinesin stepped at significant rates with as many as 6 additional AA in its NL . In contrast to DmK-WT , which is not sped up by assisting loads , the reduction in unloaded velocity produced by NL extension could be recovered by applying larger assisting loads ( ∼20 pN ) . These data are consistent with the explanation that inter-head tension , which can promote rear-head release , is effectively abolished by extending the NL past its WT length , but that the reduced tension can be restored by applying assisting load , which places differential stress on the rear head . However , the finding that assisting load fails to appreciably increase the WT velocity suggests that the rate of rear-head release by WT kinesin must be substantially higher than the rate-limiting step ( s ) of the mechanochemical cycle . If the rate of rear-head release were instead rate-limiting , then any acceleration of that step would have manifested as an increase in motor velocity . 10 . 7554/eLife . 07403 . 008Figure 5 . Assisting load can rescue the velocity of mutants with extended NLs . Velocity ( mean ± SE; N = 49‒818 ) as a function of force for constructs ( solid circles; color-coded according to the legend ) . Data were collected under the same conditions as Figure 3 . DmK-WT velocity was not affected by assisting loads , but the velocities of all mutant constructs could be increased by larger assisting loads . Solid lines show the global fit to a minimal 3-state model ( inset ) for WT ( blue ) and mutant ( red ) constructs , with parameters in Table 2 . Data sets for both DmK-WT and DmK-3AA under hindering loads are from ( Andreasson et al . , 2015 ) . Force–velocity data are compared to other mutant constructs in Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 008 To gain a quantitative understanding of the velocities of mutant constructs , and to gain additional insight into how inter-head tension affects kinesin gating , the force–velocity data ( Figure 5 ) were fit to a minimal , 3-state model of the mechanochemical cycle ( Figure 5 , inset ) . The first transition in this cycle consists of the force-producing mechanical step , [1] → [2] , which is modeled by a force-dependent rate , k1=k10exp[Ftrapδ1/kBT] where k10 is the unloaded rate constant , Ftrap is the applied force , δ1 is a characteristic distance parameter , and kBT is Boltzmann's constant times the absolute temperature . Because kinesin is in a 1-HB ATP-waiting state [1] prior to this transition , the inter-head tension , Fi , has no effect on k1 . The following transition , [2] → [3] , consists of ATP hydrolysis and related events—including the completion of NL docking by the MT-bound head , front-head binding to the MT , and ADP release by the new front head—that induce kinesin to enter the strained , 2-HB state [3] . Because this transition corresponds to the completion of the key biochemical steps in the cycle , the associated rate constant , k2 , is modeled as being independent of load . From the 2-HB state [3] , the final transition of the cycle consists of rear-head release ( which is also load-dependent ) and is modeled as k3=k30exp[ ( Ftrap+Fi ) δ3/kBT] , where k30 is the unloaded rate constant and δ3 is the associated distance parameter . The expression for k3 is a function of both Ftrap and Fi , and accounts for the roles of both internal and external tension in enhancing the dissociation of the rear head . A previously described analytical method was adopted to derive the expression for velocity as a function of force based on this model ( Chemla et al . , 2008 ) . This model represents a minimal extension of previously published models ( Schnitzer et al . , 2000; Block et al . , 2003 ) . The 3-state model was fit globally to our data ( Figure 5 ) , and the resulting parameters and uncertainties are given in Table 2 . Because the NL-insert mutants exhibited indistinguishable force-dependent velocities , we took the inter-head tension in these constructs to be negligible compared to the WT ( Fi , mutant ≈ 0 pN ) . With this assumption , we determined the inter-head tension in WT kinesin ( Fi , WT ) to be 26 ± 3 pN in the 2-HB state , a value that agrees well with molecular dynamics simulations , which predicted that it would lie in the range of 15–35 pN ( Hariharan and Hancock , 2009 ) . The model fit also supplies an estimate of the unloaded rate of rear-head release , 260 ± 10 s−1 . With the derived values for Fi , WT and the load-dependence of rear-head unbinding , δ3 , the expression for k3 shows that inter-head tension alone is capable of increasing the rate of rear-head release by an order of magnitude ( Table 2 ) . Moreover , the modeled rate of ATP hydrolysis ( plus the other primarily biochemical events , above ) , 95 ± 1 s−1 , is consistent with previous estimates of the hydrolysis rate from both biochemical and single-molecule studies ( Ma and Taylor , 1997; Gilbert et al . , 1998; Schnitzer et al . , 2000; Farrell et al . , 2002 ) . Under all practical loads , hydrolysis is substantially slower than rear-head release . 10 . 7554/eLife . 07403 . 009Table 2 . Kinetic parameters from a global fit of the 3-state model to the force–velocity data of Figure 5DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 009ParameterParameter descriptionValue*k10 ( F ) Rate of ATP binding†; mechanical step4900 ± 300 s−1δ1 , WTDistance parameter ( wild-type ) 4 . 6 ± 0 . 1 nmδ1 , mutantDistance parameter ( mutant constructs ) 4 . 0 ± 0 . 1 nmk2Rate of ATP hydrolysis; biochemical events95 ± 1 s−1k30 ( F ) Rate of rear-head release260 ± 10 s−1δ3Distance parameter ( rear-head release ) 0 . 35 ± 0 . 02 nmFi , WTInter-head tension ( wild-type ) 26 ± 3 pNFi , mutantInter-head tension ( mutant constructs ) 0 pN ( fixed ) *Parameter values correspond to mean ± SE . †Rate for saturating ATP conditions ( 2 mM ) . We note that any force-dependent transition in the kinesin cycle must be associated with a corresponding movement of a head , or other subdomain , of the motor . Previous work has shown that the main force-dependent transition occurs from a 1-HB state after ATP binding ( Schnitzer et al . , 2000; Block et al . , 2003 ) . The velocity decrease under hindering loads is associated with this transition , which we model as k1 , with a correspondingly large distance parameter ( δ1 ≥ 4 nm; Table 2 ) , equivalent to nearly half the size of the kinesin step . This transition becomes rate limiting at moderate hindering loads , and will therefore dominate any other possible force-dependent steps at high loads . Under assisting loads , k1 is exceedingly fast ( Table 2 ) and does not appreciably contribute to the completion time of the overall cycle—and therefore to the velocity—in this regime . However , for the mutant constructs , we are able to detect a second force-dependent contribution that is manifested as increased velocity under assisting loads ( Figure 5 ) . This force dependence , δ3 , is necessarily linked to some motion of the motor , and the load dependence of the associated rate , k3 , could be attributed either to rear head release when kinesin is in the 2-HB state or to binding of the forward-positioned tethered head of a 1-HB motor . We strongly favor the former possibility , because it also helps to explain the observed decrease in run length under assisting load , as well as the force dependence of exponential fits to the run length data ( Figure 3 and Table 1 ) . Put another way , if load-dependent rate k3 was instead to correspond to tethered-head binding , then an increase in k3 under assisting load would increase the likelihood of completing a step , and thereby increase run lengths in this force regime , contrary to observation ( Figure 3 ) . With these assignments , k2 remains the sole force-independent rate , corresponding primarily to biochemical , and not mechanical , events . In the 2-HB state [C] , inter-head tension exerts not only a forward load on the rear head , but also imposes a matching rearward load on the front head ( Figure 1B ) . Having determined values for the inter-head tension and the force-dependent rate of rear-head release ( Table 2 ) , it is also necessary to quantify the force-dependent rate of front-head release in order to model how inter-head tension affects the unbinding gate . The application of hindering loads to the kinesin stalk places a rearward load on its front head , and therefore can serve as a proxy for estimating the rate of detachment of the front head at the unbinding gate . Previously , we showed that large superstall forces imposed by an optical trap ( exceeding −7 pN ) slow down the forward stepping pathway to such an extent that unbinding proceeds through a competing ( slow ) ATP-dependent pathway ( Clancy et al . , 2011 ) . To quantify the rate of front-head detachment , we examined the unbinding of DmK-WT from the MT under load in the presence of saturating levels of ATP ( 2 mM ) . Measurements of unbinding rates under load ( Figure 6 ) were fit to the exponential function , koff=koff0 exp[|Ftrap|δoff/kBT] where koff0 is the unloaded rate and δoff is the associated distance parameter . The unbinding rate for all hindering loads ( koff− ) is determined by koff−0=1 . 11±0 . 03 s−1 and δoff−=0 . 60±0 . 01 nm ( fitting the unbinding rates only to data from superstall forces yielded nearly identical parameter values ) . From these numbers , it follows that the 26-pN inter-head tension in WT kinesin can enhance the rate of front-head release at the unbinding gate by a factor of ∼50 . 10 . 7554/eLife . 07403 . 010Figure 6 . Kinesin unbinding rates are asymmetric with respect to the direction of load . Single-molecule measurements of the rate of MT unbinding for DmK-WT ( mean ± SE; N = 75‒818 ) at 2 mM ATP ( solid circles ) . The unloaded release rate ( koff ) and the associated distance parameter ( δoff ) were obtained from exponential fits to unbinding data acquired under hindering loads ( − ) , −25 to 0 pN , and assisting loads ( + ) , +2 to +20 pN . Fits ( solid lines ) and associated parameters ( legend; mean ± SE ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 010 For assisting loads , we can estimate the unbinding characteristics of kinesin dimers from the MT ( Figure 6 ) by dividing the force-dependent velocities ( Figure 5 ) by their corresponding run lengths ( Figure 3 ) , yielding koff+0=7 . 4±0 . 5 s−1 and δoff+=0 . 32±0 . 02 nm . Taken together , the values of koff0 and δoff for assisting and hindering loads , which are also useful for modeling transport by multiple kinesin motors , express the asymmetry in kinesin velocity and run length with respect to the direction of load . For assisting loads , the distance parameter , δoff+ , is controlled by the run length force dependence , and we note that its value is similar to the distance parameter derived from fits to the velocity data , δ3 , suggesting that both processes are governed by the detachment of the rear head . The force–velocity relation for DmK-6AA ( Figure 5 ) , which carries the insert LQASQT in its NL , differs significantly from the corresponding relation for a human ( HsK ) cysteine-light ( CL ) construct ( Rice et al . , 1999 ) with the insert AEQKLT , HsK-CL-6AA ( Clancy et al . , 2011 ) ( Figure 7 ) . Specifically , DmK-6AA exhibits a substantially greater velocity—which is also less force dependent—than HsK-CL-6AA ( Figure 7A ) . Whereas HsK-CL-6AA was capable of undertaking many rearward steps when subjected to hindering loads beyond the stall force ( Clancy et al . , 2011 ) , no such processive backstepping was observed for DmK-6AA , nor for any other Drosophila constructs examined that retained native cysteine residues . This finding is also consistent with the proportionally greater mantADP release by HsK-CL-6AA ( Clancy et al . , 2011 ) than DmK-6AA ( Figure 2A ) . By contrast , the wild-type forms of Drosophila ( DmK-WT ) and human ( HsK-WT ) kinesin exhibit nearly identical force–velocity relations under otherwise identical conditions ( Figure 7A ) . The question therefore arises whether the discrepancy between the behavior of DmK-6AA and HsK-CL-6AA is a consequence of: ( i ) one or more of the seven mutations used to produce the CL construct , ( ii ) some intrinsic difference between human and Drosophila kinesin , ( iii ) a sequence-specific effect of the NL insert , or ( iv ) some combination of these . 10 . 7554/eLife . 07403 . 011Figure 7 . Kinesin motility characteristics depend upon parent species , NL length , NL sequence , and cysteine mutations . Force–velocity relations of constructs that differ by parent species , NL length , NL insert sequence , or cysteine mutations ( mean ± SE; N = 25–818; solid circles , color-coded according to the legends ) . ( A ) DmK-6AA exhibited lower velocity under all loads than WT , but was faster and less force-dependent than HsK-CL-6AA , a human CL construct with a different 6-AA NL insert . ( B ) Side-by-side comparison of human and Drosophila CL mutants , along with corresponding WT constructs . Unloaded velocities of CL constructs with WT-length NL domains were similar , but CL constructs were systematically slower than WT under hindering loads . Under assisting loads , only the HsK-CL construct could be sped up beyond WT velocities . ( C and D ) Comparisons of constructs carrying NL insert sequences LQASQT ( C ) and AEQKLT ( D ) . Force–velocity data for WT Drosophila and human kinesin were indistinguishable , but human constructs with 6-AA extensions of the NL moved at lower velocities than corresponding Drosophila motors under all loads . ( E and F ) Comparisons of all Drosophila ( E ) and human ( F ) constructs . Constructs with the NL insert AEQKLT were slower for all forces than those with LQASQT , for both Drosophila and human kinesin . HsK-CL-6AA data ( Clancy et al . , 2011 ) and hindering-load velocities for DmK-WT ( Andreasson et al . , 2015 ) are reproduced from previous work . Fits to DmK-WT and DmK-6AA data sets ( solid lines; color-coded according to the legends ) correspond to the model of Figure 5 , with parameters values from Table 2 . The remaining force–velocity data were fit to polynomials ( solid and dashed lines; color-coded according to legends ) , provided to guide the eye . DOI: http://dx . doi . org/10 . 7554/eLife . 07403 . 011 Recombinant mutant constructs have been widely used to investigate kinesin function . In particular , CL constructs have been developed to attach fluorescent labels , and it has been commonly assumed that such motors exhibit WT behavior , based on the similarity of their velocities measured under unloaded conditions ( Rice et al . , 1999 ) . Here , we assayed the velocity of a widely used truncated human CL construct ( HsK-CL ) ( Rice et al . , 1999 ) and a Drosophila CL construct ( DmK-CL ) ( Fehr et al . , 2009 ) over the full range of forces ( Figure 7B ) . Under hindering loads , both HsK-CL and DmK-CL were appreciably slower , and also more sensitive to load , than either DmK-WT or the corresponding truncated human construct , HsK-WT . For the case of the human CL construct , the entire shape of the force–velocity relation is affected , because this motor can be sped up under assisting loads by as much as 200 nm/s . WT motors do not exhibit this effect , and it suggests that important steps in the mechanochemical cycle have been altered in the mutant . Both CL constructs carry , among other changes , a point mutation of a conserved cysteine in the NL domain , namely C338S in DmK-CL ( Fehr et al . , 2009 ) and C330S in HsK-CL ( Rice et al . , 1999 ) . We determined the force–velocity curve for a Drosophila construct containing only the C338S mutation ( DmK-C338S ) . Comparison shows that this single mutation can account for some , but not all , of the change in force dependence under hindering loads ( Figure 7B ) . This indicates that a single CL mutation in the NL cannot alone account for the differences between WT and CL motors . Although the force–velocity curves for DmK-WT and HsK-WT are practically indistinguishable , extending the NL of these motors by 6 AA , using either the sequence LQASQT or AEQKLT , showed that with the same insert sequences , human constructs were systematically slower than the corresponding Drosophila constructs ( Figure 7C , D ) . Furthermore , NL extensions based on AEQKLT produced motors that were slower than the corresponding constructs based on LQASQT , for both DmK and HsK ( Figure 7E , F ) . These findings indicate that the sequences of the inserts play some role in determining motor velocity . Nevertheless , regardless of species or insert sequence , none of the mutants with extended NL domains that retained their native cysteine residues exhibited either processive backstepping under superstall loads or velocities surpassing the WT under assisting loads . We note that because DmK-CL requires two mutations ( Fehr et al . , 2009 ) , whereas HsK-CL necessitates seven ( Rice et al . , 1999 ) —only one of which is common between the two constructs—a more direct comparison of Drosophila and human CL constructs with extended NL domains is not practical .
The extent of mantADP release by the rear head while kinesin is in the ATP-waiting state [A] sheds light on the role of NL length in inhibiting rear-head rebinding , an essential feature of the stepping gate . Because WT and mutant constructs with as many as three additional AA in the NL can maintain a stable 1-HB ATP-waiting state , our results show that the stepping gate is not strongly influenced by NL length . Although the stepping gate is clearly compromised in constructs with longer NL inserts ( 4–6 AA ) , a substantial portion of the population retains mantADP in its rear head ( Figure 2A ) , suggesting that gating is not abolished even in these motors . This notion is reinforced by results showing that mantADP exchange rates for the extended-NL mutants ( Figure 2B ) are several orders of magnitude lower than MT-stimulated ADP release rates by WT monomers , or ATP-stimulated ADP release rates by MT-bound WT dimers ( Hackney , 2002 , 2005; Hackney et al . , 2003 ) . The results are fully consistent with the high degree of unidirectionality observed for all constructs , even when subjected to high hindering loads . As reported previously for WT kinesin ( Milic et al . , 2014 ) , the run lengths of constructs with extended NL domains are highly asymmetric with respect to the direction of applied load ( Figure 3 ) . Evidently , the binding gate is substantially more effective at maintaining processivity under unloaded or hindering-force conditions than under assisting-load conditions . Assisting loads as low as 2 pN produced a dramatic reduction in the run length , suggesting that any mechanism responsible for the binding gate becomes ineffective when kinesin is subjected to forces aligned with its overall motion . We note that models of transport by multiple kinesin motors have not yet taken load asymmetry into account ( Klumpp and Lipowsky , 2005; Muller et al . , 2008 ) , but it may be important to do so explicitly , given the magnitude of the effect . Here , we found that lengthening the NL by a single AA significantly decreased the unloaded run length , whereas further NL extensions led to negligible additional reductions ( Figure 3 ) . Furthermore , previous work has shown that a kinesin construct with a NL shortened by a single AA becomes non-processive ( Shastry and Hancock , 2010 , 2011 ) . We conclude that the length of the wild-type NL domain may be near optimal for maximizing motor processivity . We speculate that the decrease in run length associated with extended NLs ( Figure 3 ) arises from the corresponding increase in the diffusive space that the tethered head must explore before reaching its next MT binding site . Enlarging that space extends the time that the tethered head takes to reach the forward site , thereby increasing the probability that the partner head may release the MT prematurely , enhancing dimer dissociation ( Figure 1B ) . This proposal is also consistent with the effect of added Pi in increasing the processivity of DmK-WT and DmK-6AA ( Figure 4 ) : Pi acts to stabilize the bound head , thereby reducing dimer dissociation . The finding that normalized run lengths for the NL constructs were identical under all conditions tested ( Figure 4B ) suggests that the rates of any biochemical events associated with the bound head , when the kinesin cycle is at the binding gate , are independent of the NL length . The absolute run lengths measured for DmK-6AA were systematically lower than those of the WT motor ( Figure 4A ) , as anticipated from the increased time required for the tethered head to reach the forward MT binding site . Extending the NL by a one AA reduced kinesin velocity under all loads , whereas additional NL extensions , up to 6 AA , produced force–velocity curves that were indistinguishable from DmK-1AA ( Figure 5 ) . We surmise that inserting a single AA into the NL introduces sufficient slack into the inter-head linkage that any further extensions become superfluous . A relatively high inter-head tension ( 26 ± 3 pN; Table 2 ) may be advantageous in tuning WT kinesin for increased velocity . However , because constructs with up to 6 AA inserted in the NL retained significant functionality ( Figure 5 ) , we infer that inter-head tension plays only a modulatory role in kinesin motility , a conclusion that gains support from the modeled force–velocity relations ( Figure 5 ) . The fit parameters indicate that WT levels of inter-head tension can amplify the rate of rear-head release by an order of magnitude beyond its unloaded rate . However , because this unloaded rate is already more than double the rate of the hydrolysis-associated events of the kinesin cycle , k2 , the rate-determining step of the full cycle cannot be rear-head release . We find instead that the 2-HB state of kinesin is primarily front-head gated , in the sense that the kinetic cycle of the front head is slowed while the rear head remains MT-bound . Put another way , the gating mechanism responsible for processivity in the 2-HB state does not require inter-head tension , but rather occurs through a modulation of the rates of biochemical processes occurring at the front head , including ATP hydrolysis , and possibly productive ATP binding ( Clancy et al . , 2011 ) . Although an inter-head tension of 26 pN can enhance rear-head release 10-fold , unbinding measurements performed under hindering loads ( Figure 6 ) —which serve as a proxy for front-head detachment rates at the unbinding gate [C]—reveal that such an inter-head tension would be expected to enhance front-head release 50-fold . Despite the front-head release rate being two orders of magnitude lower than the rear-head release rate in the absence of inter-head tension , the fact that the load-sensitivity for detachment by the front head ( captured by the distance parameter , δoff−=0 . 60±0 . 01 nm; Figure 6 ) is nearly double that of the rear head ( δ3=0 . 35±0 . 02 nm; Table 2 ) implies that increases in inter-head tension preferentially promote the unbinding of the front head . Kinesin is , in this sense , negatively rear-head gated at the 2-HB state: if anything , the presence of inter-head tension serves to undermine any contribution of the unbinding gate in maintaining processive motility . Because inter-head tension adversely affects processivity from the 2-HB state , and because the rate of rear-head release is not rate limiting for the kinesin cycle even in the absence of inter-head tension , we conclude that it is unlikely to function as the primary mechanism for head coordination in the kinesin cycle . This conclusion seemingly runs contrary to previous publications , implicating inter-head tension as the preferred gating mechanism ( Rosenfeld et al . , 2003; Guydosh and Block , 2006 , 2009; Toprak et al . , 2009; Shastry and Hancock , 2011 ) . However , as we have previously argued ( Clancy et al . , 2011 ) , results from the earlier studies can be reconciled with our own by invoking a front-head gating mechanism for kinesin where it is the spatial orientation of the NL ( i . e . , pointing forward in the docked state and rearward when undocked ) , rather than inter-head tension , per se , that facilitates coordination of biochemical states , as discussed ( Asenjo et al . , 2003; Hahlen et al . , 2006; Clancy et al . , 2011 ) . In WT kinesin , an essential difference between the 2-HB state [C] and the 1-HB state with the tethered head positioned in front of the bound head [B] ( Figure 1B ) is that inter-head tension can develop only when both heads are strongly bound . Although tension may accelerate the rear-head release , no increase in this rate is necessary for gating , so long as the biochemical events in the front head of a 2-HB motor cannot proceed . In other words , gating at the 2-HB state may be accomplished not by accelerating rates associated with the rear head , but by inhibiting—via the spatial orientation of the NL—the ATP hydrolysis cycle at the front head , until the rear head detaches . Evidently , the species of origin , the mutation of cysteines in the catalytic domain , as well as the sequence and length of the NL , can individually affect kinesin motility ( Figure 7A–F ) . Although the ability of certain mutants to backstep processively under superstall loads was previously attributed to an increase in NL length ( Yildiz et al . , 2008; Clancy et al . , 2011 ) , the dramatic difference between DmK-6AA and HsK-CL-6AA shows that extension of the NL alone cannot account for the behavior of HsK-CL-6AA ( Clancy et al . , 2011 ) and related constructs ( Yildiz et al . , 2008 ) . The AEQKLT insert produced similar velocities for HsK-CL-6AA and the non-CL version of the human construct , HsK-6AA ( AEQKLT ) ( Figure 7C‒F ) , but only the CL version could backstep processively . Based on these findings , we conclude that the processive backstepping behavior of HsK-CL-6AA arises from some combination of NL extension and the introduction of one or more of the seven mutations used to produce human CL kinesin . Further work will be required to understand the detailed basis of any such effects . Although we stress that our specific findings on kinesin gating , together with the general gating framework presented here , are not based on data from CL mutants , numerous publications have made extensive use of these ( Rice et al . , 1999; Case et al . , 2000; Tomishige and Vale , 2000; Peterman et al . , 2001; Rosenfeld et al . , 2001 , 2002; Sosa et al . , 2001; Sindelar et al . , 2002; Asenjo et al . , 2003; Rice et al . , 2003; Rosenfeld et al . , 2003; Naber et al . , 2003a , 2003b; Yildiz et al . , 2004; Asenjo et al . , 2006; Milescu et al . , 2006; Tomishige et al . , 2006; Mori et al . , 2007; Sindelar and Downing , 2007 , 2010; Verbrugge et al . , 2007; Dietrich et al . , 2008; Yildiz et al . , 2008; Asenjo and Sosa , 2009; Toprak et al . , 2009; Wong et al . , 2009; Verbrugge et al . , 2009a , 2009b , 2009c; Clancy et al . , 2011; Naber et al . , 2011; Mattson-Hoss et al . , 2014 ) . In light of the unusual behavior displayed by certain CL mutants , it remains to be determined which previous findings are applicable to kinesin in its WT form . To avoid artifacts in future work , it would be prudent to explore alternative linking chemistries for site-specific labels that do not require the elimination of native cysteine residues . Furthermore , wild-type behavior for mutant constructs should no longer be assumed simply on the basis of their unloaded velocities or ATP hydrolysis rates . The force-dependent velocities for the various Drosophila constructs with NL extensions ( DmK-1AA through DmK-6AA ) were apparently independent of the sequence used to extend the NL ( Figure 5 ) . However , different 6-AA NL insert sequences ( LQASQT vs AEQKLT ) exhibited different motile properties in both Drosophila and human constructs ( Figure 7C‒F ) . The reason for these differences is unclear , but we conjecture that it may be due to the presence of the positively charged lysine in the AEQKLT insert . In all constructs , NL extensions were introduced at the junction of NL and the stalk to minimize any disruption to the interaction between the remaining , native NL sequence and the catalytic head domain . It is conceivable that insertions at this position nevertheless affect kinesin in some ways beyond the reduction in tension arising from increased NL length , but we consider that possibility less likely , because the data in Figures 2‒5 can be well modeled by mechanical effects . We also favor the explanation that extensions of the NL lead to negligible inter-head strain in the 2-HB state , because extensions from 1 to 6 AA increase the NL length by over 30% , yet generate no corresponding change in velocity ( Figure 5 ) . The gating framework ( Figure 1B ) and associated conclusions would remain intact even in the presence of residual inter-head tension in mutants , however . In that case , the value for inter-head tension in the WT ( Fi , wt; Table 2 ) would be interpreted as the increase in tension ( relative to the mutants ) produced by its shorter NL . The general framework ( Figure 1B ) provides a structure for evaluating the relative contributions of individual gates to the kinesin cycle . Because the rate of unbinding of the front head increases faster than that of the read head under increasing inter-head tension , the unbinding gate functions most efficiently at lower tension , or in its absence . Extensions of the NL abolished inter-head tension in kinesin , yet motors remained active , albeit at reduced speeds , and their stepping gates remained fully functional , suppressing rear-head rebinding even in constructs with NLs lengthened up to three AA . These considerations , taken on their own , suggest that NL extensions from 1 to 3 AA might lead to more effective gating , and consequently to greater processivity . Instead , the most pronounced effect of NL extension was a dramatic decrease in processivity . An explanation is that the dominant effect of lengthening the NL is to compromise the binding gate , and that the deleterious effect of NL length at this gate outweighs any compensating effects of reduced inter-head tension at the unbinding gate . We conclude that the key role played by NL length in kinesin gating is twofold: ( 1 ) to enhance processivity by facilitating binding of the tethered head to the forward MT site at the binding gate , and ( 2 ) to increase velocity by accelerating rear-head release from the 2-HB state at the unbinding gate . In general , the kinesin-1 nanomechanical properties appear to be selected for processivity and velocity . Other cytoskeletal motors , including members of the kinesin , myosin , and dynein superfamilies , may be optimized to emphasize different combinations of useful characteristics , including stall force , load-bearing ability , velocity , processivity , directionality , or the ability to work cooperatively in groups , depending upon their roles in cells . Using the general gating framework ( Figure 1B ) , the strategies adopted to optimize for a specific functionality could be characterized in terms of the extent to which each gate contributes to the overall cycle of a given motor . Because gating considerations unify the coupling of biochemical states and mechanical events , these may be of value in framing future investigations , as well as in re-evaluating previous work , on how dimeric motors are coordinated ( Block , 2007; Gennerich and Vale , 2009; Kull and Endow , 2013 ) .
All Drosophila melanogaster recombinant constructs , with the exception of CL mutants , were derived from a truncated sequence , comprising the first 559 AA of the Drosophila kinesin-1 heavy chain ( KHC ) with eGFP and a 6xHis-tag engineered at the C-terminus , hereafter referred to as DmK-WT ( Shastry and Hancock , 2010; Milic et al . , 2014 ) . One or more rounds of site-directed mutagenesis ( QuikChange , Agilent Technologies , Santa Clara , CA ) were applied to produce a series of constructs with lengthened NL domains , extended by 1 ( L; DmK-1AA ) , 2 ( HV; DmK-2AA ) , 3 ( DAL; DmK-3AA ) , 4 ( LAST; DmK-4AA ) , 5 ( LASQT; DmK-5AA ) , or 6 AA ( LQASQT; DmK-6AA ) . A version of DmK-6AA with the NL insert AEQKLT was also generated . These insertions were all introduced at the junction of the NL and the coiled-coil stalk , between residues T344 and A345 ( Shastry and Hancock , 2010 ) . The procedures for the expression of proteins in Escherichia coli and subsequent purification via nickel column chromatography were described previously ( Uppalapati et al . , 2009; Shastry and Hancock , 2010 ) . A C338S point mutant ( DmK-C338S ) and a CL ( DmK-CL ) version with 2 mutations ( C45S and C338S ) were generated from a Drosophila KHC truncated at residue 401 , with a C-terminal 6xHis-tag ( Fehr et al . , 2009 ) . These mutants were expressed and purified as described ( Fehr et al . , 2009 ) . A round of site-directed mutagenesis was used to produce DmK-C338S , based on the DmK-CL construct . Human non-CL kinesin-1 motors were based on a truncated construct ( HsK-WT ) containing the first 595 residues of KIF5B ( Navone et al . , 1992 ) , with a C-terminal 6xHis-tag . The expression plasmid for HsK-WT ( pAF4 ) was created by replacing the sequence coding for a truncated Drosophila kinesin-1 in plasmid pCA1 ( Asbury et al . , 2003 ) with one corresponding to the first 595 residues of KIF5B , via cassette mutagenesis . Multiple rounds of site-directed mutagenesis were used to generate two constructs with 6 AA inserted into the NL region ( LQASQT and AEQKLT; HsK-6AA ) . The inserted residues were positioned at the junction of the NL and common coiled-coil stalk , between residues T336 and A337 , respectively . The non-CL human constructs were expressed in E . coli and purified on a nickel column , as described ( Fehr et al . , 2009 ) . A truncated CL human kinesin-1 construct ( HsK-CL ) containing the first 560 residues of the motor domain and 7 mutations ( C7S , C65A , C168A , C174S , C294A , C330S , C421A ) ( Rice et al . , 1999; Rosenfeld et al . , 2002; Clancy et al . , 2011 ) was also used in our motility experiments ( a gift of S Rosenfeld , Cleveland Clinic ) . Stopped-flow fluorescence experiments were performed in BRB80 buffer containing 1 mM MgCl2 . Measurements were made on an Applied Physics SX20 spectrofluorimeter equipped with a 356 nm excitation filter and an HQ480SP emission filter , as previously described ( Chen et al . , 2015 ) . The fraction of mantADP release from kinesin dimers upon MT binding was computed from the amplitudes of fluorescence signal decreases for MT binding ( AMT ) and sequential release ( ASR ) experiments . For MT binding experiments , a 30 μl solution of 0 . 05 μM dimeric kinesin , pre-incubated with 0 . 5 μM mantADP , was mixed via stopped-flow with an equivalent volume of 4 μM MTs and 10 μM taxol . The decrease in the fluorescence signal after mixing reflects the amount of mantADP released . The total amount of mantADP available for release was determined from sequential release experiments , where a 30 μl solution of 0 . 05 μM dimeric kinesin , pre-incubated with 0 . 5 μM mantADP , was mixed with an equivalent volume of 4 μM MTs , 10 μM taxol , and 2 mM ATP . The amplitudes for each fluorescence decay measurement were determined by the sum of the two amplitudes obtained from a fit of the sum of two exponentials . The fraction of mantADP release upon MT binding was computed as AMT/ASR . MantADP·kinesin·MT complexes were generated by pre-incubating 0 . 2 μM dimeric kinesin with 0 . 5 μM mantADP , 4 μM MTs , and 10 μM taxol . Fluorescence changes were recorded following the introduction of 30 μl of BRB80 buffer to an equivalent volume of the mantADP·kinesin·MT complex via stopped-flow . Records were fit to single exponentials , from which the rate of mantADP exchange was obtained . The single-molecule motility assay used in this study has been described ( Milic et al . , 2014 ) . Motility buffers consisted of BRB80 ( 80 mM Pipes , 1 mM EGTA , 4 mM MgCl2 ) at pH 6 . 9 , with 2 mM DTT , 10 μM Taxol ( Paclitaxel ) , and 2 mg·ml−1 BSA . An oxygen scavenging system with final concentrations of 50 μg·ml−1 glucose oxidase , 12 μg·ml−1 catalase , and 1 mg·ml−1 glucose was added to motility buffers before introduction into flow cells . An optical force clamp was implemented to acquire run length and velocity data under controlled , external loads ( Valentine et al . , 2008; Clancy et al . , 2011 ) . Data for DmK-5AA and all constructs with 6-AA NL inserts were collected using a force clamp with an improved detection scheme ( Milic et al . , 2014 ) . Unloaded run length and velocity data were acquired by video tracking ( Clancy et al . , 2011 ) . Unbinding rates under assisting and moderate hindering loads , ranging from −6 to +20 pN , were obtained directly from velocity and run length records . Unbinding rates under large hindering loads , ranging from −25 to −7 pN , were estimated separately , based on the average dwell time of the last step prior to dissociation ( see below ) . The starting and ending points for each single-molecule record were identified by inspection . For DmK with 1–6 AA NL inserts , the mean run length under unloaded conditions , L , was obtained from an exponential fit to the histogram of individual runs , where the first bin , and all bins with less than 6 counts , were excluded from fits ( Clancy et al . , 2011; Milic et al . , 2014 ) . As previously described ( Milic et al . , 2014 ) , mean run lengths under hindering loads were determined based on the number of runs , N1 , 2 , that fell into one of two bounded intervals: x1 < x < x2 , and x > x2 , respectively . Assuming that run lengths are exponentially distributed , the mean run lengths were computed from the relation L= ( x2–x1 ) /ln ( N1/N2+1 ) , along with an estimated standard error , σL=LN1/ ( N2 ( N1+N2 ) ) /ln ( N1/N2+1 ) . For run length data collected under loads ranging from −5 to −1 pN , the lower bound was set to x1 = 30 nm , and the upper bound to x2 = 150 nm . Because run lengths in the presence of a −6-pN force are exceedingly short , the corresponding bounds were set to x1 = 15 nm and x2 = 50 nm , respectively . Runs under assisting load conditions for DmK-5AA and DmK-6AA were collected using an improved force clamp ( Milic et al . , 2014 ) . Mean run lengths for these two constructs were obtained using the identical two-bin method , where the limits were x1 = 30 nm and x2 = 150 nm for data obtained under +1 to +9 pN loads , and x1 = 15 nm and x2 = 50 nm for loads greater than +9 pN . For DmK-1AA , DmK-2AA , and DmK-3AA , respectively , the mean run lengths under assisting-load conditions were determined by a maximum-likelihood estimator , as described ( Milic et al . , 2014 ) . Velocity measurements were collected and analyzed as described ( Milic et al . , 2014 ) . For DmK-WT , unbinding rates under force-clamped conditions from −6 to +20 pN were obtained by dividing the velocity at a given force by the corresponding run length . At high hindering loads , where kinesin stalls ( −25 to −7 pN ) , the unbinding rates were calculated by fitting an exponential distribution to histograms of the kinesin residence times on the MT . The unbinding rate , koff , as a function of assisting or hindering load ( Figure 6 ) , was fit to the function koff=koff0 exp[|Ftrap|δoff/kBT] where koff0 is the unloaded release rate , δoff is the characteristic distance parameter for unbinding , Ftrap is the force applied by the optical trap , and kBT is Boltzmann's constant times the absolute temperature . As previously ( Clancy et al . , 2011 ) , we implemented a formalism ( Chemla et al . , 2008 ) to derive an analytical expression ( using Mathematica 8 , Wolfram Research ) for velocity , ν , as a function of load , Ftrap , based on the 3-state model with 2 force-dependent transitions ( inset , Figure 5 ) :v ( Ftrap ) =dstepk10k20k30eFtrapδ1+ ( Ftrap+Fi ) δ3kBTk10k20eFtrapδ1kBT+k30e ( Ftrap+Fi ) δ3kBT ( k10eFtrapδ1kBT+k20 ) . Here , dstep is the kinesin step size ( fixed at 8 . 2 nm ) , Fi is inter-head tension , kBT is Boltzmann's constant times the absolute temperature , kn0 are the unloaded rates of the 3-state model ( Figure 5 , inset ) , and δn are the corresponding characteristic distance parameters for these rates . The seven free parameters ( Table 2 ) were determined by a global fit to 84 velocity data points ( Figure 5 ) as previously described ( Clancy et al . , 2011 ) , using Igor Pro 6 ( Wavemetrics ) . Fi for mutant constructs ( Fi , mutant ) was set to 0 pN ( see text ) . | In cells , molecules are moved from one location to another by motor proteins . Kinesins are a large family of such motors that transport their cargos along long filaments known as microtubules . Most kinesin molecules are formed from two identical protein chains . Each chain has a motor region at one end ( called the head ) that can attach to microtubules . The other end of each protein chain wraps around its partner to form a common stalk region ( called the tail ) that links to the cargo being carried . The two kinesin heads are connected to the tail via a ‘neck linker’ region , and they advance along the microtubule in strict alternation , similar to the way our legs move when walking . During each step , the front head remains tightly associated with the filament as the trailing head releases itself , advances beyond the front head , and reattaches to become the new leading head . The two heads need to coordinate their activities , so that at any given time , they're not at the same stage in the process . For example , if both heads remained bound to the microtubule at the same time , the motor would not be able to advance . If they both released , the motor would fall off the filament and diffuse away . However , the process by which the heads coordinate is not fully understood , and different models for how this process works have been proposed . Now , Andreasson , Milic et al . have examined the role played by the neck linker in coordinating the motor's movement using a technique known as ‘optical trapping’ . The experiments involved attaching microscopic beads to the motor proteins , which serve as markers that can be tracked . The beads can also be used to exert controlled forces on the kinesin molecules , to see how they respond to different loads . Andreasson , Milic et al . extended the length of neck linker by inserting extra amino acids ( which are the building blocks of proteins ) into this region of the protein . It was found that kinesins can still walk even when each neck linker was extended by up to six additional amino acids . However , introducing even a single amino acid into the linker relaxed the normal tension that exists between the heads when these are both bound to the filament . This resulted in slowed speeds , shorter distances of travel , and less ability to sustain loads . The experimental results suggest that the length of the neck linker in naturally occurring kinesins may be optimized to support maximum movement . Based on their data , Andreasson , Milic et al . propose a general framework for understanding the communication that needs to take place between the heads in order to walk in a coordinated manner . Further work is required to understand if motor proteins other than kinesins can also be understood with this same framework . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Examining kinesin processivity within a general gating framework |
Poor study methodology leads to biased measurement of treatment effects in preclinical research . We used available sunitinib preclinical studies to evaluate relationships between study design and experimental tumor volume effect sizes . We identified published animal efficacy experiments where sunitinib monotherapy was tested for effects on tumor volume . Effect sizes were extracted alongside experimental design elements addressing threats to valid clinical inference . Reported use of practices to address internal validity threats was limited , with no experiments using blinded outcome assessment . Most malignancies were tested in one model only , raising concerns about external validity . We calculate a 45% overestimate of effect size across all malignancies due to potential publication bias . Pooled effect sizes for specific malignancies did not show apparent relationships with effect sizes in clinical trials , and we were unable to detect dose–response relationships . Design and reporting standards represent an opportunity for improving clinical inference .
Preclinical experiments provide evidence of clinical promise , inform trial design , and establish the ethical basis for exposing patients to a new substance . However , preclinical research is plagued by poor design and reporting practices ( van der Worp et al . , 2010; Begley , 2013a; Begley and Ioannidis , 2015 ) . Recent reports also suggest that many effects in preclinical studies fail replication ( Begley and Ellis , 2012 ) . Drug development efforts grounded on non-reproducible findings expose patients to harmful and inactive agents; they also absorb scarce scientific and human resources , the costs of which are reflected as higher drug prices . Several studies have evaluated the predictive value of animal models in cancer drug development ( Johnson et al . , 2001; Voskoglou-Nomikos et al . , 2003; Corpet and Pierre , 2005 ) . However , few have systematically examined experimental design—as opposed to use of specific models—and its impact on effect sizes across different malignancies ( Amarasingh et al . , 2009; Hirst et al . , 2013 ) . A recent systematic review of guidelines for limiting bias in preclinical research design was unable to identify any guidelines in oncology ( Henderson et al . , 2013 ) . Validity threats in preclinical oncology may be particularly important to address in light of the fact that cancer drug development has one of the highest rates of attrition ( Hay et al . , 2014 ) , and oncology drug development commands billions of dollars in funding each year ( Adams and Brantner , 2006 ) . In what follows , we conducted a systematic review and meta-analysis of features of design and outcomes for preclinical efficacy studies of the highly successful drug sunitinib . Sunitinib is a multi-targeted tyrosine kinase inhibitor sunitinib ( SU11248 , Sutent ) and is licensed as monotherapy for three different malignancies ( Chow and Eckhardt , 2007; Raymond et al . , 2011 ) . As it was introduced into clinical development around 2000 and tested against numerous malignancies , sunitinib provided an opportunity to study a large sample of preclinical studies across a broad range of malignancies—including several supporting successful translation trajectories .
Our screen from database and reference searches captured 74 studies eligible for extraction , corresponding to 332 unique experiments investigating tumor volume response ( Figure 1 , Table 1 , Table 1—source data 1E ) . Effect sizes ( standardized mean difference [SMD] using Hedges' g ) could not be computed for 174 experiments ( 52% ) due to inadequate reporting ( e . g . , sample size not provided , effect size reported as a median , lack of error bars , Figure 1—figure supplement 1 ) . Overall , 158 experiments , involving 2716 animals , were eligible for meta-analysis . The overall pooled SMD for all extracted experiments across all malignancies was −1 . 8 [−2 . 1 , −1 . 6] ( Figure 2—figure supplement 1 ) . Mean duration of experiments used in meta-analysis ( Figures 2–4 ) was 31 days ( ±14 days standardized deviation of the mean ( SDM ) ) . 10 . 7554/eLife . 08351 . 003Figure 1 . Descriptive analysis of ( A ) internal , construct , and ( B ) external validity design elements . External validity scores were calculated for each malignancy type tested , according to the formula: number species used + number of models used; an extra point was assigned if a malignancy type tested more than one species and more than one model . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00310 . 7554/eLife . 08351 . 004Figure 1—source data 1 . ( A ) Coding details for IV and CV categories . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00410 . 7554/eLife . 08351 . 005Figure 1—figure supplement 1 . Descriptive analysis of ( A ) internal , construct , and ( B ) external validity design elements for all experiments ( n = 332 ) extracted for validity data parameters . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00510 . 7554/eLife . 08351 . 006Table 1 . Demographics of included studiesDOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00610 . 7554/eLife . 08351 . 007Table 1—source data 1 . ( C ) Search Strategies . ( D ) PRISMA Flow Diagram . ( E ) Demographics of included studies at qualitative level . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 007Study level demographicsIncluded studies ( n = 74 ) Conflict of interest Declared19 ( 26% ) Funding statement* Private , for-profit44 ( 59% ) Private , not-for-profit35 ( 47% ) Public37 ( 50% ) Other2 ( 3% ) Recommended clinical testing Yes37 ( 50% ) Publication date 2003–200613 ( 18% ) 2007–200917 ( 23% ) 2010–201344 ( 59% ) *Does not sum to 100% as many studies declared more than one funding source . 10 . 7554/eLife . 08351 . 008Figure 2 . Summary of pooled SMDs for each malignancy type . Shaded region denotes the pooled standardized mean difference ( SMD ) and 95% confidence interval ( CI ) ( −1 . 8 [−2 . 1 , −1 . 6] ) for all experiments combined at the last common time point ( LCT ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00810 . 7554/eLife . 08351 . 009Figure 2—source data 1 . ( B ) Heterogeneity statistics ( I2 ) for each malignancy sub-group . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 00910 . 7554/eLife . 08351 . 010Figure 2—figure supplement 1 . Effect sizes for all included experiments ( n = 158 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 01010 . 7554/eLife . 08351 . 011Figure 3 . Relationship between study design elements and effect sizes . The shaded region denotes the pooled SMD and 95% CI ( −1 . 8 [−2 . 1 , −1 . 6] ) for all experiments combined at the LCT . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 01110 . 7554/eLife . 08351 . 012Figure 4 . Funnel plot to detect publication bias . Trim and fill analysis was performed on pooled malignancies , as well as the three malignancies with the greatest study volume . ( A ) All experiments for all malignancies ( n = 182 ) , ( B ) all experiments within renal cell carcinoma ( RCC ) ( n = 35 ) , ( C ) breast cancer ( n = 32 ) , and ( D ) colorectal cancer ( n = 29 ) . Time point was the LCT . Open circles denote original data points whereas black circles denote ‘filled’ experiments . Trim and fill did not produce an estimate in RCC; therefore , no overestimation of effect size could be found . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 012 Effects in preclinical studies can fail clinical generalization because of bias or random variation ( internal validity ) , a mismatch between experimental operations and the clinical scenario modeled ( construct validity ) , or idiosyncratic causal mediators in an experimental system ( external validity ) ( Henderson et al . , 2013 ) . We extracted design elements addressing each using consensus design practices identified in a systematic review of validity threats in preclinical research ( Henderson et al . , 2013 ) . Few studies used practices like blinding or randomization to address internal validity threats ( Figure 1A ) . Only 6% of experiments investigated a dose–response relationship ( 3 or more doses ) . Concealment of allocation or blinded outcome assessment was never reported in studies that advanced to meta-analysis . It is worth noting that one research group employed concealed allocation and blinded assessment for the many experiments it described ( Maris et al . , 2008 ) . However , statistics were reported in a way that did not align with those we needed to calculate SMD . We found that 58 . 8% of experiments included active drug comparators , thus , facilitating interpretation of sunitinib activity ( however , we note that in some of the experiments , sunitinib was an active comparator in a test of a different drug or drug combination ) . Construct validity practices can only be meaningfully evaluated against a particular , matched clinical trial . Nevertheless , Figure 1A shows that experiments predominantly relied on juvenile , female , immunocompromised mouse models , and very few animal efficacy experiments used genetically engineered cancer models ( n = 4 ) or spontaneously arising tumors ( n = 0 ) . Malignancies generally scored low ( score = 1 ) for addressing external validity ( Figure 1B ) , with breast cancer studies employing the greatest variety of species ( n = 2 ) and models ( n = 4 ) . Implementation of internal validity practices did not show clear relationships with effect sizes ( Figure 3A ) . However , sunitinib effect sizes were significantly greater when active drug comparators were present in an experiment compared to when they were not ( −2 . 2 [−2 . 5 , −1 . 9] vs −1 . 4 [−1 . 7 , −1 . 1] , p-value <0 . 001 ) . Within construct validity , there was a significant difference in pooled effect size between genetically engineered mouse models and human xenograft ( p-value <0 . 0001 ) and allograft ( p-value 0 . 001 ) model types ( Figure 3B ) . For external validity ( Figure 3C ) , malignancies tested in more and diverse experimental systems tended to show less extreme effect sizes ( p < 0 . 001 ) . For the 158 individual experiments , 65 . 8% showed statistically significant activity at the experiment level ( p < 0 . 05 , Figure 2—figure supplement 1 ) , with an average sample size of 8 . 03 animals per treatment arm and 8 . 39 animals per control arm . Funnel plots for all studies ( Figure 4A ) , as well as our renal cell carcinoma ( RCC ) subset ( Figure 4B ) suggest potential publication bias . Trim and fill analysis suggests an overestimation of effect size of 45% ( SMD changed from −1 . 8 [−2 . 1 , −1 . 7] to −1 . 3 [−1 . 5 , −1 . 0] ) across all indications . For high-grade glioma and breast cancer , the overestimation was 11% and 52% , respectively . However , trim and fill analysis suggested excellent symmetry for the RCC subgroup , suggesting coverage of the overall effect size and confidence intervals and not overestimation of effect size . Every malignancy tested with sunitinib showed statistically significant anti-tumor activity ( Figure 2 ) . Though we did not perform a systematic review to estimate clinical effect sizes for sunitinib against various malignancies , a perusal of the clinical literature suggests little relationship between pooled effect sizes and demonstrated clinical activity . For instance , sunitinib monotherapy is highly active in RCC patients ( Motzer et al . , 2006a , 2006b ) and yet showed a relatively small preclinical effect; in contrast , sunitinib monotherapy was inactive against small cell lung cancer in a phase 2 trial ( Han et al . , 2013 ) , but showed relatively large preclinical effects . Using measured effect sizes at a standardized time point of 14 days after first administration ( a different time point than in Figures 2–4 to better align our evaluation of dose–response ) , we were unable to observe a dose–response relationship over three orders of magnitude ( 0 . 2–120 mg/kg/day ) for all experiments ( Figure 5A ) . We were also unable to detect a dose–response relationship over the full dose range ( 4–80 mg/kg/day ) tested in the RCC subset ( Figure 5B ) . The same results were observed when we performed the same analyses using the last time point in common between the experimental and control arms . 10 . 7554/eLife . 08351 . 013Figure 5 . Dose–response curves for sunitinib preclinical studies . Only experiments with a once daily ( no breaks ) administration schedule were included in both graphs . Effect size data were taken from a standardized time point ( 14 days after first sunitinib administration ) . ( A ) Experiments ( n = 158 ) from all malignancies tested failed to show a dose–response relationship . ( B ) A dose–response relationship was not detected for RCC ( n = 24 ) . ( C ) Dose–response curves reported in individual studies within the RCC subset showed dose–response patterns ( blue diamond = Huang 2010a [n = 3] , red square = Huang 2010d [n = 3] , green triangle = Ko 2010a [n = 3] , purple X = Xin 2009 [n = 3] ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08351 . 013
Preclinical studies serve an important role in formulating clinical hypotheses and justifying the advance of a new drug into clinical testing . Our meta-analysis , which included malignancies that respond to sunitinib in human beings and those that do not , raises several questions about methods and reporting practices in preclinical oncology—at least in the context of one well-established drug . First , reporting of design elements and data was poor and inconsistent with widely recognized standards for animal studies ( Kilkenny et al . , 2010 ) . Indeed , 98 experiments ( 30% of qualitative sample ) could not be quantitatively analyzed because sample sizes or measures of dispersion were not provided . Experimenters only sporadically addressed major internal validity threats and tended not to test indication-activity in more than one model and species . This finding is consistent with what others have observed in experimental stroke and other research areas ( Macleod et al . , 2004; van der Worp et al . , 2005; Kilkenny et al . , 2009; Glasziou et al . , 2014 ) . Some teams have shown a relationship between failure to address internal validity threats and exaggerated effect size ( Crossley et al . , 2008; Rooke et al . , 2011 ) ; we did not observe a clear relationship . Consistent with what has been reported in stroke ( O'Collins et al . , 2006 ) , our findings suggest that testing in more models tends to produce smaller effect sizes . However , since a larger sample of studies will provide a more precise estimate of effect , we cannot rule out that the trends observed for external validity reflect a regression to the mean . Second , preclinical studies for sunitinib seem to be prone to publication bias . Notwithstanding limitations on using funnel plots to detect publication bias ( Lau et al . , 2006 ) , our plots were highly asymmetrical . That all malignancy types tested showed statistically significant anti-cancer activity strains credulity . Others have reported that far more animal studies report statistical significance than would be expected ( Wallace et al . , 2009; Tsilidis et al . , 2013 ) , and our observations that two thirds of individual studies showed significance extends these observations . Third , we were unable to detect a meaningful relationship between preclinical effect sizes and known clinical behavior . Although a full analysis correlating trial and preclinical effect sizes will be needed , we did not observe obvious relationships between the two . We also did not detect a dose–response effect over three orders of magnitude even within an indication—RCC—known to respond to sunitinib and even when different time points were used . It is possible that heterogeneity in cell lines or strains may have obscured the effects of dose . For example , experimenters may have delivered higher doses to xenografts known to show slow tumor growth . However , RCC patients—each of whom harbors genetically distinct tumors—show dose–response effects in trials ( Faivre et al . , 2006 ) and between trials in a meta-analysis ( Houk et al . , 2010 ) . It is also possible that the toxicity of sunitinib may have limited the ability to demonstrate dose response , though this contradicts demonstration of dose response within studies ( Abrams et al . , 2003; Amino et al . , 2006; Ko et al . , 2010 ) . Finally , the tendency for preclinical efficacy studies to report drug dose , but rarely drug exposure ( i . e . , serum measurement of active drug ) , further limits the construct validity of these studies ( Peterson and Houghton , 2004 ) . One explanation for our findings is that human xenograft models , which dominated our meta-analytic sample , have little predictive value , at least in the context of receptor tyrosine kinase inhibitors . This is a possibility that contradicts other reports ( Kerbel , 2003; Voskoglou-Nomikos et al . , 2003 ) . We disfavor this explanation in light of the suggestion of publication bias; also , xenografts should show a dose–response regardless of whether they are useful clinical models . A second explanation is that experimental methods are so varied as to mask real effects . However , we note that the observed patterns on experimental design are based purely on what was reported in ‘Materials and methods’ section . Third , experiments assessing changes in tumor volume might only be interpretable in the context of other experiments within a preclinical report , such as with mechanistic and pharmacokinetic studies . This explanation is consistent with our observation that studies testing effect along a causal pathway tended to produce smaller effect sizes . A fourth possible explanation for our findings is that the predictive value of a small number of preclinical studies was obscured by inclusion of poorly designed and executed preclinical studies in our meta-analysis . Quantitative analysis of preclinical design factors that confer greater clinical generalizability awaits side-by-side comparison with pooled effects in clinical trials . Finally , it may be that design and reporting practices are so poor in preclinical cancer research as to make interpretation of tumor volume curves useless . Or , non-reporting may be so rampant as to render meta-analysis of preclinical research impossible . If so , this raises very troubling questions for the publication economy of cancer biology: even well-designed and reported studies may be difficult to interpret if their results cannot be compared to and synthesized with other studies . Our systematic review has several limitations . First , we relied on what authors reported in the published study . It is possible certain experimental practices , like randomization , were used but not reported in methods . Further to this , we relied only on published reports , and restriction of searches to the English language may have excluded some articles . In February of 2012 , we filed a Freedom of Information Act request from the Food and Drug Administration ( FDA ) for additional preclinical data submitted in support of sunitinib's licensure; nearly 4 years later , the request has not been honored . Second , effect sizes were calculated using graph digitizer software from tumor volume curves: minor distortion of effect sizes may have occurred but were likely non-differential between groups . Third , subtle experimental design features—not apparent in ‘Materials and methods’ sections—may explain our failure to detect a dose–response effect . For instance , few reports provide detailed animal housing and testing conditions , perhaps leading to important inter-laboratory differences in tumor growth . It should also be emphasized that our study was exploratory in nature; findings like ours will need to be confirmed using prespecified protocols . Fourth , our study represents analysis of a single drug , and it may be our findings do not extend beyond receptor tyrosine kinase inhibitors , or sunitinib . However , many of our findings are consistent with those observed in other systematic reviews of preclinical cancer interventions ( Amarasingh et al . , 2009; Sugar et al . , 2012; Hirst et al . , 2013 ) . Fifth , our analysis does not directly address many design elements—like duration of experiment or choice of tissue xenograft—that are likely to bear on study validity . Finally , we acknowledge that there may be funding constraints that limit implementation of validity practices described above . We note , nevertheless , that other realms , in particular , neurology , have found ways to make such methods a mainstay . Numerous commentators have raised concerns about the design and reporting of preclinical cancer research ( Sugar et al . , 2012; Begley , 2013b ) . In one report , only 11% preclinical cancer studies submitted to a major biotechnology company withstood in-house replication ( Begley and Ellis , 2012 ) . The Center for Open Science and Science Exchange has initiated a project that will attempt to reproduce 50 of the highest impact papers in cancer biology published between 2010 and 2012 ( Morrison , 2014 ) . In a recent commentary , Smith et al . fault many researchers for performing in vitro preclinical tests using drug levels that are clinically unachievable due to toxicity ( Smith and Houghton , 2013 ) . Unaddressed preclinical validity threats like this—and the ones documented in our study—encourage futile clinical development trajectories . Many research areas , like stroke , epilepsy , and cardiology , have devised design guidelines aimed at improving the clinical generalizability of preclinical studies ( Fisher et al . , 2009; Galanopoulou et al . , 2012; Curtis et al . , 2013; Pusztai et al . , 2013 ) ; and the ARRIVE guidelines ( Kilkenny et al . , 2010 ) for reporting animal experiments have been taken up by numerous journals and funding bodies . Our findings provide further impetus for developing and implementing guidelines for the design , reporting , and synthesis of preclinical studies in cancer .
To identify all in vivo animal studies testing the anti-cancer properties of sunitinib ( ‘efficacy studies’ ) , we queried the following databases on 27 February 2012 using a search strategy adapted from Hooijmans et al . ( 2010 ) and de Vries et al . ( 2011 ) : Ovid MEDLINE In-Process & Other Non-Indexed Citations and Ovid MEDLINE ( dates of coverage from 1948 to 2012 ) , EMBASE Classic and EMBASE database ( dates of coverage from 1974 to 2012 ) and BIOSIS Previews ( dates of coverage from 1969 to 2012 ) . Search results were entered into an EndNote library and duplicates were removed . Additional citations were identified during the screening of identified articles . See Table 1—source data 1C , D for detailed search strategy and PRISMA flow diagram . Screening was performed at citation level by two reviewers ( CF and VCH ) , and at full-text by one reviewer ( VCH ) . Inclusion criteria were ( a ) original reports or abstracts , ( b ) English language , ( c ) contained at least one experiment measuring disease response in a live , non-human animals , and ( d ) employed sunitinib in a control , comparator , or experimental context , ( e ) tested anti-cancer activity . To avoid capturing the same experiment twice , in rare cases where the same experiment was reported in different articles , the most detailed and/or recent publication was included . All included studies were evaluated at the study-level , but only those with eligible experiments ( e . g . , those evaluating the effect of monotherapy on tumor volume and that were reported with sample sizes and error measurements ) were forwarded to experiment-level extractions . We excluded experiments when they had been reported in a previous publication after specifically searching for duplicates during screening and analysis . For each eligible experiment , we extracted experimental design elements derived from a prior systematic review of validity threats in preclinical research ( Henderson et al . , 2013 ) . Details regarding the coding of internal and construct validity categories are given in Figure 1—source data 1A . To score for external validity , we created an index that summed the number of species and models tested for a given malignancy and awarded an extra point if more than one species and model was tested . For example , if experiments within a malignancy tested two species and three different model types , the external validity score would be 4 ( 1 point for the second species , one point for the second model type , one point for the third model type , and an extra point because more than one model and species were employed ) . Our primary outcome was experimental tumor volume and we extracted necessary information ( sample size , mean measure of treatment effect , and SDM/SEM ) to enable calculation of study and aggregate level effect sizes . Since the units of tumor volume were not always consistent between experiments , we extracted those experiments for which a reasonable proxy of tumor volume could be obtained . These included physical caliper measurements ( often reported in mm3 or cm3 ) , tumor weights ( often reported in mg ) , optical measurements made from luminescent tumor cell lines ( often reported in photons/second ) , and fold differences in tumor volumes between the control and treatment arms . We extracted experiments of both primary and metastatic tumors , but not experiments where tumor incidence was reported . To account for these different measures of tumor volume , SMDs were calculated using Hedges' g . Hedges' g is a widely accepted standardized measure of effect in meta-analyses where units are not always identical . For experiments where more than one dose of sunitinib was tested against the same control arm , we created a pooled SMD to adjust appropriately for the multiple use of the same control group . Data were extracted at baseline ( Day 0 and defined as the first day of drug administration ) , Day 14 ( the closest measured data point to 14 days following first dose ) , and the last common time point ( LCT ) between the control group and the treatment group . The LCT was variable between experiments and the last time point for which we could calculate SMD and often represented the point at which the greatest difference was observed between the arms . Data presented graphically were extracted using the graph digitizer software GraphClick ( Arizona Software ) . Extraction was performed by four independent and trained coders ( VCH , ND , AH , and NM ) using DistillerSR . There was a 12% double-coding overlap to minimize inter-rater heterogeneity and prevent coder drift . Discrepancies in double coding were reconciled through discussion , and if necessary , by a third coder . The gross agreement rate before reconciliation for all double-coded studies was 83% . Effect sizes were calculated as SMDs using Hedges' g with 95% confidence intervals . Pooled effect sizes were calculated using a random effects model employing the DerSimonian and Laird ( 1986 ) method , in OpenMeta[Analyst] ( Wallace et al . , 2009 ) . We also calculated heterogeneity within each malignancy using I2 statistics ( Figure 2—source data 1B ) . To assess the predictive value of preclinical studies in our sample , we calculated pooled effect sizes for each type of malignancy . Subgroup analyses were performed for each validity element . p-values were calculated by a two-sided independent group T-test . Statistical significance was set at a p-value <0 . 05; as this was an exploratory study we did not adjust for multiple analyses . Funnel plots to assess publication bias and Duval and Tweedie's trim and fill estimates were generated using Comprehensive Meta Analyst software ( Dietz et al . , 2014 ) . Funnel plots were created for all experiments in aggregate , and for the three indications for which greater than 20 experiments were analyzable . Dose–response curves are a widely used tool for testing the strength of causal relationships ( Hill , 1965 ) , and if preclinical studies indicate real drug-responses , we should be able to detect a dose–response effect across different experiments . Dose–response relationships were found in post-analysis of sunitinib clinical studies in metastatic RCC and Gastrointestinal stromal tumour ( GIST ) ( Houk et al . , 2010 ) . We tested for all indications in aggregate , as well as for RCC , an indication known to respond to sunitinib in human beings ( Motzer et al . , 2006a , 2006b , 2009 ) . To eliminate variation at the LCT between treatment and control arms , dose–response curves were created using data from a time point 14 days from the initiation of sunitinib treatment . Experiments with more than one treatment arm were not pooled as in other analyses , but expanded out so that each treatment arm ( with it's respective dose ) could be plotted properly . As we were unable to find experiments that reported drug exposure ( e . g . , drug serum levels ) , we calculated pooled effect sizes in OpenMeta[Analyst] and plotted against dose . To avoid the confounding effect of discontinuous dosing , we included only experiments that used a regular administration schedule without breaks ( i . e . , sunitinib administered at a defined dose once a day instead of experiments where sunitinib was dosed more irregularly or only once ) . As this meta-analysis was exploratory and involved development of methodology , we did not prospectively register a protocol . | Developing a new drug can take years , partly because preclinical research on non-human animals is required before any clinical trials with humans can take place . Nevertheless , only a fraction of cancer drugs that are put into clinical trials after showing promising results in preclinical animal studies end up proving safe and effective in human beings . Many researchers and commentators have suggested that this high failure rate reflects flaws in the way preclinical studies in cancer are designed and reported . Now , Henderson et al . have looked at all the published animal studies of a cancer drug called sunitinib and asked how well the design of these studies attempted to limit bias and match the clinical scenarios they were intended to represent . This systematic review and meta-analysis revealed that many common practices , like randomization , were rarely implemented . None of the published studies used ‘blinding’ , whereby information about which animals are receiving the drug and which animals are receiving the control is kept from the experimenter , until after the test; this technique can help prevent any expectations or personal preferences from biasing the results . Furthermore , most tumors were tested in only one model system , namely , mice that had been injected with specific human cancer cells . This makes it difficult to rule out that any anti-cancer activity was in fact unique to that single model . Henderson et al . went on to find evidence that suggests that the anti-cancer effects of sunitinib might have been overestimated by as much as 45% because those studies that found no or little anti-cancer effect were simply not published . Though it is known that the anti-cancer activity of the drug increases with the dose given in both human beings and animals , an evaluation of the effects of all the published studies combined did not detect such a dose-dependent response . The poor design and reporting issues identified provide further grounds for concern about the value of many preclinical experiments in cancer . These findings also suggest that there are many opportunities for improving the design and reliability of study reports . Researchers studying certain medical conditions ( such as strokes ) have already developed , and now routinely implement , a set of standards for the design and reporting of preclinical research . It now appears that the cancer research community should do the same . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"medicine"
] | 2015 | A meta-analysis of threats to valid clinical inference in preclinical research of sunitinib |
Mammalian spermatozoa gain competence to fertilize an oocyte as they travel through the female reproductive tract . This process is accompanied by an elevation of sperm intracellular calcium and a membrane hyperpolarization . The latter is evoked by K+ efflux; however , the molecular identity of the potassium channel of human spermatozoa ( hKSper ) is unknown . Here , we characterize hKSper , reporting that it is regulated by intracellular calcium but is insensitive to intracellular alkalinization . We also show that human KSper is inhibited by charybdotoxin , iberiotoxin , and paxilline , while mouse KSper is insensitive to these compounds . Such unique properties suggest that the Slo1 ion channel is the molecular determinant for hKSper . We show that Slo1 is localized to the sperm flagellum and is inhibited by progesterone . Inhibition of hKSper by progesterone may depolarize the spermatozoon to open the calcium channel CatSper , thus raising [Ca2+] to produce hyperactivation and allowing sperm to fertilize an oocyte .
Mammalian spermatozoa are unable to fertilize the oocyte immediately after their deposit into the female reproductive tract . Instead , they have to undergo a final maturation termed capacitation , during which spermatozoa gain competence to fertilize ( Chang , 1951; Austin , 1952 ) . Early stages of capacitation include the bicarbonate-mediated acceleration of sperm beat frequency and an increase in linear motility ( Visconti et al . , 1995a , 1995b , 1999 , 2002; Chen et al . , 2000; Wennemuth et al . , 2003; Wandernoth et al . , 2010; Mannowetz et al . , 2011 ) . Late stages of capacitation comprise—amongst others—intracellular alkalinization ( Meizel and Deamer , 1978 ) , elevation of intracellular Ca2+ ( Visconti et al . , 2002 ) , and membrane hyperpolarization ( Zeng et al . , 1995; Arnoult et al . , 1996; Demarco et al . , 2003 ) . These interdependent processes are regulated by sperm ion channels , of which Hv1 and CatSper ( Cation channel of sperm ) were identified as the major H+ and Ca2+ channels of human spermatozoa ( Ren et al . , 2001; Kirichok et al . , 2006; Lishko and Kirichok , 2010; Lishko et al . , 2010 , 2011; Ren and Xia , 2010; Strunker et al . , 2011; Lishko et al . , 2012 ) . However , the identity of the principal human K+ channel remained elusive . Potassium channels are indispensable for normal sperm physiology , since they regulate membrane potential and cell motility . Recently , an alkalinization-sensitive sperm K+ channel , encoded by the kcnu1 gene ( also known as Slo3 ) , was shown to be essential for male fertility in mice ( Schreiber et al . , 1998; Navarro et al . , 2007; Santi et al . , 2010; Zeng et al . , 2011 ) . It has been assumed , but never proven , that the K+ channel of human sperm has a similar molecular identity . The Slo gene family is represented by Slo1 , Slo2 , and Slo3 ( Wei et al . , 2005 ) . These channels possess seven transmembrane helices S0–S6 , with the S1–S6 helices exhibiting homology to classic voltage-gated K+ channels . They are tetramers of α subunits , with the K+-selective pore formed by S5 and S6 ( Adelman et al . , 1992; Butler et al . , 1993; Diaz et al . , 1998; Cui and Aldrich , 2000 ) . In addition , the Slo1 channel contains a large cytosolic C-terminus with two intracellular regulators of K+ conductance ( RCK ) , both of which contain high affinity Ca2+ binding sites ( Jiang et al . , 2001; Yuan et al . , 2010 ) . These structural elements give Slo1 channels the ability to sense changes in both voltage and intracellular Ca2+ concentrations ( Marty , 1981; Pallotta et al . , 1981; Barrett et al . , 1982; Latorre et al . , 1982; Schreiber et al . , 1999 ) . Due to their large single-channel conductance of 60–270 pS , Slo1 channels are also known as big potassium ( BK ) or maxi K channels ( Atkinson et al . , 1991; Kaczorowski et al . , 1996; Salkoff et al . , 2006 ) . Slo3 channels , on the other hand , lack the Ca2+ bowl ( Schreiber et al . , 1999; Xia et al . , 2004 ) , but are sensitive to intracellular alkalinization ( Schreiber et al . , 1998; Zhang et al . , 2006a , 2006b; Navarro et al . , 2007 ) . The pore-forming α subunits of Slo channels are associated with auxiliary β- and γ-subunits ( Behrens et al . , 2000; Brenner et al . , 2000; Uebele et al . , 2000; Yan and Aldrich , 2010 , 2012; Yang et al . , 2011 ) , which interact with the S0 segment of the α subunit . Several studies demonstrate that the association with different subunits impacts channel pharmacological and gating properties . In addition , splice variants of the Slo1 mRNA contribute to the functional diversity of BK channels ( Fodor and Aldrich , 2009; Johnson et al . , 2011 ) . Apart from responding to different stimuli , Slo1 and Slo3 channels are distributed discretely within the body as shown in numerous animal studies . Slo1 is detectable in excitable tissues , such as in hippocampus ( Hicks and Marrion , 1998 ) , smooth muscle cells ( Knaus et al . , 1994a , 1994b ) and adrenal chromaffin cells ( Solaro and Lingle , 1992 ) , whereas Slo3 transcripts are exclusively expressed in male germ cells ( Schreiber et al . , 1998 ) . Male Slo1−/− animals are able to produce offspring when paired with Slo+/+ females . However , the litter size was normal only in 10% of the matings ( Meredith et al . , 2004 ) . Abolishing the Slo3 gene results in more dramatic changes in testicular spermatozoa , such as morphological abnormalities after capacitation , reduced progressive motility , impaired acrosome reaction , and membrane depolarization during capacitation ( Schreiber et al . , 1998; Santi et al . , 2010; Zeng et al . , 2011 ) . These data indicate that Slo channels are essential for male fertility in mice , which makes them possible candidates for being the major K+ channel of human sperm . The goal of our work was to resolve the identity of the major K+ channel in human ejaculated spermatozoa . By applying the patch-clamp technique to ejaculated and epididymal human sperm cells , we found that human K+ currents are insensitive to intracellular alkalinization but are dependent on intracellular [Ca2+] . We furthermore demonstrate that the human sperm potassium ( hKSper ) current is inhibited by three known Slo1 channel inhibitors: charybdotoxin ( Anderson et al . , 1988; MacKinnon and Miller , 1988 ) , iberiotoxin ( Galvez et al . , 1990; Candia et al . , 1992; Giangiacomo et al . , 1992 ) and paxilline ( Knaus et al . , 1994c; Sanchez and McManus , 1996; Zhou et al . , 2010 ) , as well as by the micromolar concentrations of progesterone . Taking together our electrophysiological , biochemical , and immunocytochemistry data , we conclude that the Slo1 protein constitutes a major potassium channel of human spermatozoa . Therefore , the molecular identity of human KSper is distinct from that of murine KSper , which is represented by the Slo3 protein .
The flagellar pH-dependent Ca2+ channel CatSper is indispensable for male fertility . However , to gain its full activity several events must be met: intracellular alkalinization , presence of progesterone and membrane depolarization ( Ren et al . , 2001; Kirichok et al . , 2006; Lishko and Kirichok , 2010; Lishko et al . , 2011; Strunker et al . , 2011 ) . Since K+ channels are involved in the regulation of membrane potential , we hypothesized that the human KSper current ( IKSper ) also originates from the sperm tail to support CatSper activity . To address this question , we recorded from both whole sperm cells and isolated sperm flagella ( Figure 1 ) . To isolate IKSper from ICatSper , we recorded K+ currents in a potassium methanesulfonate-based solution in the presence of 0 . 1–1 mM extracellular Ca2+ . When divalent cations are absent from the extracellular solution , so called divalent free ( DVF ) condition , CatSper is able to conduct monovalent ions , such as K+ . However , in the presence of 0 . 1–1 mM external Ca2+ , ICatSper is effectively blocked ( Kirichok et al . , 2006; Lishko et al . , 2011; Smith et al . , 2013 ) , thus leading to pure K+ conductance through K+ channels . As shown in Figure 1A , B , K+ currents elicited under DVF conditions were approximately four times larger than the current recorded in the presence of 1 mM Ca2+ . The larger potassium currents ( gray bars ) represent a mixture of the K+ efflux through CatSper and KSper while only KSper current remains in the presence of external calcium ( red bars ) . Similar amplitudes of KSper currents recorded from whole sperm cells or sperm flagella indicate that IKSper originate primarily from the sperm flagellum , in the same manner as does ICatSper ( Figure 1B; Lishko et al . , 2011 ) . 10 . 7554/eLife . 01009 . 003Figure 1 . hKSper currents originate from the sperm tail . ( A ) IKSper was recorded in response to a voltage ramp as shown . Shown are representative traces from whole spermatozoon ( left panel; recordings are from the same cell ) and sperm tail ( right panel; recordings are from the same flagellum ) . Black traces represent currents in divalent free conditions , which allow K+ current through CatSper . Red traces show true IKSper . Latter was recorded in the presence of 1 mM extracellular Ca2+ , which inhibits monovalent currents through CatSper . ( B ) Current densities were obtained at +120 mV and presented as mean ± SEM . ( n ) , number of experiments . Four different sperm cells ( or four different sperm flagella ) of two different human donors were used . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 003 Sperm intracellular alkalinization was shown to be essential for murine KSper ( Slo3 ) activation ( Navarro et al . , 2007; Santi et al . , 2010; Zeng et al . , 2011 ) . Recently , recombinant human Slo3 co-expressed with a γ-subunit was also shown to exhibit pH-dependency ( Leonetti et al . , 2012 ) . However , its pH-sensitivity was shifted toward a more acidic pH range than that of mouse Slo3 . Therefore , we decided to test whether human KSper exhibits the same pH sensitivity and recorded K+ currents from human spermatozoa under conditions when intracellular pH ( pHi ) was held either at pH 7 . 4 or 5 . 5 , and external pH kept at 7 . 4 . To evoke intracellular alkalinization , 10 mM of NH4Cl was added to the external ( bath ) solution , which is a standard technique to effectively and quickly raise an intracellular pH ( Babcock et al . , 1983; Kirichok et al . , 2006; Navarro et al . , 2007 ) . In the experiments with a pHi of 5 . 5 , 1 mM of Zn2+ was added to the bath solution to inhibit sperm voltage-gated proton channel ( Hv1 ) activity ( Lishko et al . , 2010 ) . As shown in Figure 2A , human IKSper remained unaffected by intracellular alkalinization at both pHi 7 . 4 ( upper left panel ) and pHi 5 . 5 ( lower left panel ) . However , K+ currents were greatly potentiated by intracellular alkalinization in the absence of divalent cations ( Figure 2A , B , right panels ) , which was primarily due to the activation of pH-dependent K+ efflux through CatSper channels . Note that in DVF conditions , control currents were larger than control KSper currents , due to the efflux of potassium ions through both KSper and CatSper channels . Intracellular alkalinization up-regulates CatSper channel activity , therefore increasing the potassium efflux through it , while KSper currents remain unchanged . 10 . 7554/eLife . 01009 . 004Figure 2 . hKSper currents are insensitive to intracellular alkalinization . ( A ) Representative KSper currents were recorded from sperm cells in response to voltage ramps as shown . Recordings were done with various pHi as indicated . The bath solution containing 1 mM Ca2+ was used to inhibit K+ current through CatSper ( left panels ) . Right panels show traces in divalent free conditions , which allow K+ current through CatSper . Intracellular alkalinization was evoked by addition of 10 mM NH4Cl to the bath ( red traces ) . A weak intracellular buffer ( 5 mM of HEPES or MES ) allowed instantaneous pH changes . Zn2+ was used to block H+ currents via Hv1 at acidic intracellular pH . The upper panels and the lower panels are recordings from two different sperm cells . ( B ) KSper and CatSper/KSper current densities ( CDs ) recorded from sperm cells as shown in ( A ) . At pHi 7 . 4 KSper CDs were: 45 ± 3 pA/pF ( control ) and 44 ± 4 pA/pF ( plus NH4Cl ) . These values were similar at pHi 5 . 5: CDs were: 51 ± 8 pA/pF ( control ) and 53 ± 4 pA/pF ( plus NH4Cl ) . However , under DVF conditions that permit K+ efflux through CatSper , CDs at pHi 7 . 4 were: 116 ± 11 pA/pF ( control ) and 231 ± 26 pA/pF ( plus NH4Cl ) . At pHi 5 . 5 , CDs were: 85 ± 5 pA/pF ( control ) and 341 ± 21 pA/pF ( plus NH4Cl ) . Shown are CDs acquired at +120 mV and presented as mean ± SEM; n = 4–6 independent experiments with cells from four different human donors . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 004 To exclude that components of the seminal plasma may alter human sperm K+ channel behavior with regard to pH sensitivity , we also recorded K+ currents from human epididymal spermatozoa from a fertile patient undergoing vasectomy reversal . As shown in Figure 3 , IKSper did not change upon intracellular alkalinization . This indicates that the lack of hKSper pH-sensitivity is not due to the effect of seminal plasma , but rather it is an intrinsic property of the human sperm potassium channel . This particular feature of human KSper differentiates it from mouse KSper and suggests that the molecular identities of the channels are different . 10 . 7554/eLife . 01009 . 005Figure 3 . hKSper currents from human epididymal spermatozoa are insensitive to intracellular alkalinization . The upper panel shows representative IKSper traces recorded from human epididymal spermatozoa ( whole sperm cell ) in the control ( black ) and in the presence of 10 mM NH4Cl ( red ) . The lower panel presents mean currents acquired at +120 mV; ( n ) , number of experiments . IKSper did not change upon intracellular alkalinization with current densities averaging at 67 ± 10 pA ( control ) and 62 ± 11 pA ( after addition of 10 mM NH4Cl ) . Three epididymal spermatozoa were tested . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 005 As mentioned earlier , the pH-dependent mouse KSper channel is encoded by the Slo3 gene , while other members of the Slo family , such as Slo1 are not pH-dependent , but rather Ca2+-dependent . To determine if intracellular calcium affects human IKSper , we recorded K+ currents under different intracellular free Ca2+ concentrations: 0 , 0 . 1 or 50 μM ( Figure 4 ) with 0 . 1 mM Ca2+ in the bath solution . As illustrated in Figure 4A , B , the outward IKSper was slightly increased with [Ca2+]i = 0 . 1 μM compared to the control ( zero calcium ) . Under these conditions , IKSper exhibited outward rectification . However , with a [Ca2+]i = 50 μM , not only was the outward current potentiated twofold , but an inward potassium current was also present . Interestingly , intracellular calcium also notably decreased the activation time for human KSper ( Figure 4A , lower panel ) . However , the quantitative measurements of activation time constant in the presence of calcium were hindered by a fast channel kinetics that overlapped with capacitance artifacts . 10 . 7554/eLife . 01009 . 006Figure 4 . hKSper is activated by intracellular calcium . ( A ) Upper panels: representative IKSper recorded with various intracellular [Ca2+]free as indicated , in response to a voltage ramp . Lower panels: corresponding representative IKSper elicited by a step protocol from a holding potential of −80 mV to +120 mV with 20 mV increments . For clarity , traces at −80 mV , 0 mV , and +120 mV are labeled in blue , magenta , and green , respectively . Representative traces were obtained from three different sperm cells ( upper and lower panels ) . ( B ) Current–voltage ( I–V ) relationship in response to 0 μM ( black ) , 0 . 1 μM ( gray ) , and 50 μM ( red ) intracellular [Ca2+]free . At a membrane potential ( Vm ) of −80 mV , potassium currents were: -1 . 2 ± 0 . 5 pA ( [Ca2+]i = 0 ) , -0 . 8 ± 0 . 2 pA ( [Ca2+]i = 0 . 1 μM ) , and -18 . 5 ± 2 . 6 pA ( [Ca2+]i = 50 μM ) . At Vm = +120 mV , IKSpers were 57 ± 8 pA ( [Ca2+]i = 0 ) , 87 ± 5 pA ( [Ca2+]i = 0 . 1 μM ) , and 122 ± 8 pA ( [Ca2+]i = 50 μM ) . Data are shown as means ± SEM; n = 6–11 independent experiments with cells from six different donors . Data are from whole sperm cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 006 Regulation by Ca2+ is a hallmark behavior of Slo1 , but not Slo3 channels . Our results indicate that intracellular Ca2+ , and not pH , is a driving force for the opening of the human KSper channel and suggest that the molecular identity of hKSper might be the Slo1 protein rather than Slo3 . To verify the molecular identity of human KSper , we applied three of the known Slo1 channel blockers to the bath solution: charybdotoxin ( ChTX ) ( Anderson et al . , 1988; MacKinnon and Miller , 1988 ) , iberiotoxin ( IbTX ) ( Galvez et al . , 1990; Candia et al . , 1992; Giangiacomo et al . , 1992 ) , and paxilline ( Knaus et al . , 1994c; Sanchez and McManus , 1996; Zhou et al . , 2010 ) . Figure 5A , B shows a potent and reversible inhibition of human IKSper by 93% in the presence of 1 μM ChTX . Human K+ currents were also effectively blocked by both 100 nM IbTX ( Figure 6A , B ) and 100 nM paxilline ( Figure 7A , B ) with 87% and 62% inhibition , respectively . To verify that this pharmacological profile was specific to Slo1 , we also recorded K+ currents from mouse sperm , which express Slo3 ( Schreiber et al . , 1998; Zhang et al . , 2006a; Navarro et al . , 2007; Santi et al . , 2010; Zeng et al . , 2011 ) . It was previously reported that mouse Slo3 is insensitive to ChTX , IbTX , and paxilline ( Tang et al . , 2010 ) , and indeed Figures 5C , D , 6C , D , and 7C , D demonstrate that mouse K+ currents remained unaffected upon stimulation with 1 μM ChTX , 100 nM IbTX , or 500 nM paxilline . The fact that human , but not mouse , KSper is sensitive to Slo1-specific channel blockers strongly suggests that Slo1 forms the potassium channel in human sperm . 10 . 7554/eLife . 01009 . 007Figure 5 . Human , but not mouse KSper is sensitive to the Slo1 channel blocker charybdotoxin ( ChTX ) . ( A ) Representative human IKSper traces under control conditions ( black ) and in the presence of 1 μM ChTX ( red ) elicited in response to the given voltage ramp . ( B ) Mean current densities ( CDs ) ± SEM calculated at +120 mV . CDs ( human ) were: 65 ± 10 pA/pF ( control ) , 5 ± 1 pA/pF ( ChTX ) , and 49 ± 5 pA/pF ( washout ) . ( C ) Representative mouse IKSper traces under control conditions ( black ) and in the presence of 1 μM ChTX ( red ) elicited in response to the voltage ramp as shown in ( A ) . ( D ) CDs ( mouse ) were: 205 ± 8 pA/pF ( control ) vs 196 ± 8 pA/pF ( ChTX ) . ( n ) , number of experiments . Three human and three mouse sperm cells were used . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 00710 . 7554/eLife . 01009 . 008Figure 6 . Human , but not mouse KSper is sensitive to the Slo1 channel blocker iberiotoxin ( IbTX ) . ( A ) Representative human IKSper traces under control conditions ( black ) and in the presence of 100 nM IbTX ( red ) elicited in response to the shown voltage ramp . ( B ) Mean current densities ( CDs ) ± SEM calculated at +120 mV . CDs ( human ) were 74 ± 8 pA/pF ( control ) , 9 ± 1 pA/pF ( IbTX ) , and 70 ± 19 pA/pF ( washout ) . ( C ) Representative mouse IKSper traces under control conditions ( black ) and in the presence of 100 nM IbTX ( red ) elicited in response to the voltage ramp as in ( A ) . ( D ) CDs ( mouse ) were 244 ± 11 pA/pF ( control ) and 235 ± 4 pA/pF ( IbTX ) . ( n ) , number of experiments . Four human and three mouse sperm cells were used . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 00810 . 7554/eLife . 01009 . 009Figure 7 . Human , but not mouse KSper is sensitive to the Slo1 channel blocker paxilline ( Pax ) . ( A ) Representative human IKSper traces under control conditions and in the presence of paxilline elicited in response to the indicated voltage ramp . ( B ) Mean current densities ( CDs ) ± SEM calculated at +120 mV . Cells from three donors were used . CDs ( human ) were: 53 ± 4 pA ( control ) , 20 ± 2 pA ( 100 nM paxilline ) , and 69 ± 6 pA ( washout ) . ( C ) Representative mouse IKSper traces under control conditions and in the presence of paxilline elicited in response to the voltage ramp as in ( A ) . ( D ) CDs ( mouse ) were: 119 ± 5 pA/pF ( control ) and 113 ± 10 pA/pF ( 500 nM paxilline ) . ( n ) , number of experiments . Four human and three mouse sperm cells were used . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 009 Mouse KSper appeared to have both notably larger current amplitudes and current densities ( Figures 5–7 ) . Mouse spermatozoa are twice larger than human sperm cells: human sperm capacitance is usually within 1 pF , while the capacitance of mouse sperm is about 2 . 5 pF ( Kirichok et al . , 2006; Lishko et al . , 2010 , 2011 ) . However , the fact that KSper current densities ( pA/pF ) are still larger in mouse sperm than in human spermatozoa indicates the potential differences in KSper expression and distribution along the sperm flagella . We and others previously have shown that the sperm-specific calcium channel CatSper is activated by progesterone ( Lishko et al . , 2011; Strunker et al . , 2011 ) . Progesterone ( P ) shifts CatSper activation to more physiological , hyperpolarized , membrane potentials ( Lishko et al . , 2011 ) . To test whether progesterone has any effect onto hKSper , we recorded IKSper in the presence of different progesterone concentrations in the bath solution . Figure 8A , C shows that hKSper outward currents were blocked by progesterone in a dose-dependent manner . We have determined that progesterone’s half-maximum inhibitory concentration ( IC50 ) for hKSper is 7 . 5 ± 1 . 3 μM ( Figure 8B ) . Moreover , mouse Slo3 turned out to be insensitive to 10 µM of progesterone ( Figure 9 ) , which is above the IC50 for human KSper . Since potassium channels are well known to regulate the membrane potential ( Navarro et al . , 2007 ) , it is likely that the inhibition of human KSper by P will produce membrane depolarization and create favorable conditions for opening of CatSper . CatSper activation , in turn , will result in an elevation of intracellular [Ca2+] and trigger hyperactivated motility . To test this hypothesis we selectively blocked human KSper by adding 100 nM ChTX to high saline ( HS ) bath solution in which sperm cells are usually kept , and recorded any changes in sperm motility . As evident from Video 1 sperm motility was symmetrical in the absence of the Slo1 inhibitor ( ChTX ) . However , incubation in 100 nM ChTX for 25 min resulted in sperm cells exhibiting an asymmetrical motility pattern similar to hyperactivation ( Video 2 ) . The normal , symmetrical motility was resumed after a prolonged washout ( data not shown ) . 10 . 7554/eLife . 01009 . 010Figure 8 . hKSper is blocked dose-dependently by progesterone ( P ) . ( A ) Representative IKSper recordings from two sperm cells ( left and right panel ) in response to the given voltage ramp protocol under control conditions ( black ) , 0 . 5 μM P ( gray ) , 5 μM P ( magenta ) , 10 μM P ( red ) , and 30 μM P ( blue ) . ( B ) Dose-dependent inhibition of human IKSper by progesterone . Human IKSper amplitudes were acquired at +80 mV at the end of the voltage ramps , as shown in ( A ) . Current amplitudes in the presence of indicated progesterone concentrations were normalized onto control amplitudes ( in the absence of progesterone ) . Remaining IKSper in the presence of 0 . 5 μM , 5 μM , 10 μM and 30 μM of P was: 90 ± 2% , 60 ± 4% , 47 ± 3% and 11 ± 1% , respectively . Data were fitted with the Hill equation . Data shown are means ± SEM of 4–10 sperm cells from three different donors . ( C ) Representative IKSper traces elicited by the given voltage step protocol of the control ( left panel ) and in the presence of 10 μM P ( middle panel ) and 30 μM P ( right panel ) . Recordings are from the same cell as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 01010 . 7554/eLife . 01009 . 011Figure 9 . Mouse KSper is insensitive to progesterone ( P ) . The left panel shows representative traces of mouse IKSper of the control ( black ) and in the presence of 10 μM P ( red ) . The right panel shows current densities ( CDs ) acquired at +120 mV presented as mean ± SEM . CDs were: 119 ± 5 pA/pF ( control ) and 118 ± 8 pA/pF ( 10 μM P ) . ( n ) , number of experiments . Three sperm cells were used . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 01110 . 7554/eLife . 01009 . 012Video 1 . Inhibition of hKSper induces a hyperactivation- like motility pattern . Normal motility of human spermatozoa in the control HS solution . Scale bar is 5 mm . Recording was slowed down five times . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 01210 . 7554/eLife . 01009 . 013Video 2 . Inhibition of hKSper induces a hyperactivation- like motility pattern . Motility of human spermatozoa is altered after incubation in HS solution , which contained 100 nM of charybdotoxin ( ChTX ) . Scale bar is 5 mm . Recording was slowed down five times . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 013 To confirm that the Slo1 protein is actually present in human spermatozoa , we performed immunostaining with anti-Slo1 specific antibodies . Figure 10A demonstrates that the antibody selectively stained the principal piece of the sperm flagellum , the same compartment where other sperm ion channels , such as Hv1 and CatSper , reside . The head and the flagellar midpiece region showed no signals ( Figure 10A , middle and left panel ) . Furthermore , the presence of the Slo1 protein was confirmed by Western blotting ( Figure 10B ) . Immunoreactive bands in the range of 110–130 kDa were detectable in human spermatozoa and in mouse brain , which served as the positive control . 10 . 7554/eLife . 01009 . 014Figure 10 . Slo1 protein is present in human spermatozoa . ( A ) Human sperm immunostaining with primary polyclonal anti-Slo1 antibodies and Cy3-conjugated secondary antibodies . Left and middle panels show Slo1 staining localized to the principal piece of human sperm flagellum . Left panel: nuclei are stained by DAPI . Right panel: DIC image of the same cells . Scale bar is 5 mm . ( B ) Representative immunoblot of the mouse brain ( positive control ) and human spermatozoa from two different donors ( donor 1 and donor 2: D1 and D2 , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 014 We also tested the presence of Slo1 transcripts in human sperm cells . Indeed , both Slo1 α ( kcnma1 ) and Slo1 β3 ( kcnmb3 ) transcripts were amplified from the total RNA isolated from human ejaculated sperm ( Figure 11 ) . Interestingly , Slo1 was shown to have a decreased sensitivity to ChTX in the complex with different auxiliary subunits ( Xia et al . , 1999 ) . For example , the presence of the β3 subunit in the Slo1 complex requires micromolar , but not nanomolar concentrations of ChTX to completely inhibit channel activity ( Xia et al . , 1999 ) . Indeed , the β3 subunit of Slo1 was shown to be expressed in testis ( Uebele et al . , 2000 ) , and we also found transcripts of β3 ( kcnmb3 ) from sperm RNA ( Figure 11 , right panel ) . Therefore , a reduced sensitivity of human KSper to ChTX ( nearly complete inhibition of activity was achieved only with 1 μM ChTX ) is likely due to the presence of the β3 auxiliary subunit in human sperm . 10 . 7554/eLife . 01009 . 015Figure 11 . Slo1 transcripts are present in human spermatozoa . PCR bands of the portion of the translated region of kcnma1 ( left panel; 1433–3554 bp , corresponding to the coding sequence of splice isoform1; UniProt # Q12791 ) , and of the translated region of kcnmb3 ( right panel; 529–829 bp of the coding sequence of splice isoform 3d , Uniprot # Q9NPA1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 015
Potassium channels are indispensable for sperm physiology and are essential for membrane hyperpolarization upon sperm capacitation—the final sperm maturation in the female reproductive tract . Capacitation is also associated with intracellular alkalinization , which , in turn , has been shown to activate the calcium channel CatSper in human spermatozoa and the potassium channel KSper in murine sperm ( Zeng et al . , 1995; Arnoult et al . , 1996; Kirichok et al . , 2006; Navarro et al . , 2007; Lishko et al . , 2011 ) . In mouse sperm , K+ currents originate mainly from the Slo3 channel , which is alkalinization-activated , calcium-insensitive potassium channel ( Schreiber et al . , 1998; Zhang et al . , 2006a , 2006b; Navarro et al . , 2007; Santi et al . , 2010; Yang et al . , 2011; Zeng et al . , 2011 ) . The currents we recorded from human sperm , however , show very different properties . According to our data , KSper currents recorded either from human epididymal or ejaculated spermatozoa were alkalization-independent producing the same current amplitudes at pHi = 5 . 5 and 7 . 4 . However , we found that human KSper instead is sensitive to intracellular calcium . Capacitation also results in the elevation of intracellular calcium ( Visconti et al . , 2002 ) . The C-terminus of Slo1 potassium channel possesses RCK domains with high-affinity Ca2+ binding sites ( Moss et al . , 1996; Schreiber and Salkoff , 1997; Jiang et al . , 2001; Yuan et al . , 2010 ) . This raises the possibility that Slo1 may represent human KSper . Associated γ- ( leucine-rich repeat-containing proteins , LRRCs ) and β-subunits further modulate channel behavior in response to calcium . Subunits γ1 ( LRRC26 ) , γ2 ( LRRC52 ) , γ3 ( LRRC55 ) , and γ4 ( LRRC38 ) produce a shift towards more hyperpolarized membrane potentials , even in the absence of intracellular calcium and transcripts for all four subunits are detectable in human testis ( Yan and Aldrich , 2010 , 2012; Yang et al . , 2011 ) . So far , four β subunits ( β1–4 ) have been identified and are expressed in a tissue-specific manner . Subunits β2–4 are mainly expressed in brain and neurons , β3 is also detectable in testis , whereas the β1 subunit is preferentially found in smooth muscle cells ( Knaus et al . , 1994b; Behrens et al . , 2000; Brenner et al . , 2000; Uebele et al . , 2000 ) . hSlo1 activation time is reduced when the α subunit is co-expressed with β3 ( Brenner et al . , 2000 ) . Inward currents with increased concentrations of calcium ( 10 , 60 , and 300 μM ) occur when the α-subunit is expressed alone and are potentiated in the presence of subunit β1 and β3 ( Xia et al . , 2000 ) . Keeping the [Ca2+]i = 10 μM , an inward current becomes apparent when Slo1 α is co-expressed with β1 and β4 , whereas β2 and β3 show no effect ( Brenner et al . , 2000; Lippiat et al . , 2003 ) . These data can be explained by the presence of β3 splice variants ( β3a–d ) arising from four different exons with each of them encoding for an alternative N terminus ( Zeng et al . , 2008 ) . One study so far showed that β3b , β3c , and β3d transcripts are present in human testis , with β3d showing the greatest expression ( Uebele et al . , 2000 ) . It is possible that IKSper of human spermatozoa originates from Slo1 α-subunits , which are in the complex with γ , β , or even both auxiliary subunits . Indeed , according to our data , β3 transcripts are present in the RNA pool isolated from mature ejaculated human sperm cells . Moreover , our data show that elevated intracellular calcium strongly potentiates outward current and results in the appearance of an inward current . These hKSper properties ( calcium sensitivity and pH-insensitivity ) favor the idea that Slo1 represents the potassium channel in human spermatozoa . Slo1 channels are potently blocked by the scorpion peptide toxins charybdotoxin ( ChTX ) ( Miller et al . , 1985; Anderson et al . , 1988; MacKinnon and Miller , 1988 ) and iberiotoxin ( IbTX ) ( Galvez et al . , 1990; Candia et al . , 1992; Giangiacomo et al . , 1992 ) . However , there is evidence that all four β-subunits can confer resistance to these toxins . It has been reported that IbTX effectively blocks recombinant hSloα with an IC50 of 33 nM , but IbTX inhibition of recombinant hSloα in a complex with β1 raises IC50 to 371 nM ( Lippiat et al . , 2003 ) . Also , subunit β2 greatly reduces the sensitivity of the α subunit to ChTX ( IC50 = 1 nM vs 58 nM ) ( Wallner et al . , 1999 ) . Another study revealed that 20 nM ChTX was sufficient to block recombinant hSloα , whereas 100 nM of toxin was required to inhibit hSloα + β1 . Moreover , even 100 nM of ChTX was insufficient to effectively block hSloα + β3 ( Xia et al . , 1999 ) . Furthermore , slower blocking kinetics for ChTX and IbTX have been shown in hSloα + β4 constructs ( Meera et al . , 2000 ) . Northern blot analyses demonstrate that mRNA for subunits β3 and β4 is detectable in human testis ( Brenner et al . , 2000 ) , and it has been shown that the resistance to IbTX and ChTX is determined by the large extracellular loop of the β4 subunit ( Meera et al . , 2000 ) . Since we observed different blocking kinetics with ChTX ( 93% reduction with 1 μM ChTX ) and IbTX ( 87% reduction with 100 nM IbTX ) , it seems likely that in human sperm , the Slo1 channel is associated with β subunits that modulate channel behavior in response to these toxins . Paxilline , a fungal indole alkaloid , has also been shown to inhibit Slo1 channels ( Knaus et al . , 1994c; Sanchez and McManus , 1996 ) . Interestingly , Slo3 is paxilline insensitive , and recently it has been demonstrated that paxilline binds to a glycine residue at position 311 within the S6 segment of the α-subunit ( Zhou et al . , 2010 ) . This effect is independent of auxiliary subunits . However , the same study elucidated other factors within the Slo channel structure , which seem to be important for paxilline block . First , the turret region could determine the effectiveness of paxilline block . The turret is the extracellular loop between S5 and the pore domain , which contains more residues in Slo channels as compared to other K+ channels ( Carvacho et al . , 2008; Giangiacomo et al . , 2008; Latorre et al . , 2010 ) . Replacing the first half of the mSlo1 pore loop with the corresponding mSlo3 sequence leads to a five times greater paxilline inhibition and increases inhibition and washout rates ( Zhou et al . , 2010 ) . A second source for altered abilities of paxilline inhibition is the pore loop region in the S6 segment , which differs in 10 residues between Slo1 and Slo3 . When we applied 100 nM paxilline to human spermatozoa , only 62% of hKSper reduction was observed , whereas even a fivefold higher concentration did not affect currents recorded from mouse sperm . From these data , we conclude that in human spermatozoa , either the turret region or the S6 segment of BK channels show properties that do not allow a complete block of the K+ currents by paxilline . We and others previously have shown that progesterone is a potent non-genomic activator of CatSper with an EC50 of 8 nM ( Lishko et al . , 2011; Strunker et al . , 2011 ) . But as apparent from this study , progesterone also blocks human KSper with an IC50 of around 8 μM . Moreover , murine KSper is not affected by 10 μM progesterone . Together , the data from steroid and toxin treatment indicate that pharmacological properties of human and mouse KSper channels are quite different . In conclusion , we show that human IKSper originated from the sperm flagellum , the same compartment where also CatSper and Hv1 channels reside ( Lishko et al . , 2010 , 2011 ) . Human KSper is a pH-independent , calcium-sensitive potassium channel sensitive to selective Slo1 inhibitors , such as charybdotoxin , iberiotoxin and paxilline , and is inhibited by micromolar concentrations of progesterone . Apart from its localization in sperm flagella , mouse KSper lacks all earlier-mentioned properties . Taken together , these results indicate that the human sperm potassium channel comprises the Slo1 protein and not Slo3 . In addition , we propose the following model: the functional proximity of KSper to other sperm ion channels helps temporally coordinate their actions in a concerted manner during capacitation ( Figure 12 ) . In the uterus and the Fallopian tube , intracellular alkalinization is evoked by Hv1 , thus activating CatSper channels . However , CatSper will not be fully active , as hKSper channels function as feedback regulators in response to calcium influx , thus retaining the membrane potential in a hyperpolarized state . In close proximity to the oocyte however , sperm encounter high concentrations of progesterone , which , in turn , will block hKSper , leading to membrane depolarization opening CatSper channels , which will become fully potentiated by the presence of progesterone . These events will lead to elevated levels of intracellular calcium in sperm , thereby initiating calcium-dependent processes such as hyperactivity and the acrosome reaction making the fertilization event possible . 10 . 7554/eLife . 01009 . 016Figure 12 . Role of human KSper ( Slo1 ) in sperm physiology . In the uterus and fallopian tube , CatSper is partially activated due to the intracellular alkalinization evoked by proton extrusion through Hv1 and picomolar- to nanomolar progesterone ( P ) concentrations . However , to achieve full activation of CatSper , flagellar plasma membrane must be depolarized . This is achieved by the inhibition of sperm KSper , the channel responsible for membrane hyperpolarization . In close proximity to the oocyte , spermatozoa encounter micromolar concentrations of P , which inhibit hKSper , resulting in membrane depolarization . These events allow full activation of CatSper , trigger sperm hyperactivation , allow spermatozoa to penetrate through the egg protective vestment , and make fertilization possible . DOI: http://dx . doi . org/10 . 7554/eLife . 01009 . 016
A total of 19 healthy fertile volunteers aged 21−38 years were recruited for this study . The study was conducted with approval of the Committee on Human Research at the University of California , Berkeley ( protocol 10-01747 , IRB reliance #151 ) , and University of California , San Francisco ( protocol 10-04868 ) . Informed consent was obtained from all participants . Ejaculates were obtained by masturbation and spermatozoa were purified following the swim-up protocol as previously described ( Lishko et al . , 2011 ) . Men with proven fertility who were undergoing sperm retrieval procedures or a vasectomy reversal in the UCSF Center for Reproductive Health were also included in this study . As part of the ongoing IRB-approved LIFE ( Lifestyle , Fertility , and Evaluation ) study , men who agreed to participate donated portions of surgical specimens . All men enrolled in the present study had a documented history of prior paternity and had undergone a vasectomy in the past . As part of routine clinical care , these men elected to undergo a sperm retrieval procedure ( microscopic epididymal sperm aspiration , MESA , or percutaneous epididymal sperm aspiration , PESA ) combined with in vitro fertilization ( IVF ) or a vasectomy reversal . An aliquot of epididymal fluid was used for the present study with patient consent . Male C57BL/6 mice were purchased from Harlan Laboratories ( Livermore , CA ) and were kept in the Animal Facility of the University of California , Berkeley . All experiments were performed in strict accordance with the NIH Guidelines for Animal Research and approved by UC Berkeley Animal Care and Use Committee , the approved protocol MAUP #R352-012 . Animals were killed by CO2 asphyxiation and cervical dislocation , and sperm were collected as described previously ( Wennemuth et al . , 2003 ) . Progesterone was purchased from CalBiochem ( EMD Millipore , Darmstadt , Germany ) , charybdotoxin and iberiotoxin from Tocris Bioscience ( Bristol , UK ) , and all other compounds were obtained from Sigma ( St . Louis , MO , USA ) . Gigaohm seals were formed at the cytoplasmic droplet ( Cooper , 2011 ) of highly motile cells or separated flagella in standard high saline ( HS ) buffer containing ( in mM ) 130 NaCl , 20 HEPES , 10 lactic acid , 5 glucose , 5 KCl , 2 CaCl2 , 1 MgSO4 , 1 sodium pyruvate , pH 7 . 4 adjusted with NaOH , 320 mOsm/l as reported in Lishko et al . ( 2010 , 2013 ) . Transition into whole-cell mode was achieved by applying voltage pulses ( 499–611 mV , 1 ms ) and simultaneous suction . Cells were stimulated every 5 s , data were sampled at 10 kHz and filtered at 1 kHz and access resistance was 21–57 MΩ . Pipettes ( 13–16 MΩ ) were filled with 130 mM KMeSO3 , 20 mM HEPES , 4 mM KCl , 10 mM EGTA , 1 mM EDTA , and pH 7 . 4 was adjusted with KOH , 330 mOsm/l . In the experiments with NH4Cl , pipette solutions were of similar composition , but contained just 5 mM HEPES to allow efficient intracellular pH changes . The nominal free bath solution ( NMF ) consisted of ( in mM ) 140 KMeSO3 , 20 HEPES , and pH 7 . 4 was adjusted with KOH , 320 mOsm/l . To inhibit monovalent currents through CatSper channels ( Smith et al . , 2013 ) , 0 . 1–1 mM Ca2+ was added to the NMF solution , as indicated . To elicit potassium currents through CatSper , currents were recorded in a K+-based divalent free bath solution ( K-DVF ) containing ( in mM ) 140 KMeSO3 , 45 HEPES , 1 EDTA , 7 . 4 adjusted with KOH , 320 mOsm/l . Inside ( pipette ) solutions with different concentrations of free Ca2+ contained ( in mM ) 130 KMeSO3 , 20 HEPES , 4 KCl and 1 BAPTA , 1 EDTA , 1 EGTA ( for 100 nM Ca2+ ) or 1 HEDTA ( for 50 μM Ca2+ ) , respectively . CaCl2 was added according to WinMAXC32 version 2 . 51 ( Chris Patton , Stanford University ) . Since changing of the pipette solution cannot be easily done on one cell , the data obtained with different intracellular pH or different intracellular [Ca2+] are a combination of recordings from multiple cells . However , since the changing of bath solution can be easily accomplished on the same cell , the experiments with different bath solutions ( addition of EDTA , extracellular calcium , NH4Cl , ChTX , IbTX , Paxilline , progesterone , etc ) were performed on the same sperm cell ( flagellum ) : before and after addition of the above-mentioned compound . Data were analyzed with Clampfit 10 . 3 ( Molecular Devices , Sunnyvale , CA , USA ) and OriginPro 8 . 6 ( OriginLab Corp . , Northampton , MA , USA ) . Statistical data are presented as mean ± standard error of the mean ( SEM ) , and n indicates number of experiments . All electrophysiology experiments were performed at 24°C . Cells were seeded onto cover slips in HS solution and allowed to adhere for 30 min at room temperature ( RT ) . Cells were fixed with ice cold methanol for 1 min , washed in PBS , and subsequently permeabilized with PBS/0 . 1% Triton ( PBS-T ) with 5% BSA for 1 hr at room temperature . Incubation with the primary antibody ( rabbit anti-Maxi K+ alpha , 1:100 in PBS-T and BSA , Thermo Scientific , # PA1-923 ) was performed at 4°C overnight . Cells were then washed in PBS-T and incubated with Cy3 conjugated goat anti-rabbit IgG for 45 min at room temperature . After washing , the samples were mounted with ProLong Gold antifade reagent . Images were taken on a Zeiss LSM 710 microscope ( Carl Zeiss Microscopy , Oberkochen , Germany ) and processed with the Zeiss ZEN 2010B imaging software . Purified human spermatozoa were centrifuged and re-suspended in lysis buffer containing 0 . 1% SDS , 0 . 5% sodium deoxycholate , 50 mM dithiothreitol ( DTT ) , 20 mM EDTA , 4 M urea and protease inhibitor cocktail ( Roche ) . After sonication for 5 min , the cell suspension was mixed 1:0 . 5 with water , then mixed 1:1 with sample buffer , and sonicated again for 5 min β-mercaptoethanol ( 5% ) was added to each sample and after boiling ( 5 min , 100°C ) samples were transferred to a 4–12% polyacryalamide gel and blotted on PVDF membranes . Membranes were fixed with methanol followed by blocking with 3% IgG-free BSA in PBS containing 0 . 1% Tween ( PBS-T ) for 30 min at room temperature . Subsequently , membranes were incubated with 1 μg of mAb anti-Slo1 ( clone L6/60 ) ( UC Davis/NIH NeuroMab Facility , Davis , CA , USA ) overnight at 4°C . After washing three times with PBS-T , the membranes were incubated with HRP-conjugated anti-mouse IgG ( 1:20 , 000 dilution ) for 1 hr at room temperature . Protein bands were detected by enhanced chemiluminescence on a Fluor Chem M imaging system ( Protein Simple ) . Spermatozoa were purified by swim-up procedure and total donor-specific RNA was extracted from purified spermatozoa using a Qiagen RNeasy mini kit followed by cDNA synthesis with a Phusion RT-PCR kit ( Finnzymes , MA , USA ) . The donor- specific translated regions of kcnma1 between 1433 bp and 3554 bp ( corresponding to the canonical coding sequence of Slo1 α isoform1; UniProt id Q12791 ) and kcnmb3 between 529 bp and 829 bp of the canonical coding sequence ( Slo1β isoform 3d , Uniprot id Q9NPA1 ) were amplified using the following primers: 5′-ATGCCTCGAATATCATGAGAG-3′ ( kcnma1 , forward ) , 5′-TATATTGGTTGATCTGGTTAGCC-3′ ( kcnma1 , reverse ) ; 5′-CTCGCCTAGGTTCTTCGATCACAAAAATGG-3′ ( kcnmb3 , forward ) , and a reverse 5′-ATCGCTCGAGCTGCTCTTCCTTTGCTCCT-3′ ( kcnmb3 , reverse ) . All PCRs were carried out for 40 cycles of replication and had annealing temperatures of 61°C . The obtained PCR products were gel-purified and sequence-verified ( Sequetech , Mountain View , CA , USA ) . Purified spermatozoa were plated onto 5-mm coverslips in HS solution . Sperm movement was recorded within the first 3 hr after sperm retrieval with a high speed GX-1 Memrecam camera ( NAC Image Technology ) attached to an Olympus IX71 microscope ( Olympus Corp . , Central Valley , PA , USA ) . The recording speed was 960 frames per second ( fps ) , and videos were slowed down to playback at 200 fps where indicated . | The sperm cells that are released into the female reproductive tract when a mammal ejaculates , are not capable of fertilizing an egg right away , so they must go through a process called maturation . The early stages of this process involve interactions with the seminal fluid that increase the motility of the sperm cells , and the latter stages involve interactions with the walls of the reproductive tract and vaginal secretions to ensure that the sperm cells move toward the egg . Many of these interactions involve positive ions entering and leaving the sperm cells via ion channels . The properties of the ion channels that allow protons and calcium ions to move into and out of human sperm cells are well understood , but little is known about the channels that control the movement of the potassium ( K ) ions . At first it was assumed that the molecular structure of these channels was similar to that of the Slo3 potassium channel in mouse sperm , but crucial differences between human and mouse sperm cells have been reported in recent years . Now Mannowetz et al . have shown that the potassium channel in human sperm is opened by increased levels of calcium ions inside the sperm cells . Moreover , the pH inside the sperm cells had no influence on this process . Furthermore , the channel was blocked by three toxins that have no effect on the Slo3 potassium channels in mice , but are known to block a type of potassium channel known as Slo1 . Mannowetz et al . then used a technique called Western blotting to confirm the presence of Slo1 potassium channels in the tails of human sperm cells . Mannowetz et al . also showed that the Slo1 potassium channel can be blocked by the female hormone progesterone . This is important because blocking the potassium channels causes the calcium ion channels in the cells to open fully , and the resulting influx of calcium ions triggers a process called sperm hyperactivation that makes it possible for the sperm cell to fertilize the egg . By clearly showing the fundamental differences between human sperm cells and mouse sperm cells , this work stresses the need to exercise caution in using mice as a model of male fertility in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Slo1 is the principal potassium channel of human spermatozoa |
All organisms live within a given thermal range , but little is known about the mechanisms setting the limits of this range . We uncovered cellular features exhibiting signature changes at thermal limits in Caenorhabditis elegans embryos . These included changes in embryo size and shape , which were also observed in Caenorhabditis briggsae , indicating evolutionary conservation . We hypothesized that such changes could reflect restricted aerobic capacity at thermal limits . Accordingly , we uncovered that relative respiration in C . elegans embryos decreases at the thermal limits as compared to within the thermal range . Furthermore , by compromising components of the respiratory chain , we demonstrated that the reliance on aerobic metabolism is reduced at thermal limits . Moreover , embryos thus compromised exhibited signature changes in size and shape already within the thermal range . We conclude that restricted aerobic metabolism at the thermal limits contributes to setting the thermal range in a metazoan organism .
All organisms live within a given thermal range , beyond which growth and fecundity decrease ( Pörtner et al . , 2006 ) . Partly as a result , organisms tend to distribute in the ocean and on land according to latitude as well as depth and altitude , although other elements such as availability of food and light also play a role in shaping preferred habitats ( Pörtner , 2002; Pörtner et al . , 2006; Prasad et al . , 2011 ) . Despite their importance , the mechanisms that set the thermal limits remain incompletely understood . A mismatch between oxygen supply and demand has been suggested to play a role in setting thermal limits in multicellular organisms . This hypothesis , referred to as the oxygen- and capacity-limited thermal tolerance ( OCLTT ) , derives in part from the observation that oxygen partial pressure in aquatic organisms is constant within a given thermal range and decreases both below the lower thermal limit and above the upper thermal limit ( Pörtner , 2002; Pörtner et al . , 2006 ) . In agreement with this hypothesis , the metabolic status of some aquatic organisms has been shown to peak at a given temperature and to decrease both below and above that ( Melzner et al . , 2006; Wittmann et al . , 2008 ) . Interestingly too , tolerance to high temperatures is increased in an amphibian crab when the animal is in the air compared to when it is in water , reflecting the reduced cost of oxygen supply in air ( Giomi et al . , 2014 ) , again supporting the OCLTT hypothesis . Overall , these data suggest that thermal limits in complex organisms are characterized by a mismatch in oxygen supply and demand , which would result in reduced energy production and thus limit reproduction and growth ( Pörtner , 2002; Pörtner et al . , 2006 ) . Intriguingly , the temperature-dependence of oxygen diffusion is significantly lower than that of metabolism ( Woods , 1999 ) , raising the question of how oxygen supply and demand can be matched , even within the thermal range . One possible solution is suggested by the observation that body size decreases with augmented temperature in the vast majority of ectotherms ( ‘temperature-size rule’ ) ( Atkinson , 1994; Forster et al . , 2012 ) , thereby increasing surface to volume ratio and thus potentially oxygen availability . In support of this , the slope of this ‘temperature-size rule’ is steeper for aquatic organisms than terrestrial organisms , in agreement with the lower availability of oxygen in water compared to air ( Forster et al . , 2012 ) . This has led to the suggestion that alterations in cell size in response to changes in temperature within the thermal range are adaptive responses to preserve aerobic capacity , which has been dubbed the MASROS hypothesis ( Maintain Aerobic Scope—Regulate Oxygen Supply ) ( Atkinson et al . , 2006 ) . What happens beyond the thermal limits within this conceptual framework ? One might expect that thermal limits could be characterized by further changes in cell size and potentially also cell shape , in an attempt to increase the available surface area and thus maximize oxygen availability . Furthermore , the MASROS hypothesis predicts that aerobic metabolism , measured as respiration , should decrease beyond both the lower and the upper thermal limit as compared to within the thermal range , once the organism can no longer compensate for the insufficient oxygen availability . To our knowledge , these central predictions of the MASROS hypothesis have not been challenged experimentally in an integrative fashion . Therefore , the extent to which restricted aerobic metabolism is a general principle characterizing thermal limits remains unclear .
We determined embryonic viability in a range of temperatures for Caenorhabditis elegans and Caenorhabditis briggsae and operationally defined the thermal limits as the upper and lower edges of the temperature range within which >90% of embryos hatched . We thus found that the thermal limits of C . elegans were of 12°C and 25°C ( Figure 1A ) , and those of C . briggsae of 14°C and 27°C ( Figure 1B ) , in line with the fact that C . briggsae usually lives in warmer climates than C . elegans ( Prasad et al . , 2011 ) . The thermal range defined by these upper and lower limits ensures robust propagation of the population and is narrower than merely the reproductive range for C . elegans ( 9°C–26°C [Anderson et al . , 2011] ) or C . briggsae ( 14°C–30°C [Anderson et al . , 2007; Prasad et al . , 2011] ) . 10 . 7554/eLife . 04810 . 003Figure 1 . Defining the thermal range and quantifications . ( A and B ) Progeny tests were performed on acclimated C . elegans worms from 7 . 5°C to 27°C and C . briggsae worms from 9°C to 30°C . Dotted line highlights 90% embryonic viability . Temperatures below 20°C exhibiting less than 90% viability are shown in cyan , temperatures above 20°C exhibiting less than 90% viability in magenta . Between panels A and B , we show the thermal range of each species . Error bars show SEM . ( C–F ) Stills from a time-lapse temperature-controlled DIC microscopy recording of a first-cell stage embryo at the indicated stages ( G–J ) Examples of feature quantification at the different cellular stages ( 24°C ) : female pronucleus speed ( G ) , pronuclei position during centration-rotation ( H ) , spindle pole oscillations ( I ) , as well as areas of the AB ( anterior ) and P1 ( posterior ) daughter cells ( J ) . See ‘Materials and methods’ for details on the quantifications . Figure 1—figure supplement 1 shows the temperature control setup . Figure 1—source data 1 lists all the quantified features and their thermal response within and beyond the thermal range . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 00310 . 7554/eLife . 04810 . 004Figure 1—source data 1 . Quantified features . List of features that were quantified and their thermal responses within and beyond the thermal range for C . elegans ( N2 ) . Within the thermal range , features were categorized as ‘temperature-dependent’ if the Pearson correlation p-value was below 0 . 0014 = 0 . 05/35 ( see ‘Materials and methods’ for Bonferroni correction; ‘temperature-independent’ is shown underlined ) . Beyond the thermal limit , we performed an F-test to determine if the thermal response of the feature was changing compared to within the thermal range ( see ‘Materials and methods’; we indicated a change in thermal response when the F-test p-value was below 0 . 0014 , highlighted in bold ) . Abbreviations: PC: pseudo-cleavage , PM: pronuclear meeting , ME: mitotic entry , T: temperature , C/R: centration-rotation , MT: microtubules . The following features were also quantified but displayed no consistent thermal response both within and beyond the thermal range and hence were not included in the table: anterior-most position at the end of C/R , number of anterior and posterior oscillations , spindle position at the onset of oscillations . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 00410 . 7554/eLife . 04810 . 005Figure 1—figure supplement 1 . Temperature-control setup . ( A ) The temperature within the sample was measured using a thin thermocouple of type K ( see ‘Materials and methods’ ) , connected to a temperature controller . This temperature feedback was used to control the temperature of the air blown on the sample and objective , so that sample and set temperatures always matched . ( B ) We verified that the temperature controller was well calibrated over the range of temperatures of interest by checking the temperature on the sample with yet another thermocouple connected to a separate external thermometer at the same time . We report the measured temperature on the sample by the temperature controller and external thermometer . The measured imprecision was of the order of 0 . 2°C at all temperatures . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 005 We reasoned that identifying cellular features that operate differently beyond the thermal limits defined above , as compared to within the thermal range , might reveal critical mechanisms acting at these limits . In order to systematically identify such limit-sensitive features , we first analyzed cellular processes within the thermal range . We conducted this analysis initially in C . elegans embryos , but then also studied embryos of C . briggsae , which has been estimated to have diverged from C . elegans 18–100 million years ago ( Stein et al . , 2003; Cutter , 2008 ) , thus probing evolutionary conservation of putative limit-sensitive features . Using temperature-controlled time-lapse DIC ( Differential Interference Contrast ) microscopy and semi-automated quantifications of the resulting movies with in-house scripts ( Figure 1C–J , Figure 1—figure supplement 1 , ‘Materials and methods’ , and Source code 1 ) , we measured 35 cellular features that describe the main events of the first cell cycle of C . elegans embryos ( Figure 1C–F ) . In brief , after fertilization , the female pronucleus migrates towards the male pronucleus ( Figure 1C ) . After their meeting , the pronuclei move to the embryo center whilst undergoing a 90°C rotation ( Figure 1D ) . The nuclear envelopes then break down , followed by assembly of the mitotic spindle , which moves slightly to the posterior during the remainder of mitosis whilst oscillating perpendicular to the anterior–posterior axis ( Figure 1E ) . This results in the asymmetric division of the one-cell stage embryo into a larger anterior cell and a smaller posterior one ( Figure 1F ) . Our analysis established that the vast majority of the monitored features were temperature-dependent within the thermal range ( Figure 1—source data 1 ) . Interestingly , some features , including the fraction of time spent in mitosis ( Figure 2B ) and cell division asymmetry ( Figure 2C ) , exhibited a temperature-independent behavior , suggesting that temperature-compensation mechanisms are also at play . 10 . 7554/eLife . 04810 . 006Figure 2 . C . elegans thermal responses . ( A ) Cell cycle duration as a function of temperature ( error bars show SEM ) . ( B ) Relative cell cycle duration as a function of temperature ( error bars show SEM ) . ( C ) Relative size of the AB blastomere as a function of temperature . Dotted line represents the average relative size within the thermal range ( 57 . 4% ) . ( D ) Maximum amplitude of posterior pole oscillations during anaphase . ( E ) Embryo size as a function of temperature . Dotted line shows a linear regression of the data within the thermal range ( white boxes ) . ( F ) Embryo shape , measured as the ratio of embryo length over embryo width , as a function of temperature . See main text for p-values . Color code ( for the whole figure ) : white bars show data within the thermal range . Colored bars show data below ( cyan ) and above ( magenta ) the thermal limit . Boxplots show median as well as 25th and 75th percentiles . Whiskers extend to the most extreme points not considered outliers ( i . e . , within 99 . 3% coverage ) . Note that the variance of cellular features does not increase beyond the thermal limits as compared to within the thermal range . Figure 2—figure supplement 1 depicts embryo size and shape at various temperatures . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 00610 . 7554/eLife . 04810 . 007Figure 2—figure supplement 1 . Embryo size and shape at various temperatures , exaggerating the actual differences for visualization purposes . We show a cartoon embryo below the lower thermal limit ( A ) , which has the same shape as the embryo within the thermal range ( B ) , that is , a/b is equal in both cases . Above the upper thermal limit ( C ) , the embryo elongates ( i . e . , a/b increased ) . In ( D ) , we show an overlay of the embryos in ( A–C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 007 We then were in a position to identify cellular features that might operate differently beyond the thermal limits compared to within the thermal range . We found that although some features exhibited the same thermal response as within the thermal range , others responded differently , suggesting that they were sensitive to the thermal limits ( Figure 1—source data 1 ) . Thus , the duration of mitosis , which decreased monotonically with increasing temperatures within the thermal range , plateaued beyond both lower and upper thermal limits in C . elegans ( Figure 2A ) . Moreover , although C . briggsae can develop at warmer temperatures than C . elegans ( Figure 1A–B ) ( Prasad et al . , 2011 ) , we found that cell cycle duration was not faster in C . briggsae than in C . elegans at any temperature ( Figure 3A , compare with Figure 2A ) . Interestingly , cell cycle duration within the thermal range was well described by Arrhenius kinetics in C . elegans ( 92% of explained variance; ‘Materials and methods’ ) ( Arrhenius , 1915 ) . In C . briggsae , by contrast , the data beyond 25°C reduced the explained variance from 86% to 39% , suggesting that cell cycle duration plateaued already below the upper thermal limit in this species , underscoring the fact that mitosis duration is a limit-sensitive feature . 10 . 7554/eLife . 04810 . 008Figure 3 . Thermal responses in C . briggsae . See legend of Figure 2 and main text for p-values . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 008 We also observed that the asymmetry of the first cell division in C . elegans , which was constant within the thermal range , decreased below 12°C ( F-test p = 0 . 0005 ) and increased above 25°C ( F-test p < 10−7 ) ( Figure 2C; see ‘Materials and methods’ for statistics ) . In C . briggsae , the asymmetry of the first cell division also increased beyond the upper thermal limit , at both 28°C and 29°C ( F-test p-value < 10−10 ) ( Figure 3C ) . However , a reciprocal decrease was not observed at the lower thermal limit in this species , perhaps because spindle pole oscillations are weaker in C . briggsae than in C . elegans ( Riche et al . , 2013 ) ( compare panel D in Figures 2 , 3 ) , potentially limiting the dynamic range over which asymmetry can be tuned . Our analysis also revealed interesting alterations in embryo geometry at the upper and at the lower thermal limits . Thus , embryo size was larger at the lower end of the thermal range ( i . e . , 12°C for C . elegans , Figure 2E; 14°C for C . briggsae , Figure 3E ) , and tended to decrease with increasing temperature within the thermal range . This is in line with the ‘temperature-size rule’ observed in the vast majority of ectotherms ( Atkinson , 1994; Forster et al . , 2012 ) , and in agreement with previously reported data for C . elegans at 10°C vs 20°C ( Van Voorhies , 1996 ) . Strikingly , below the lower thermal limit , embryo size was actually significantly reduced in both C . elegans and C . briggsae ( Figures 2E , 3E; F-test p < 10−7 and p = 0 . 001 , respectively ) . Such a reversal of the temperature size rule below the lower thermal limit has also been reported in protists and in Drosophila ( Karan et al . , 1998; Atkinson et al . , 2003 ) . These observations are compatible with the MASROS hypothesis , which posits that such a size decrease below the lower thermal limit may reflect cold-inhibited mitochondrial function ( Atkinson et al . , 2006 ) . Beyond the upper thermal limit , we observed a plateau in the size of both C . elegans and C . briggsae embryos ( Figures 2E–3E ) . Interestingly , however , we observed that embryos in both species were more elongated beyond the upper thermal limit ( Figures 2F–3F; F-test p = 0 . 0004 and F-test p < 10−10 , respectively ) . Such an elongation results in an increase of the surface area , thus potentially augmenting its availability for oxygen diffusion ( Figure 2—figure supplement 1 ) . Overall , these results reveal that changes in cell size and shape are signature hallmarks of the thermal limits . Do the observed changes in embryo size below the lower thermal limit and of shape above the upper thermal limit reflect an adaptation to reduced aerobic metabolism at those temperatures ? We set out to explore this possibility by determining the extent of respiration at different temperatures in wild-type C . elegans embryos . As shown in Figure 4A , we found that respiration increased exponentially within the thermal range , as predicted by Arrhenius-like kinetics ( Arrhenius , 1915 ) . Strikingly in addition , this analysis uncovered that respiration departed from Arrhenius-like kinetics both below the lower thermal limit ( F-test p-value < 10−4 ) and above the upper thermal limit ( F-test p-value < 10−10 ) , in support of reduced respiration at those temperatures ( Figure 4A ) . Although we do not know whether the observed relative reduction in aerobic capacity beyond both thermal limits as compared to within the thermal range contributes to increased lethality at those limits , our results show a clear correlation between these features . 10 . 7554/eLife . 04810 . 009Figure 4 . Restricted aerobic metabolism at the thermal limits . ( A ) Oxygen consumption ( y-axis displays the logarithm of O2 flow per volume ) in embryos at different temperatures from 9°C to 28°C . Pooled data from two biological replicates , each with two technical replicates ( see ‘Materials and methods’ ) . Error bars represent the SEM . Note that respiration increases exponentially between 12°C and 24°C ( white discs ) , as shown by the linear increase in log-scale ( gray dashed line shows exponential fit between 12°C and 24°C ) . Note also that respiration decreases beyond both thermal limits ( cyan and magenta discs , respectively ) , and no longer follows the exponential trend observed within the thermal range . ( B ) Color-code for panels ( B–F ) : white ( wild-type ) , blue ( atp-2 ( RNAi ) ) , orange ( cyc-1 ( RNAi ) ) , green ( nuo-1 ( RNAi ) ) . Progeny tests on atp-2 ( RNAi ) embryos . ( C ) Same as B for cyc-1 ( RNAi ) . ( D ) Same as B for nuo-1 ( RNAi ) . ( E ) Embryo size as a function of temperature . ( F ) Embryo shape as a function of temperature . See main text for p-values . Figure 4—figure supplement 1 shows the RNAi feeding times as a function of temperature . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 00910 . 7554/eLife . 04810 . 010Figure 4—figure supplement 1 . RNAi feeding time as a function of temperature . Reported durations of embryogenesis ( blue crosses ) were fitted by an Arrhenius-like model ( dashed green line ) ( Gillooly et al . , 2002 ) . The same activation energy was used to fit reported RNAi induction times ( black circles and fitted red solid line ) . The latter fit was used to extrapolate RNAi feeding times at the temperatures of interest . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 01010 . 7554/eLife . 04810 . 011Figure 4—figure supplement 2 . Progeny tests in air-1 ( RNAi ) . ( A ) air-1 ( RNAi ) is 100% embryonic lethal at 12°C , 20°C and 24°C , as anticipated ( Schumacher et al . , 1998; Hannak et al . , 2001 ) . ( B ) In order to titrate the phenotype , we performed double RNAi by mixing bacteria expressing dsRNA against air-1 with bacteria expressing dsRNA against gfp in a 1:3 ratio . We found that lethality is greater at 12°C and at 24°C than at 20°C , indicating that the results we uncovered when targeting mitochondrial respiratory chain components ( Figure 4B–D ) are not due to a general RNAi temperature-dependent response . DOI: http://dx . doi . org/10 . 7554/eLife . 04810 . 011 One possibility to interpret these data is that the energetic needs of the embryo are not satisfied beyond the thermal limits due to insufficient aerobic metabolism . Another possibility is that these needs are actually fulfilled to some extent despite reduced respiration , either because other metabolic routes are used to a larger relative extent or because embryos are metabolically depressed at the thermal limits and thus require less energy altogether . We reasoned that if aerobic metabolism became insufficient beyond the thermal limits , then further compromising mitochondrial activity should have more of an impact at the thermal limits than within the thermal range . By contrast , if energetic needs could be fulfilled at the least to some extent despite reduced respiration beyond the thermal limits , then further compromising mitochondrial activity should have less of an impact at the thermal limits than within the thermal range . Therefore , to distinguish between these two possibilities , we depleted three components of the mitochondrial respiratory chain using RNAi: the beta-subunit of ATP synthase ATP-2 ( Tsang et al . , 2001 ) , a complex V component , the subunit of the mitochondrial complex I NUO-1 ( Tsang et al . , 2001 ) , and the component of the mitochondrial complex III CYC-1 ( Dillin et al . , 2002 ) . We ascertained that embryonic respiration was reduced in cyc-1 ( RNAi ) embryos , reaching on average 56% ± 13% of the wild-type levels under the assay conditions ( t-test p-value < 10−3; see ‘Materials and methods’ ) . Since ATP-2 and NUO-1 are part of complex V and I , respectively , respiration may still occur upon their depletion , even if the mitochondrial respiratory chain is compromised ( Braeckman et al . , 2009 ) , so that respiration measurements may not have been telling in these cases . Importantly , we found that all three RNAi conditions were embryonic lethal to some extent ( Figure 4B–D ) , probably owing to decreased energy production through respiration , although we cannot exclude that the observed lethality stems from changes in pH or increased reactive oxygen species . Importantly , in addition , this analysis uncovered that embryonic lethality was reduced towards the lower thermal limit as compared to within the thermal range in all three cases , as well as towards the upper thermal limit in both cyc-1 ( RNAi ) and nuo-1 ( RNAi ) ( Figure 4B–D ) . These results offer strong experimental support to the notion that the capacity of the mitochondrial respiratory chain is restricted beyond both thermal limits , and raise the possibility that other metabolic routes are used to a larger relative extent at those temperatures in the face of reduced respiration . Following up on this result , we set out to test whether the changes in size or shape observed beyond the thermal limits in the wild-type reflect an adaptation response to restricted aerobic capacity . We reasoned that if this were the case , then such changes should occur already within the thermal range of embryos in which components of the mitochondrial respiratory chain are compromised . Interestingly , we found that whereas embryo size was not significantly affected upon RNAi-mediated depletion of atp-2 ( Figure 4E ) , these embryos were more elongated at both 12°C and 16°C ( Figure 4F , U-test p ( 12°C ) = 0 . 0034 , p ( 16°C ) = 0 . 0039 ) . In nuo-1 ( RNAi ) , embryo size was significantly reduced at both temperatures ( U-test p ( 12°C ) = 0 . 02 , U-test p ( 16°C ) < 10−3 , Figure 4E ) , whereas a similar response was observed in cyc-1 ( RNAi ) embryos at 12°C ( U-test p ( 12°C ) < 10−3 , Figure 4E ) . While it remains to be investigated why the cellular consequences of depleting these three components differ to some extent , remarkably , they share the net result of increasing surface to volume ratio within the thermal range , thus mimicking the situation in the wild-type beyond the thermal limits . Therefore , these results strongly support the notion that the uncovered cellular hallmarks observed at the thermal limits of wild-type embryos reflect restricted aerobic capacity . In this work , we assessed the thermal response of cellular features during the first cell cycle of C . elegans and C . briggsae embryos . Interestingly , we uncovered that the thermal response of select cellular features changed precisely at the limit temperatures defined by embryonic viability tests ( see Figure 1A–B ) . While we do not know whether these cellular hallmarks are responsible for the observed increased lethality beyond the thermal limits , we note that a mere 10% decrease in embryonic viability is associated with readily observable cellular changes during the first cell cycle . Importantly , experiments in which mitochondrial respiration is compromised revealed that aerobic metabolism plays a smaller relative role towards the thermal limits than within the thermal range , raising the possibility that other metabolic routes are favored to produce energy . Furthermore , these experiments uncover that the changes in size and shape observed beyond the thermal limits in the wild-type can be recapitulated within the thermal range by impairing aerobic metabolism , strongly supporting the view that these changes arise in response to restricted aerobic metabolism . Together , our work provides critical experimental evidence supporting the notion that restricted aerobic metabolism is a general principle characterizing thermal limits in multicellular organisms in water and on land . Other elements contribute to setting boundary conditions within which a thermal range can be envisaged . Thus , cold-induced increase in unsaturated fatty acids in cyanobacteria , Arabidopsis ( Hazel , 1995 ) and C . elegans contributes to setting the lower thermal limit ( Svensk et al . , 2013 ) , although it only accounts for 16% of the observed difference in cold tolerance at 10°C vs 25°C in C . elegans ( Murray et al . , 2007 ) . Moreover , warm-induced increase in post-translational glycosylation also contributes to setting the upper thermal limit in Drosophila melanogaster , Danio rerio and C . elegans ( Radermacher et al . , 2014 ) . In addition , defects in synaptonemal complex assembly ( Bilgir et al . , 2013 ) and in sperm ( Harvey and Viney , 2007 ) contribute to setting the upper organismal limit in C . elegans . The restricted aerobic metabolism experimentally uncovered here is another important piece of the puzzle that contributes to defining both thermal limits . Another study investigating the temperature dependence of cell division processes in C . elegans and C . briggsae was published whilst the present manuscript was under consideration ( Begasse et al . , 2015 ) .
All the strains were maintained according to standard procedures ( Brenner , 1974 ) in incubators set at the temperature at which embryos would then be imaged . Note , however , that since C . elegans was not fully viable above 25°C ( see Figure 1A ) , worms were kept at the imaging temperature for only 6–24 hr prior to imaging . Embryos were dissected in 1× M9 medium tempered at the culture temperature , mounted on slides , placed under a coverslip and imaged using time-lapse DIC microscopy . Considering the crowded compressive environment of the uterus in the intact animal , and considering furthermore that the same mounting procedure was followed for all specimens at all temperatures , we surmise that the observed alterations in thermal response of embryo size and shape at given temperatures are not due to the mounting procedure . However , we cannot totally exclude that the observed changes in embryo size and shape may result from differential resilience to pressure of the cover slip used for imaging at the various temperatures . The recording rate was adjusted as follows ( we also mention the number n of embryos that were imaged from each condition ) : C . elegans ( N2 ) : 10°C ( 9 s , n = 9 ) , 12°C ( 8 s , n = 12 ) , 14°C ( 6 s , n = 11 ) , 16°C ( 6 s , n = 9 ) , 20°C ( 5 s , n = 20 ) , 24°C ( 4 s , n = 19 ) , 25°C ( 4 s , n = 15 ) , 26°C ( 4 s , n = 16 ) , 27°C ( 2 s , n = 10 ) . C . briggsae ( AF16 ) : 12°C ( 8 s , n = 8 ) , 14°C ( 7 s , n = 9 ) , 16°C ( 6 s , n = 16 ) , 20°C ( 4 . 5 s , n = 21 ) , 25°C ( 3 s , n = 14 ) , 28°C ( 2 s , n = 16 ) , 29°C ( 1 . 5 s , n = 15 ) . atp-2 ( RNAi ) : 12°C ( n = 14 ) , 16°C ( n = 16 ) . In this condition , only few embryos were imaged over the whole first cell cycle ( n ( 12°C ) = 3 , n ( 16°C ) = 6 ) . cyc-1 ( RNAi ) : 12°C ( n = 15 ) , 16°C ( n = 14 ) . In this condition , only few embryos were imaged over the whole first cell cycle ( n ( 12°C ) = 2 , n ( 16°C ) = 4 ) . nuo-1 ( RNAi ) : 12°C ( n = 8 ) , 16°C ( n = 15 ) . In this condition , only few embryos were imaged over the whole first cell cycle ( n ( 12°C ) = 1 , n ( 16°C ) = 4 ) . While imaging , the temperature was regulated by an air-blower that cooled/heated both sample and objective , and which was feedback-controlled by a thermocouple ( LABFACILITY ZO-PFA-K-1 ) inserted next to the embryo ( Figure 1—figure supplement 1A ) . We also ensured that the device was well calibrated in the experimental thermal range [8 , 32]°C ( Figure 1—figure supplement 1B ) . Prior to imaging , we made sure that embryos did not touch each other in order to facilitate segmentation . All DIC recordings were analyzed in a semi-automated fashion using Matlab . The analysis pipeline consisted of the following steps:We automatically segmented the eggshell contour using ASSET ( Blanchoud et al . , 2010 ) . All the measured positions were then automatically corrected at each time frame by the centroid of the egg in the same frame . This was an important step because the air-blow from the temperature controller displaced the embryos during the recordings . We detected by careful visual inspection the onset of pseudo-cleavage ( deepest furrow ) , mitotic entry ( nuclear envelope breakdown ) and cytokinesis ( start of membrane invagination ) . Cytokinesis onset defined time 0; hence , all the times in our analysis were negative . We automatically detected the migrating pronuclei using a custom segmentation algorithm based on ( Hamahashi et al . , 2005 ) . The exact timing of pronuclear meeting was then corrected by manual inspection . The speed of the female pronucleus was computed using its movement along the x-axis . After pronuclear meeting , the spindle poles were manually tracked until completion of centration-rotation . The angular and spatial trajectories were then fitted with the following model:x=A·|t|nKn+|t|n+cte , where x is the mid-position of the spindle poles or the angle they make with respect to the A-P axis . K represents the time at which centration ( resp . rotation ) is midway to completion . A relates to the initial position ( resp . angle ) . cte is an offset and n relates to the steepness of the profile . The velocity can then be computed using ν = dx/dt . After mitotic entry , the spindle poles were manually tracked until oscillations had dampened out . The position along the x-axis was used to compute spindle pole elongation speed towards the anterior and posterior poles , while positions along the y-axis monitored spindle oscillations . In order to retrieve the oscillation frequency , amplitude and duration , we first identified the dominant angular frequency ω by fast Fourier transform . We then applied a low-pass filter with threshold 3 2π ω , to remove the noise , followed by a high-pass filter to remove any drift of the oscillations ( with threshold 0 . 5 2π ω ) . Note that these filters did not change the dominant frequency of the signal , but were useful to better detect the peaks and measure the amplitude of the oscillations . Since the oscillations envelope was not always well fit by a sinusoidal function , we determined the duration of the oscillations by manual inspection of the oscillations profile ( after filtering ) . In order to determine embryo size , the embryo was manually contoured just before cytokinesis onset in order to extract its area . The area of each daughter cell was manually contoured at time t≅0 . 25·tPM , where tPM is the duration of the first cell cycle , defined as the time from pronuclear meeting to cytokinesis onset . In order to perform progeny tests , five to ten young adults were placed on a plate with a 5 µl drop of OP50 and left to lay eggs at the temperature of interest ( at least in triplicates ) . After 2–4 hr , we removed all the adults and counted the number of embryos on the plate ( generally between 30 and 100 embryos , except at extreme temperatures beyond the thermal limits where few or no embryos were laid ) . After a few days at the temperature of interest , we assessed the number of larvae that had hatched . Unsynchronized embryos were obtained by bleaching adult wild-type C . elegans worms . The number of embryos per μl was then assessed by optical density ( OD595 nm ) . We measured the respiration of wild-type C . elegans embryos from 9°C to 28°C using the Oroboros Oxygraph-2k , following the manufacturer's instructions . Prior to the experiment , a calibration was performed with 1× M9 buffer at 20°C in each chamber . We then dispensed 100 , 000 embryos in four chambers containing M9 buffer ( i . e . , 25 , 000 embryos/chamber ) : two chambers were used to go down in temperature from 20°C to 9°C , and two chambers were used to go up from 20°C to 28°C . The data from each chamber was normalized to its respiration rate at 20°C . We also repeated the same experiment using 35 , 000 embryos per chamber ( i . e . , 140 , 000 embryos in total ) . In order to measure respiration at 20°C upon CYC-1 depletion , we dispensed 2000 wild-type embryos/plate on 16 large Petri dishes with OP50 bacteria as food source . After 28 hr at 20°C , all the resulting larvae were collected by centrifugation and washed three times to remove the OP50 . Half of the collected larvae was re-suspended and distributed in 16 large OP50 plates , the other half in 16 large cyc-1 ( RNAi ) IPTG feeding plates ( prepared the day before and left at room temperature ) . After 44 hr at 20°C , embryos were collected in both control and cyc-1 ( RNAi ) conditions by bleaching adult worms . Respiration was measured in two chambers as follows: after calibrating the machine with 1× M9 buffer at 20°C , 35 , 000 wild-type embryos were dispensed in each chamber and their respiration measured at that temperature . The chamber was then washed , and we dispensed 35 , 000 embryos from the cyc-1 ( RNAi ) condition and likewise measured their respiration . For each chamber , we compared the respiration of cyc-1 ( RNAi ) embryos over wild-type . The experiment was repeated once using 35 , 000 embryos in the four chambers . Note that the lethality incurred following cyc-1 ( RNAi ) in these experiments was less pronounced than that reported in Figure 4C , probably owing to the need to scale up to assess respiration in a large number of embryos , such that the reported diminution of respiration is likely an underestimate of the actual impact . The C . elegans ORFeome RNAi library was a gift from Jean-François Rual and Marc Vidal , Harvard Medical School , Boston , USA ( Rual et al . , 2004 ) . Bacterial RNAi feeding strains were prepared as described ( Kamath et al . , 2001 ) . RNAi was performed by feeding early L3 larvae at temperature T with bacteria expressing dsRNA against the target gene for N hours at temperature T . The required feeding durations at each temperature were determined by fitting the duration of embryogenesis at 16°C ( 29 hr ) , 20°C ( 18 hr ) and 25°C ( 14 hr ) ( Epstein and Shakes , 1995 ) with the following equation ( Gillooly et al . , 2002 ) :t ( T ) =A/exp ( α·Tc1+TcT0 ) , where T0 = 273 K and Tc is the temperature in °C , yielding α = 0 . 1 , in agreement with ( Gillooly et al . , 2002 ) ( Figure 4—figure supplement 1 ) . We therefore used this value of α to fit the reported feeding durations from the literature ( ∼72 hr at 15°C [Ahringer , 2006] , ∼47 hr at 20°C [Afshar et al . , 2005] and ∼38 hr at 22°C [Ahringer , 2006] ) , yielding the following feeding durations: 12°C ( 90 hr ) , 16°C ( 65 hr ) , 20°C ( 44 hr ) , 24°C ( 31 hr ) . In order to verify that the results we uncovered in Figure 4B–D did not result from a general temperature-dependency in the effectiveness of the RNAi response , we also performed RNAi directed against AIR-1 , a serine/threonine kinase required for spindle assembly ( Hannak et al . , 2001 ) , a process not known to be related to metabolic status ( Figure 4—figure supplement 2 ) . Thermal response within the thermal range was assessed by Pearson correlation . A feature was considered to be temperature-dependent within the thermal range if its Pearson p-value was below 0 . 0014 ( which assumes a Bonferroni correction for multiple-testing 35 features , i . e . , 0 . 05/35 = 0 . 0014; c . f . Figure 1—source data 1 ) . Beyond the thermal range ( TR ) , we assessed if there was a significant change in the thermal response of the features by F-test ( nested-model analysis ) . For temperature-independent features , our first model was a simple regression y = mean ( feature within TR ) . This model was nested within our second model y = mean ( feature within TR ) + β ( T − 25 ) ( T ≥ 25°C ) which accounted for the potential change in the feature's thermal response beyond 25°C ( a similar model was implemented for C . elegans lower thermal limit , as well as for C . briggsae at its respective thermal limits ) . We then determined if the second model ( which has more parameters and therefore always fits the data better ) significantly improved the fit of the data using an F-test . The F statistic is given by:F= ( RSS1−RSS2p2−p1 ) ( RSS2n−p2 ) , where RSSi is the residual sum of squares of model i , pi the number of parameters of model i and n the number of data points . Under the null hypothesis , F follows an F-distribution with ( p2 − p1; n − p2 ) degrees of freedom . For temperature-dependent features , the first model was an exponential fit of the data within the thermal range: for example , for C . elegans , y=y0·exp ( α·T ( T∈[12° , 25°] ) ) . The nested model included a linear regression for the data above 25°C ( or similar for C . elegans lower thermal limit , as well as for C . briggsae at its respective thermal limits ) :y=y0·exp ( α·T ( T∈[12° , 25°] ) ) +β ( T−25 ) ( T≥25°C ) . In some temperature-dependent cases , the data within the thermal range was better fitted by linear regression ( the exponential and linear model having the same number of parameters , the model with smallest sum of residuals was considered to be the best ) . In those cases , we used the following first model: y=y0+α·T ( T∈[12° , 25°] ) . And the nested model: y=y0+α·T ( T∈[12° , 25°] ) +β ( T−25 ) ( T≥25°C ) . In all cases ( temperature-dependent or temperature-independent ) , the thermal response of a feature was considered to change beyond the upper thermal limit if the F-test p-value was below 0 . 0014 ( threshold 0 . 05 corrected for multiple testing 35 features , i . e . , 0 . 05/35 = 0 . 0014; c . f . Figure 1—source data 1 ) . If the p-value was above 0 . 0014 , the feature's thermal response above the upper thermal limit was tagged as unchanged in Figure 1—source data 1 . We also performed Mann Whitney U-tests to test if two distributions were different ( Figure 4E–F ) . Cell cycle durations within the thermal range were fitted to the following Arrhenius-like model ( Arrhenius , 1915 ) , to determine if the pace of cell division increased exponentially with temperature , as would be expected by Arrhenius kinetics:duration=A exp ( −EkBT ) , where kB is the Boltzmann constant , T is the temperature ( in [K] ) , E is the activation energy describing the thermal dependence ( in [eV] ) and A is a normalization constant . | An organism can thrive within a certain range of temperatures , beyond which it is less able to grow and reproduce . Different species are adapted to live in environments of different temperatures and this influences where on the planet they can be found . Researchers have suggested that the optimum temperature range for an organism may be influenced by its oxygen supply . The cells of most organisms need oxygen to produce chemical energy from sugars in a process called respiration . Within their normal temperature range , cells in the body of an organism are adapted to be able to take up sufficient oxygen to produce the energy they need . However , if the temperature rises above or falls below the limits of this range , the uptake of oxygen into cells may work less efficiently . Neves et al . tested this idea by studying the embryos of two different species of nematode worms grown at the limits of their respective temperature ranges . This led to several changes in the appearance of the embryos . For example , the embryos were larger than normal when grown at the lower end of their temperature ranges , but were smaller when grown at temperatures close to their upper limit . Shape changes were also seen: the embryos of both species were longer when grown at higher temperatures , a change that increases their surface area relative to their volume and may improve their ability to take up oxygen . Further experiments showed that disrupting respiration in the worms could lead to similar size and shape changes within the thermal range . Neves et al . 's findings provide experimental support that respiration plays an important role in setting the temperature ranges in which organisms can live . The next challenge will be to identify the genes that influence the capacity of respiration in these cells , which may help to explain how particular species have adapted to specific environments . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"cell",
"biology"
] | 2015 | Cellular hallmarks reveal restricted aerobic metabolism at thermal limits |
Ornithischian dinosaurs were ecologically prominent herbivores of the Mesozoic Era that achieved a global distribution by the onset of the Cretaceous . The ornithischian body plan is aberrant relative to other ornithodiran clades , and crucial details of their early evolution remain obscure . We present a new , fully articulated skeleton of the early branching ornithischian Heterodontosaurus tucki . Phase-contrast enhanced synchrotron data of this new specimen reveal a suite of novel postcranial features unknown in any other ornithischian , with implications for the early evolution of the group . These features include a large , anteriorly projecting sternum; bizarre , paddle-shaped sternal ribs; and a full gastral basket – the first recovered in Ornithischia . These unusual anatomical traits provide key information on the evolution of the ornithischian body plan and suggest functional shifts in the ventilatory apparatus occurred close to the base of the clade . We complement these anatomical data with a quantitative analysis of ornithischian pelvic architecture , which allows us to make a specific , stepwise hypothesis for their ventilatory evolution .
AM , Albany Museum , Makanda , Eastern Cape , South Africa; NCSM , North Carolina Museum of Natural Sciences , Raleigh , North Carolina; SAM , Iziko South African Museum , Cape Town , South Africa .
The osteology of H . tucki has been described elsewhere ( Crompton and Charig , 1962; Santa Luca et al . , 1976; Sereno , 2012; Galton , 2014 ) ; we focus instead on novel anatomical features preserved in AM 4766 ( Figure 1A ) . Visualization of these features was made possible by a high-resolution , phase-contrast enhanced SRμCT with a bespoke reconstruction algorithm developed for this particular specimen ( elaborated further in Appendix 1 ) that has recently been used elsewhere ( Cau et al . , 2017 ) . Approximately 18 gastralia are present in total and would have produced two longitudinal rows with each containing 9 gastralia . The gastralia follow the ventral abdominal midline , from the posterior margin of the sternal plates to the level of the distal ends of the pubes ( Figure 1B ) . The first two pairs of gastralia have slightly thickened medial facets that are absent from all subsequent pairs ( Figure 1C ) , with the overall thickness of gastralia diminishing posteriorly . The gastralia are autapomorphic among non-avian dinosaurs in lacking a lateral segment ( Claessens , 2004; Fechner and Gößling , 2014; Barrett et al . , 2019 ) , which is retained in even the diminished gastral basket of early branching avialans ( O’Connor et al . , 2015 ) . Fragments associated with the H . tucki specimen SAM-PK-K1332 are of comparable dimensions to the gastralia in AM 4766; however , they have been removed from context and could potentially represent displaced ossified tendons or posteriormost dorsal ribs . We tentatively identify the long , narrow bone fragments on either side of the proximal femur in the holotype specimen of Tianyulong confuciusi STMN 26-3 ( Zheng et al . , 2009; Appendix 1—figure 3 ) as gastralia based on their similarity with AM 4766 . Two separate sternal plates are present , although only the left one is complete . The sternal plates are sub-rectangular , their long axes are oriented anteroposteriorly , and they are dorsoventrally thickest on their lateral margin and progressively thin medially ( Figure 2 ) . The fenestra that perforates the centre of the sternal plate preserved in SAM-PK-K1332 , identified by Sereno , 2012 , is also present in AM 4766 . The left sternal plate of AM 4766 bears an autapomorphic , anteromedially projecting , tongue-shaped process that projects abruptly from the anterolateral portion of the sternal plate ( Figure 2B–E ) . The proximal portion of this process is partially visible in SAM-PK-K1332 , but most of it is still obscured by matrix . The exact nature and function of the tongue-shaped process is currently unknown , but , when paired , they likely buttressed the region between coracoids . The posterolateral corner of the sternal plate of SAM-PK-K1332 has a small but distinct protuberance , identified as an articulation for the sternal ribs ( Sereno , 2012 ) , herein referred to as a costal process ( Figure 2C , D ) . While this structure appears to be missing from AM 4766 , its absence cannot be confidently confirmed as the resolution in this region of the SRμCT data is diminished by metallic inclusions obscuring boundaries between bones and matrix . Three pairs of sternal ribs are preserved in AM 4766 , each similar in size and morphology ( Figures 3 and 4 ) . The sternal ribs have a spatulate morphology , with an elongate and semi-cylindrical anterior half , and an abruptly dorsoventrally expanded posterior half that thins to a mediolaterally compressed , sheet-like structure , similar to the avialan Jeholornis prima ( Zheng et al . , 2020; Figure 4G ) . A thickened nub on the posterior apex of this sheet-like portion forms a monocondylar sternocostal articulation with the distal end of the corresponding dorsal rib . A similar articular relationship is present between the sternal , intermediate , and dorsal ribs of extant crocodilians ( Claessens , 2009; Brocklehurst et al . , 2017 ) . Although the gross morphology of their sternal ribs differs , the sternal and dorsal ribs of pterosaurs ( e . g . , Rhamphorhynchus muensteri , Figure 4F ) also bear monocondylar sternocostal joints that are strikingly similar to those of H . tucki ( Claessens et al . , 2009 ) . The sternal ribs of AM 4766 contrast markedly with the few other ornithischian examples: for example , in Thescelosaurus neglectus ( NCSM 15728 ) ( Figure 4E ) and Nanosaurus agilis ( BYU 163 ) ( identified as ‘costal cartilage’ in Carpenter and Galton , 2018 ) , the sternal ribs are comparatively shorter , subrectangular , and bearing broad butt joints rather than condylar articulations at their distal and proximal ends . The paired clavicles ( Figure 5A , B ) are preserved in life position anterior to the scapulocoracoid and are proportionally long , thin , and bowed posteriorly . The proximal end of the left clavicle is marginally thicker than the rest of this element and gently tapers laterally . Among ornithischians , clavicles are mostly known in basal ceratopsians and neoceratopsians from the Cretaceous , for example , Psittacosaurus mongoliensis ( Fairfield , 1924; Sereno , 1990 ) , Psittacosaurus sibiricus ( Averianov et al . , 2006 ) , Auroraceratops rugosus ( Morschhauser et al . , 2018a ) , Leptoceratops gracilis ( Sternberg , 1951 ) , Montanoceratops cerorhynchos ( Chinnery and Weishampel , 1998 ) , and Protoceratops andrewsi ( Brown and Schlaikjer , 1940 ) but are also present in the basal thyreophoran Scelidosaurus harrisonii ( Norman , 2020 ) as well as a new , undescribed taxon that is purported to be at the base of Ornithopoda ( Spencer et al . , 2020 ) . The clavicles of H . tucki are similar to those of ceratopsians in contouring the anterior margin of the scapulocoracoid , with no apparent contact present between the clavicles along their length . The clavicles of AM 4766 differ from those of ceratopsians in overall size . AM 4766 has clavicles that are ~60% the length of the anterior margin of the body of the scapulocoracoid ( i . e . , excluding the scapular blade ) , where the clavicles of ceratopsians are approximately 40% of the length of the anterior margin of the scapulocoracoid . The suprascapula ( Figure 5 ) is sub-trapezoidal in shape , being broader distally than it is proximally . The distal margin of the suprascapula is concave and articulates with the proximal , convex margin of the scapular blade . A suprascapula is also present in SAM-PK-K1332 and was originally described by Santa Luca et al . , 1976 as a ‘cartilaginous extension’ that capped the dorsal margin of the scapula . At present , we are unable to eliminate the possibility that this structure is indeed cartilaginous , but it is undistorted and has clearly defined margins that allow us to tentatively consider the suprascapula as an ossification ( rather than a chondrification ) . An ossified suprascapula has only been described in one other dinosaurian taxon , the Cretaceous neornithischian Parksosaurus warreni ( Sternberg , 1936 ) . Synchrotron scanning of AM 4766 permitted the reconstruction of vertebral morphology previously obscured in specimen SAM-PK-K1332 . Notably , the parapophyses migrate anterodorsally as the vertebral series progresses posteriorly ( Figure 6 ) : on the first , second , and third dorsal vertebrae , the parapophyses are located immediately ventral to the diapophyses; the fourth and fifth dorsal vertebrae mark the transition where the parapophyses migrate dorsally from the centrum and onto the neural arch; and from the sixth dorsal vertebrae and in all more posterior dorsals , the parapophyses are anterior to the diapophyses and on the same horizontal level . Our reconstruction based on SRμCT data shows that the entire internal structure of the cervical , dorsal , sacral , and proximal caudal vertebral column lacks pneumatic chambers or fossae , including those areas often implicated in the early evolution and development of PSP ( Wedel , 2006; Butler et al . , 2009; Benson et al . , 2012 ) , conclusively showing that early ornithischians lacked PSP . Our measured variables show strong ( log10 pubic rod length; r2 = 0 . 858 ) to very strong ( log10 APP length , ischial length; r2 > 0 . 96 ) correlations with body size ( represented in our analysis by log10 femur length , see Materials and methods , and Appendix 1—figures 4 and 5 ) . We therefore corrected for phylogenetic and allometric effects by using the residuals of phylogenetically corrected generalized least squares ( pGLS ) regressions of each variable against log10 femoral length . The residuals from our pGLS regression of each of our three variables showed poor correlation with log10 femoral length . This indicates that changes in pubic and ischial dimensions are largely dissociated from the allometric effects of body size ( see Appendix 1—figure 5A–F ) . Optimizing residuals on the phylogeny ( Figure 7 ) shows that later branching taxa have APPs that are elongate relative to early branching taxa . APP elongation occurs at the base of Genasauria , and within this clade it is modified comparatively little over its subsequent history . There are generally declining rates of change in APP length in later-branching lineages and temporally later-appearing tips of the tree , with exactly zero known instances of reversion to the plesiomorphic relative length . Derived hadrosaurs and neoceratopsians apparently appear to have slightly shorter APPs relative to earlier-diverging taxa of their respective clades; however , it should be noted that these taxa dorsoventrally expand the APP independently , significantly increasing the surface area for muscle attachment . Early branching ornithischians have long pubic rods , which subsequently shorten independently in ornithopods and marginocephalians well after the APP begins to elongate on the tree ( i . e . , after the major splits in Genasauria ) . Ischial length shows a more complex pattern , with most ornithischians retaining the plesiomorphic proportional length , but with stegosaurs showing large decreases in relative length and ornithopods and certain ceratopsians showing modest increases . We used each set of residuals as continuous characters for an evolutionary model testing analysis using phylogenetic comparative methods ( see Materials and methods ) . Among these , ‘Early Burst’ is strongly preferred for the evolution of APP length and performs better than other competing models ( Akaike Information Criterion [AICc] weight: 99 . 99% , likelihood ratio test p<0 . 01; see Table 1 ) . The ‘Early Burst’ model posits declining evolutionary rates over time , that is , expected variance is higher between earlier-branching taxa ( Harmon et al . , 2010 ) , matching our qualitative observations from mapping residuals on the tree ( Figure 7 ) . Pubic rod length is best modelled by a ‘Drift’ model ( AICc weight: 82 . 34% , p=0 . 01 ) , and ‘Stasis’ is more strongly , but non-significantly preferred for ischial length ( AICc weight: 59 . 37%; p=1 ) .
The nearly sequentially complete gastral basket of AM 4766 is the first known in ornithischian dinosaurs , and the tentative identification of gastralia in the holotype of the Chinese taxon T . confuciusi suggests that gastralia may have been present in all heterodontosaurids . With gastralia being plesiomorphically ubiquitous across a range of tetrapod clades , discovering gastralia in Heterodontosauridae is not surprising as this clade is consistently recovered as the earliest branching lineage of ornithischian dinosaurs ( Butler et al . , 2008; Boyd , 2015 ) . It is more surprising , however , that these gastralia are retained in H . tucki despite its typical ornithischian retroverted pubis . Three-dimensional reconstruction of our SRμCT data clearly demonstrate gastralia in close association with the distalmost point of the pubes , indicating that their complete retroversion ( opisthopuby ) was achieved with the gastralia still intimately coupled . Together , these observations contest previous hypotheses that reasoned that a divorce of the gastral basket from the pubis was a necessary prerequisite for ornithischian pubic retroversion ( Rasskin-Gutman and Buscalioni , 2001 ) . Furthermore , the association of the gastralia with the distal end of the pubic rod indicates that the latter structure is homologous to the pubic shaft/apron of other archosaurs ( Galton , 1970 , contra references therein ) , and that the APP is a de novo feature . The presence of sternal ribs in H . tucki extends the occurrence of these bones from late diverging taxa like T . neglectus and other relatively late-branching , small-bodied Late Jurassic and Cretaceous neornithischians ( Carpenter and Galton , 2018; Butler and Galton , 2008 ) to the basalmost members of Ornithischia . This broader distribution strongly implies that the presence of sternal ribs may optimize as an ornithischian plesiomorphy . However , the sternal ribs we describe in H . tucki are autapomorphic in morphology , differing markedly from those of other ornithischians , and showing clear evidence of being mobile about their dorsal rib and sternal plate joints ( Figure 4A–C ) . The dorsoventrally expanded dorsal and ventral margins of these ribs were likely attachment sites for intercostal musculature and in this way perhaps analogous to similar projections ( sternocostapophyses ) on the sternal ribs of pterosaurs ( Claessens et al . , 2009 ) , the uncinate processes of maniraptorans ( Tickle et al . , 2012; Codd et al . , 2008 ) , and the remarkably similar sternal ribs of the ornithothoracine J . prima ( Zheng et al . , 2020 ) – all of which are adaptations hypothesized to increase lever-arm potential and facilitate efficient deformation of the body wall to drive ventilation . The complex sternal plates of AM 4766 are distinct from the comparatively simple ‘hatchet-shaped’ sternal plates of iguanodontians and the ‘kidney-shaped’ ( reniform ) sterna of other neornithischians but are not unique among Dinosauria . Instead , the complex sternal plates of AM 4766 bear similarities with early- and late-diverging theropods such as Tawa hallae ( Bradley et al . , 2019 ) and various avialans ( Zheng et al . , 2012; O'Connor et al . , 2015 ) , respectively . Features like the tongue-shaped process of AM 4766 and the coracoid facet of T . hallae are strikingly similar in their dimensions , location , and abrupt change in orientation relative to the posterior half of their respective sternal plates . Further similarities include the single knob-like costal process of AM 4766 exhibiting a similar morphology to the series of costal processes of T . hallae , and the analogous position of the lateral tubercula of enantiornithines ( Zheng et al . , 2012 ) . It is unclear whether these sternal similarities are homologous , but they are likely functionally analogous . The nature of change in the relative length of the APP is conspicuous from both qualitative and quantitative analyses of pelvic evolution . Innovation in APP length occurred early in ornithischian evolution , before the diversification of genasaurians , and after this significant early burst ( see Figure 7 , Table 1 ) , modifications of the APP were generally restricted to gross shape differences that do not affect the relative length: derived ornithopods evolved a large , lobate APP with derived neoceratopsians evolving an APP that fanned-out anteriorly . We interpret these results as rapid switching of optimal phenotypes for APP size; from the plesiomorphically small condition in H . tucki to a proportionally long APP that is then maintained across all later-branching ornithischian lineages . We consider this as evidence that the APP was involved in a major macroevolutionary shift in ornithischian dinosaurs , which occurred immediately prior to their radiation in the Jurassic . Moreover , the timing of this change in the relative length of the APP cooccurs with the reduction and loss of the gastralia and possibly the loss or reduction of sternocostal mobility . These results are consistent with , but greatly expand upon , Brett-Surman’s hypothesis ( Brett-Surman , 1989 ) that the enlarged APP is an adaptation of ornithischians involved in driving considerable changes in abdominothoracic volume . The pelves of published ankylosaur specimens are often obscured or incomplete and do not contain sufficient measurement information to include in this analysis ( Kirkland and Carpenter , 1994; Carpenter et al . , 2013; Arbour and Currie , 2013; Xu et al . , 2001 ) . Nevertheless , we observe that despite highly derived pubic rod morphologies , including near loss in late branching taxa like Euoplocephalus tutus , the APP is retained as a process fused to the ventral surface of the ilium ( Carpenter et al . , 2013 ) . This strongly implies that the APP was subject to a constraint that favoured its retention when the rest of the pubis was made redundant . This pattern of pubic evolution is best explained by a ‘Drift’ evolutionary hypothesis and suggests a trend away from elongate pubes . The reduction of the pubic rod is pervasive in ornithischians as it is independently lost in derived iguanodontians , neoceratopsians , and pachycephalosaurs ( as well as ankylosaurs; Carpenter et al . , 2013 ) . This likely signifies a relaxing of a plesiomorphic constraint between the hypaxial abdominal musculature and the pubis . This reduction of the pubic rod is decoupled temporally and phylogenetically from the loss of gastralia and the expansion of the APP , strongly suggesting that the pubic rod was rendered vestigial . It is possible that modifications like the bowed ischium of neoceratopsians and the ‘ischial boot’ of some hadrosaurs are responses to the ischium subsuming the myological role previously played by the pubis . Ischial residuals are harder to interpret , and despite being best explained by a ‘Stasis’ model , there is no statistical significance between this explanatory model or any other evolutionary models we tested . Qualitatively , most ornithischian dinosaurs have similar-length ischia relative to their body size . The most conspicuous departures from this are in late-branching ornithopods , where the ischium is elongated ( i . e . , strong positive residuals ) , and in stegosaurs where the ischium is shortened ( i . e . , strong negative residuals ) . That the pubic rods of stegosaurs remained elongate and the ischia short and robust almost certainly indicates that some other selective pressure such as tail-driven defence ( Mallison , 2011; Carpenter et al . , 2005 ) was imposed on the ischium that prevented it from supplanting the role of the pubis in anchoring abdominal musculature . Nevertheless , both ischial modifications occur temporally and phylogenetically well after the increase in relative APP length and the loss of gastralia . The anatomical features presented here provide consilient evidence that H . tucki preserves morphologies that reflect early steps in the evolution of a novel means of lung ventilation in ornithischian dinosaurs . Below , we review this evidence and propose a potential model for the ornithischian ventilation system . Gastralia are widespread among Palaeozoic and Mesozoic tetrapods but have never before been unambiguously reported in Ornithischia . Among other dinosaurian lineages , theropods retain their gastral basket until the evolution of neornithine birds , and sauropodomorphs lose their gastralia relatively late in their evolutionary history at the base of Eusauropoda – potentially retaining gastralia even during the emergence of Neosauropoda ( Tschopp and Mateus , 2013 ) . The reduction or loss of the gastralia independently occurs in other major tetrapod lineages like stem turtles ( Lyson et al . , 2013; Schoch and Sues , 2020 ) , and eutheriodont therapsids including mammals ( Cisneros et al . , 2015 ) . Interestingly , specialized lung ventilatory mechanisms are present in all extant clades that have lost ( birds , mammals ) or co-opted the gastralia ( turtles ) . Some workers have explicitly linked the loss of gastralia and subsequent thoracic and lumbar differentiation of therocephalian and cynodont therapsid axial skeletons to the evolution of a mammalian-style , diaphragm-driven ventilatory arrangement ( Perry et al . , 2010; Brink , 1956 ) . The specialized facets on the medial gastralial elements of non-avian theropods have led multiple authors ( Carrier and Farmer , 2000; Claessens , 2004; Fechner and Gößling , 2014; Codd et al . , 2008; Lambe , 1917 ) to hypothesize that gastralia played an important role in ventilating the lungs , a mechanism Carrier and Farmer , 2000 termed ‘cuirassal breathing’ that was inherited from a common non-dinosaurian archosaurian ancestor . These hypotheses posit the gastralia would have facilitated expansion and contraction of the body wall to facilitate volumetric changes in the thoracoabdominal cavity ( Carrier and Farmer , 2000; Claessens , 2004; Codd et al . , 2008; Lambe , 1917 ) . Although in extant crocodilians the gastralia themselves only contribute a relatively small amount to such volumetric changes in isolation ( Claessens , 2009 ) , the gastral basket is integral in bridging the sternocostal complex and mobile pubis , serving as an attachment site for muscles fundamental to body wall deformation and function of the ‘hepatic piston’ . In archosaurian ventilation models , the involvement of the pelvis is ubiquitous , ranging from pelvic rocking in birds ( Baumel et al . , 1990 ) , to the hepatic piston in crocodilians ( Farmer and Carrier , 2000 ) , to the prepubis of pterosaurs ( Claessens et al . , 2009 ) . Anterior bony projections of the pubic region are key components of these models , including the mobile pubis of crocodilians , and the prepubis and puboiliac complex in pterosaurs . Carrier and Farmer , 2000 highlighted the APP as the integral locus for interpreting ornithischian lung ventilation , focusing their hypothesis on the major genasaurian clades Neoceratopsia , Ornithopoda , and Stegosauria . Macaluso and Tschopp , 2018 hypothesized that pubic retroversion in dinosaurs is linked to the evolution of an innovative ventilatory mechanism , arguing that the plesiomorphic cuirassal breathing proposed by Carrier and Farmer , 2000 constrains the pubis into the propubic condition , and that evolution of mesopubic and opisthopubic conditions indicates a relaxing of those constraints as a new mechanism evolves . The role of the APP in ventilation has been contentious , however , with other authors assigning it a locomotory function ( as the origin of the ambiens [Maidment and Barrett , 2011] or pubotibialis [Galton , 1969] muscles ) . Although the locomotory and ventilatory explanations are not mutually exclusive , the evidence gathered here makes us consider the locomotory role to be a poor explanation for APP changes for the following reasons . First , our evolutionary analysis clearly shows that the major changes in the APP residuals are phylogenetically and temporally divorced from the independent acquisitions of quadrupedality and major postural changes ( Maidment and Barrett , 2011; Maidment and Barrett , 2012a; Maidment and Barrett , 2012b; Barrett and Maidment , 2017 ) . Second , although there is little available data from extant taxa , the ambiens muscle appears to have weak negative allometry in Dromaius novaehollandiae ( Lamas et al . , 2014 ) , suggesting that body size increases in ornithischian lineages would not drive a trend of disproportional APP increase ( additionally , our analysis of pubic measurements using residuals precludes this ) . Third , the APP is subparallel to the vertebral column and medial to the ribcage , thus precluding it from being a major driver of hindlimb extension or retraction when body wall musculature is reconstructed ( principally M . obliquus abdominus externus; Fechner and Gößling , 2014; Fechner and Schwarz‐Wings , 2013 ) . These lines of evidence together indicate that the factors driving the evolution of APP length and shape are distinct from locomotory influences . In total , our observations here show that H . tucki has reduced gastralia , an apomorphically elaborate sternum , well-developed and mobile sternal ribs , an incipient APP , and completely lacks PSP . We propose a single explanatory model for these observations: that H . tucki is a transitional animal preserving the early steps in the evolution of a unique ventilation mechanism in ornithischian dinosaurs . We name this model the ‘pelvic bellows’ and elaborate on it below . This model does not require us to make ad hoc assumptions about airflow direction ( i . e . , unidirectional versus tidal ) , but phylogenetic bracketing predicts intrapulmonary unidirectional airflow in Ornithischia and is fully compatible with our model ( Schachner et al . , 2014; Cieri et al . , 2014; Farmer and Sanders , 2010; O'Connor and Claessens , 2005 ) . Stem dinosauromorphs ( Figure 8A; Kammerer et al . , 2020 ) ( or stem sulcimentisaurians; Müller and Garcia , 2020 ) bear the plesiomorphic archosaurian condition of a typical gastral basket connecting to a propubic pelvis , and they lack both PSP ( Butler et al . , 2009 ) and an APP . This points to cuirassal ventilation as the primary means of volume change . Interestingly there is evidence of a bipartite , semi-compliant lung in Silesaurus opolensis ( Schachner et al . , 2011 ) , adding potential support to a hypothesized but controversial relationship between silesaurids and Ornithischia ( Müller and Garcia , 2020; Ferigolo and Langer , 2007 ) . In early branching ornithischians , exemplified by H . tucki ( Figure 8B ) , the retroverted pubes and the reduced gastralia indicate that the cuirassal breathing mechanism ( Carrier and Farmer , 2000 ) is still present but has reduced capacity to affect changes in volume . The sternal ribs of H . tucki would have facilitated the pivoting and leveraging of the sternum and aided in its posteroventral contraction , providing substantial volumetric changes . The small APP would have served as a nascent area for the origination of a muscle analogous to the dorsal component of M . diaphragmaticus in living crocodilians , which we term the ‘puboperitoneal muscle’ . We hypothesize that the puboperitoneal muscle would have functioned as an accessory lung ventilator in early ornithischians , similar to the accessory ventilatory function provided by the iliocostalis musculature of some crocodilians ( Codd et al . , 2019 ) . The puboperitoneal muscle would have provided an additional anteroposterior vector to the dorsoventral displacement already afforded by the cuirassal and sternocostal mechanisms . Upon inspiration , contraction of M . rectus abdominus would have distended the gastral basket and the sternal complex posteroventrally , the puboperitoneal musculature simultaneously contracting to generate negative pressure in the posteriorly compliant half of the lung . During expiration , the abdomen , sternal complex , and puboperitoneal muscle relaxed and would rebound anterodorsally to force air out . Gastralia are seemingly lost amongst early branching ornithischians , and the cuirassal breathing mechanism is no longer present in Genasauria ( Figure 8C ) . However , the pubic rod is still long and M . rectus abdominus is likely still present , not entirely precluding the possibility of body wall deformation by contractions of hypaxial musculature . The sternal ribs are greatly simplified or lost . When they are present , broad , immobile butt joints replace the condylar joints between the sternum and sternal ribs . Additionally , the well-developed processes and eminences that were features of the sternum and sternal ribs of H . tucki are lost , simplifying the sternal complex in all subsequent clades . Together , this simplification of the sternum decreases the degrees of freedom for associated skeletal components , reducing both the range of motion of the sternocostal complex ( relative to the plesiomorphic condition ) and its ability to contribute to changes in abdominothoracic volume . The APP is prominent and anteriorly elongated , with anteroposteriorly oriented muscle scars present on the dorsal and medial surfaces ( Morschhauser et al . , 2018a; Owen , 1875 ) . In Genasauria , the puboperitoneal muscle is now the major contributor to changes in volume , with the sternum and the abdominal musculature relegated to a secondary role . Finally , in deeply nested ornithischians ( Figure 8D ) , the gastralia remain absent , with the pubic rod shortening to a spur ( derived ornithopods ) or a tab ( derived neoceratopsians ) , indicating that abdominal musculature now attaches to the ischium and mainly functions to support the viscera . The sternal plates , where present , are relatively small , show no evidence of dorsal rib interaction , and differ markedly in morphology between clades , suggesting that they played no constrained role in ventilation . Convergently , APP area is substantially enlarged and develops clade-specific morphologies . At this stage , the puboperitoneal muscle served as the main ventilatory apparatus , sternal movements contribute little to no volumetric changes , and the non-puboperitoneal abdominal musculature functions mainly as support for viscera . The changing vertebrocostal orientations along the axial column of H . tucki observed here ( Figure 6 ) support the bipartite and dorsally immobile lung previously reconstructed in ornithischians and silesaurids ( Schachner et al . , 2011; Brocklehurst et al . , 2018 ) . Considering that a dorsally immobilized , anatomically and functionally heterogeneous lung has been reconstructed for all of Ornithischia ( Schachner et al . , 2011; Brocklehurst et al . , 2018 ) , and that the M . diaphragmaticus of extant crocodilians is coupled with a flexible lung and shifting viscera , the proposed ventilatory mechanism for Ornithischia was likely functionally distinct from the hepatic piston model present in crocodilians , although the two may have been anatomically convergent . The crocodilian M . diaphragmaticus originates on both the pelvis anterior to the acetabulum and the gastralia ( or pubic apron , depending upon the taxon ) ( Gans and Clark , 1976 ) . It then fans out anteriorly , encapsulates all of the abdominal viscera ( dorsally , laterally , and ventrally ) , and inserts on the liver , with fibres occasionally extending to the pericardium ( Gans and Clark , 1976 ) . The proposed puboperitoneal muscle in ornithischians originating on the APP ( Figure 8B–D ) is reconstructed here as travelling anteriorly , and inserting on any potential number of anatomical structures , including the dorsal surface of the liver , pulmonary septa , posteriorly positioned air sacs ( or non-invasive pulmonary diverticula emerging from the lung ) , or even the posterior aspect of the lung itself if no pulmonary diverticula existed . This putative mechanism would be distinctly different from that of extant crocodilians , particularly in the larger , later-branching ornithischian taxa in that there would be no ventral attachment due to a loss of gastralia and the shortened pubis ( Figure 8D ) . Without bundling the abdominal viscera into a fusiform tube , there would be no anterior-posterior translation of the entire visceral mass within the thoracocoelomic cavity . Additionally , ventilation of the anteriorly immobilized respiratory parenchyma by a posterior/ventral flexible region ( whether air sacs , or just a flexible sac-like expansion ) would not theoretically cause the same shifts in centre of mass that the crocodilian hepatic-piston mechanism does ( see , e . g . , Uriona and Farmer , 2008 ) , and may be more functionally analogous to the complementary integration of pelvic musculature as observed in birds ( e . g . , Columba livia; Baumel et al . , 1990 ) . This hypothesis posits that only the flexible regions of the lung linked to the pelvic bellows would be stretching with contraction of the muscle , while the anterior and dorsal regions of the lung containing the respiratory parenchyma could remain fixed and immobilized to the adjacent skeletal tissues . This type of pulmonary heterogeneity is well documented in other sauropsids outside of birds ( e . g . , varanids [Schachner et al . , 2014] , chameleons [Klaver , 1973] , snakes [Wallach , 1998] ) , where there is an extreme separation of the respiratory parenchyma and more flexible sac-like structures in the posterior region of the lung , and thus supports the possibility of these characters independently evolving in this lineage if this ventilatory mode is truly divergent from other dinosaurs . Investigation into dinosaur respiration has focused on PSP , using its presence or absence as a sole proxy for avian-like ventilation and physiology ( O'Connor and Claessens , 2005; O'Connor , 2006; Butler et al . , 2012; Wedel , 2003 ) . Recent studies showing the remarkable multiplicity of respiratory systems employed by living reptiles ( Owerkowicz et al . , 1999; Cieri et al . , 2018; Claessens , 2009; Brocklehurst et al . , 2017; Baumel et al . , 1990; Farmer and Carrier , 2000; Lyson et al . , 2014 ) show that PSP is only one component of a complex suite of features that coevolve to enable lung ventilation across a swathe of tetrapod lineages . This recent research shows that some presumed ‘bird-like’ respiratory features , such as unidirectional air flow , are actually plesiomorphies characterizing much larger groups ( Schachner et al . , 2014; Farmer and Sanders , 2010 ) and highlights the diversity of ways in which multiple anatomical systems interlink to effectively ventilate the lungs . New work on the pulmonary anatomy of the ostrich ( Struthio camelus ) has demonstrated that PSP relationships with the respiratory system in extant birds may not be as straightforward as previously thought ( Schachner et al . , 2021 ) , and reconstructions of dinosaur lungs that directly follow a standardized avian bauplan may need to be reconsidered . Additionally , primitive features like gastralia , simple sterna , and ‘propubic’ pelves impede attempts at completely superimposing the highly derived physiology of birds onto comparatively less-specialized clades like non-avian theropods and sauropods . Inquiry into ornithischian breathing has been stunted by virtue of their phylogenetic position between clades that have received more thorough respiratory evolution investigation ( O'Connor and Claessens , 2005; Wedel , 2006; Butler et al . , 2009; Claessens et al . , 2009; Fechner and Gößling , 2014; Zheng et al . , 2020; Tickle et al . , 2012; Codd et al . , 2008; Wedel , 2003 ) paired with inferences informed by phylogenetic bracketing ( Witmer , 1995 ) . As discussed here , ornithischians are outliers among ornithodirans for many reasons – in particular , their unique lung structure and lineage-wide lack of PSP contradict the more parsimonious inferences made about their respiratory anatomy ( i . e . , that ornithischians are predicted to have conspicuous air sacs ) . Archosaurs likely demonstrate a remarkably labile respiratory evolution that has yet to be fully appreciated , and future inquiry is at risk of overlooking a variety of ventilatory mechanisms that are obfuscated by more parsimonious explanations . This suggests that dinosaur , and archosaur , breathing should be investigated with a more nuanced view of evolution; a paradigm that is informed by extant respiratory diversity and that is simultaneously willing to risk relaxing an insistence on phylogenetic bracketing in an attempt to capture increased ventilatory diversity in extinct lineages . Similar trappings will inevitably extend beyond the topic of respiratory evolution , with the evolution and homology of archosaurian integumentary structures being an additional area that will likely struggle from comparable oversimplifications . The success of ornithischians is remarkable , and the reason for the marked differences between their body plans and those of other dinosaurs remains enigmatic . The diverging ventilatory adaptations hypothesized here provide an overarching explanation for a wide range of skeletal modifications , and perhaps accompanying metabolic and physiological changes , that shaped the lineage for 130 million years . It is likely no coincidence that the evolution of the APP and its hypothesized role in ventilation precede dramatic and conspicuous increases in ornithischian diversity and disparity .
To quantify pelvic evolution in Ornithischia , we measured femoral length , APP ( from the anterior margin of the acetabulum ) , pubic rod ( from in line with the anterior margin of the acetabulum to the distalmost tip ) , and ischial ( the contour of the posterior surface that initiates on the iliac peduncle and terminates at the middle of the distalmost point ) lengths for a phylogenetically broad sample of ornithischian taxa ( Appendix 1—table 1 ) through direct measurements of specimens , high-resolution photos , and published sources . We measured specimens either by hand , using digital callipers and measuring tapes , digitally from 3D SRμCT data , or from high-resolution photos of specimens where scale bars were available and accurate . To normalize scale , we log10-transformed all measurements . We selected pubic and ischial measurements because of their hypothesized relationship with lung ventilation ( in particular , plesiomorphic models like cuirassal breathing; Carrier and Farmer , 2000 ) . We chose femoral length as a proxy for body mass , even though it has lower correlation coefficients than femoral circumference for body mass estimation ( Campione and Evans , 2012; Anderson et al . , 1985 ) . We used it here because circumference measurements were unavailable for most of our specimens and because femoral length is frequently used in the literature and therefore practical to collect ( e . g . , Christiansen and Fariña † , 2004 ) . Although it is unlikely that this choice greatly affects the results we present here , stegosaurs appear to have apomorphically long femora that are likely to affect body mass corrections for these taxa specifically; we accept this localized trade-off in error for the benefit of standardized measurements across our sampling of Ornithischia . We analysed these data using scripts written in the R statistical software language ( R Development Core Team , 2013 ) and its associated packages ‘ape’ ( Paradis et al . , 2004 ) , ‘ggplot2’ ( Wickham , 2016 ) , ‘phytools’ ( Revell , 2012 ) , ‘strap’ ( Bell and Lloyd , 2015 ) , ‘geiger’ ( Harmon et al . , 2008 ) , and ‘nlme’ ( Pinheiro et al . , 2012 ) . To investigate evolutionary patterns in the APP , pubic rod , and ischium , we used pGLS regressions of these pelvic measurements against femoral length and calculated residuals from these regressions . This is a common means of assessing phylogenetic and size-corrected variance in morphological datasets and can be used together with comparative phylogenetic methods ( Revell , 2009; Hunt and Carrano , 2010 ) . We used the residuals as continuous characters to both qualitatively map on the ornithischian tree and to assess the fit of a variety of macroevolutionary models implemented in the R package Geiger ( Harmon et al . , 2008 ) . We used the corrected AICc and computed likelihood ratio tests to assess whether the preferred model is significantly better than the next-best model . ‘Source code 1’ is R code to reproduce statistical analysis; ‘Source code 2 and 3’ are phylogenetic tree files in . phy and . nex formats , respectively; and ‘Source code 4’ is ‘First Appearance Date’ and ‘Last Appearance Date’ data of taxa analysed , obtained from the Paleobiology Database ( paleobiodb . org ) . AM 4766 was recovered from the upper Elliot Formation ( uEF ) in strata that correlate with the Massospondylus Assemblage Zone ( Viglietti et al . , 2020 ) and is likely Sinemurian in age ( Bordy et al . , 2020 ) . The specimen was recovered from a light red , clast-rich , very fine-grained sandstone that is consistent with palaeo-environmental reconstructions of the uEF as a seasonally wet , fluvio-lacustrine system ( Bordy et al . , 2004a ) . Further details of the geological context are elaborated in Appendix 1 and figured in Appendix 1—figures 1 and 2 . Volume files of AM 4766 were reconstructed using a combination of manual and semi-automated tools ( i . e . , pen tool , interpolate function ) in Avizo Lite version 9 . 0 ( FEI Visualization Sciences Group , Merignac , France ) , with various segmented regions stitched together in VGStudio Max version 3 . 2 ( Volume Graphics , Heidelberg Germany ) . Detailed synchrotron scanning and data processing protocols are outlined in Appendix 1 . An exceptionally preserved specimen of the ornithischian dinosaur H . tucki reveals novel features of the anatomy of this taxon . Some of these features were previously unknown in Ornithischia , including a complete gastral basket with thin , single-element gastralia; bizarre , paddle-like sternal ribs with prominent condylar articular surfaces; and an apomorphic pair of well-developed sternal plates . Other features present in the specimen have a sporadic distribution in Ornithischia , including an ossified suprascapula and clavicles . These findings support the basal position of H . tucki and further support its importance as a transitional taxon showing the early evolution of iconic ornithischian anatomical features . Using SRμCT scans of the specimen , we were also able to observe in H . tucki the lack of PSP in the vertebral column , a smooth posterior thoracic ceiling , and a relatively small APP . We conducted a quantitative analysis of relative sizes of pelvic girdle elements and showed that the APP alone evolved in a manner consistent with the predictions of an ‘Early-Burst’ model , increasing markedly in proportional size early in the diversification of Ornithischia , and then remaining relatively large in all ornithischian lineages , with lower rates of change . These results are explained by a model for the evolution of the ornithischian ventilatory apparatus , in which the lineage undergoes a shift from a hypaxial-dominated system of volumetric change to a system where the lungs are ventilated by a novel pelvic muscle attached to the APP – a muscle functionally analogous to the dorsal component of M . diaphragmaticus of extant crocodilians . H . tucki preserves evidence for a critical transition in dinosaurs , demonstrating how key innovations evolve , showing how they can have pervasive effects on multiple anatomical systems , and providing a possible explanation for the success and longevity of a major lineage of dinosaurian herbivores . | The fossilised skeletons of long extinct dinosaurs are more than just stones . By comparing these remains to their living relatives such as birds and crocodiles , palaeontologists can reveal how dinosaurs grew , moved , ate and socialised . Previous research indicates that dinosaurs were likely warm-blooded and also more active than modern reptiles . This means they would have required breathing mechanisms capable of supplying enough oxygen to allow these elevated activity levels . So far , much of our insight into dinosaur breathing biology has been biased towards dinosaur species more closely related to modern birds , such as Tyrannosaurus rex , as well as the long-necked sauropods . The group of herbivorous dinosaurs known as ornithischians , which include animals with head ornamentation , spikes and heavy body armour , like that found in Triceratops and Stegosaurus , have often been overlooked . As a result , there are still significant gaps in ornithischian biology , especially in understanding how they breathed . Radermacher et al . used high-powered X-rays to study a new specimen of the most primitive ornithischian dinosaur , Heterodontosaurus tucki , and discovered that this South African dinosaur has bones researchers did not know existed in this species . These include bones that are part of the breathing system of extant reptiles and birds , including toothpick-shaped bones called gastralia , paired sternal bones and sternal ribs shaped like tennis rackets . Together , these new pieces of anatomy form a complicated chest skeleton with a large range of motion that would have allowed the body to expand during breathing cycles . But this increased motion of the chest was only possible in more primitive ornithischians . More advanced species lost much of the anatomy that made this motion possible . Radermacher et al . show that while the chest was simpler in advanced species , their pelvis was more specialised and likely played a role in breathing as it does in modern crocodiles . This new discovery could inform the work of biologists who study the respiratory diversity of both living and extinct species . Differences in breathing strategies might be one of the underlying reasons that some lineages of animals go extinct . It could explain why some species do better than others under stressful conditions , like when the climate is warmer or has less oxygen . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2021 | A new Heterodontosaurus specimen elucidates the unique ventilatory macroevolution of ornithischian dinosaurs |
Adenylate cyclases convert intra- and extracellular stimuli into a second messenger cAMP signal . Many bacterial and most eukaryotic ACs possess membrane anchors with six transmembrane spans . We replaced the anchor of the AC Rv1625c by the quorum-sensing receptor from Vibrio harveyi which has an identical 6TM design and obtained an active , membrane-anchored AC . We show that a canonical class III AC is ligand-regulated in vitro and in vivo . At 10 µM , the cholera-autoinducer CAI-1 stimulates activity 4 . 8-fold . A sequence based clustering of membrane domains of class III ACs and quorum-sensing receptors established six groups of potential structural and functional similarities . The data support the notion that 6TM AC membrane domains may operate as receptors which directly regulate AC activity as opposed and in addition to the indirect regulation by GPCRs in eukaryotic congeners . This adds a completely novel dimension of potential AC regulation in bacteria and vertebrates .
In 1958 Sutherland and Rall reported the structure of a second messenger , cyclic 3‘ , 5‘-adenosine monophosphate ( cAMP ) which was generated upon incubation of a liver extract with the first messengers epinephrine or glucagon ( Sutherland and Rall , 1958 ) . Since then cAMP has been demonstrated to be a universal second messenger translating a variety of extracellular stimuli into a uniform intracellular chemical signal . The enzymes responsible for biosynthesis of cAMP from ATP , adenylate cyclases ( ACs ) , have been biochemically and genetically identified in most bacterial and eukaryotic cells ( Khandelwal and Hamilton , 1971; Linder and Schultz , 2003; 2008 ) . To date , sequencing has identified six classes of ACs . The small-sized class I ( enterobacterial ACs ) , class II ( toxin class ) and the minor classes IV-VI are restricted to bacteria ( Bârzu and Danchin , 1994; Linder and Schultz , 2003 ) . The class III ACs are ubiquitous , albeit with differing domain architectures ( Linder and Schultz , 2008 ) . The catalytic domains share sequence and structural similarities , yet minor , characteristic sequence peculiarities have resulted in a division into four subclasses , a-d ( Linder and Schultz , 2008; Tesmer et al . , 1997; Tews et al . , 2005 ) . A fundamental difference between bacterial and eukaryotic ACs is that the former are monomers which require homodimerization for activity . The mammalian congeners , exclusively class IIIa , present themselves as pseudoheterodimers composed of two concatenated ‘bacterial’ monomers with slightly diverged , yet complementary domains ( Guo et al . , 2001 ) . Accordingly , mammalian class III ACs are anchored to the membrane by two putative 6TM bundles , one in each of the concatenated repeats . Our knowledge about regulation of bacterial class III ACs is limited . Apart from a few soluble ACs which appear to be regulated by carbon dioxide or pH near to nothing is known ( Kleinboelting et al . , 2014; Steegborn et al . , 2005; Tews et al . , 2005 ) . The established regulation of the nine mammalian , membrane-delimited AC isoforms is indirect . Stimulation of G-protein-coupled receptors ( GPCRs ) by extracellular ligands releases Gsα intracellularly which binds to ACs and activates . A potentially direct ligand-regulation of class III ACs via their large membrane anchors remains a genuine possibility . The 6TM membrane anchors of bacterial ACs are obviously structural analogs of the 6TM bundles in mammalian ACs . In the past , we replaced the membrane anchor of the mycobacterial AC Rv3645 by the E . coli chemotaxis receptors for serine , Tsr , and aspartate , Tar ( Kanchan et al . , 2010 ) . Tsr/Tar and Rv3645 have a signal-transducing HAMP domain in common and both require dimerization . The chimeras were regulated in vitro and in vivo by serine or aspartate ( Kanchan et al . , 2010; Mondéjar et al . , 2012; Winkler et al . , 2012 ) , i . e . a 2TM receptor , Tsr or Tar , with an extensive periplasmic ligand-binding domain replaced a 6TM AC membrane anchor which lacks periplasmic loops . The data demonstrated that in principle direct regulation of a class III AC via an extracellular ligand is a possibility . The question whether also a 6TM receptor might directly regulate a 6TM AC remained open . This question is addressed here . Membrane anchors with 6TMs are present in many proteins . Often they have short transmembrane-spanning α-helices and short connecting loops , e . g . in bacterial and mammalian ACs ( Krupinski et al . , 1989; Linder and Schultz , 2003 ) , in the cytochrome subunits of succinate dehydrogenases and fumarate reductases ( Hederstedt , 1998; Yankovskaya et al . , 2003 ) , ABC transporters ( Chang and Roth , 2001 ) , in bacterial HdeD proteins ( Mates et al . , 2007 ) , six transmembrane epithelial antigen of the prostate ( STEAP , Kleven et al . , 2015 ) , or quorum-sensing ( QS ) receptors from Vibrio and Legionella which have His-kinases as cytosolic effectors ( Ng and Bassler , 2009 ) . For the latter lipophilic ligands have been identified ( Ng et al . , 2011; 2010 ) . This has opened the opportunity to replace the 6TM anchor of the mycobacterial class IIIa AC Rv1625c which is considered to be an ancestral form of mammalian ACs ( Guo et al . , 2001 ) , by the prototypically identical 6TM QS-receptor CqsS from V . harveyi ( Figure 1 ) . Here , as a proof of principle , we demonstrate that a 6TM receptor not only substitutes a membrane-anchoring function , but also confers direct regulation of a class IIIa AC via an extracellular ligand with nanomolar potency . Taken together with a bioinformatic analysis the data indicate that the 6TM AC membrane anchors probably have an additional receptor function . This is a first and decisive step to add a novel dimension of direct regulation of class IIIa AC activities . 10 . 7554/eLife . 13098 . 003Figure 1 . Two-dimensional models of the canonical class IIIa adenylate cyclase Rv1625c from M . tuberculosis ( left ) and the quorum-sensing receptor from V . harveyi ( right ) in the membrane . Both proteins require dimerization to be catalytically active . The alignment below covers the amino acid sequences at the exit of TM6 of both proteins . The most efficient functional linkage of the CqsS receptor to the catalytic domain of Rv1625c is indicated by an arrow . The H-Box of the histidine-kinase domain is underlined . The numbering for CqsS and Rv1625c is indicated above and below the respective sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 003
Many bacterial and the nine membrane-bound mammalian class IIIa ACs possess 6TM modules as membrane anchors ( Guo et al . , 2001; Linder and Schultz , 2003; 2008 ) . They comprise >40% of the proteins . A function of these membrane domains beyond fixation in the membrane is unknown . Recently we became aware of the QS-receptors from Vibrio and Legionella , CqsS and LqsS , which feature a membrane anchor of an essentially identical design as the aforementioned ACs , i . e . minimal-length α-helices and short connecting linkers which presumably severely restrict conformational possibilities ( see Figure 1 for a 2D representation; Ng et al . , 2010; Tiaden and Hilbi , 2012 ) . For both QS-receptors highly lipophilic ligands have been identified such as CAI-1 , the Cholerae AutoInducer-1 , ( S ) -3-hydroxy-tridecan-4-one and LAI-1 , the Legionella AutoInducer-1 , ( S ) -3-hydroxy-pentadecan-4-one ( Ng et al . , 2010; Spirig et al . , 2008 ) . The QS-receptors from Vibrio and Legionella are homodimers linked to histidine-kinases as cytosolic effector domains ( Ng and Bassler , 2009 ) . This is comparable to bacterial class III ACs which are homodimers ( Linder and Schultz , 2003 ) . These superficial observations suggested that by swapping the membrane anchor/receptor between a 6TM AC and CqsS from Vibrio one might generate a 6TM AC which is regulated by the QS-ligand CAI-1 . We chose the mycobacterial class IIIa AC Rv1625c for this investigation because of its similarity to the mammalian congeners ( Guo et al . , 2001 ) and the QS-receptor from V . harveyi . The success of generating a functionally productive chimera between the AC and the QS-receptor hinges on the precondition that a suitable point of transition between both proteins can be found which allows signal propagation from the CqsS receptor to the AC effector . The cytosolic aa sequences exiting from the respective last TMs have no conspicuous complementarity which would indicate a self-evident point of connection ( Figure 1 ) . Therefore , 15 different points of connection between the CqsS receptor and the catalytic domain of Rv1625c were probed . The first points of transition tested were the arginine residues present at the cytosolic membrane exit in both proteins ( Figure 1 ) . The chimera CqsS1-168Rv1625c203-443 was active ( 16 . 2 nmol cAMP·mg-1·min-1 ) , but unregulated . The basal AC activity of this chimera was comparable to that of the membrane-bound Rv1625c holoenzyme ( Guo et al . , 2001 ) , i . e . generally the two disparate domains were functionally fully compatible with each other . In the cyanobacterial class IIIa AC CyaG from Arthrospira platensis a distinct N-terminal domain that starts with RSEELL , was required for a functional interaction with the chemotaxis receptor Tsr ( Winkler et al . , 2012 ) . A comparison between CyaG and Rv1625c AC sequences revealed a similar domain in Rv1625c beginning with RSEALL ( Figure 1 ) . Hence , for the Rv1625c AC this point of transition to CqsS was chosen . In the CqsS His-kinase the auto-phosphorylated histidine is part of the canonical H-box ( underlined in Figure 1; Grebe and Stock , 1999 ) . However , several chimeras of CqsS and Rv1625c linked in this region were unaffected by CAI-1 . With transition points closer to the membrane exit of the CqsS receptor , e . g . at Val172 , Ala181 or Gly185 , AC activities were reproducibly stimulated by CAI-1 ( not shown ) . For further experiments we linked Ala181 of the QS-receptor to Arg218 of the Rv1625c AC , generating CqsS1-181-Rv1625c218-443 ( abbreviated CqsS-Rv1625c; Figure 1 ) because it responded maximally . The chimera CqsS-Rv1625c was stimulated by 85% with 10 µM CAI-1 ( Figure 2 ) . The response was concentration-dependent and the EC50 concentration for CAI-1 was 400 nM ( Figure 2 ) . Discrimination between CAI-1 with a C9 and LAI-1 with a C11 lipid tail was absent . In both cases maximal stimulation at 10 µM CAI-1 and the EC50 concentrations were identical ( Figure 2 ) . These parameters perfectly matched the concentrations of 400 nM required to observe phenotypic responses from the CqsS-His-kinase in V . cholerae ( Ng et al . , 2011 ) . 10 . 7554/eLife . 13098 . 004Figure 2 . Stimulation of the chimera CqsS1-181-Rv1625c218-443 by the QS-ligands CAI-1 or LAI-1 . Basal activity was 5 . 5 nmol cAMP·mg-1·min-1 . The EC50 concentrations were 400 nM . Filled squares , CAI-1 ( n = 5–12; ± S . E . M . ) ; open squares , LAI-1 ( n= 1–2 ) . CAI-1 stimulations were significant starting at 100 nM ligand . The insert shows a Western blot of the expression product with MW standards indicated at the side . The structure of the ligands is depicted at right . The catalytic domain of Rv1625c alone was not affected by CAI-1 or LAI-1 ( not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 004 In vivo characterization of CqsS from V . cholerae identified Cys170 at the exit of TM6 as critical for signaling ( Ng et al . , 2011 ) . In CqsS from V . harveyi the corresponding residue is Phe166 ( Figure 1 ) . We examined its role by substituting it with all 19 proteinogenic amino acids . Substitutions by large flexible hydrophobic residues , i . e . Leu , Ile , or Met enhanced CAI-1 stimulation up to 480% compared to 185% with the parent Phe166 construct ( compare Figures 2 and 3 ) . Furthermore , a high discrimination between CAI-1 and LAI-1 was effectuated and the potency ratio CAI-1/LAI-1 at 10 µM increased from 1 to 3 ( see Figures 2 and 3 ) . Evidently , the F166L CqsS sensor domain in the chimera shows high selectivity between CAI-1 and LAI-1 . Our findings were in line with earlier results that the natural CqsS receptor preferably detects ligands with C10 or C8 but not shorter ( C6 , C4 ) or longer ( C12 , e . g . LAI-1 ) alkyl tails ( Tiaden and Hilbi , 2012 ) . 3 , 4-tridecanediol which has not been reported as a natural ligand stimulated by about 10% of CAI-1 at 10 μM ( Figure 3 ) . Other amino acid substitutions at position 166 of CqsS did not affect the extent of AC stimulation with the exception of tryptophan which abrogated stimulation ( data not shown ) . Activation enhanced Vmax from 14 to 42 nmol cAMP·mg-1·min-1 whereas Km for substrate ATP was not significantly affected ( 233 and 163 µM ATP , respectively ) . An up to 104-fold activation of CqsS-stimulated reporter gene transcription was earlier observed in V . harveyi and V . cholerae ( 21 ) . The 5-fold activation of CqsS-F166L reported here appears comparatively small . This discrepancy can be explained by the fact that the amplification systems used differ profoundly . Earlier studies investigated the quorum-sensing system in vivo with a reporter gene transcription/bioluminescence read-out which is much more sensitive than an in vitro AC assay with isolated cell membranes used in the present study ( Ng et al . , 2011 ) . In addition , the coupling of CqsS to its authentic effector might well be more stringent than that attainable in a chimera with an exogenous Rv1625c AC output domain . 10 . 7554/eLife . 13098 . 005Figure 3 . Stimulation of the chimera CqsS1-181F166L-Rv1625c218-443 by the QS-ligands CAI-1 or LAI-1 . Basal activity was 4 nmol cAMP·mg-1·min-1 . Filled squares , CAI-1 ( n = 5–11; ± S . E . M . ) ; open squares , LAI-1 ( n=2 ) ; open circles , 3 , 4-tridecanediol . The EC50 concentrations were 400 nM CAI-1 , 900 nM LAI-1 , and 2000 nM 3 , 4-tridecanediol . CAI-1 stimulations were significant starting at 100 nM ligand . Insert: Western blot of expression product . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 005 Next we examined whether the CqsS-Rv1625c AC chimera is operational in vivo . Use of maltose by E . coli requires activation of the maltose operon via the cAMP/CRP signaling system . Maltose fermentation produces organic acids which are visualized on MacConkey plates by the pH indicator phenol red . We used the AC-deficient E . coli cya-99 strain with a high affinity CRP variant ( Garges and Adhya , 1985 ) . It cannot metabolize carbohydrates for lack of cAMP . When grown on MacConkey agar , colonies appear whitish . We transformed the CqsS-F166L-Rv1625c into E . coli cya-99 , plated the cells on MacConkey maltose agar and induced AC expression by a filter strip soaked with 30 mM IPTG . The reddish zone along the filter strip which is indicative of maltose metabolism was expanded at the side where 10 µl of a 100 µM CAI-1 solution was applied ( Figure 4 ) , clearly demonstrating a CAI-1 stimulated cAMP production in vivo . 10 . 7554/eLife . 13098 . 006Figure 4 . CAI-1 stimulates cAMP formation in vivo . A MacConkey maltose agar plate with E . coli cya-99 crp*144 transformed with CqsS1-181F166L-Rv1625c218-443 was induced by a filter strip soaked with 1 mM IPTG ( running from top to bottom in the middle ) . 10 µl of 100 µM CAI-1 in DMSO/water was spotted at the asterisk , the plate was tipped and the solution was allowed to move left . As a surface active compound it regularly spread over a large area . Note that the bacterial lawn at left was not induced . Picture was taken from the bottom of the Petri-dish ( three independent experiments were carried out , each with at least three agar plates and different concentrations of CAI-1 and IPTG; controls with solvent were negative ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 006 The monomeric bacterial class III ACs require homodimerization ( Linder and Schultz , 2003 ) . Similarly , bacterial His-kinases of two-component systems are homodimers in which a His-residue of the H-box is phosphorylated either in cis or trans ( Casino et al . , 2014; 2009 ) . Here we examined whether dimerization of the QS-receptors is required for AC regulation . The CqsS receptor was connected to known inactive Rv1625c AC point mutants , Rv1625cD300A and Rv1625cR376A . These point mutants complement each other and the dimer is catalytically active ( Guo et al . , 2001 ) . Thus , CqsS-Rv1625cD300A was inserted into pETDuet-3 alone or in combination with CqsS-Rv1625cR376A . When CqsS-Rv1625cD300A and CqsS-Rv1625cR376A were jointly expressed , robust AC activity and regulation by CAI-1 was observed ( Figure 5A ) . This demonstrated that the membrane anchors of CqsS dimerize and allow individually inactive AC domains to form an active , ligand-regulated dimer . Expectedly , the expression product of CqsS-Rv1625cD300A was inactive ( Figure 5B ) . The experiment did not answer the question whether ligand binding requires a CqsS dimer . This was addressed by linking the inactive Rv1625c monomers to the 6TM anchors of either Rv1625c or the Legionella LqsS QS-receptor . First , CqsS-Rv1625cD300A and the full-length Rv1625cR376A were jointly expressed in pETDuet-3 . AC activity was observed , yet CAI-1 did not regulate ( Figure 5C ) . Second , the QS-receptor from LqsS was joined with the inactive Rv1625cR376A in a similar manner . CqsS-Rv1625cD300A and LqsS-Rv1625cR376A were concomitantly expressed . As before , AC activity was restored , yet regulation by CAI-1 was absent ( Figure 5D ) . This indicated that the membrane anchors were close enough to enable productive heterodimerization; however , a regulatory ligand-binding site was absent . Possibly one ligand molecule binds at the interface of a receptor homodimer as is the case in the chemotaxis receptors Tsr or Tar or Ni2+-binding in PhoQ ( Cheung et al . , 2008; Gardina and Manson , 1996; Kanchan et al . , 2010; Mowbray and Koshland , 1990; Yang et al . , 1993 ) . Because the highly lipophilic ligand does not allow meaningful receptor binding studies , presently this cannot be examined any further . 10 . 7554/eLife . 13098 . 007Figure 5 . Homodimerization of the CqsS receptor is required for signaling . ( A ) with complementary Rv1625c point mutations Rv1626cD300A and Rv1625cR376A a regulated dimeric chimera was generated ( *p<0 . 05 compared to respective basal activity ) . ( B ) as a control the construct CqsS-Rv1625cD300A was expressed alone . It was inactive . ( C , D ) complementing mutants with differing membrane domains were active , yet unregulated . Basal activity of construct B significantly differed from those in constructs A , C , and D ( †p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 007 The QS-system of Vibrio controls virulence . At high cell density , CAI-1 is produced , released and binds to the extracellular CqsS receptor ( Wei et al . , 2012 ) . In a multistep intracellular process this results in reduced production of virulence factors and allows the pathogen to escape from the host , thus spreading disease ( Higgins et al . , 2007; Rutherford and Bassler , 2012 ) . Because extracellular loops for ligand binding are absent in CqsS and because of the lipophilicity of the ligand , CAI-1 may bind within the membrane segments of the receptor dimer . Therefore , the question whether CAI-1 stimulation is reversible was examined next . Membranes were stimulated with sub-saturating ( 300 nM ) and saturating ( 10 µM ) concentrations of CAI-1 for 10 min . The reactions were rapidly stopped by cooling to 0°C and the pre-stimulated membranes were re-isolated by ultracentrifugation . Those membranes were then stimulated again by CAI-1 ( Figure 6 ) . Membranes pre-treated with 300 nM CAI-1 had an elevated ‘basal’ AC activity equivalent to the previous 0 . 3 µM CAI-1 stimulation ( Figure 6 , left ) . Accordingly , re-stimulation by 0 . 3 µM CAI-1 failed whereas addition of 10 µM CAI-1 activated to the maximal possible extent . Membranes which were initially exposed to 10 µM CAI-1 remained almost fully activated and were completely refractory to re-stimulation ( Figure 6 ) . In this context we checked whether ligand was specifically binding at the receptor sites or remained unspecifically associated with the membranes . The supernatants of the ultracentrifugation steps of the pre-incubated membranes stimulated naïve membranes according to the previously tested CAI-1 concentrations . This excluded unspecific association of the lipophilic ligand with the membrane or adherence to the reaction vessels . Therefore , we can conclude that CAI-1 stimulation was irreversible . This is slightly reminiscent on the biochemistry of rhodopsin in the mammalian eye . There , retinal is even covalently bound into the membrane-segments of opsin , a GPCR with scant extra-membrane loops . After excitation receptor regeneration requires ligand removal by an enzymatic process and transport to the retinal pigment epithelium which , notably , contains exclusively adenylate cyclase type VII ( Völkel et al . , 1996 ) . How in Vibrio signal termination is accomplished remains to be investigated . Possibilities are an inactivating metabolism of CAI-1 or proteolysis . 10 . 7554/eLife . 13098 . 008Figure 6 . CAI-1 ligand binding to the CqsS QS-receptor is irreversible . Membranes containing CqsS1-181F166L-Rv1625c218-443 were stimulated with 0 . 3 or 10 µM CAI-1 ( left ) , re-isolated and re-stimulated with 0 . 3 and 10 µM CAI-1 . Only the stimulations marked with an asterisk differed significantly from the respective unstimulated controls . White bars represent basal AC activities , gray bars on top represent additive CAI-1-stimulated activities ( S . E . M . , n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 008 The above data demonstrated that the homodimeric catalytic domain of the canonical class IIIa Rv1625c AC was capable to decode the ligand-initiated conformational signal of a 6TM QS-receptor and translate it into a change in AC activity . The functional membrane anchor replacement accompanied by a gain of a novel physiological function suggests that the Rv1625c 6TM anchor actually constitutes an orphan receptor , i . e . receptor without known ligand . Therefore , we examined the relationship between the QS-receptors and the 6TM AC membrane anchors by a focused bioinformatic approach . From the Uniprot Reference Proteomes databank we obtained all class III ACs using the highly conserved catalytic domain for data mining ( see Materials and methods ) . The catalytic domains were stripped; the putative 6TM domains were extracted and retained . Similarly , we extracted 6TM modules related to the CqsS quorum-sensing receptor . The combined data set comprising a total of 1616 6TM modules was subjected to a cluster analysis using the CLANS software ( Figure 7A; Frickey and Lupas , 2004 ) . In total , six distinct clusters were observed , five of which were interconnected to different extents , among them the cluster of QS-receptors ( Figure 7A ) . This clear-cut separation of 6TM modules is surprising and revealing . Transmembrane spanning α-helices have a strong predilection for hydrophobic residues with Leu , Ile , Val , Ala , Gly and Phe comprising two-thirds ( Senes et al . , 2000 ) . Yet , the unequivocal clustering indicates a differentiated pattern of conserved 6TM variants which is at odds with an assignment of simply an anchoring device . We propose that the diversity of conserved patterned subtypes mirrors a succinct adaptation to particular physiological functions . By far the majority of 6TM modules were eukaryotic ACs , from the slime mold Dictyostelium , the extant coelacanth up to man . The preponderance of eukaryotic sequences is , at least in part , due to the fact that in vertebrates nine distinct membrane-delimited pseudoheterodimeric AC isoforms exist , each with two TM modules , TM1 and TM2 . The observed clusters were identified as follows: ( 1 ) An isolated , unconnected cluster of 30 bacterial class IIIb ACs is characterized by a ferredoxin between the membrane and the catalytic domain . A representative example is an AC from Rhodopseudomonas palustris ( Uniprot Q132R4 ) which intracellularly appends to a ferredoxin module . The membrane anchors show sequence similarity to the cytochrome b subunits present in the fumarate reductase/succinate dehydrogenase protein families to an extent that they can be matched by simple BLAST searches . These 6TM modules have four precisely spaced intra-membrane histidine residues which coordinate heme as an electron carrier ( Einsle et al . , 2000; Hederstedt , 1998; Kern et al . , 2010a; 2010b ) . It can confidently be predicted that the AC 6TM anchors of this cluster will turn out to contain two heme entities as prosthetic groups . ( 2 ) A cluster of 118 bacterial class IIIb ACs is characterized by a signal-transducing HAMP domain between membrane anchor and catalytic domain such as the well-studied mycobacterial AC Rv3645 ( Hazelbauer et al . , 2008; Hulko et al . , 2006; Kanchan et al . , 2010; Mondéjar et al . , 2012 ) . ( 3 ) A cluster of 42 bacterial class IIIb ACs shows partly high local similarity to ACs of cluster ( 2 ) as visualized by the number and intensity of the gray colored connecting lines between both clusters ( Figure 7A ) . Yet , ACs in cluster ( 3 ) lack a HAMP domain . To our knowledge , none of these ACs has ever been interrogated experimentally . ( 4 ) A cluster of 35 membrane anchors from bacterial class IIIa ACs as prototypically represented by the mycobacterial AC Rv1625c used here . It is predominantly connected to the cluster with mammalian TM modules . ( 5 ) Two tightly connected groups are combined here into one cluster of 677 TM1 and 653 TM2 modules which are derived from eukaryotic , mostly vertebrate , pseudoheterodimeric class IIIa ACs . The difference in the number of TM1 and TM2 units is due to individual sequence divergences of the TM2 domains , database miss-annotations , and problems in the automated prediction of transmembrane spans . ( 6 ) A cluster of 56 6TM modules corresponds to quorum-sensing receptors of the CqsS type . 10 . 7554/eLife . 13098 . 009Figure 7 . ( A ) Cluster map of 6TM domains of adenylate cyclases and CqsS-like sensory receptors . A comprehensive set of 6TM AC anchors was extracted from 408 eukaryotic and 1456 bacterial proteomes and clustered in CLANS ( HHsearch p-value cutoff 5E-4 , attraction value 10 , repulsion value 5 ) ; outliers were removed . Each dot represents a single 6TM domain . Above threshold HHsearch hits are shown as connecting lines between AC pairs in different clusters; the darker line color , the more similar the protein sequences . 6TM anchors of ACs form five clusters of high pairwise sequence similarity: cluster ( 1 ) , anchors of bacterial class IIIb ACs characterized by the presence of a cytosolic ferredoxin domain ( 30 sequences from the α and β branches of proteobacteria ) . Cluster ( 2 ) , anchors of bacterial class IIIb ACs characterized by a signal-transducing HAMP domain ( 118 sequences mainly from Actinobacteria , but also from α-proteobacteria , δ-proteobacteria , Chlorobia and Thermoleophilia ) . Cluster ( 3 ) , anchors of bacterial class IIIb ACs similar to HAMP-associated anchor domains but which lack a HAMP domain ( 42 sequences mainly from α-proteobacteria ) . Cluster ( 4 ) , anchors from bacterial class IIIa ACs prototypically represented by the mycobacterial AC Rv1625c ( 35 from many different phyla of bacteria , including Actinobacteria , Proteobacteria , Chlorophyta , Spirochaetes , and Bacteriodetes ) . The enlarged asterisk denotes the position of the mycobacterial AC Rv1625c . Cluster ( 5 ) , anchors of the pseudoheterodimeric eukaryotic class IIIa ACs ( TM1 677 sequences , TM2 653 sequences ) . CqsS Cluster ( 6 ) , 6TM domains of sensory His-Kinases similar to CqsS ( from Bacteriodetes , Chlorobia , α- , β- , and γ-proteobacteria ) . ( B ) Length comparisons of the transmembrane helices and loops of 6TM membrane anchors/receptors/sensors . The data sets from clusters 4 , 5 and 6 from Figure 7A were used supplemented with 250 cytochrome b561 proteins . In case no S . E . M . is visible the size of the vertical bar is within the line thickness of the respective bar . ‘α’-Numbering denotes the consecutive TM helices starting from the N-terminus , 'out' and 'in’ denote sequential extra- and intracellular loop sequences . The two horizontal lines are at the 5 and 20 aa level . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 00910 . 7554/eLife . 13098 . 010Figure 7—figure supplement 1 . Length comparisons of α-helices and loops of 6TM modules from adenylate cyclases and quorum sensors . Data are means ± S . E . M . of the sequence groups of all clusters shown in Figure 7A as indicated below . The color-coding used in Figure 7A has been conserved . The designations ‘in’ and ‘out’ indicate intra- and extracellular loops . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 01010 . 7554/eLife . 13098 . 011Figure 7—figure supplement 2 . The 12 sequences used for the alignment were from Agrobacterium albertimagni , WP_006725538; Arthrospira maxima , B5VUZ0; Arthrospira platensis , D5A5G2; Beggiatoa , A7BXS6; Dechloromonas , Q47AI8; Hyphomicrobium , C6QBG1; Lyngbya , A0YQ82; Mesorhizobium , LSHC420B00; Microcoleus sp . , PCC_7113; Oscillatoria acuminata , PCC_6304; Nostoc , YP_001866931; Mycobacterium tuberculosis , ALB18789 ( Rv1625c ) . The alignment was made by ClustalW and adjusted and converted into a ‚bargraph ‘style alignment in Genedoc . Black lines reflect sequence identity , shades of grey different degrees of similarity , and white patches denote sequence diversity . Note the striking dissimilarity of the TM domains and the conservative nature of the catalytic domains . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 01110 . 7554/eLife . 13098 . 012Figure 7—figure supplement 3 . AC_1 Sequences used for the alignment: Macaca fasci , XP_005551359 . 1; Bos taurus , NP_776654 . 1; Anas platyrhynchos , XP_005014606 . 1; Danio rerio , NP_001161822 . 1; Mesocricetus auratus , XP_005083033 . 1; Pseudopodoces humilis , XP_005525280 . 1; Gallus gallus , XP_418883 . 4; Homo sapiens , NP_066939 . 1; Mouse , GI:62512159; Heterocephalus glaber , XP_004840600 . 1; Orcinus orca , XP_004283063 . 1; Pan troglodytes , XP_519081 . 3; Ficedula albicollis , XP_005041477 . 1; Sarcophilus harrisii , XP_003762599 . 1; Odobenus rosmarus divergens , XP_004409496 . 1; Melopsittacus undulatus , XP_005148091 . 1 . AC_2 sequences used for the alignment: Jaculus jaculus , XP_004670691 . 1; Ictalurus punctatus , AHH38946 . 1; Myotis brandtii , EPQ02509 . 1; Anas platyrhynchos , XP_005024930 . 1; Monodelphis domestica , XP_001363692 . 1; Ornithorhynchus anatinus , XP_001519046 . 2; Macaca mulatta , NP_001252581 . 1; Rattus norwegicus , NP_112269 . 1; Rabbit , XP_002721866 . 1; Latimeria chalumnae , XP_006007585 . 1; Mouse , Q80TL1; Callithrix jacchus , XP_002745178 . 1; Homo sapiens , Q08462 . 5; Ficedula albicollis , XP_005041747 . 1; Columba livia , XP_005508897 . 1; Dog , XP_535798 . 3; Bos taurus , XP_587884 . 4; Myotis brandtii , EPQ02509 . 1; Alligator mississippiensis , XP_006274555 . 1 . On top are the joint alignments of TM1 and TM2 of AC1 and AC2 , respectively . Similarities are limited . Deconstruction of the AC1 and AC2 alignment results in individual AC1 and AC2 alignments . The similarities within one isoform throughout evolution are striking . DOI: http://dx . doi . org/10 . 7554/eLife . 13098 . 012 The cluster of the eukaryotic TM1 domain ( from cluster 5 ) was connected to the clusters of the bacterial type IIIa ( 4 ) and type IIIb ( 3 ) by a large number of pairwise matches , typically covering all six transmembrane spans . In contrast , the CqsS cluster had significantly fewer and only local matches , aligning quorum-sensing sequences specifically to transmembrane spans 1 and 2 of individual eukaryotic AC TM1 sequences ( all from the bony fish Osteichtyes ) , and to TM 3 , 4 and 5 of the bacterial type IIIa ACs , each at approximately 20–25% identity . Notably , the average sequence identity within the clusters themselves is approximately 40% . Thus , currently we can only speculate whether the observed similarities indicate remote homology at the limit of recognition or convergence due to identical structural and functional constraints . Visual inspection of sequence properties of single TMs indicated a remarkable shortness of α-helices and respective interconnecting loops already apparent in the 2D presentations ( Figure 1 ) . Notably , the loops are by far shorter than those of comparable sensory proteins , such as mammalian GPCRs with seven TMs , bacterial chemotaxis receptors , His-kinases and ACs with 2 TMs or 4 TMs such as the chase domain ( cyclase/histidine kinases associated sensory extracellular ) . To support this notion we investigated the length parameters of the transmembrane segments and loops of AC and CqsS anchor types which were used in the cluster analysis , and compared them to 250 orthologues of cytochrome b561 ( Figure 7B and Figure 7—figure supplement 1 ) . The latter protein was selected because a high-resolution structure is available , i . e . they form a 2x6TM homodimer of short TM spans and loops analogous to those from Rv1625c and CqsS ( Lu et al . , 2014 ) . The lengths of the six TM segments and of the five loops were well conserved between orthologues of the same anchor type and highly similar between anchors of the bacterial AC classes IIIa and IIIb , eukaryotic TM1 , and , most surprisingly , CqsS-like QS-receptors ( Figure 7B ) . On the other hand , the eukaryotic TM2 anchors , those of the ferredoxin-associated class IIIb , and anchors coupled to a cytosolic HAMP-domain ( class IIIb ACs in cluster 2 ) had one or even several elongated loops or α-helices ( Figure 7 and Figure 7—figure supplement 1 ) . The short helices of around 20 residues in this 6TM module design must cross the 30 Å thick lipid bilayer almost orthogonally and the short connecting loops restrict the number of positional permutations foreshadowing a compact packing . Therefore , the structures of transmembrane domains of eukaryotic AC TM1 , the bacterial class IIIa ACs such as Rv1625c , and QS-receptors of the CqsS type can be reasoned to possess overall structural similarities with that of cytochrome b561 .
In 1989 the analysis of the first aa sequence of a mammalian AC suggested that the two membrane anchors , each consisting of six transmembrane segments , might carry a transporter or channel function ( Krupinski et al . , 1989 ) . Since then , our knowledge concerning the membrane anchors has not advanced; to date no function for approximately 40% of a class IIIa AC protein sequence has been identified which goes beyond a mere membrane fixation . After 26 years we demonstrate for the first time that a canonical class IIIa AC with a 6TM membrane anchor is directly regulated by a membrane receptor of an identical design , yet with a known ligand , the QS-receptor from V . harveyi . Our data raise novel questions concerning the evolution of the regulation of class III ACs with 6TM membrane-anchoring modules . In absence of bacterial G-proteins bacterial class III ACs probably will turn out to be directly regulated via their prominent membrane anchors with the above mentioned R . palustris AC as an emerging example . In fact the membrane anchors of the class IIIa bacterial ACs are highly diverged whereas the catalytic domains are conserved ( Figure 7 and Figure 7—figure supplement 2 ) . This suggests that in bacteria the TM domains have evolved independently and very rapidly relative to the catalytic domains . This is in agreement with the general observation that mutations in upstream regulatory domains mostly are neutral and over time a variety of ligand specificities have evolved by chance mutations . At the same time even slightly detrimental mutations in downstream effector domains are not tolerated , thus resulting in sequence and functional conservation ( Schultz and Natarajan , 2013 ) . In such an evolutionary scenario the mechanisms of signal transduction per se remain intact and allow revealing combinatorial diversity as demonstrated here ( Schultz et al . , 2015 ) . In contrast , in vertebrates the membrane-delimited ACs are regulated indirectly by GPCR activation which intracellularly results in release of Gα proteins . Is this the result of a loss of direct ligand regulation during evolution while indirect GPCR regulation evolved or has ligand regulation of mammalian ACs been missed so far ? Because the AC membrane bundles of vertebrates are highly conserved in an isoform-specific manner from the coelacanth to man ( Figure 7 and Figure 7—figure supplement 3 as an example of mammalian AC_I and AC_II subtypes ) one can reasonably assume that they have an indispensable physiological function . This could be a particular compartmentalized membrane localization e . g . described in ( Crossthwaite et al . , 2005 ) or , in our view more likely and in accordance with our findings an as yet hidden receptor function , as a direct ligand-binding module or in conjunction with an accessory membrane protein that operates as the true sensor . The second alternative would suggest that in vertebrates regulation of intracellular cAMP concentrations is subject to an interaction between direct , ligand-mediated and indirect GPCR-Gsα-regulated effects . A lasting receptor occupation by a ligand during different physiological states might well set the responsiveness of mammalian AC isoforms to a transient GPCR-Gsα activation . This would allow that lasting and transient physiological conditions converge via direct and indirect regulation in a central second messenger system , a situation absent in bacteria . This concept poses the questions whether in vertebrates suitable extracellular signals exist and whether the molecular provisions for a direct signal transduction through the 9 AC isoforms have been evolutionarily conserved . We are currently exploring these questions . Recently we demonstrated that we could functionally replace the bacterial AC 6TM anchors by the E . coli chemotaxis receptors for serine or aspartate ( Kanchan et al . , 2010; Schultz et al . , 2015; Winkler et al . , 2012 ) . By analogy , the results supported the hypothesis of a receptor function for the AC membrane anchors . A less plausible interpretation would have been that the functional coupling might just be a manifestation of the modular composition of signaling proteins . The compatibility may only have required that a satisfactory domain order is maintained in such chimeras without invoking a functional relatedness . The data reported here add an entirely novel dimension to the working hypothesis that AC membrane anchors in bacterial homodimeric as well as in mammalian pseudoheterodimeric ACs function as ligand receptors . The 2D models of the membrane anchors of Rv1625c and CqsS are similar ( see Figure 1 ) . This should allow an almost isosteric replacement . Competent QS-receptor and cyclase chimeras were dependent on the point of linkage indicating that the connecting sites had been evolutionarily predesigned , were conserved and fully operational for signal transduction between functionally differing proteins ( Schultz and Natarajan , 2013 ) . Implicitly this supports the prediction that the 6TM anchors of such ACs have a receptor function for which stimuli have yet to be identified . Because the ligands for CqsS and LqsS are known , the direct regulation of AC activity then is no real surprise . Admittedly , based alone on the biochemical data one might again take pains to argue that the exchangeability of 6TM-anchors just extends the range of signaling modules which can structurally and possibly functionally replace each other . This alternative interpretation is rather implausible in our view . First mixing and matching TM domains of eukaryotic AC is impossible without loss of activity ( Seebacher et al . , 2001 ) . Second , the cluster analysis visualizes similarities between individual pairs of CqsS receptor type modules and membrane anchors from class IIIa ACs . Thus it supports the hypothesis that the membrane anchors of class III ACs , bacterial and mammalian alike have a function beyond membrane fixation . In this context cluster ( 1 ) , which is unrelated to all other clusters , is particularly interesting . It demonstrates that comparable transmembrane architectures can result in significantly different sequence patterns notwithstanding the similar amino acid composition commonly shared by all membrane domains . Currently , it is impossible to speculate about or even predict the nature of ligands for the 6TM AC modules examined here . One might expect , however , that they will be closely associated with the intra-membrane space as extra-membranous loops for ligand-binding are noticeably absent in this type of 6TM bundles .
In CqsS-Rv1625c chimeras the following CqsS receptor length variants were probed in different combinations: F168; V172; K177; A181; S183; G185; G187; I188; H190; P195;L196 . For AC Rv1625c catalytic domains the tested length variants were: A201 , L202 , R203 and R218 . Standard molecular biology methods were used for DNA manipulations ( primer sequences are in Supplementary file 1 ) . DNA fragments and vectors were restricted at their 5´BamHI or EcoRI and 3´HindIII sites and inserted into pQE80L ( Δ XhoI; Δ NcoI ) . When appropriate , silent restriction sites were introduced . All constructs carried an N-terminal His6-tag for detection in Western blots . In the pETDuet-3 vector the first MCS was been replaced by that of pQE30 introducing an N-terminal His6-tag . The second MCS in pETDuet-3 carried a C-terminal S-tag for Western blotting . The fidelity of all constructs was confirmed by double-stranded DNA sequencing . Constructs were transformed into E . coli BL21 ( DE3 ) . Strains were grown overnight in LB medium ( 20g LB broth/l ) at 37°C containing 100 μg/ml ampicillin . 200 ml LB medium ( with antibiotic ) was inoculated with 5 ml of a preculture and grown at 37°C . At an A600 of 0 . 3 , the temperature was lowered to 22°C and the expression was started with 500 μM isopropyl thio-β-D-galactoside ( IPTG ) for 2 . 5–5 hrs . Cells were harvested by centrifugation , washed once with 50 mM Tris/HCl , 1mM EDTA , pH 8 and stored at -80°C . For preparation of cell membranes cells were suspended in lysis buffer ( 50 mM Tris/HCl , 2 mM 3-thioglycerol , 50 mM NaCl , pH 8 ) containing complete protease inhibitor cocktail ( Roche Molecular , Mannheim , Germany ) and disintegrated by a French press ( 1100 p . s . i . ) . After removal of cell debris ( 4 . 300 x g , 30min , 4°C ) membranes were collected at 100000 x g ( 1h at 4°C ) . Membranes were suspended in buffer ( 40 mM Tris/HCl , pH8 , 1 . 6 mM 3-thioglycerol , 20% glycerol ) and assayed for AC activity . A more detailed description of protein expression and membrane preparation is available as described in detail at Bio-protocol ( Beltz and Schultz , 2016 ) . Adenylyl cyclase activity was determined for 10 min in 100 μl at 37°C ( Salomon et al . , 1974 ) . The reactions contained 5 µg protein , 50 mM Tris/HCl pH 8 , 22% glycerol , 3 mM MnCl2 , 6 mM creatine phosphate and 230 µg creatine kinase , 75 μM [α-32P]-ATP , and 2 mM [2 , 8-3H]-cAMP to monitor yield during cAMP purification . Substrate conversion was kept below 10% . The integrity of expressed recombinant membrane proteins was probed by Western blotting . Sample buffer was added to the membrane fractions and applied to SDS-PAGE ( 12 or 15% ) , in which proteins were separated according to size . For Western blot analysis , proteins were blotted onto PVPF membrane and examined with an RGS-His4-antibody ( Qiagen , Hilden , Germany ) or S-tag antibody ( Novagen R&D systems , Darmstadt , Germany ) and a 1:2500 dilution of the fluorophore conjugated secondary antibody Cy3 ( ECL Plex goat-α-mouse IgG-Cy3 , GE Healthcare , Freiburg , Germany ) . Detection was carried out with the Ettan DIGE Imager ( GE Healthcare ) . In general , proteolysis of expressed proteins was not observed . The helix and loop lengths were measured using the sequences taken from the respective clusters of the CLANS analysis and cytochrome b561 homologs . Predicted transmembrane spans were based on manually refined Polyphobius predictions . All experiments were repeated at least thrice . Data are presented as means ± S . E . M . when applicable . Student’s t test was used . | Cells are surrounded by a membrane that separates the inside of the cell from the external environment . To communicate information across the cell membrane , cells often employ a relay system . In this system , receptor proteins on the surface of the cells sense information about the environment and trigger the production of a chemical signal inside the cell . Certain receptors activate enzymes called adenylate cyclases , which reside just inside the cell , to produce a chemical signal . In some human and bacterial adenylate cyclases , about 40% of the protein is anchored in the membrane , far more than is necessary to hold the protein in place . It is therefore possible that this “membrane anchor” region plays an additional role , perhaps even detecting external signals . A “quorum sensing” receptor protein that was recently discovered embedded in the membrane of a species of marine bacteria called Vibrio harveyi has a similar structure to the membrane anchor of adenylate cyclases . Beltz et al . have now replaced the adenylate cyclase membrane anchor with a V . harveyi receptor . This produced a hybrid protein that could both receive and translate signals from the membrane receptor . A computational analysis of the membrane anchors of adenylate cyclases showed that they have striking similarities to quorum sensors . Furthermore , the membrane anchors of different types of adenylate cyclase have diverse structures that may have helped the cyclases to adapt to different environments and biological requirements . Overall , Beltz et al . ’s results suggest the adenylate cyclase membrane anchor is a new type of cell surface receptor . In the future it will be important to identify the environmental signal that activates adenylate cyclases , both in bacteria and mammals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | Regulation by the quorum sensor from Vibrio indicates a receptor function for the membrane anchors of adenylate cyclases |
The Late Jurassic ‘Solnhofen Limestones’ are famous for their exceptionally preserved fossils , including the urvogel Archaeopteryx , which has played a pivotal role in the discussion of bird origins . Here we describe a new , non-archaeopterygid avialan from the Lower Tithonian Mörnsheim Formation of the Solnhofen Archipelago , Alcmonavis poeschli gen . et sp . nov . Represented by a right wing , Alcmonavis shows several derived characters , including a pronounced attachment for the pectoralis muscle , a pronounced tuberculum bicipitale radii , and a robust second manual digit , indicating that it is a more derived avialan than Archaeopteryx . Several modifications , especially in muscle attachments of muscles that in modern birds are related to the downstroke of the wing , indicate an increased adaptation of the forelimb for active flapping flight in the early evolution of birds . This discovery indicates higher avialan diversity in the Late Jurassic than previously recognized .
The Mörnsheim Formation is a unit of the southern German Weißjura Group , a package of mainly calcareous marine sediments that is widely distributed in Bavaria and Baden-Württemberg . In the southern Franconian Alp in Bavaria , the Weißjura Group in the region between Weißenburg and Regensburg is famous for often laminated , very fine-grained limestones of late Kimmeridgian to early Tithonian age , often collectively called ‘Solnhofen limestones’ , and the fossils from these rocks are accordingly referred to as ‘Solnhofen fossils’ . A large number of often local or regional names have been proposed for the different units that make up the ‘Solnhofen limestones’ , but a recent overview of the lithostratigraphy of the area helped to clarify the nomenclature and correlations of the formations ( Niebuhr and Pürner , 2014 ) . Thus , the ‘Solnhofen limestones’ sensu stricto are now included in the Altmühltal Formation and restricted to the area northwest of Ingolstadt , whereas more eastern occurrences of contemporaneous plattenkalks are included in the Painten Formation . Both of these formations underly the Mörnsheim Formation . Biostratigraphic dating with the help of ammonites has furthermore shown that the Altmühltal Formation spans from the uppermost horizon of the Beckeri zone of the latest Kimmeridgian over five ammonite horizons to the rueppelianus horizon of the Hybonotum zone of the Early Tithonian , and that the lithographic limestones within this formation in the areas of Solnhofen and Eichstätt are not synchronous , but the Eichstätt Member is somewhat older ( Schweigert , 2007; Schweigert , 2015; Niebuhr and Pürner , 2014 ) . Furthermore , in more eastern areas , important vertebrate fossils have also been found in the upper part of the Torleite Formation , which underlies the Painten Formation . In the light of this geological and stratigraphic complexity , the traditional habit to talk about all the fossils from these diverse units as the ‘fauna of the Solnhofen limestones’ has been abandoned in favour of the expression ‘fauna of the Solnhofen Archipelago’ in recent years , to distinguish the regional palaeoecological setting from the concrete geological unit that the different fossils are derived from ( e . g . Röper , 2005; López-Arbarello and Schröder , 2014; Rauhut et al . , 2017 ) . The Mörnsheim Formation has its best outcrops in the areas between Mörnsheim , Solnhofen , Monheim , and Daiting ( Figure 1 ) . Lithologically , the Mörnsheim Formation differs from the Altmühltal Formation in the considerably higher amount of silicified limestones ( ‘Kieselplattenkalke’ ) , especially in its lower part . Biostratigraphically , this unit represents the uppermost horizon of the Hybonotum zone , the moernsheimensis horizon , and is thus slightly younger than the Upper Solnhofen Member of the Altmühltal Formation . Fossils have been known for a long time from the Mörnsheim Formation , mainly from the locality of Daiting ( Tischlinger , 2001 ) , but the fauna of this formation has only partially been explored so far , mainly due to the fact that this formation has not been quarried extensively . The specimen described here comes from the Schaudiberg , near Mühlheim , close to Mörnsheim ( Figure 1A ) . Two quarries are currently exposed at the Schaudiberg , both owned by the Grundstücksgemeinschaft Pöschl/Leonhardt , a public quarry for fossil collectors , and the Old Schöpfel Quarry , which is being systematically excavated for fossils ( see Heyng et al . , 2015 ) . The lower part of the Mörnsheim Formation has a total thickness of approximately 50 m at the Schaudiberg , but only parts of this are exposed in the two quarries . Some 8 m of the lowermost Mörnsheim Formation are exposed in the Old Schöpfel Quarry , with the boundary to the underlying Altmühltal Formation at the base of the section being currently covered . Thus , the currently exposed section starts some 4 m above this boundary with silicified laminated limestones and intercalated thick layers of massive limestones and silicified limestones ( Heyng et al . , 2015 ) . In the higher part of the profile and the visitors quarry , the section becomes more dominated by laminated limestones and intercalations of laminated marly limestones and clays . The new urvogel specimen comes from a thin marly intercalation within the lowermost 2 . 5 m of the section in the Old Schöpfel Quarry . It was found in 2017 by Roland Pöschl , who leads the systematic excavations in this quarry . The Mörnsheim Formation at the Schaudiberg is very fossiliferous , with the most common fossils being strongly compressed ammonites , and a rich invertebrate and vertebrate fauna is present , but remains largely unstudied so far . In contrast to the underlying Altmühltal Formation , most vertebrate fossils in the Mörnsheim Formation are at least partially disarticulated and often fragmentary . In the Schaudiberg quarries , fishes are represented by chondrichthyans , including well-preserved specimens of Asteracanthus ( Pfeil , 2011 ) , actinopterygians ( e . g . Schröder and López-Arbarello , 2013 ) , and mainly isolated remains of coelacanths . Tetrapods are represented by unstudied turtles , rhynchocephalians , marine crocodiles , and pterosaurs ( Heyng et al . , 2011; Heyng et al . , 2015; Rauhut et al . , 2011; Rauhut et al . , 2012; Moser and Rauhut , 2011; Rauhut , 2012 ) , with the only formally studied taxon being the unusual rhynchocephalian Oenosaurus muehlheimensis Rauhut et al . , 2012 , so far . As noted above , the only theropod specimen reported from the Mörnsheim Formation so far is the fragmentary holotype of Archaeopteryx albersdoerferi , which comes from the outcrop area of Daiting ( Tischlinger , 2009; Kundrát et al . , 2019 ) . Traditionally , the different specimens found in the Kimmeridgian-Tithonian limestones of Bavaria have been numbered according to the time they were first described as Archaeopteryx ( see Wellnhofer , 2008; Wellnhofer , 2009; Rauhut and Tischlinger , 2015 ) . Thus , the London specimen is usually called the first skeletal specimen , the Berlin specimen the second and so on . Accordingly , more recently , Foth et al . , 2014 described a new specimen that is hitherto simply known as the 11th specimen , and Rauhut et al . , 2018 described the 12th specimen . However , the story becomes more complicated if one accepts that some specimens might actually not belong to the genus Archaeopteryx , such as the Solnhofen specimen , which was argued to represent a distinct genus , Wellnhoferia ( Elzanowski , 2001; Elzanowski , 2002; though see Mayr et al . , 2007 , and Wellnhofer , 2008 , Wellnhofer , 2009 ) , or the Haarlem specimen , which was recently referred to a distinct genus of anchiornithids , Ostromia ( Foth and Rauhut , 2017 ) . Thus , in this case , theoretically , the numbering of Archaeopteryx specimens would need to be revised , with the 11th specimen becoming the 10th specimen and so forth . However , we propose to retain the original numbering of specimens , even if one accepts the different generic assignments , in order to avoid confusion between the recent and older literature . Given the gradual assembly of the avialan body plan ( Brusatte et al . , 2014; Cau et al . , 2015 ) and the general similarity of the basalmost members of this clade , it might be justified to simply talk about 'urvogel specimens' instead of using the generic name Archaeopteryx , to thus accommodate the taxonomic uncertainty . Accordingly , the specimen described here should be regarded as the 13th urvogel specimen from the Solnhofen Archipelago . Apart from the numbers , most specimens also have informal , but widely used names , usually based on their repository ( ‘London specimen’ , ‘Berlin specimen’ , ‘Eichstätt specimen’ , etc . ) . In contrast to many other fossil taxa , it is these names , often in combination with the numbering outlined above , rather than specimen numbers that are usually used to identify the different specimens of Archaeopteryx and possible other basal avialans from the Kimmeridgian/Tithonian of Bavaria , even in the technical scientific literature ( e . g . Elzanowski , 2002; Mayr et al . , 2007; Wellnhofer , 2008; Wellnhofer , 2009; Rauhut et al . , 2018 ) . Unfortunately , however , the most recently described specimens have so far only been identified by their numbers , and no names have been proposed . Thus , in order to facilitate communication about the specimens , we here propose the following names for specimens 11 to 13: 11th specimen: This specimen represents an almost complete postcranial skeleton and parts of the skull of Archaeopteryx , preserved in articulation and with exceptionally detailed feather impressions ( Foth et al . , 2014 ) . The specimen most probably comes from Upper Eichstätt Member of the Altmühltal Formation of the Schernfeld/Blumenberg area , close to Eichstätt ( M . Röper , pers . com . to OR , 05 . 2018 ) and is currently on exhibition at the Museum Solnhofen . However , as neither the exact provenance nor the final repository of this specimen are certain by now , we propose to refer to it as the ‘Altmühl specimen’ , referring to both its general area of provenance close to the Altmühl river and the geological unit it is derived from . 12th specimen: This specimen is an almost complete , although partially poorly preserved , largely articulated skeleton of Archaeopteryx ( Rauhut et al . , 2018 ) . It comes from the lowermost parts of the Painten Formation of the Gerstner Quarry in Schamhaupten and is currently on exhibition at the Dinosaurierpark Altmühltal in Denkendorf . Due to its provenance , we propose to name this specimen the ‘Schamhaupten specimen’ . 13th specimen: This specimen is described here . It comprises an associated right wing of a large basal avialan from the Mörnsheim Formation of Mühlheim , close to Mörnsheim , Bavaria . The specimen belongs to the collections of the Bayerische Staatssammlung für Paläontologie und Geologie in Munich . As there already is a ‘Munich specimen’ , we propose to refer to this specimen as the ‘Mühlheim specimen’ . AMNH , American Museum of Natural History , New York , USA; IGM , Institute of Geology , Ulan Bataar , Mongolia; IVPP , Institute of Vertebrate Paleontology and Paleoanthropology , Beijing , China; JME , Jura-Museum Eichstätt , Germany; JZT , Jizantang Paleontological Museum , Chaoyang City , China; MCF , Museo Carmen Funes , Plaza Hiuncul , Argentina; SNSB-BSPG , Staatliche naturwissenschaftliche Sammlungen Bayerns , Bayerische Staatssammlung für Paläontologie und Geologie , Munich , Germany .
Traditionally , all paravian specimens from the late Kimmeridgian - early Tithonian laminated limestones of southern Germany have been identified as Archaeopteryx , and this would thus be an obvious identification for the new specimen as well . However , as noted in the introduction , recent discoveries of basal paravian and even avialan theropods , also from the Jurassic , have made the distinction of Archaeopteryx from other basal avialans ( and some small , more basal paravians , such as Microraptor ) difficult , and there is no reason for an a priori assumption that all paravian specimens from this area should represent a single genus or even a single lineage ( Foth and Rauhut , 2017 ) . Unfortunately , neither the recent diagnosis of the genus Archaeopteryx by Rauhut et al . ( 2018 ) nor that by Kundrát et al . ( 2019 ) includes any forelimb characters , and due to the great similarity of the forelimbs of many non-ornithothoracan paravians , detailed comparisons are necessary to approach the taxonomic identity of SNSB-BSPG 2017 I 133 . These comparisons are further complicated by the fact that the forelimb bones in almost all specimens of Archaeopteryx , and many other relevant taxa known from flattened specimens in matrix slabs , are exposed in dorsal view , while they are exposed in ventral view in the current specimen . Another problem in comparing the new specimen with specimens of Archaeopteryx is that it is considerably larger in size than any of the other specimens ( Table 1 , see Mayr et al . , 2007 and Rauhut et al . , 2018 for comparison ) . Thus , based on comparisons of the length of the ulna , the only long bone that can be measured with certainty , the new urvogel specimen is more than 220% of the size of the smallest known Archaeopteryx , the Eichstätt specimen , and still c . 111% the size of the largest specimen , the Solnhofen specimen . Compared with the only Archaeopteryx known from the Mörnsheim Formation , the ulna of SNSB-BSPG 2017 I 133 is almost 175% of this specimen . Thus , possible allometric and/or ontogenetic changes have also to be taken into account . This is not only true for proportions , as muscle insertion areas often also become more conspicuous with age and size in vertebrates ( e . g . Hübner , 2010 ) . As far as proportions can be evaluated , the new specimen is generally closely comparable to specimens that can certainly be identified as Archaeopteryx , especially in the ratio of the ( estimated ) length of humerus versus ulna , ulna versus maximal length of metacarpus , and metacarpal II versus length of various phalanges . However , the significance of this similarity is unclear , as these proportions are also comparable in a wide variety of other basal paravians , including Anchiornis , Sapeornis , and , at least for several of these ratios , also Microraptor ( Hwang et al . , 2002; Pei et al . , 2014 ) . Interestingly , however , differences in proportions are found in a few ratios ( see Table 2 ) , most notably in the length of the manual unguals . When compared to the ulna length , manual unguals are relatively smaller than in specimens of Archaeopteryx . This is most marked in comparison with unguals I and III of the largest specimen of Archaeopteryx , the Solnhofen specimen: although , as noted above , the ulna of this specimen is about 10% shorter than that of SNSB-BSPG 2017 I 133 , its unguals I and III are even slightly longer than in the new specimen . Based on a one-sample t test ( see Table 2 ) , the new specimen differs significantly from Archaeopteryx in the following ratios: manual phalanx I-1 vs . ulna , manual phalanx II-1 vs . manual phalanx I-1 , manual phalanx II-2 vs . manual phalanx I-1 , manual phalanx III-1 vs . manual phalanx I-1 , manual ungual II vs . manual phalanx II-1 , and manual digit I vs . manual digit II . When the juvenile Eichstätt specimen is excluded the ratios of manual phalanx III-2 vs . ulna , and manual phalanx III-2 vs . manual phalanx III-1 are also significant different from each other . A further significant difference from specimens of Archaeopteryx might be the length of metacarpal I , which seems to be considerably shorter in the new specimen , under the assumption that the position of the proximal end of the first manual phalanx indicates the length of this bone . However , as the distal end of metacarpal I is not preserved , this cannot be established with any certainty . Likewise difficult to establish are probable differences in relative robusticity of structures in comparison with Archaeopteryx , mainly because of the strong compression of the new specimen . One striking feature of the new specimen is the width of the deltopectoral crest of the humerus , which seems to considerably exceed the width of the humeral shaft , similar to Confuciusornis ( Chiappe et al . , 1999 ) and Ichthyornis ( Clarke , 2004 ) . This is an unusual feature not seen in specimens of Archaeopteryx , but some uncertainty remains due to the strong compression of the bone . Also unusual in the humerus is the angle at which the proximal part that bears the deltopectoral crest is offset from the distal shaft ( Figure 11 ) . This angle is below 30° in dromaeosaurids , such as Microraptor ( Hwang et al . , 2002; Pei et al . , 2014 ) , Zhenyuanlong ( Lü and Brusatte , 2015 ) , and Deinonychus ( Ostrom , 1969 ) , and most specimens of Anchiornis ( Hu et al . , 2009; Pei et al . , 2017 ) , and varies between 30° and 33° in specimens of Archaeopteryx ( Figure 11B–D; see Wellnhofer , 2008 , Wellnhofer , 2009 ) . However , this angle is 38° in the Mühlheim specimen , which is close to Confuciusornis ( 36°: SNSB-BSPG 1999 I 15; 38°: JME 1996/15 , 1997/1; Figure 11H ) and some other more derived avialans such as Sulcavis ( O’Connor et al . , 2013 ) , Archaeorhynchus ( e . g . Zhou et al . , 2013 ) , Yanornis ( Zhou and Zhang , 2001 ) or Gansus ( Wang et al . , 2016 ) . In the manus , the extreme robusticity of metacarpal II in comparison with the other metacarpals is striking . As noted in the description , the proximal part of this metacarpal is almost twice as wide as the proximal articular surface of metacarpal I . However , this metacarpal is not only robust in comparison to the other metacarpals , but also in itself: whereas the length of metacarpal II exceeds ten times the maximal width of the bone in specimens of Archaeopteryx , it is only around 8 . 2 times the maximal width of this element in SNSB-BSPG 2017 I 133 ( see Figure 12C , E ) . As the metacarpus is preserved in articulation , it seems very unlikely that this robusticity of metacarpal II is entirely due to preservation , and thus probably represents a true difference of the new specimen from specimens of Archaeopteryx ( Figure 12 ) . On the other hand , this feature resembles the condition of many basal birds from the Jehol Group , for example Jeholornis ( Zhou and Zhang , 2002; Lefèvre et al . , 2014 ) and Archaeorhynchus ( Zhou et al . , 2013 ) , which highlight the transition of an individual metacarpal II to the major element of the fused capometacarpus of modern birds . Another element in the manus that is remarkably robust is the first phalanx of the second digit . In non-avialan paravians and specimens that can securely be referred to Archaeopteryx ( e . g . Figure 12A–C ) , this phalanx is only slightly more robust than phalanx II-2 or III-1 ( e . g . 1 . 25 times the width of phalanx II-2 at the proximal shaft in the Berlin specimen ) , but it is more than 1 . 7 times the width of the widest part of phalanx II-2 and more than twice the width of phalanx II-1 in the Mühlheim specimen . This pronounced robusticity of phalanx II-1 in comparison with Archaeopteryx ( and more basal paravians ) probably represents an apomorphic character shared by SNSB-BSPG 2017 I 133 and more derived avialans , as a widened phalanx proximalis digiti majoris is a general character of avialans more derived than Archaeopteryx , and is present e . g . in basal forms , such as Jeholornis ( Lefèvre et al . , 2014 ) , Sapeornis ( Yuan , 2008; Provini et al . , 2009 ) , Chongmingia ( Wang et al . , 2016 ) , and Confuciusornis ( Figure 12D; Chiappe et al . , 1999 ) . The phalanx in the Mühlheim specimen is plesiomorhic in comparison to most more derived avialans ( i . e . , Ornithothoraces ) in that it is not dorsoventrally flattened , but seems to be rather robust also in this plane . In this respect it is similar to the robust phalanx in Jeholornis ( Lefèvre et al . , 2014 ) , Sapeornis ( Yuan , 2008; Provini et al . , 2009 ) and Confuciusornis . In the latter , this element is dorsoventrally robust at least along its anterior edge , whereas the posterior edge is somewhat flattened ( Figure 12D; SNSB-BSPG 1999 I 15 , JME 1997/1 , 2005/1 ) . Apart from these morphometric differences , there are also several qualitative characters that differ between the new specimen and specimens of Archaeopteryx . One of these characters concerns the insertion of the m . pectoralis on the deltopectoral crest of the humerus . This facet is not especially marked , or only indicated by slight thickening of the apex of the crest in most non-avialan theropods . In Archaeopteryx , the anterior side of the deltopectoral crest is only exposed in the London , Thermopolis and , partially , the Maxberg specimens ( pers . obs . ; see de Beer , 1954; Heller , 1959; Mayr et al . , 2007; Wellnhofer , 2008; Wellnhofer , 2009 ) . Although the edge of the deltopectoral crest is slightly damaged in the London and Thermopolis specimens , it can be established that these specimens follow the general theropodan condition of not showing a marked facet ( Figure 13 ) . In contrast , SNSB-BSPG 2017 I 133 shows a well-developed , anteromedially inclined , elongate oval facet for the insertion of the m . pectoralis on the deltopectoral crest ( Figure 3 ) . Again , this character is present in many basal avialans from the Jehol group ( e . g . Figure 14 ) and seems to be a derived character in comparison with Archaeopteryx ( see below ) . Another striking feature of SNSB-BSPG 2017 I 133 is the development of the tuberculum bicipitale radii as a raised crest on the proximal radius . However , due to preservation the exact development of this tuberculum in specimens of Archaeopteryx is difficult to establish . Several specimens , including the Munich , Altmühltal and Ottmann and Steil ( "chicken wing“ ) specimens show that this tuberculum is present in this taxon . Although a direct comparison of this structure is difficult due to preservation , these specimens seem to show a rounded to triangular expansion that is different from the crest-like , more rectangular tubercle in the new specimen . In those avialan specimens from the Jehol Group where this characters can be evaluated it varies between a triangular and crest-like state ( Chiappe et al . , 1999; Provini et al . , 2009; Zhou et al . , 2013 ) . If the interpretation of longitudinal furrows on the radius and phalanx I-1 as original features is correct , these are further differences from Archaeopteryx . As noted above , such furrows are observed in some other paravian theropods in variable elements ( see e . g . Chiappe and Walker , 2002; Sanz et al . , 2002; Hu et al . , 2015; Foth and Rauhut , 2017; Xu et al . , 2017 ) , but the combination of such furrows in the radius and only one manual phalanx might be unique for SNSB-BSPG 2017 I 133 . However , as such features might be easily overlooked , more studies of these morphologies are needed . Finally , apart from being relatively smaller , the manual unguals also show differences from those of specimens of Archaeopteryx . Especially the shape , position and prominence of the flexor tubercle seems to differ . In general the flexor tubercles in the new specimen seem to be placed slightly more proximally than in specimens of Archaeopteryx and the tubercles are more pointed , that is the angle between their proximal and distal margin are sharper; this is especially marked in manual ungual II . Likewise , the transverse distopalmar expansion of the flexor tubercle , which is present in all three unguals of the Mühlheim specimen , is not present in specimens of Archaeopteryx . The comparison with the anchiornithid Ostromia ( Foth and Rauhut , 2017 ) is more difficult due to the fragmentary nature of both specimens . Most of the forelimb bones of Ostromia are actually preserved as imprints , so that only parts of the manus can be used for comparison . Like Ostromia , SNSB-BSPG 2017 I 133 seems to possess a longitudinal furrow along the manual phalanx I-1 , a character that is also shared with Sinornithosaurus and Jianianhualong ( Foth and Rauhut , 2017; Xu et al . , 2017 ) . However , Ostromia shows similar grooves also in manual phalanx III-3 , while the corresponding element is smooth in Alcmonavis . Striking differences between Alcmonavis and Ostromia are present in the size and shape of the manual unguals . When compared to the length of the manual phalanx I-1 , the first ungual of Ostromia is much smaller than that of Alcmonavis . In contrast , the third ungual is much smaller in Alcmonavis when compared to the size of the first ungual , while in Ostromia they almost have the same size . While some basal avialans like Jeholornis and Sapeornis from the Jehol Group show enlarged manual unguals , they also show a clear size reduction from manual ungual I to III ( Zhou and Zhang , 2002; Zhou and Zhang , 2003a; Provini et al . , 2009; Gao et al . , 2012 ) . The flexor tubercles of the manual unguals in Ostromia are much more distally displaced than in SNSB-BSPG 2017 I 133 ( but also Archaeopteryx ) . As described above , the flexor tubercles of SNSB-BSPG 2017 I 133 are very prominent and pointed ventrally , while in Ostromia they are relatively low , forming a plateau-like ventral apex ( see Wellnhofer , 2008: Figure 5 . 79B , C; Foth and Rauhut , 2017: Figures 3B and 4 ) . Furthermore , like Archaeopteryx , Ostromia lacks the transverse distopalmar expansion of the flexor tubercle and a notably flattened palmar margin of the bony ungual . While the manual unguals of Confuciusornis , Jeholornis and Sapeornis still bear prominent flexor tubercles ( Chiappe et al . , 1999; Zhou and Zhang , 2002; Zhou and Zhang , 2003b ) , they become reduced within Ornithothoraces ( e . g . , Zhou et al . , 2013 ) . In summary , despite the overall similarity and very similar proportions , the new specimen shows numerous small differences from Archaeopteryx , precluding a referral to this taxon . Several characters , including the markedly concave proximal articular surface of the ulna , the very massive phalanx II-1 and the marked , anteromedially inclined facet for the attachment of the m . pectoralis in the humerus , are shared with more derived avialans , and indicate that the Mühlheim specimen represents a third , phylogenetically slightly more crownward taxon of avialans from the Tithonian limestones of southern Germany . From the description and comparisons of the specimen it is furthermore clear that SNSB-BSPG 2017 I 133 also cannot be referred to Ostromia or to any other known theropod taxon . We thus opt to describe this rather incomplete specimen as a new genus and species of 'Urvogel' . Theropoda Marsh , 1881 Maniraptora Gauthier , 1986 Avialae Gauthier , 1986 Alcmonavis poeschli gen . et sp . nov . urn:lsid:zoobank . org:act:668F42B6-5BDC-4ADF-B271-36C6A43C7DB3 From Alcmona , the old Celtic name of the Altmühl River , which flows through the principal region in which the famous ‘Solnhofen limestones’ are exposed , and avis , from the Greek ‘aves’ for bird . The species name honours Roland Pöschl , who leads the excavations at the Schaudiberg and found the specimen . SNSB-BSPG 2017 I 133 , an almost complete , partly disarticulated skeleton of the right wing ( Figure 2 , see Tab . 1 for measurements of SNSB-BSPG 2017 I 133 ) . Old Schöpfel Quarry at the Schaudiberg , Mühlheim , close to Mörnsheim , Bavaria . Mörnsheim Formation , moernsheimensis ammonite horizon of the Hybonotum zone of the Early Tithonian . The specimen comes from a thin layer of marly laminated limestone some 6 m above the contact with the underlying Altmühltal Formation . Alcmonavis poeschli differs from all other theropods ( including birds ) in the following combination of characters: humerus with large deltopectoral crest , with a maximal expansion that exceeds the width of the humeral shaft; proximal part of humerus strongly angled at approximately 38° in respect to distal shaft; ulna with well-defined , single , oval , concave proximal cotyla and small lateral tubercle; distal end of ulna slightly asymmetrically expanded; large , crest-like biceps tubercle on the proximal radius; longitudinal groove along the medial side of the radial shaft; metacarpal II considerably more robust than metacarpal I and III; phalanx I-1 with longitudinal groove; phalanx II-1 very robust , but with rounded , rather than flattened cross-section; phalanx II-1 slightly twisted; manual unguals with strongly developed and palmarly transversely expanded flexor tubercles . The phylogenetic analysis ( see Materials and methods ) resulted in more than 99 , 999 trees with a length of 2690 steps . The strict consensus ( Figure 15—figure supplement 1 ) is rather well resolved and includes monophyletic Maniraptora , Paraves and Avialae with equivalent taxonomic contents to other recent analyses . Areas with lack of resolution include a polytomy between therizinosauroids , oviraptorosaurs and paravians , two larger polytomies at the base of Deinonychosauria , and a polytomy at the base of Ornithothoraces , as well as minor polytomies in the higher nodes or Alvarezsauridae , Oviraptorosauria , and Dromaeosauridae . Reduced consensus methods recovered a number of problematic taxa ( Albinykus , Byronosaurus , Balaur , Citipati , Hesperonychus , Jinfengopteryx , Pyroraptor , Xixiasaurus , Yixianosaurus , and a Vorona-Liaoningornis clade ) , the a postiori pruning of which further increased resolution . In contrast to other recent iterations of this matrix ( Foth et al . , 2014; Foth and Rauhut , 2017 ) , oviraptorosaurs and therizinosaurs were found in a monophyletic clade in the reduced consensus tree ( Figure 15 , Figure 15—figure supplement 1 ) , as in several earlier phylogenetic analyses ( e . g . Makovicky and Sues , 1998; Holtz , 2000; Rauhut , 2003; Turner et al . , 2011 ) , and the analysis recovered a monophyletic Deinonychosauria , including Troodontidae and Dromaeosauridae as sister groups , as in most analyses of coelurosaur interrelationships . Epidexipteryx , often considered to be a basal paravian ( e . g . Turner et al . , 2012; Godefroit et al . , 2013a; Godefroit et al . , 2013b; Xu et al . , 2015 ) or even avialan theropod ( e . g . Xu et al . , 2011; Foth et al . , 2014 ) is here recovered as a basal oviraptorosaur , as in Agnolín and Novas , 2013 . As originally proposed by Csiki et al . , 2010 , Turner et al . , 2012 and Brusatte et al . , 2013 , Balaur is placed within Dromaeosauridae in the current analyses and not at the base of Avialae ( see Godefroit et al . , 2013a; Foth et al . , 2014; Cau , 2018; Foth and Rauhut , 2017 ) . The controversial Late Jurassic paravians Anchiornis , Xiaotingia , Eosinopteryx , Pedopenna , and Ostromia were found as basal avialans ( as in Godefroit et al . , 2013b; Foth et al . , 2014; Foth and Rauhut , 2017 ) but a monophyletic Anchiornithidae ( as defined by Xu et al . , 2016; see also Foth and Rauhut , 2017 ) is restricted to the genera Eosinopteryx , Ostromia and Anchiornis , whereas Pedopenna and Xiaotingia form sister taxa just basal to Archaeopteryx . The new taxon , Alcmonavis , was found crownwards to Archaeopteryx , thus representing the most derived avialan known from the Jurassic so far . The phylogenetic position of Alcmonavis is supported by four synapomorphic characters: char . 217 , the distal humeral condyles are positioned on anterior surface ( 1 ) ; char . 562 , the attachment for m . pectoralis on the deltopectoral crest of the humerus is marked as an elongate oval , anteromedially inclined facet on the mediodistal surface of the deltopectoral crest ( 1 ) ; char . 563 , the proximal articular surface of the ulna is developed as an oval concavity with slightly raised rims ( 1 ) ; and char . 565 , manual phalanx II-1 is strongly broadened , more than 1 . 5 times the width of phalanx II-2 and phalanges of digit III ( 1 ) . Clade support is low for most clades , as is expected in a matrix with numerous very incomplete taxa and an average amount of missing codings of c . 60% . However , whereas most clades have Bremer support values of 1 , the phylogenetic position of Alcmonavis is supported by a Bremer value of 2 ( Figure 15 , 15—figure supplement 1 ) . In order to evaluate support for some of the results relevant to the question of bird origins , we ran several constrained analyses . Both the monophyly of Anchiornithidae as proposed by Foth and Rauhut , 2017 and a position of troodontids closer to Avialae than to dromaeosaurids only requires one additional step , so neither of these possibilities can currently be excluded . Placing anchiornithids in Troodontidae , as argued by Hu et al . , 2009 and subsequent authors , requires at least 10 additional steps , and placing both anchiornithids and Archaeopteryx in Deinonychosauria , as proposed by Xu et al . , 2011 leads to trees that are 21 steps longer than the most parsimonious trees , making this arrangement rather unlikely . Concerning the phylogenetic position of Alcmonavis , both a placement below Archaeopteryx , and as sister taxon to the latter require four additional steps . Given that only 79 of 565 characters ( less than 14% ) can be coded for the new taxon , this difference indicates that the position retrieved for Alcmonavis is rather robust . The implied weight analysis retrieved 405 equally parsimonious trees with a score of 114 . 762 . The general results of this weighted analysis are similar to those obtained from the analysis under equal weights ( Figure 15—figure supplement 3 ) , with one notable exception: therizinosaurs are here recovered as the most basal clade of maniraptorans , followed by alvarezsaurids and then oviraptorosaurs , as in Senter , 2007 . Most importantly , however , the weighted analysis supports the relationships between anchiornithids , Archaeopteryx , and Alcmonavis . Alcmonavis shows several notable characters that might help to elucidate the early steps of the osteological evolution of the bird wing . Current discussions of the evolution of flight capabilities have focused on feather evolution and arrangement ( e . g . Clarke et al . , 2006; Chiappe et al . , 2014; Foth et al . , 2014; O'Connor et al . , 2013; O'Connor et al . , 2016; O’Connor and Chang , 2015; Sullivan et al . , 2017; Saitta et al . , 2018 ) , whereas the flight musculature and its osteological correlates has received comparatively little attention recently , with the exception of the study of the shoulder girdle and supracoracoideus muscle of Mesozoic birds by Mayr , 2017 . The identification of several muscle attachment areas in the forelimb bones of Alcmonavis has implications for our understanding of the early evolution of avialan flight musculature . In recent birds , the most important flight muscles are the m . pectoralis ( also named m . pectoralis major or m . pectoralis superficialis ) , which is the main muscle in the downstroke of the wing , and the m . supracoracoideus ( m . pectoralis minor; m . pectoralis profundus ) , which lifts the forelimb and has an important role in the rotation of the humerus ( Dial , 1992; Ostrom et al . , 1999; Baier et al . , 2007; Biewener , 2011; Tobalske , 2016 ) . The supracoracoideus muscle attaches on the anterodorsal edge of the proximal humerus ( on the tuberculum dorsale; Baumel and Witmer , 1993; see also Jasinoski et al . , 2006 ) ; as this part of the humerus is not preserved in Alcmonavis , nothing can be said about this muscle in this taxon . More general discussions of the evolution of the supracoracoideus muscles can be found in Ostrom , 1976b , Ostrom et al . , 1999 , Baier et al . , 2007 and Mayr , 2017 . The deltopectoral crest and attachment area for the pectoralis muscle , is , however , generally well-preserved in the new taxon . Dinosaurs in general are noteworthy for having a well-developed deltopectoral crest on anterolateral side of the proximal end of the humerus . This crest serves as attachment site for several proximal forelimb muscles that extend from the shoulder girdle to the humerus . In crocodiles , these muscles are mainly the m . coracobrachialis brevis , which attaches on a large area on the medial side of the crest , the m . deltoideus clavicularis , the attachment of which covers most of the lateral side of the crest , and the m . supracoracoideus , which attaches on the apex of the crest , whereas the m . pectoralis only has a rather small attachment on the mediodistal side of the crest , just below the apex ( Meers , 2003 ) . This general arrangement seems to have been retained in sauropodomorphs ( Remes , 2008; Otero , 2018 ) and basal theropods ( Burch , 2014 ) . In more derived theropods , the attachment site of the pectoralis muscle is enlarged , but not specifically marked on the medial side of the deltopectoral crest ( Jasinoski et al . , 2006 ) , whereas the apex of the crest , potentially still serving for the attachment of the supracoracoides muscle , might be slightly expanded transversely , as in Ornitholestes ( pers . obs . on AMNH 619 by CF and OR; Figure 16A ) and the basal paravian Mei ( IVPP V 12733 ) . In most basal paravians ( Figure 16B ) and Archaeopteryx , the deltopectoral crest is large , but thin , and lacks an enlarged and marked attachment area for the pectoralis muscle ( London , Maxberg and Thermopolis specimens; Figure 13 ) . In contrast , later avialans , such as Sapeornis ( Figure 14; Provini et al . , 2009 ) , Jeholornis ( Lefèvre et al . , 2014 ) , Jixiangornis ( Chiappe and Meng , 2016 ) , Confuciusornis ( e . g . JME 1997/1; Chiappe et al . , 1999 ) , Enantiornithes ( Walker et al . , 2007 ) , Ichthyornis ( Clarke , 2004 ) , and many modern birds ( Figure 16D ) have a well-developed , anteromedially facing facet for the insertion of this most important flight muscle on the mediodistal part of the deltopectoral crest . Alcmonavis also shows such a marked facet on the medial side of the deltopectoral crest ( Figures 3 and 16C ) , which is smaller than in many modern birds , but comparable in development to the facet seen in Confuciusornis , Jeholornis , Jixiangornis and Sapeornis . This indicates an increase in importance of the pectoralis muscle in avialan evolution after Archaeopteryx . Another marked muscle attachment in Alcmonavis is the tuberculum bicipitale radii on the proximal radius . This tubercle is the attachment site of one of the branches of the m . biceps brachii , which extends from its origin on the coracoid ( and anterior side of the proximal humerus in many groups ) to insert on the proximal ends of radius and ulna in amniotes generally ( Remes , 2008 ) . In most dinosaurs , the insertion of the m . biceps brachii on the proximal radius is only noted by a rugose patch , but no marked tubercle is present . In contrast , birds usually have a well-developed tubercle on the anteromedial side of the proximal radius , as it is also present in Alcmonavis ( Figure 7A , B ) and other Mesozoic birds ( e . g . Confuciusornis , SNSB-BSPG 1999 I 15 , Figure 15C; Sapeornis , Provini et al . , 2009; Archaeorhynchus , Zhou et al . , 2013 ) . Although this structure has not been described in Archaeopteryx , a small , triangular tubercle can actually be identified in several specimens , including the Munich ( Figure 17A ) , Altmühltal and Ottmann and Steil ( Figure 17B ) specimens . We could not identify this structure in any specimen of Anchiornis we have seen or that is illustrated in the literature , nor in any other anchiornithid . Outside avialans , a marked tuberculum bicipitale radii is only present in Microraptor ( IVPP V 13352 ) and , apparently in hypertrophied form , in Bambiraptor ( Burnham , 2004 ) . Thus , with the exception of Bambiraptor , a marked tubercle for the insertion of m . biceps brachii seems only to be present in volant forms . Both the presence of a marked attachment area for the pectoralis muscle and the well-developed tubercle for the insertion of m . biceps brachii might thus have implications for the early evolution of flight . Whereas the role of the pectoralis muscle as the main downstroke muscle has been rather well studied , the function of other forelimb muscles , such as the m . biceps brachii , are less well understood ( Biewener , 2011; Tobalske , 2016 ) . As for the latter muscle , its primary function is usually considered to be the flexion of the forearm and the stabilization of the elbow joint , especially during the downstroke ( Dial , 1992; Biewener , 2011; Robertson and Biewener , 2012 ) . However , a recent study of muscle activity during flight in pigeons ( Robertson and Biewener , 2012 ) also shows that this muscle has its highest activity patterns during take-off , indicating that it might be important for flapping take-off in modern birds in general . The development of a pronounced tubercle for the insertion of the m . biceps brachii on the radius in volant basal avialans is thus consistent with the idea that these animals used a primitive form of flapping flight ( Carney , 2016; Heers and Carney , 2017 ) , probably starting as burst fliers , as recently suggested for Archaeopteryx by Voeten et al . , 2018 , based on cross-sectional geometry of the long bones of the forelimb . Thus , although a potential flight performance of the most basal avialans like Anchiornis is controversial ( Evangelista et al . , 2014; Dececchi et al . , 2016; Pan et al . , 2019 ) , active flapping flight might have originated early within the lineage ( see also Meseguer et al . , 2012; Dececchi et al . , 2016 ) . If this primitive flapping flight was preceded by an intermediate gliding stage cannot be evaluated for the moment and requires more detailed studies on the ecology and life style of Anchiornis and its closest relatives . However , the appearance of a pronounced attachment site for the m . pectoralis and m . biceps brachii in Alcmonavis ( in comparison with Archaeopteryx ) , a phylogenetically next step in the evolution towards modern birds , then possibly indicates an early improvement in flapping flight capabilities already in the Late Jurassic . This is also in accordance with the increased robusticity of metacarpal II and phalanx II-1 , which form the digitus majoris in more derived birds , which serves as the attachment for the flight primaries .
The primary specimen described here , SNSB-BSPG 2017 I 133 , was discovered in the old Schöpfel Quarry at the Schaudiberg , near Mühlheim , by R . Pöschl in 2017 . The specimen was mechanically prepared by U . Leonhardt and subsequently purchased by the State of Bavaria , where it will be permanently stored at the Bayerische Staatssammlung für Paläontologie und Geologie in Munich . Numerous specimens of paravian theropods were studied for comparison , including all other urvogel specimens from the Solnhofen Archipelago ( the currently missing Maxberg specimen could only be studied on the basis of a high quality cast at the BSPG ) and several specimens of anchiornithids and basal avialans from China . The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature , and hence the new names contained herein are available under that Code from the electronic edition of this article . This published work and the nomenclatural acts it contains have been registered in ZooBank , the online registration system for the ICZN . The ZooBank LSIDs ( Life Science Identifiers ) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ‘http://zoobank . org/” . The LSID for this publication is: urn:lsid:zoobank . org:act:668F42B6-5BDC-4ADF-B271-36C6A43C7DB3 . The electronic edition of this work was published in a journal with an ISSN . In accordance with Wilson , 2006 , we generally use the anatomical terms and orientation of bones as commonly used in the palaeontological literature on dinosaurs , rather than rigorously applying the skeletal terms proposed in the Nomina Anatomica Avium ( Baumel and Witmer , 1993 ) . However , the latter are used in respect to anatomical features that are typical for birds , but not present in non-avialan dinosaurs . Thus , we prefer the terms 'anterior' and 'posterior' over 'cranial' and 'caudal' ( as the latter might be confused with anatomical regions of the skeleton ) , and refer to the different sides of long bones according to their orientation in the resting pose in a theropod dinosaur . Many fossils from the Upper Jurassic plattenkalks of southern Germany are fluorescent under artificial ultraviolet light ( UV ) which allows a more precise investigation of morphological details of skeletal remains as well as of soft parts . Since each fossil fluoresces slightly differently , a variety of filters and high performance UV-A lamps is required for investigation and imaging ( Tischlinger and Arratia , 2013; Tischlinger , 2015 ) . For our investigation of the 13th urvogel we used different UV-lamps with wavelengths of 312 nanometers ( UV-B ) and 365–366 nm ( UV-A ) . During the UV pictorial documentation of the 13th urvogel best results were obtained with a wavelength of 365–366 nanometers ( long-wave radiation , UV-A ) . The following UV lamps were used: 3 Benda UV lamps: type N , 16 Watt , UV-A , 366 nanometers ( size of filter 200 mm x 50 mm ) ; 1 Labino UV lamp: UV-Spotlight S135 , 35 Watt , UV-A , peak at 365 nm: spotlight ( >50 . 000 microwatts per cm² at 30 cm distance ) plus midlight reflector replacement ( >8 . 000 microwatts per cm² at 30 cm distance ) . The visibility of details under UV was enhanced considerably by an established filtering technique , crucial for the photographic documentation . The application of different filters allowed a selective visualisation of peculiar fine structures . Color compensation filters ( yellow , cyan and magenta of different types and densities ) were adjusted in front of the camera lens or under the microscope objective lens ( if pictures were taken through the microscope ) . The optimum number and specification of the compensation filters was tested in a series of experiments . The predominant color of luminescence was of minor importance . In fact , the crucial decision on the amount of filtering was the optimal visibility of details and their differentiation from surrounding structures and the matrix ( Tischlinger and Arratia , 2013 ) . In order to test the phylogenetic position of the new taxon , we included it in a modified version of the phylogenetic data matrix used in Foth and Rauhut , 2017 . We furthermore checked codings for several taxa , added the recently described dromaeosaurid Zhenyuanlong ( Lü and Brusatte , 2015 ) , the troodontid Jianianhuanlong ( Xu et al . , 2017 ) and the avialan Jinguofortis ( Wang et al . , 2018 ) and four new characters . These characters are: Character 562 . Attachment for m . pectoralis on deltopectoral crest of humerus: not specifically marked , distal edge of deltopectoral crest might be slightly expanded ( 0 ) ; marked as an elongate oval , anteromedially inclined facet on the mediodistal surface of the deltopectoral crest ( 1 ) . See discussion of this character above and Figures 14 and 16 . Character 563 . Proximal articular surface of ulna: anteroposteriorly concave and flat or slightly convex transversely ( 0 ) ; developed as a round to oval concavity with slightly raised rims ( 1 ) . In basal theropods and basal coelurosaurs , the ulna has a single proximal articular surface that is developed as an anteroposteriorly concave and transversely slightly convex to flat facet , in which the margins of the articular surface are not specifically marked ( Figure 18A ) . Although the proximal end of the ulna is poorly exposed in the available specimens of Archaeopteryx , this also seems to be the condition in this taxon . In contrast , in birds , including Alcmonavis and other Mesozoic birds , such as Confuciusornis , there is a pronounced , round to oval concavity with slightly raised margins for the articulation with the ulnar condyle of the humerus ( Figure 18B ) . Character 564 . Tuberculum bicipitale radii on the proximal radius: absent or indistinct ( 0 ) ; pronounced as a marked tubercle or crest ( 1 ) . See discussion of this character above and Figure 17 . Character 565 . Manual phalanx II-1: not significantly broadened when compared to other manual phalanges ( 0 ) ; strongly broadened , more than 1 . 5 times the width of phalanx II-2 and phalanges of digit III ( 1 ) . See discussion of this character above and Figure 12 . The complete matrix had 136 taxa scored for 565 characters ( see supplementary files ) . The matrix was analysed using the software TNT 1 . 5 ( Goloboff et al . , 2008; Goloboff and Catalano , 2016 ) under equally weighted parsimony through a heuristic search of 10 , 000 replicates of Wagner trees followed by TBR branch swapping . In order to further test the position of the new taxon , we ran a second analysis using implied weights ( K = 12; Goloboff et al . , 2018 ) . This analysis was also carried out using TNT with 1000 replicates of Wagner trees followed by TBR branch swapping . In order to statistically evaluate the significance of differences in proportions , we performed one sample parametric t-test ( all samples show normal distribution ) for the ratios of metacarpal III/metacarpal II , manual phalanx I-1/ulna , manual ungual I/ulna , manual phalanx II-1/manual phalanx I-1 , manual phalanx II-1/manual ungual II , manual phalanx II-2/manual phalanx I-1 , manual phalanx III-1/manual phalanx I-1 , manual phalanx III-2/ulna , manual phalanx III-2/manual phalanx III-1 , manual phalanx III-3/manual phalanx III-2 , and manual digit I/manual digit II , testing the Mühlheim specimen against those specimens that can be classified as Archaeopteryx based on the diagnosis provided by Rauhut et al . , 2018 . Ratios including metacarpal I were not included due to uncertainties in the length of the bone . The one sample t-test compares the value in question with the range of the comparative statistical population of Archaeopteryx to evaluate the probability that this value represents the same population . The test was performed with help of the software PAST 3 . 21 ( Hammer et al . , 2001 ) . | The origin of birds and their flight has been heavily debated in the field of evolutionary biology since the late nineteenth century . Birds are the only living descendants of dinosaurs and , for paleontologists , the famous Archaeopteryx has played a pivotal role in this discussion . Living during the Jurassic period about 150 million years ago in what is now southern Germany , Archaeopteryx is generally accepted as the oldest known flying bird . Yet , with the discovery of other bird-like dinosaurs from the same period , a question has arisen as to whether Archaeopteryx is the only flying bird from the Jurassic . To answer this question , Rauhut et al . carefully examined a fossil of an isolated wing skeleton that was recently discovered in the same region of Germany where Archaeopteryx was found . The new specimen shows several characteristics that are otherwise only found in modern birds and not seen in Archaeopteryx . As such , this fossil represents a new species and the most bird-like bird discovered from the Jurassic . Rauhut et al . named the species Alcmonavis poeschli , after the ancient name for a river that flows near the discovery site , the Greek word for bird , and Roland Pöschl – the collector who found the specimen . The wing of Alcmonavis also shows several features related to the attachment of flight muscles that suggest it was better adapted for flapping flight than Archaeopteryx . Together these findings are mostly consistent with the hypothesis that birds first started flying by flapping their wings rather than starting from a gliding stage . However , more detailed studies of the anatomy of primitive birds and their close relatives are needed to further test this hypothesis . | [
"Abstract",
"Introduction",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2019 | A non-archaeopterygid avialan theropod from the Late Jurassic of southern Germany |
Photorhabdus is a highly effective insect pathogen and symbiont of insecticidal nematodes . To exert its potent insecticidal effects , it elaborates a myriad of toxins and small molecule effectors . Among these , the Photorhabdus Virulence Cassettes ( PVCs ) represent an elegant self-contained delivery mechanism for diverse protein toxins . Importantly , these self-contained nanosyringes overcome host cell membrane barriers , and act independently , at a distance from the bacteria itself . In this study , we demonstrate that Pnf , a PVC needle complex associated toxin , is a Rho-GTPase , which acts via deamidation and transglutamination to disrupt the cytoskeleton . TEM and Western blots have shown a physical association between Pnf and its cognate PVC delivery mechanism . We demonstrate that for Pnf to exert its effect , translocation across the cell membrane is absolutely essential .
Bacteria belonging to the Enterobacteriaceae genus Photorhabdus exist in a symbiotic partnership with entomopathogenic Heterorhabditis sp . nematodes . This Entomopathogenic Nematode complex ( EPN ) comprises a highly efficient symbiosis of pathogens that is commonly used as a biological agent to control crop pests ( Forst et al . , 1997 ) . The Photorhabdus bacteria are delivered into the hemocoel of the insect , after regurgitation from the worm , where they resist the insect immune response and rapidly kill the host via septicaemic infection . Insect tissues are subsequently bio-converted into a dense soup of Photorhabdus bacteria , which provide a food source to support the replication of the nematode . As food resources are depleted Photorhabdus re-associates with infective juvenile nematodes , and together they emerge from the insect cadaver able to re-infect a new host ( Ciche et al . , 2008; Somvanshi et al . , 2012 ) . Three major species have been formally recognized to date within the genus - P . luminescens , P . asymbiotica , and P . temperata . It should be noted however that with increasing numbers of Photorhabdus genome sequences becoming available , the genus structure is under revision ( Machado et al . , 2018 ) . In addition to the normal insect life cycle , P . asymbiotica is also the etiological agent of a serious human infection termed Photorhabdosis , which is associated with severe ulcerated skin lesions both at the initial infection foci and later at disseminated distal sites ( Gerrard et al . , 2004; Gerrard et al . , 2006; Gerrard et al . , 2003a; Gerrard et al . , 2003b ) . The Photorhabdus genome encodes a diverse repertoire of virulence genes encoding for protein toxins , proteases and lipases for combating diverse hosts , that can be found in chromosomally encoded pathogenicity islands ( Waterfield et al . , 2009a; ffrench-Constant , 2007; Waterfield et al . , 2002; Duchaud et al . , 2003; Waterfield et al . , 2009b; Wilkinson et al . , 2009 ) . In addition , the bacteria also secrete a potent cocktail of other biologically active small molecules to preserve the insect cadaver in the soil from competing saprophytes and microbial predators such as amoeba ( Cai et al . , 2017; Bozhüyük et al . , 2017 ) . Several classes of Photorhabdus protein insecticidal toxins have now been well characterised including the Toxin Complexes ( ffrench-Constant and Bowen , 2000; Waterfield et al . , 2001a; Waterfield et al . , 2001b; Meusch et al . , 2014; Erickson et al . , 2007; Waterfield et al . , 2007; Hares et al . , 2008 ) , the binary PirAB toxins ( Waterfield et al . , 2005; Ahantarig et al . , 2009; Sirikharin et al . , 2015 ) and the large single polypeptide Mcf ( ‘makes caterpillars floppy’ ) toxins ( Daborn et al . , 2002; Waterfield et al . , 2003; Dowling et al . , 2007 ) . A fourth class of highly distinct toxin delivery systems first identified in Photorhabdus are the ‘Photorhabdus virulence cassettes’ , or PVCs ( Yang et al . , 2006 ) . These represent operons of around 16 , conserved , structural and synthetic genes ( from hereon just described as the structural genes ) encoding for a phage ‘tailocin’ like structure ( Ghequire and De Mot , 2015 ) and one or more tightly linked downstream toxin-effector like genes . Genomic analysis of multiple strains of Photorhabdus revealed they often encode up to five or six copies of the operon , each with unique downstream effector genes ( Hapeshi and Waterfield , 2017 ) . It should be noted that PVC-like elements are not restricted to Photorhabdus as a well-characterized homologous operon can also be found on the pADAP plasmid of the insect pathogenic bacteria Serratia entomophila ( Hurst et al . , 2004 ) . This system has been named the anti-feeding prophage ( AFP ) , as it is responsible for the cessation of feeding in the New Zealand grass grub host . Recent cryo-electron microscopy studies have revealed that , morphologically , AFP resembles a simplified version of the sheathed tail of bacteriophages such as T4 , including a baseplate complex . It also shares features with type-VI secretion systems , with the central tube of the structure having a similar diameter and axial width to the Hcp1 hexamer of P . aeruginosa T6SS ( Heymann et al . , 2013 ) . One important difference between the PVC and T6SS machinery is that the T6SS relies upon direct contact between host and bacterial cell , and is anchored in to the membrane by a substantial membrane complex whose structure is still being elucidated ( Durand et al . , 2015 ) , whereas the PVC needle complex is freely released into the surrounding milieu and so can act at a distance . Furthermore , recent reports have indicated that other more diverse bacteria can also make similar needle complexes for manipulation of eukaryotic hosts . A well-characterized example is the production of analogous devices by the marine bacterium Pseudoalteromonas luteoviolacea ( Figure 1A ) . These structures are involved in the developmental metamorphosis of the larvae of the tubeworm Hydroides elegans , and they are deployed in outward-facing arrays comprising about 100 contractile structures , with baseplates linked by tail fibres in a hexagonal net ( Shikuma et al . , 2014 ) . Interrogation of sequence databases with PVC protein sequences suggests many other more diverse tailocin-like systems are yet to be characterized ( Sarris et al . , 2014 ) . These include operons closely related to the PVCs in Xenorhabdus bovienii CS03 , Yersinia ruckeri ATCC29473 and Vibrio campbellii AND4 . In addition , evidence of more diverse elements , like that of P . luteoviolacea , can also be seen . To address this , we have recently performed an exhaustive analysis of all available prokaryotic and archaeal genome sequences in the public databases to look at the distribution of pvc-like elements ( unpublished data ) . This suggests that PVC-like nano-syringes and their distant cousins are of enormous ecological and perhaps biomedical significance . Here we focus on a single Photorhabdus pvc operon ( which elaborates the PVCpnf needle complex ( Yang et al . , 2006 ) to understand the relationship between the structural genes and the tightly linked effector gene , pnf . We confirm in vivo expression during insect infection and reveal a high level of population heterogeneity of expression in vitro . We demonstrate for the first time the physical association of the Pnf effector toxin protein with the secreted structural needle complex using Western blot and electron microscopy . Furthermore , we prove that the cognate Pnf effector needs to be delivered into the eukaryote cell cytoplasm to exert any measurable effect and confirm its predicted activity targeting small Rho-GTPase target proteins . Taken together this work describes an important new class of protein toxin secretion and injection delivery systems which , unlike the well-described Types III , IV and VI systems , can act ‘at a distance’ , requiring no intimate contact between bacteria and host cells .
A comparison of pvc structural operons identified in the genome sequences of Photorhabdus and certain members of other genera , available at the time of publication ( Duchaud et al . , 2003; Wilkinson et al . , 2009; Wilkinson et al . , 2010 and our unpublished data ) , allowed us to define three distinct genetic sub-types . The PVCpnf operon belongs to class I , which has 16 structural genes and three translationally coupled gene blocks , and is of the type typically seen in non-Photorhabdus genera . Class II and III operons differ in the number of structural genes and translationally coupled gene blocks ( Figure 1B ) . Given the diversity of pvc-operons , and their typically poor annotation in genome sequences , it is necessary here to define a nomenclature protocol to allow reference to any given operon . An example of the method we have adopted is as follows; [Pa ATCC43949 PVCpnf] , where Pa ATCC43949 is species and strain , in this case Photorhabdus asymbiotica strain ATCC43949 and PVCpnf is the specific operon within that genome with the suffix referring to one of the tightly linked effectors , in this case the pnf effector gene . We will also include gene identifiers for either end of the operon where appropriate , which in this case would be PAU_03353-PAU_03332 , which are the genes for pvc1 and pnf respectively . With reference to published literature and a detailed bioinformatic analysis of promoter regions upstream of the pvc1 genes , we can identify two distinct , potential cis-operator sequences . Firstly operons belonging to classes I ( e . g . PVCpnf ) and III ( e . g . PVCPaTox ) typically encode the highly conserved RfaH operator sequence , GGCGGTAGNNT ( Belogurov et al . , 2009 ) . RfaH is a conserved anti-termination protein that is known to regulate large operons encoding for extracellular factors in E . coli . It is also believed to be important in ensuring appropriate transcriptional control of horizontally acquired operons ( Belogurov et al . , 2009 ) . Many of the pvc-operons in Photorhabdus and members of other genera ( including Xenorhabdus and Yersinia ) encode this operator sequence . An unusual example is the pADAP plasmid encoded Serratia entomophila anti feeding prophage ( afp ) ( Hurst et al . , 2004 ) . While the afp promoter also encodes an RfaH operator sequence , it has been demonstrated that it is positively regulated by a tightly linked specific regulator protein , AnfA1 ( Hurst et al . , 2007; Hurst et al . , 2003 ) . This protein is a distant homologue of RfaH suggesting that other class I or III pvc operons are not necessarily under the regulation of the chromosomal RfaH orthologue , but might also be controlled by other diverse regulators that utilise this same operator sequence ( Carter et al . , 2004 ) . Unfortunately , we were not able to directly test the role of RfaH in pvc-regulation as attempts to knock out the single P . luminescens TT01 rfaH gene were not successful , suggesting an essential role . Secondly , all class II operons ( e . g . PVClopT ) and certain class I operons ( e . g . PVCunits1-4 ) encode a minimal cryptic conserved sequence motif , CAGGTTGXTGCGGTAGCTAT . In both cases these conserved cis-encoded sequences are located between the pvc1 gene and the transcription start sites , as defined by previous RNA-seq analysis ( Mulley et al . , 2015 and unpublished data ) . We speculate that this motif , which we have so far only found in pvc loci , may represent an operator sequence for a cryptic regulator . Although there is no experimental evidence , several observations suggest that horizontal gene transfer may have been responsible for the dissemination of many observed pvc-operons . These include; the presence of four pvc operons in tandem in P . luminescens TT01 ( directly adjacent to a type IV DNA conjugation pilus operon ) , the presence of multiple pvc operons in any given genome , the suggestion that several operons are regulated by RfaH and the remnants of YhgA-like IS elements and inverted repeats ( IR ) associated with many of the operons ( e . g . the IR sequence 5’-TTATATTGAA ( t/g ) GAATATTAAGCAAGAAAC-3’ flanking [PlTT01PVCu4] ) . Nevertheless , automatic prediction of horizontal gene transfer regions ( HGTs ) using Alien Hunter 1 . 7 ( Vernikos and Parkhill , 2006 ) either did not detect any HGT elements spanning the structural regions of PVCs or in the cases where such an element was detected it was assigned a low confidence score ( Figure 1—figure supplement 1 ) . An analysis of the conservation of individual genes across different pvc operons at both DNA and protein sequence levels suggests that either recombination or diversifying selection is more likely to have occurred in the more 3’ regions of the operons ( Figure 1—figure supplement 2 ) . This is perhaps no surprise as each pvc operon can be seen to encode different effector genes in the 3’ payload region of the operons . An analysis of conservation of protein sequences of the pvc operons showed that within pvc-operons a good deal of variability is possible while presumably retaining the ability to produce a similar macromolecular structure ( Figure 1—figure supplement 3 ) . This is supported by HHPRED structural homology comparisons for equivalent PVC proteins across different operons , despite often-variable primary amino acid sequences ( data not shown ) . We note that the most diverse protein seen in pvc-operons is that of the predicted tail fibre proteins , Pvc13 , which we may expect if different pvc-operons are adapted for different host cell targets . Paralogous genes within pvc-operons include pvc1 and pvc5 which encode homologs of Hcp , the inner tube protein of contractile tube mechanisms such as T6SS and phage protein Gp27 and pvc2 , −3 and −4 which encode homologues of the outer sheath proteins of phage ( Leiman et al . , 2010 ) and T6SS ( Russell et al . , 2014 ) . Figure 1—figure supplement 4 illustrates the organisation of the [PaATCC43949 PVCpnf] operon used as a model system in our experimental studies described here , showing the top HHPRED structural homology hits and predicted roles for each encoded protein at the time of writing . A comparison of the 3’ effector ‘payload regions’ of different pvc operons reveals a large diversity of effector genes , with a range of predicted activities , covering a large range of sizes and isoelectric point values ( data not shown ) . Some operons encode only a single putative effector , for example [PaATCC43949 PVCPaTox PAU_02249–02230] while others have several , either tandem homologues of one another , for example [PaATCC43949 PVCu4 PAU_02790–02808] or entirely unrelated putative effector genes , for example [PaATCC43949 PVClopT PAU_02112–02095] . Many effector genes are also tightly linked to transposase gene remnants suggesting they are typically exchanged by horizontal acquisition . This is further supported by the observation that orthologous pvc-operons in the same chromosomal context may have different effector genes in different strains . A good example of this being the unrelated effector genes seen in the orthologous structural ‘PVCpnf’ operon loci of PaKingscliff and PaATCC43949 which carry a tyrosine glycosylase and Pnf ( this paper ) respectively . Analysis with Alien Hunter 1 . 7 , suggests that certain pvc-operon/effector associations are ancestral to any given species . For example , the association of the pvc17 effector with PVCu4 , and the multiple linked effectors with the PVClopT operon in both PaATCC43949 and PlTT01 . Conversely other pvc-operons show evidence of recent horizontal acquisition of their 3’linked effectors , for example PVCcif and PVCpnf ( not shown ) . A previous RNA-seq analysis of global transcription in three strains; P . asymbiotica ATCC43949 ( Mulley et al . , 2015 ) , P . asymbioticaKingscliff and P . luminescens TT01 ( unpublished ) showed condition dependent expression of certain pvc-operons but not all . Therefore , due to the diversity of pvc operons and effectors in Photorhabdus , we focused on a single model class I pvc operon , [PaATCC43949 PVCpnf] , to elucidate the relationship between the conserved structural and effector proteins . This operon was selected as it elaborates a well-defined needle complex structure ( as observed by electron microscopy ) which has potent insect killing activity when heterologously expressed in E . coli ( Yang et al . , 2006 ) . This operon has two putative effector genes in the downstream ‘payload region’ , PAU_03337 , which shows similarity to adenylate cyclase toxins ( e . g . the anthrax Edema Factor and Pseudomonas ExoY toxin ) and pnf ( PAU_03332 ) . While the predicted activity of PAU_03337 has not been tested directly , when expressed in the NIH-3T3 cell cytoplasm ( in transient transfection experiments ) it did produce a highly unusual cytoskeleton phenotype ( Yang et al . , 2006 ) . Pnf ( Photorhabdus necrosis factor ) is a homologue of the active site domain of the Yersinia CNF2 ( Cyto Necrosis Factor 2 ) toxin , which has small-GTPase deamidase and transglutaminase activities ( Knust and Schmidt , 2010 ) . In order to confirm the expression of this model pvc-operon in Photorhabdus during an insect infection we constructed transcription-translation reporter plasmids in which the promoter regions and the first 150 bp of coding sequence of pvc1 , pnf [both from PaATCC43949 PVCpnf] and the P . asymbiotica chromosomal rpsM ribosomal ‘housekeeping’ gene ( as a positive control ) were genetically fused in frame to a gfpmut2 gene with no start codon ( referred to hereon as pvc1::gfp , pnf::gfp and rpsM::gfp reporters ) . Note , the genomic context and our previous unpublished RT-PCR studies suggested that pnf had its own promoter and could be transcribed independently of the pvc structural genes . As we are unable to transform PaATCC43949 itself , these plasmids were transformed into the well-characterised and genetically tractable strain P . luminescensTT01 to provide suitable reporter strains for in vitro and in vivo expression studies . For in vitro studies we cultured the bacteria in LB medium supplemented with Manduca sexta clarified hemolymph and grown to late stationary phase , before microscopic examination ( Figure 2A ) . For in vivo studies , we injected a low inoculum ( c . a . 100 CFU ) of the reporter strains into M . sexta , and allowed the infection to establish for 12 hr before macroscopic examination of insect tissues in situ using a ( fluorescence ) dissecting microscope . We also took hemolymph samples from these insects and visualised the hemocytes and bacteria microscopically using confocal microscopy . This approach provides confidence that the cells showing gfp expression are those in the process of a normal productive infection rather than those carried through in the inoculum . Figure 2A shows expression of GFP reporter from the rpsM positive control and both the pvc1::gfp and pnf::gfp reporters in LB supplemented with M . sexta hemolymph , although not in all cells of the bacterial population ( see below ) . Furthermore , we also saw expression in bacteria in the ex vivo hemolymph samples taken during infection of live insects ( Figure 2A ) . It was also possible to confirm expression of pnf::gfp in bacteria attached to the insect trachea in localised putative biofilm masses . In this case , while the expected insect melanisation immune response could be seen to have occurred elsewhere on the trachea , it was notably absent from the pnf expressing bacterial biomass ( Figure 2B ) . We subsequently expanded this analysis to include a panel of transcription-translation reporter plasmids for different PVC operons from two different species of Photorhabdus . We cloned around 500 bp of the 5’ regions ( containing the promoter , translation initiation region and first ATG of pvc1 translationally fused to gfp ) of P . luminescensTT01; PVCunit1 , PVCunit4 , PVClopT and PVCcif and from P . asymbiotica PB68 . 1; PVCpnf , PVCunit , PVClopT and PVCcif ( see supplementary file ‘Reporter construct primers’ ) . Each of the eight reporter constructs were then transformed into the relevant Photorhabdus strain and examined using fluorescent microscopy to assess the expression patterns across growth phases , when grown in LB with aeration and maintaining plasmid marker selection . We initially focused on the P . asymbiotica PB68 . 1 [PVCpnf] reporter strain as it represents an orthologue of the P . asymbiotica ATCC43949 PVCpnf operon , which is the model operon central to this report . We observed a high level of population heterogeneity in expression , usually with only very few cells expressing GFP at any one time ( Figure 2—figure supplement 1 ) . Image analysis provides an objective assessment of heterogeneity , and small percentage of cells expressing Gfp , and thus the pvc-operon ( Figure 2—figure supplement 2 ) . Interestingly , a similar pattern of heterogeneity in expression was seen for the other seven operon reporter constructs from both PaPB68 and PlTT01 ( Figure 2—figure supplements 3 and 4 ) . Note in certain cases we saw no expression at all ( e . g . the PVClopT operon of PaPB68 . 1 ) , which fits with the low or lack of mRNA expression seen in a previous in vitro RNAseq study for the majority of pvc-operons ( upublished data ) . We also assessed expression in biofilms grown statically on glass slides and observed the same pattern , though with even fewer cells seen to express GFP ( not shown ) . In order to corroborate the observations made using the plasmid-based reporter constructs in P . luminescens TT01 during infection we also performed RT-PCR analysis of transcription of the PVCpnf chromosomal operon in the original PaATCC43939 strain . This confirmed transcription across the operon in vitro when the bacteria were grown at either 28°C or 37°C , although transcription of certain genes was difficult to detect in vivo during Manduca sexta infections ( Figure 3 ) . We investigated if the Pnf effector protein actually becomes physically associated with the pvc-encoded needle complex we had previously visualised by electron microscopy ( Yang et al . , 2006 ) . While a recent publication has described a cryo-EM structure of this particular PVC needle complex ( Jiang et al . , 2019 ) , they did not include expression of an effector . In order to investigate if the complex does indeed physically associate with a cognate effector , we raised anti-peptide antibodies against synthetic peptides representing amino acids 206–219 of effector Pnf ( TGQKPGNNEWKTGR ) and amino acids 130–143 ( DGPETELTINGAEE ) of the predicted outer sheath protein Pvc2 . We then used these antibodies to probe PVCpnf needles expressed from cosmid clones in E . coli . Previously we used 2D-SDS PAGE analysis of PVCpnf needle complex produced by these same cosmid clones to confirm the presence of Pvc2 , along with Pvc1 , 3 , 5 , 11 , 14 and 16 proteins ( Yang et al . , 2006 and unpublished data ) . We confirmed specificity of the Pnf antibody using western blot analysis of extracts of E . coli heterologously expressing Pnf alone . We first used the anti-Pnf peptide antibody to test for the presence of Pnf protein in supernatants from the native bacterial strain PaATCC43949 . We initially tested for the presence of Pnf in clarified supernatants and particulate preparations of stationary phase cultures . Previous studies reporting successful detection of Photorhabdus secreted insecticidal toxins under similar conditions . We could detect Pnf in preparations enriched for the complexes but not in clarified supernatants . More specifically , the Pnf protein could only be detected in the needle complex fraction of wild-type PaATCC43949 which was isolated from 250 ml overnight cultures using DEAE-Sepharose chromatography but not in concentrated supernatant collected from a culture grown from the same starter and under the same conditions . The toxin could only be detected if the needle complexes were first either chemically or physically disrupted before electrophoresis ( Figure 4B ) . Taken together these findings are consistent with the hypothesis that the Pnf protein is sequestered inside the needle complex or in some other configuration such that the TGQKPGNNEWKTGR epitope is physically hidden from access by the antibody . There is still the potential that the toxin is released in the supernatant but at not detectable amounts in these conditions . Secondly , we enriched needle complexes from insect toxic supernatants of an E . coli cosmid clone that encodes the PaATCC43949 PVCpnf operon , as previously described ( Yang et al . , 2006 ) . The anti-Pnf antibody was used for in situ labelling of Pnf on Transmission Electron Microscopy grids , visualised with negative staining and an anti-rabbit gold-conjugate secondary antibody . It was only possible to detect Pnf protein near the ends of either contracted or damaged needle complexes ( Figure 4A ) . Note we saw no non-specific labelling when the gold-conjugate secondary antibody was used alone . As the full cryo-EM structure of this model PVCpnf was recently published , it allowed us to examine the location of the epitope we used to raise and anti-peptide antibody to the main outer sheath component Pvc2 . While the epitope should indeed be surface exposed based on our analysis , we nevertheless only saw an anti-Pvc2 signal associated with what appeared to be disrupted fragments of needle complexes . We therefore surmise that the secondary structure of Pvc2 in the intact needle abolishes antibody binding to the epitope we selected . In a previous publication we reported that injection of an enriched PaATCC43949 PVCpnf needle complex preparation; heterologously produced by an E . coli cosmid clone , caused melanisation and death of Galleria mellonella larvae within 30 min . In addition , microscopic analysis of phalloidin stained hemocytes taken from these dying animals revealed the cells were shrunken with highly condensed cytoskeletons , and likely already dead . This effect was abolished by heat denaturing the preparation . In this same publication ( Yang et al . , 2006 ) we demonstrated that transient cytoplasmic expression of the Pnf protein caused extensive cytoskeleton re-arrangement and likely cell death in cultured Human HeLa ATCC CCL2 cells , similar to that observed in the ex vivo G . mellonella hemocytes . In an attempt to directly visualise the interaction of the heterologously produced PVCpnf needle complex with insect hemocytes and to determine the initial effects on the cellular morphology , we injected intact or heat denatured PVCpnf needle complex preparations into 5th instar Manduca sexta larvae before bleeding the animals and preparing their circulating hemocytes for surface examination by cryo-SEM . The surface of hemocytes taken from insects that had been injected with intact complex showed membrane ruffling consistent with the predicted mode of action of the Pnf protein ( see below ) . Furthermore , we could also see linear structures approximately 150 nm in length on the surface of the cells near the sites of membrane ruffles consistent with attached needle complexes . The surface of the control hemocytes taken from insects that had been injected with heat-denatured complex remained relatively smooth and homogeneous and we saw no equivalent linear structures . Figure 5 shows representative images from these experiments . We wished to know if the Pnf effector could exert this toxic effect independently of the needle complex , when applied externally to eukaryotic cells . Therefore , we heterologously expressed ( in E . coli ) and purified the Pnf protein in addition to a predicted toxoid derivative . The wild-type Pnf protein and toxoid derivative were purified using HisTrap Ni2+-affinity columns with the fast phase liquid chromatography ( FPLC ) AKTA system as described in the methods . SDS-PAGE was used to confirm high levels of purity and that no obvious degradation had occurred . The toxoid was designed based on homology between Pnf and the CNF2 toxin active site , wherein we mutated the cysteine at amino acid position 190 into an alanine ( Pnf C190A ) . Firstly , neither purified wild type nor toxoid proteins had any obvious toxic effect when injected into cohorts of G . mellonella , even at high doses ( data not shown ) . We subsequently used bioPORTER , a liposome based transfection system , to introduce the purified proteins directly into cultured human cells . We visualised effects on the cytoskeleton and nucleus using TRITC-phalloidin and DAPI staining respectively . The wild type Pnf protein had a very clear effect on the cells , producing phenotypes consistent with those predicted by similarity to the CNF2 toxin . CNF2 is known to modify the cellular Rho GTPases , RhoA , Rac1 and Cdc42 . Pnf delivery as a bioPORTER formulation led to the formation of F-actin filaments within 24 hr followed by multi-nucleation by 48 hr , phenotypes consistent with the modification of the Rho GTPases . The toxoid derivative , delivered at the same dose using the same approach , produced no changes , giving cellular phenotypes consistent with that of the negative control or of the wild-type Pnf protein topically applied without the bioPORTER transfection agent ( Figure 6 ) . Computational secondary structure predictions of the toxin and toxoid were performed to assess the potential impact of the C190A active site amino acid substitution using the Phyre2 algorithm . We obtained structural predictions ( 100% confidence score ) based on similarity to the known structure of a E . coli CNF1/2 family toxin active site domain ( PDB Entry: 1hq0 ) , which revealed only very minor predicted changes ( Figure 6—figure supplement 1 ) . Based on homology to CNF2 the effect of Pnf on target cell proteins is predicted to include the modification of several Rho-family GTPases . Therefore , we used western blot assays to examine in vitro transglutamination and deamidation effects of purified heterologously produced Pnf on purified small GTPases RhoA , Rac1 and Cdc42 . The RhoA , Rac1 and Cdc42 were purified from E . coli heterologous expression strains using GSTrap HP affinity columns with the FPLC AKTA system in accordance with previously published protocols , as described in the methods . SDS-PAGE analysis was used to confirm good levels of purity and lack of any significant degradation . Transglutamination is the formation of a covalent bond between a free amine group , as may be found on a lysine residue , and the gamma-carboxamide group of glutamine . As a result protein electrophoretic mobility of the protein is altered . Deamidation is a chemical reaction in which an amide functional group is removed from the protein , which may be detected using deamidated protein specific antibodies . These experiments demonstrated that Pnf induced transglutamination and deamidation of both RhoA and Rac1 ( Figure 7 ) , although unlike the reported activity of CNF2 , had no effect on Cdc42 . As predicted the active site toxoid mutant had no enzymatic activity on any of the three Rho GTPases confirming it was a true toxoid derivative .
An analysis of the different subunit proteins of PVCs shows they share several elements in common with other contractile phage-tail derived systems , including the Type VI secretion system ( T6SS ) ( Kapitein and Mogk , 2013 ) and to a lesser extent R-type pyocins ( Taylor et al . , 2018 ) . However , PVC-like elements are distinct in two important ways . Firstly , unlike the T6SS , they require no membrane complex for anchoring and synthesis and are freely released from the producing bacterial cell . Therefore , in common with R-type pyocins , they can act at a distance . Secondly , like T6SS but unlike R-type Pyocins , they are evolved to inject bioactive protein effectors into other cells . We hypothesise that the PVCs are evolved to specifically target eukaryotic cells , unlike T6SS , which have been shown to be able to deliver to both eukaryotes and prokaryotic competitors . However , while our previous attempts to show PVCpnf attachment to a range of bacterial species from different genera showed no binding we could detect ( data not shown ) , we cannot rule out the possibility that homologues exist which are able to target prokaryotes . We speculate that these large protein complexes are costly for the cell to produce , consistent with the observation of population heterogeneity of pvc-operon expression . Indeed , uncontrolled heterologous over expression of the PaATCC43949 PVClopT and PVCpnf operons in cosmid clones in E . coli , results in deletion of regions of the pvc-operon ( confirmed by sequencing ) and significant loss of culture viability ( not shown ) . It should be noted that in a natural insect infection the vast majority of the Photorhabdus bacterial population are sacrificial . The majority of the population act as a food source for the replicating nematodes , with very few cells passing into the next generation of infective juvenile nematodes ( Ciche et al . , 2008 ) . As such the population may restrict PVC production to a limited number of sacrificial cells . We have noted the presence of holin-lysin gene homologues tightly linked to certain pvc-operons , including both 5’ and 3’ to the PaATCC43949 PVCpnf operon which we focus on in this publication . Release by cell lysis would be consistent with the proposed method of release of the related MAC arrays of Pseudoalteromonas ( Shikuma et al . , 2014 ) . Nevertheless , to date we have not directly observed any cell lysis associated with pvc expression in Photorhabdus itself . The finding that PaATCC43949 PVCpnf , and seven other pvc-operons from PaPB68 and PlTT01 ( not shown ) all show population heterogeneity in expression , at least in vitro , suggests that they are likely deployed in a highly regulated and conservative manner . While it is difficult to fully characterise this heterogeneity in vivo , the PVCpnf GFP reporter strain did show restricted expression to one specific tissue , the insect trachea , and not throughout the body of the animal . In regards to these experiments , it should be noted that we did not see any melanisation response around the bacterial biomass showing GFP . The insect melanisation immune response is typically activated at sites of encapsulation . This is mediated by the recruitment of hemocytes , surrounding and enclosing foreign bodies , and entombing them in melanin . The absence of melanisation around this GFP expressing bacterial mass is consistent with the expression of anti-hemocyte virulence genes , which are likely to include the native PlTT01 pvc-operons . Examination of the promoter regions has provided no clue as to the mechanism of population heterogeneity of expression . Nevertheless , the identification of RfaH and a second cryptic conserved potential operator sequence upstream of the pvc1 genes provides a starting point for addressing this in future . Operons containing the second cryptic putative regulatory sequence include [Pl TT01PVClopT] and [Pl TT01 PVCu4] . Analysis of the supplementary data from a recently published RNA-seq study ( Tobias et al . , 2017 ) , suggests that these operons may be dependent upon Hfq/HexA activity ( Joyce and Clarke , 2003 ) . Unlike many of the other genera in which we see pvc-like operons , Photorhabdus genomes encode multiple copies , typically around 5 to 6 , suggesting they play important and diverse roles in the life cycle . With this in mind , we examined the conservation of the different subunit genes between operons . We observe a ‘break point’ in conservation , toward the 3’ end of the operons . We postulate this may be due to imprecise recombination events in the 3’ payload regions of pvc-operons , where incoming sequences , which have a GC-content that is distinct from the host genome , gradually ‘erode’ the upstream sequence . Alternatively , it is plausible that the lower GC at the distal end of these long operons ( each of ~25 kb ) may assist in strand separation during transcription , maintaining stoichiometry for these large , multi-subunit structures . Indeed low GC stretches of DNA are common origins of replication because of their reduced strand separation energy ( Meijer et al . , 1979 ) . However , as yet we do not know whether the pvc1 promoter serves the whole operon , or if there are additional promoters internal to the operon . Each of the pvc-operons in a Photorhabdus genome encodes multiple paralogous copies of pvc1/5 and pvc2/3/4 genes . We were therefore surprised not to see any operons showing signs of genetic degradation . This suggests there is sufficient positive selection for maintaining these multiple operons , with each operon potentially adapted for a specific role . This hypothesis is supported by the high variation in the Pvc13 protein sequences , which we speculate represent the host cell binding fibres . The need to maintain multiple copies of pvc-operons may also have arisen if the structural genes for the needle complexes are specifically adapted for delivery of their cognate cis-linked effector proteins in some way . Circumstantial evidence from genomic sequences and previous work on the related AFP system of Serratia has suggested the needle complexes serve to deliver the cis-encoded effector proteins . We present here for the first time direct evidence that a linked effector protein does in fact become physically associated with the needle complex . Western blot detection of Pnf from preparations enriched for needle complexes taken from the native PaATCC43949 supernatants confirmed it was being expressed in vitro and suggested it was physically associated with the complexes . In addition , physical or chemical disruption was required to release the Pnf protein for detection . When taken alongside the immuno-gold EM observations , showing Pnf could only be seen near contracted or damaged needle complexes , it confirms the protein is either sequestered inside the complex or physically associated in such as way that the TGQKPGNNEWKTGR epitope is not solvent accessible . The anti-Pvc2 antibody is able to specifically detect the protein in Western blots , however it only showed binding to what appeared to be disrupted fragments of needle complexes , again suggesting the relevant epitope is not accessible in the intact native needle complex structure . Indeed , iTasser structural model simulations of a PVC outer sheath Pvc2 protein , using the homologous Pseudomonas 3J9Q PDB structure of an R-type pyocin outer sheath as a model ( Ge et al . , 2015 ) , supports this idea , suggesting the epitope is partially occluded between adjacent subunits . While we have not yet directly demonstrated injection of Pnf into host cells by the needle complex , the results of the topical application and bioPORTER transfection experiments confirmed that the Pnf effector absolutely requires a mechanism to facilitate entry into the host cell cytoplasm to exert its effect . We argue the evidence for injection by the needle complex is very strong , and is corroborated by the SEM visualisation of needle-like structures of the correct dimensions on the surface of intoxicated hemocytes . Finally , we have confirmed that Pnf acts in a manner similar to the Yersinia CNF2 toxin , modifying two of the same Rho-GTPases , which correlates with the observed phenotypic effects on the cell .
Manduca sexta ( Lepidoptera: Sphingidae ) were individually reared as described ( Reynolds et al . , 1985 ) . Briefly , larvae were maintained individually at 25°C under a photoperiod of 17 hr light: 7 hr dark and fed on an artificial diet based on wheat germ . Larvae 1 day after ecdysis to the 5th instar were used for all experiments . Batches of wax moth larvae ( 75 g; Livefood UK Ltd , Rooks Bridge , UK ) in their final instar stage were stored in the dark at 4°C and used within a week of receipt . DH5α E . coli ( containing various plasmid constructs ) were grown on LB agar at 37°C or in LB liquid , shaking at 200 rpm . Spontaneous rifampicin-resistant mutants of Photorhabdus asymbiotica subsp . asymbiotica Thai ( strain PB68 . 1 ) ( Thanwisai et al . , 2012 ) and Photorhabdus luminescens subsp . laumondii TTO1 ( Duchaud et al . , 2003 ) were used in these studies as hosts for reporter plasmids . Photorhabdus were routinely cultured in LB broth or on LB agar supplemented with 0 . 1% ( w/v ) pyruvate at 30°C or 37°C ( for P . asymbiotica ) . When required antibiotics were added at the following concentrations: ampicillin ( Amp ) : 100 µg ml−1 , kanamycin ( Km ) : 25 µg ml−1 , chloramphenicol ( Cm ) : 25 µg ml−1 , rifampicin ( Rif ) : 25 µg ml−1 . Human adenocarcinoma HeLa ATCC CCL2 cells ( obtained from ATCC culture collection – mycoplasma free ) were cultured for 10 passages in Dulbecco’s modified Eagle medium ( Sigma-Aldrich ) containing 4 . 5 g/L glucose ( Sigma-Aldrich ) , 10% heat-inactivated fetal bovine serum ( Sigma-Aldrich ) , 2 mM glutamine ( Sigma-Aldrich ) , 100 μg /mL penicillin , and 100 μg/mL streptomycin ( Sigma-Aldrich ) and incubated at 37°C and 5% CO2 . Translational fusions with the gfpmut2 gene were constructed by PCR in a pACYC184 vector containing the gfpmut2 ( pACYC-GFP ) ( Jacobi et al . , 1998 ) as follows . The pvc1 , pnf and rpsM genes ( consisting of promoter regions and the first 150 bp of coding sequence ) were amplified from P . asymbiotica ATCC43949 genomic DNA and cloned into pACYC-gfp to generate pACYC-afp1-gfp , pACYC-pnf-gfp and pACYC-rpsM-gfp . The constructs were further digested to release the pvc1 , pnf or rpsM genes in frame with gfp and the fusion fragments were cloned into pBBR1-MCS ( Kanter-Smoler et al . , 1994 ) to generate pBBR1-pvc1-gfp , pBBR1-pnf-gfp and pBBR1-rpsM-gfp . Mating experiments were performed as previously described ( Brillard et al . , 2002 ) to transfer plasmid constructs into P . luminescens TTO1 resulting in PlTTO1-pvc1-gfp , PlTTO1-pnf-gfp and PlTTO1-rpsM-gfp . Plasmid stability was confirmed in bacteria harbouring the various constructs isolated after in vivo passages . For the expanded panel of gfp-reporter fusions , the promoter regions for the operons selected , inclusive of the putative RfaH operator sites ( if present ) , and the native RBS and first codon of the pvc1 gene ( approximately 500 bp upstream ) , were cloned in to the pAGAG vector . pAGAG was derivatised from the promoterless pGAG1 gfp bearing plasmid . , In brief , pGAG1 was used as a template to amplify gfpmut3* without a start codon using primers pG_GFPfor ( 5’-aatgtcgaccgtaaaggagaagaacttttc ) and pG_GFPrev ( 5’-aatactagtggatctatttgtatagttcatccatg ) . The resulting product and the pGAG1 vector were cut by digestion with SalI-HF and SpeI and ligated together , thus replacing the original intact gfpmut3* gene with one that lacks a ribosome binding site and the first ATG codon . This reporter plasmid backbone was used to construct a panel of specific transcription-translation reporter plasmids for different pvc-operon 5’ promotor regions . In brief regions 5’ to the pvc1 open reading frames from each operon were PCR amplified from P . luminescens strain TT01; PVCunit1 , PVCunit4 , PVClopT and PVCcif and from P . asymbiotica strain PB68 . 1; PVCPnf , PVCunit , PVClopT and PVCcif . Primers used are shown in Supplementary file ‘Reporter construct primers’ . All upstream promoter regions were incorporated between the KpnI and BamHI sites of the pAGAG vector . After initial cloning into E . coli , correct constructs , as confirmed by DNA sequencing , were transformed into the relevant Photorhabdus strain ( P . luminescens TT01 or P . asymbiotica PB68 . 1 ) for fluorescent microscopy studies . ( Figure 2—figure supplements 1 , 2 , 3 and 4 ) . Quantitative image analysis of micrographs of P . asymbiotica PB68 . 1 harbouring the PaPB68 PVCpnf pvc1 promoter fusion construct was performed using Fiji . Brightfield images were used to automatically detect the bacteria by converting the image to binary , followed by edge detection and particle analysis with the size of the particles set to 0 . 5–7 μm2 . The resulting regions of interest when then used to measure the intensity in the corresponding green channel ( for GFP detection ) . At least four images were used per time point with a minimum total number of cells of 450 for each time point . The control consists of images taken of P . asymbiotica PB68 . 1 harbouring the negative control plasmid pAGAG and for the purposes of this analysis the mean intensity per cell calculated for each time point were combined to find a threshold value for subtraction of background fluorescence . For the calculation of the percentage number of green cells for each time point , a threshold of 1200 was used above which there were no control cells with this mean intensity value . Raw data can be seen in Figure 2—figure supplement 2—source data 1 . Cosmid libraries of P . asymbiotica ATCC43949 were prepared in E . coli EC100 and arrayed into 96-well microtiter plates by MWG Biotech , Munich , Germany , as described previously ( Waterfield et al . , 2009b; Daborn et al . , 2002 ) . A 250 ml overnight culture of E . coli with the PaATCC43949 PVCpnf cosmid ( c4DF10 ) and a wild-type PaATCC43949 were grown in LB medium supplemented with 100 µg ml−1 ampicillin ( in case of the cosmid strain ) at 28°C with aeration in the dark . The cultures were centrifuged at 6800 x g at 4°C for 30 min at 4°C . The supernatants were decanted to remove each cell pellet , and the centrifugation procedure was repeated to remove any remaining cells . Cell-free supernatants were then centrifuged , in small batches , at 150 , 000 × g for 90 min at 4°C to harvest particulate material . The particulate pellets were washed by gentle re-suspension in 1 × Phosphate Buffered Saline ( PBS ) before a second centrifugation at 150 , 000 × g for 90 min at 4°C to pellet the particulate material . Each pellet was further separated by DEAE-Sepharose chromatography . 10 ml of particulate material in ice-cold PBS were mixed with an equivalent volume of DEAE-Sepharose CL-6B anion exchanger ( in PBS ) and the preparation was incubated at room temperature for 15 min . The Sepharose resin was harvested by centrifugation ( 3 , 000 × g ) , and the supernatant was discarded . The resin was resuspended in 40 ml of ice-cold PBS and again harvested by centrifugation . This washing step was repeated another three times , and the resin was finally resuspended in 10 ml of elution buffer ( 0 . 5 M NaCl , 50 mM phosphate buffer [pH 7 . 4] ) . The resin was removed by centrifugation , and the supernatant containing the PVCs was again centrifuged at 150 , 000 × g for 90 min at 4°C to pellet the particulate material and concentrate the needle structures in 500 μl of ice-cold PBS . For transmission electron microscopy ( TEM ) pioloform-covered 300-mesh copper grids that were coated with a fine layer of carbon were used as substrates for the protein fractions . The following four aqueous negative stains were tested with the protein samples: 1% uranyl acetate , 3% ammonium molybdate , 3% methylamine tungstate , and 2% sodium silicotungstate . The preferred stain , 3% methylamine tungstate , produced acceptable contrast and minimum artefacts and was subsequently used for all samples viewed by TEM . The coated grids were exposed to UV light for 16 hr immediately prior to use to ensure adequate wetting of the substrate . A 10 μl drop was applied to the TEM grid , and the protein was allowed to settle for 5 min . Liquid was absorbed with filter paper from the edge of the grid and replaced immediately with 10 μl of filtered negative stain . The drop was partially removed with filter paper , and the grids were allowed to air dry thoroughly before they were viewed with a JEOL 1200EX transmission electron microscope ( JEOL , Tokyo , Japan ) operating at 80 kV . Production and isolation of secreted Pnf from wild-type P . asymbioticaATCC43949 was performed at the same time as the production and isolation of wild-type needle complex in a duplicate culture . Bacterial cultures were grown from the same starter culture and under the same conditions . However , for isolation of secreted proteins , the supernatant was filtered through 0 . 45 µm and 0 . 22 µm filters to remove any remaining bacteria and particulate matter . Then it was concentrated using 4% TCA overnight at 4°C and the next morning washed and resuspended in approximately 10 ml of 50 mM Tris-HCl pH 7 . 4 . The secreted proteins were further concentrated using Amicon filters with NMWL of 10 kDa to 500 µl . Pnf gene was amplified from P . asymbiotica ATCC43949 genomic DNA ( using primers Pnf_NdeI 5’-ATATATCATATGATGTTAAAATATGCTAATCCT-3’ , Pnf_BamHI 5’-ATATATGGATCCTTATAACAACCGTTTTTTAAG-3’ ) and the PCR product was purified and cloned in-frame with a His-tag into the IPTG-inducible expression plasmid pET-15b ( Novagen ) to create construct pET15b-Pnf . The clone was verified by sequencing and transformed into Arctic Express competent cells ( Agilent ) for protein expression . A site-directed mutant of Pnf ( toxoid ) was generated with the QuikChange site-directed mutagenesis kit ( Agilent ) . To construct the Pnf mutant plasmid pET15b-PnfC190A , pET15b-Pnf was amplified with FPLC-purified primers designed to generate a Cys to Ala substitution at position 190 ( PnfC190A_for 5'-TCACCGAATATACCATAGTAGCACCGCTCAATGCTCCAGAC-3’ , PnfC190A_rev 5’-GTCTGGAGCATTGAGCGGTGCTACTATGGTATATTCGGTGA-3’ ) using the following thermal profile ( step 1: 95°C for 30 s , step 2: 95°C for 30 s , 55°C for 60 s , 68°C for 6 min 45 s for 16 cycles ) . The identity of six different positive clones was confirmed by sequencing . Subsequently , −80°C glycerol stocks were used to inoculate 5 ml of fresh LB medium supplemented with 0 . 2% ( w/v ) glucose and 100 µgml−1 ampicillin . Bacteria were grown overnight at 30°C with shaking , and 1 ml of the culture was then harvested , re-suspended in 100 ml of the same medium , and incubated in an orbital incubator at 37°C until the optical density at 600 nm was 0 . 7 to 0 . 9 . Cells were then harvested at room temperature by centrifugation at 4 , 000 rpm for 10 min . The pellet was re-suspended in 100 ml of fresh LB medium supplemented with the 100 µg ml−1 ampicillin and 0 . 1 mM of the inducer isopropyl-β-d-thiogalactopyranoside ( IPTG ) . Optimized times for inductions were determined experimentally , and cells were then harvested . The bacterial cell pellet was re-suspended in 10 ml of 1x PBS and sonicated ( four 20 s sonications at 45 mA using a Branson 450 digital Sonifier fitted with a tapered probe ) . The freshly sonicated samples were then diluted in 1x PBS for injection into Galleria larvae and for SDS-polyacrylamide gel electrophoresis analysis to confirm expression of the target protein . For toxicity testing cohorts of Galleria larvae ( n = 20 ) were chilled on ice before injection with 10 μl of a dilution series ( in sterile PBS ) of sonicated cells expressing Pnf or vector control . Insects were then returned to room temperature and observed for 5 days or mortality or morbidity . ArcticExpress containing pET-15b-Pnf were initially grown in LB broth supplemented with 100 µg ml−1 ampicillin at 37°C until OD 0 . 6 when Pnf expression was induced with a final concentration of 0 . 1 mM of IPTG at 12°C for 16 hr to produce soluble Pnf . Pnf was purified over HisTrap Ni2+-affinity column with the fast phase liquid chromatography ( FPLC ) AKTA system as per the manufacturer’s protocol ( GE Healthcare ) . Plasmids pGEX-2T-wtRhoA , pGEX-2T-wtRac1 and pGEX-2T-G25K ( Cdc42 ) were gifts from Prof Alan Hall ( University College London , London , UK ) and were maintained in E . coli DH5α grown on LB agar or in LB broth supplemented with 50 µg ml−1 ampicillin . RhoA , Rac1 and Cdc42 were purified over GSTrap HP affinity columns with the FPLC AKTA system as per the manufacturer’s protocol ( GE Healthcare ) . For BioPORTER assays , 80 µl of purified wild type and mutant Pnf proteins ( 500 µg ml−1 ) , or PBS as a negative control , were added to one BioPORTER tube ( Genlantis ) and resuspended in 920 µl of DMEM . The samples were added to Human HeLa ATCC CCL2 cells ( ATCC collection ) grown in 6-well plates and incubated for 4 hr . BioPORTER/protein or PBS mixes were replaced by fresh complete medium and the cells were incubated for 20–48 hr . To visualise cell morphology and actin cytoskeleton , cells were fixed for 15 min in 4% PBS-formaldehyde , permeabilized with 0 . 1% Triton X-100 and stained with Tetramethylrhodamine B isothiocyanate ( TRITC ) -phalloidin ( Sigma ) and DAPI dihydrochloride ( Sigma ) . Images were acquired with a LSM510 confocal microscope ( Leica ) . Transglutamination assays were done as previously described ( Schmidt et al . , 1999 ) with several modifications . Briefly , a 2:1 molar ratio of RhoA to toxin was incubated in transglutamination buffer ( 50 mM Tris-HCl pH 7 . 4 , 8 mM CaCl2 , 5 mM MgCl2 , 1 mM DTT , 1 mM EDTA ) in the presence of 50 mM ethylenediamine ( which raised the pH of the buffer to 9 ) for 10 min or 1 hr at 37°C . As a negative control , RhoA was incubated with ethylenediamine but without toxin . Samples ( 0 . 25 µg RhoA/well ) were subjected to SDS-PAGE and then processed for Western blot analyses as described above . Immunoblots were probed with a mouse anti-RhoA monoclonal antibody ( 1:1500 , Santa Cruz Biotechnology ) and reactive proteins visualised with DAB after incubation with the HRP-conjugated goat anti-mouse IgG secondary antibody . DNA sequences for each of the 16 conserved structural loci were clustered syntenically ( all pvc1s , all pvc2’s etc . ) . % GC content for each CDS in each syntenic position was calculated ( up to 16 observations per locus ) , and plotted as a boxplot via ggplot2 ( Figure 1—figure supplement 2 ) . The average GC content across the full operon , as well as for the whole genome , were plotted as intervals in the plot background to show the PVC loci %GC in contrast . The breakpoint was defined by use of the ‘cumSEG’ package in R ( Muggeo and Adelfio , 2011 ) . Amino acid similarity scores ( Figure 1—figure supplement 3 ) were generated by CLUSTAL Omega ( Sievers et al . , 2011 ) multiple sequence alignment , using default parameters . Resulting pairwise alignment scores were plotted as boxplots using ggplot . For in vitro transcription analysis , overnight cultures of P . asymbiotica were sub-cultured into liquid LB medium and grown with aeration at 28°C or 37°C 200 rpm in the dark . Planktonic cultures were collected at 4 , 8 and 24 hr and mixed with a double volume of RNAlater ( Ambion ) and after 5 min incubation , bacteria were harvested by centrifugation and the pellets stored at −80°C . For in vivo transcription analysis , overnight cultures of P . asymbiotica were extensively washed in PBS and diluted in Grace’s insect media ( GIM ) to achieve 1000 bacteria per 50 μl of culture . Each M . sexta larvae was injected with 50 μl of P . asymbiotica culture and they were placed in a humid temperature-controlled room at 28°C . After 3 hr or 6 hr of incubation , insects were bled in equal volume of GIM containing 20 mM phenylthiocarbamide ( PTC ) . The sample was initially fractionated into plasma and total hemocytes by centrifugation at 200 x g at 4°C for 5 min , and plasma was further centrifuged at 6800 x g at 4°C for 5 min to form a bacterial pellet . For each condition , total RNA was extracted using the RNeasy Mini Kit ( Qiagen ) and 2 μg total RNA was treated with TURBO DNA-free Kit ( Ambion ) and subjected to RT-PCR using the Qiagen OneStep RT-PCR kit . Each RT-PCR reaction performed in a volume of 50 μl ( containing 100 ng template RNA , 1x QIAGEN OneStep RT-PCR buffer , 400 μM dNTPs , 0 . 6 μM gene specific primers , 5U RNase inhibitor and 2 μl of QIAGEN OneStep RT-PCR enzyme mix ) for 28 cycles . | Photorhabdus are the only known group of non-marine bacteria that can produce their own light . These organisms prey on insects , which then glow in the dark once infected . The group also has an unusual weapon system formed of miniscule needle-like structures that can be sent out in the environment . These ‘Photorhabdus virulence cassettes’ are loaded with toxins that are injected inside host cells; the cassettes alone can kill a caterpillar within minutes . However , it is still unclear how exactly these structures work: are they like poison darts , with the toxin on the outside , or like hypodermic needles , with the toxin within ? Photorhabdus bacteria make lots of deadly substances , so to look at the needles on their own , Vlisidou et al . had them produced by another species of bacteria that does not carry these weapons . The experiments revealed that the cassettes packaged the toxic proteins inside , like a hypodermic needle . Alone in the environment , the toxin cannot penetrate host cells . Creating the cassettes takes a lot of energy , and a closer look at Photorhabdus showed that , at any given time during an infection , only a small number of bacteria produce them . It is therefore possible that the bacteria share the high cost of making these deadly devices by using a division of labour approach . With a better understanding of the cassettes , it could one day become possible to harness these molecular machines to control insect pests or parasites . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"genetics",
"and",
"genomics"
] | 2019 | The Photorhabdus asymbiotica virulence cassettes deliver protein effectors directly into target eukaryotic cells |
Enduring memories of sensory cues associated with drug intake drive addiction . It is well known that stressful experiences increase addiction vulnerability . However , it is not clear how repeated stress promotes learning of cue-drug associations , as repeated stress generally impairs learning and memory processes unrelated to stressful experiences . Here , we show that repeated social defeat stress in rats causes persistent enhancement of long-term potentiation ( LTP ) of NMDA receptor-mediated glutamatergic transmission in the ventral tegmental area ( VTA ) . Protein kinase A-dependent increase in the potency of inositol 1 , 4 , 5-triphosphate-induced Ca2+ signaling underlies LTP facilitation . Notably , defeated rats display enhanced learning of contextual cues paired with cocaine experience assessed using a conditioned place preference ( CPP ) paradigm . Enhancement of LTP in the VTA and cocaine CPP in behaving rats both require glucocorticoid receptor activation during defeat episodes . These findings suggest that enhanced glutamatergic plasticity in the VTA may contribute , at least partially , to increased addiction vulnerability following repeated stressful experiences .
Humans with a history of stressful or traumatic experiences are more prone to develop substance use disorders ( Sinha , 2008 ) . Adverse experience recruits the hypothalamic-pituitary-adrenal ( HPA ) axis stress response , culminating in release of glucocorticoids that enables the body to cope with insults to homeostasis ( Munck et al . , 1984 ) . In rodent models , repeated activation of the stress response typically disrupts learning and cognition [e . g . , spatial learning ( Conrad et al . , 1996 ) , working memory ( Mizoguchi et al . , 2000 ) , and cognitive flexibility ( Liston et al . , 2006 ) ] . In contrast to these deficits , prior stress enhances the learning of Pavlovian cue-outcome associations driven by rewarding stimuli , assessed with conditioned place preference ( CPP ) ( Kreibich et al . , 2009; Burke et al . , 2011; Chuang et al . , 2011 ) , or aversive/stressful stimuli , assessed with fear conditioning ( Conrad et al . , 1999; Sandi et al . , 2001; Suvrathan et al . , 2014 ) . These effects of stress may have arisen from evolutionary pressure to rapidly acquire information predicting food , shelter , and predator threat during periods of duress . Augmented Pavlovian reward learning mechanisms in stressed individuals may also heighten susceptibility to addiction , as acquisition of cue-drug associations is a crucial early step in drug use , and powerful , enduring memories of drug-associated cues trigger craving and relapse as recreational use progresses to addiction ( Hyman et al . , 2006 ) . However , it is not clear how repeated stressful experience promotes the learning of cue-drug/reward associations , as repeated stress is generally detrimental to synaptic plasticity underlying learning and memory unrelated to stressful events ( Kim and Diamond , 2002; Joels et al . , 2006; Schwabe et al . , 2012; Chattarji et al . , 2015 ) . The mesolimbic dopamine system originating in the ventral tegmental area ( VTA ) is critical for reward processing . VTA dopamine neurons tonically fire action potentials ( APs ) at 1–5 Hz , while responding to unexpected rewards with phasic burst firing ( 2–10 APs at 10–50 Hz ) . These dopamine neuron responses are hypothesized to drive the learning of Pavlovian cue-reward associations ( Tsai et al . , 2009; Darvas et al . , 2014 ) . Intriguingly , over the course of repeated cue-reward pairing , dopamine neurons acquire a conditioned burst response to reward-predictive cues , which is thought to encode the positive motivational valence of those cues and to invigorate reward-seeking behavior ( Schultz , 1998; Berridge et al . , 2009; Bromberg-Martin et al . , 2010 ) . Glutamatergic inputs activating NMDA receptors ( NMDARs ) drive the transition from tonic firing to bursting in dopamine neurons ( Overton and Clark , 1997; Zweifel et al . , 2009; Wang et al . , 2011 ) ; therefore , potentiation of cue-driven NMDAR inputs may contribute to the acquisition of conditioned bursting . Indeed , NMDAR-mediated transmission undergoes long-term potentiation ( LTP ) when cue-like glutamatergic input stimulation is repeatedly paired with reward-like bursting in dopamine neurons ( Harnett et al . , 2009 ) . LTP induction requires amplification of burst-evoked Ca2+ signals by preceding activation of group I metabotropic glutamate receptors ( mGluRs; more specifically mGluR1 ) coupled to the generation of inositol 1 , 4 , 5-triphosphate ( IP3 ) . Here , IP3 receptors ( IP3Rs ) detect the coincidence of IP3 generated by glutamatergic input activating mGluRs and burst-driven Ca2+ entry . IP3 enhances Ca2+-induced activation of IP3Rs by promoting access to the stimulatory Ca2+ sites , thereby promoting Ca2+-induced Ca2+ release from intracellular stores ( Taylor and Laude , 2002 ) . In this study , we demonstrate that repeated social defeat stress ( 1 ) enhances NMDAR LTP in the VTA via an increase in IP3 sensitivity of IP3Rs and ( 2 ) promotes acquisition of cocaine CPP in behaving rats , and both of these effects require glucocorticoid action during defeat stress .
NMDAR LTP induction requires mGluR/IP3-induced facilitation of burst-evoked Ca2+ signals ( Harnett et al . , 2009 ) . Therefore , we first examined the effect of the group I mGluR agonist DHPG ( 1 µM; 5-min perfusion ) on burst-evoked Ca2+ signals , assessed by the size of Ca2+-activated SK currents ( termed burst IK ( Ca ) ) in control and stressed animals . Rats were unhandled , handled , or socially defeated ( at the end of the dark cycle ) for 1 , 5 , or 10 consecutive days , and VTA slices were prepared 1–2 days after the final handling/defeat session . The magnitude of DHPG effect on IK ( Ca ) was significantly larger in animals that underwent 5 or 10 days of defeat stress compared to unhandled and handled controls , whereas a single defeat session failed to alter the DHPG effect ( Figures 1A and B ) . There was no significant difference between unhandled and handled controls . The effect of stress plateaued by 5 days , as comparable enhancement of DHPG effect was observed after 10-day defeat . Basal burst IK ( Ca ) was consistent across groups ( Figure 1C ) , suggesting no alterations in AP-evoked Ca2+ influx . DHPG-induced inward currents , which are independent of Ca2+ signaling ( Guatteo et al . , 1999 ) , were not affected ( Figure 1D ) ; thus the stress-induced increase in IK ( Ca ) facilitation results from changes in IP3 signaling downstream of mGluRs . 10 . 7554/eLife . 15448 . 003Figure 1 . mGluR-dependent facilitation of burst-evoked Ca2+ signals is enhanced after repeated social defeat . ( A ) Example traces ( left ) and summary time graph ( right ) illustrating the facilitating effect of DHPG ( 1 µM ) on burst IK ( Ca ) in neurons from unhandled rats ( traces not shown ) , rats handled for 5 days , and rats that underwent social defeat for 1 , 5 , or 10 days . ( B ) Summary bar graph showing the magnitude of DHPG-induced burst IK ( Ca ) facilitation ( unhandled: 20 cells from 12 rats , handled: 20 cells from 13 rats , 1 day defeat: 19 cells from 11 rats , 5 day defeat: 21 cells from 13 rats , 10 day defeat: 19 cells from 10 rats; F4 , 94 = 6 . 19 , p<0 . 001 , one-way ANOVA ) . *p<0 . 05 , **p<0 . 01 ( Bonferroni post hoc test ) . ( C ) The size of basal burst IK ( Ca ) was not altered by social defeat . ( D ) DHPG-induced inward currents were not affected by social defeat . Inset: Example traces of DHPG-induced currents from unhandled and 5-day defeated rats ( 5-min DHPG perfusion at the horizontal bar ) . ( E ) Summary graph depicting DHPG effect on burst IK ( Ca ) after different intervals following 5-day social defeat . Data in 6–7 weeks old unhandled and 1–2 day interval groups were from those in panels A–D ( 6–7 weeks old unhandled: 20 cells from 12 rats , 1–2 day interval: 21 cells from 13 rats , 10-day interval: 17 cells from 8 rats , 30-day interval: 16 cells from 9 rats , 10–11 weeks old unhandled: 15 cells from 7 rats ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 003 Next , to examine the persistence of repeated stress effect , the interval between the last social defeat session and recording was prolonged to 10 and 30 days . Although stress-induced enhancement displayed gradual recovery , DHPG effect was still elevated after 30 days compared to age-matched controls ( Figure 1E ) . Subsequent electrophysiology experiments were performed in 5-day defeated rats ( with 1–2 day interval ) and controls ( unhandled and handled controls combined ) . To directly examine alterations in IP3 signaling , we applied different concentrations of IP3 ( expressed in µM·µJ; see Methods and materials ) into the cytosol using flash photolysis of caged IP3 , and IP3R-mediated Ca2+ release was assessed by flash-evoked SK currents ( IIP3 ) ( Figure 2A ) . The average IP3 concentration-response curve displayed a leftward shift in defeated rats compared to controls ( Figure 2B ) . Accordingly , the average EC50 valuewas significantly smaller in the defeated group ( Figure 2C ) . Maximal IIP3 amplitude did not differ between groups ( Figure 2D ) , indicating a change in the potency , but not the efficacy , of IP3 in eliciting Ca2+ release . 10 . 7554/eLife . 15448 . 004Figure 2 . PKA activity maintains increased IP3R sensitivity in socially defeated rats . ( A ) Example traces of IIP3 evoked by different concentrations of IP3 ( 2000 , 6000 , 16000 , 48000 , and 140000 µM·µJ ) in control and defeated rats . ( B ) Averaged IP3 concentration-response curves from control and defeated rats . IIP3 amplitudes were normalized to the maximal value ( estimated from fit to a logistic equation ) in each cell . Recordings were made with normal internal solution or with PKI ( control: 12 cells from 8 rats , defeat: 12 cells from 7 rats , control + PKI: 15 cells from 9 rats , defeat + PKI: 14 cells from 8 rats; treatment ( defeat/PKI ) : F3 , 196 = 4 . 88 , p<0 . 01; IP3 concentration: F4 , 196 = 1214 , p<0 . 001; treatment × IP3 concentration: F12 , 196 = 4 . 42 , p<0 . 001; mixed two-way ANOVA ) . ***p<0 . 001 vs . control ( Bonferroni post hoc test ) . Lines represent logistic fit to the averaged data in each group . ( C ) Summary bar graph depicting the average EC50 values ( EC50 determined in each cell ) in the 4 groups shown in ( B ) ( defeat: F1 , 49 = 5 . 11 , p<0 . 05; PKI: F1 , 49 = 5 . 11 , p<0 . 05; two-way ANOVA ) . *p<0 . 05 ( Bonferroni post hoc test ) . ( D ) The maximal IIP3 amplitude was not affected by social defeat experience or PKI during recording . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 004 Protein kinase A ( PKA ) -dependent phosphorylation of IP3Rs increases their IP3 sensitivity ( Wagner et al . , 2008 ) . To determine the involvement of PKA in stress-induced IP3R sensitization , the effect of a selective peptide inhibitor of PKA , PKI- ( 6–22 ) -amide ( PKI; 200 µM , loaded into the cytosol via the whole-cell pipette for >15–20 min after break-in ) , was tested . PKI reversed the leftward shift in IP3 concentration-response curve , and thus the decrease in EC50 value , in stressed animals , while having no significant effect on maximal IIP3 amplitude ( Figures 2B–D ) . PKI had no effect in the control group , suggesting low basal PKA activity in non-stressed animals . It should be noted that PKI eliminated the difference in IP3 potency between the two groups . These data indicate that repeated social stress sensitizes IP3Rs via a PKA-dependent mechanism . We next examined whether repeated social defeat affects NMDAR LTP induction , which requires mGluR/IP3-dependent facilitation of burst-evoked Ca2+ signals and is gated by PKA ( Harnett et al . , 2009 ) . Application of a low concentration of IP3 preceding APs can effectively facilitate IK ( Ca ) ( Cui et al . , 2007; Ahn et al . , 2010; Bernier et al . , 2011 ) . Thus , the LTP induction protocol consisted of applying a low concentration of IP3 ( 250 µM·µJ ) 50 ms prior to simultaneous pairing of a burst with a brief train of synaptic stimuli ( Figure 3A ) , the latter being necessary to activate NMDARs at stimulated synapses at the time of burst for LTP induction ( Harnett et al . , 2009; Whitaker et al . , 2013 ) . This induction protocol produced little LTP in control animals , while large LTP was induced in defeated animals ( Figure 3B and C ) . IP3 application , which caused little IIP3 by itself , caused facilitation of burst IK ( Ca ) ( assessed immediately before LTP induction ) , which was significantly larger in cells from defeated animals ( Figure 3D ) . Furthermore , the magnitude of LTP was positively correlated with that of IP3-induced facilitation of IK ( Ca ) across neurons from both groups ( Figure 3E ) . Robust LTP was induced in control rats when a higher IP3 concentration ( 500 µM·µJ ) , which produced larger IK ( Ca ) facilitation , was used during induction ( Figure 3—figure supplement 1 ) . These results suggest that the enhanced LTP in defeated rats is a consequence of increased IP3R sensitivity enabling greater facilitation of burst-evoked Ca2+ signals . 10 . 7554/eLife . 15448 . 005Figure 3 . NMDAR-mediated transmission is more susceptible to LTP induction after social defeat . ( A ) Example experiments to induce NMDAR LTP in neurons from control and defeated rats . Time graphs of NMDAR EPSCs are shown with example traces at times indicated by numbers ( baseline: gray , post-induction: black ) . The LTP induction protocol ( IP3-synaptic stimulation-burst combination; illustrated in top inset ) was delivered after 10-min baseline recording ( at arrow ) . ( B ) Summary time graph of baseline-normalized NMDAR EPSCs in LTP experiments ( control: 7 cells from 7 rats , defeat: 7 cells from 7 rats ) . ( C ) Summary of NMDAR LTP magnitude in control and defeated rats ( t12 = 3 . 93 , **p<0 . 01 , unpaired t-test ) . ( D ) Example traces ( left; from the defeated rat shown in A ) and summary ( right ) of IK ( Ca ) facilitation by IP3 assessed before LTP induction ( t12 = 4 . 65 , ***p<0 . 001 , unpaired t-test ) . ( E ) The magnitude of NMDAR LTP is plotted versus the magnitude of IP3-induced facilitation of IK ( Ca ) . Dashed line is a linear fit to all data points from both control and defeated rats . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 00510 . 7554/eLife . 15448 . 006Figure 3—figure supplement 1 . Summary time graph of baseline-normalized NMDAR EPSCs in LTP experiments using a high concentration of IP3 ( 500 μM·μJ ) during induction in control rats ( 4 cells from 4 rats ) . Facilitation of burst IK ( Ca ) by IP3 was 26 ± 5% in these cells . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 006 It has been reported that repeated stress alters NMDAR expression in certain brain areas ( Fitzgerald et al . , 1996; Yuen et al . , 2012; Costa-Nunes et al . , 2014; Chattarji et al . , 2015 ) . We found that bath application of NMDA ( 10 µM ) produced comparable inward currents in control and defeated rats ( Figure 4 ) ; thus repeated defeat stress caused no significant changes in global NMDAR-mediated excitation . 10 . 7554/eLife . 15448 . 007Figure 4 . Summary time graph depicting inward currents induced by 1-min perfusion of NMDA ( 10 μM ) in VTA dopamine neurons from control ( 8 cells from 3 rats ) and 5 day defeated rats ( 6 cells from 2 rats ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 007 Repeated stress appears to differentially modulate tonic firing of VTA dopamine neurons recorded in vivo , as both an increase and decrease have been reported with different stress paradigms ( Cao et al . , 2010; Valenti et al . , 2012; Tye et al . , 2013 ) . However , repeated social defeat failed to alter tonic firing measured in ex vivo slices ( Figure 5A ) . Furthermore , the amplitude of hyperpolarization-activated cationic currents ( Ih ) , which contribute to intrinsic dopamine neuron pacemaker activity ( Neuhoff et al . , 2002 ) , was not affected ( Figure 5B ) . 10 . 7554/eLife . 15448 . 008Figure 5 . Tonic firing is unaltered by social defeat . ( A ) Example traces ( left ) and summary ( right ) of tonic firing frequency in VTA neurons from control and defeated rats ( control: 15 cells from 3 rats , 5 defeats: 9 cells from 3 rats; t22 = 0 . 066 , p=0 . 95 , unpaired t-test ) . In these experiments , loose-patch recordings ( <20 MΩ seal ) were made using pipettes filled with 150 mM NaCl to monitor tonic pacemaker firing . ( B ) Example traces ( left; voltage step depicted at bottom ) and summary ( right ) of Ih currents recorded in cells from animals that underwent control procedures or 1 , 5 , or 10 days of defeat ( data were obtained from the same cells shown in Figures 1A–D; F4 , 94 = 2 . 01 , p=0 . 10 , one-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 008 A major consequence of stress-induced HPA axis activation is the secretion of glucocorticoids ( corticosterone in rodents ) into the blood ( Munck et al . , 1984 ) . Thus , we sought to determine whether corticosterone , which readily crosses the blood-brain barrier , is involved in the increase in IP3R sensitivity with repeated stress . Corticosterone activates both glucocorticoid receptors ( GRs ) and mineralocorticoid receptors ( MRs ) ; however , MRs are typically saturated by circadian fluctuations in corticosterone , while lower-affinity GRs are activated by levels attained with stress ( Joels and de Kloet , 1994 ) . Therefore , the role of GRs was examined by treating rats with the antagonist mifepristone ( 40 mg/kg , i . p . ) or vehicle 30 min prior to each defeat session . The DHPG effect on burst IK ( Ca ) was significantly enhanced by repeated defeat in the vehicle-treated group , while there was no effect of stress in the mifepristone-treated group ( Figure 6A ) . Thus , blockade of GRs during defeat sessions prevented IP3R sensitization . Next , rats were treated with corticosterone ( 2 . 5 , 5 , or 15 mg/kg , i . p . ; at the end of the dark cycle ) for 5 days ( no social defeat ) . The lowest dose ( 2 . 5 mg/kg ) produces elevation in blood corticosterone concentration comparable to that evoked by a moderate stressor ( Graf et al . , 2013 ) , while higher doses ( ≥10 mg/kg ) have been used to simulate severe stress levels ( Akirav et al . , 2004 ) . None of the tested doses significantly altered the DHPG effect on burst IK ( Ca ) ( Figure 6B ) . Together , these results show that GR signaling is necessary , but not sufficient , for IP3R sensitization . 10 . 7554/eLife . 15448 . 009Figure 6 . Stress-induced , but not cocaine-induced , IP3R sensitization is prevented by GR blockade . ( A ) Example traces ( left ) and summary ( right ) of DHPG-induced burst IK ( Ca ) facilitation in neurons from animals that were injected with vehicle or mifepristone before undergoing control handling or social defeat sessions ( vehicle + control: 8 cells from 5 rats , vehicle + defeat: 10 cells from 4 rats , mifepristone + control: 10 cells from 5 rats , mifepristone + defeat: 11 cells from 4 rats; defeat × mifepristone: F1 , 35 = 4 . 56 , p<0 . 05 , two-way ANOVA ) . *p<0 . 05 ( Bonferroni post hoc test ) . ( B ) Summary bar graph showing that repeated corticosterone treatment ( once daily for 5 days ) failed to affect DHPG-induced burst IK ( Ca ) facilitation ( vehicle: 8 cells from 5 rats , 2 . 5 mg/kg: 11 cells from 5 rats , 5 mg/kg: 10 cells from 4 rats , 15 mg/kg: 12 cells from 7 rats ) . ( C ) Summary bar graph demonstrating that mifepristone pretreatment failed to block the increase in DHPG effect resulting from repeated cocaine treatment ( 10 mg/kg , i . p . , once daily for 5 days ) ( saline: 17 cells from 7 rats , cocaine: 16 cells from 7 rats , mifepristone + saline: 21 cells from 8 rats , mifepristone + cocaine: 16 cells from 6 rats; cocaine: F1 , 66 = 11 . 4 , p<0 . 01 , two-way ANOVA ) . *p<0 . 05 ( Bonferroni post hoc test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 009 It has been shown that repeated psychostimulant treatment sensitizes IP3Rs and enhances NMDAR LTP ( Ahn et al . , 2010 ) . As addictive drugs , including psychostimulants , stimulate corticosterone secretion ( Armario , 2010 ) , we next examined the role of GR signaling in psychostimulant-induced IP3R sensitization . Rats were treated with mifepristone ( 40 mg/kg , i . p . ) 30 min prior to injection of cocaine ( 10 mg/kg , i . p . ) or saline for 5 days . The effect of DHPG on burst IK ( Ca ) was significantly larger in cocaine-treated animals , which was not affected by mifepristone pretreatment ( Figure 6C ) . Thus , GR signaling is not involved in psychostimulant-induced IP3R sensitization . Next , the effect of social defeat stress was tested on acquisition of cocaine CPP , in which animals learn to associate a particular context with drug reward . Acquisition of psychostimulant CPP is inhibited by mGluR1 or NMDAR antagonist in the VTA , while CPP expression is attenuated by NMDAR antagonist , but not by mGluR1 antagonist , in the VTA ( Whitaker et al . , 2013 ) , supporting the potential role of NMDAR LTP in driving CPP . Rats underwent stress or control procedures for 5 days , then underwent 1-day CPP conditioning with cocaine ( 5 mg/kg , i . p . ) . It should be noted that a single psychostimulant treatment does not cause IP3R sensitization ( Ahn et al . , 2010; Whitaker et al . , 2013 ) . Stressed rats displayed robust preference for the cocaine-paired side after 1-day conditioning , while unhandled and handled controls showed small preference ( Figure 7A ) . The 1-day CPP score was significantly larger in stressed rats compared to unhandled and handled controls ( Figure 7B ) . Control rats developed significant cocaine CPP comparable to that observed in stressed rats after 3-day conditioning with the same dose of cocaine ( Figure 7—figure supplement 1 ) . These data suggest that repeated defeat experience promotes the rate of learning of cocaine-associated cues . 10 . 7554/eLife . 15448 . 010Figure 7 . Social defeat promotes cocaine-induced CPP via a GR-dependent mechanism . ( A ) Summary of changes in the preference for the cocaine-paired side following 1-day conditioning in unhandled , handled , and defeated rats ( unhandled: t7 = 2 . 51 , p<0 . 05; handled: t7 = 1 . 90 , p=0 . 10; defeat: t6 = 11 . 0 , p<0 . 001; paired t-test ) . ( B ) Summary of 1-day cocaine CPP scores in unhandled , handled , and defeated rats ( F2 , 20 = 25 . 2 , p<0 . 001 , one-way ANOVA ) . ***p<0 . 001 ( Bonferroni post hoc test ) . ( C ) Summary of changes in the preference for the cocaine-paired side following 1-day conditioning in rats pretreated with vehicle or mifepristone 30 min prior to social defeat or handling sessions ( vehicle + defeat: t8 = 5 . 30 , p<0 . 001; mifepristone + defeat: t8 = 1 . 90 , p=0 . 09; mifepristone + handled: t7 = 0 . 95 , p=0 . 37; paired t-test ) . ( D ) Summary of 1-day cocaine CPP scores in the 3 groups shown in panel C ( F2 , 23 = 4 . 90 , p<0 . 05 , one-way ANOVA ) . *p<0 . 05 ( Bonferroni post hoc test ) . None of the treatments in this figure affected the overall activity level during the pretest ( Figure 7—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 01010 . 7554/eLife . 15448 . 011Figure 7—figure supplement 1 . Unhandled control rats developed robust CPP after 3-day conditioning with cocaine ( 5 mg/kg ) . ( A ) Summary of changes in the preference for the cocaine-paired side following 3-day conditioning in unhandled control and defeated rats ( unhandled: t5 = 10 . 1 , p<0 . 001; defeat: t5 = 7 . 73 , p<0 . 001; paired t-test ) . ( B ) Summary of 1-day ( from the data shown in Figure 7B ) and 3-day CPP scores in unhandled control and defeated rats ( defeat: F1 , 23 = 29 . 9 , p<0 . 0001; conditioning period: F1 , 23 = 7 . 95 , p<0 . 01; defeat x conditioning period: F1 , 23 = 13 . 4 , p<0 . 01; two-way ANOVA ) . ***p<0 . 001 ( Bonferroni posthoc test ) . ( C ) Summary of the overall activity level , i . e . , total number of beam breaks in the CPP box compartment , during the three cocaine conditioning sessions for the 3-day conditioning experiments shown in ( A ) and ( B ) ( defeat: F1 , 20 = 3 . 45 , p=0 . 09; defeat x conditioning day: F2 , 20 = 0 . 43 , p=0 . 66; mixed two-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 01110 . 7554/eLife . 15448 . 012Figure 7—figure supplement 2 . Summary graphs depicting the overall activity level ( i . e . , total number of beam breaks in the CPP box compartments ) during the pretest for the experiments shown in Figure 7 . ( A ) Data from the groups in Figures 7A and B ( F2 , 20 = 1 . 76 , p=0 . 20 , one-way ANOVA ) . ( B ) Data from the groups in Figure 7C and D ( F2 , 23 = 0 . 92 , p=0 . 41 , one-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 01210 . 7554/eLife . 15448 . 013Figure 7—figure supplement 3 . Photograph of CPP box compartment with color-contrasting ceramic weight . The weight serves as a discrete cue to further differentiate the two conditioning contexts . DOI: http://dx . doi . org/10 . 7554/eLife . 15448 . 013 Finally , we asked whether GR signaling , which is necessary for IP3R sensitization , also plays a role in promoting cocaine CPP . As in the electrophysiology experiments , rats were treated with mifepristone ( 40 mg/kg , i . p . ) or vehicle 30 min before each social defeat session . An additional group received mifepristone followed by control handling procedure . We found that mifepristone suppressed cocaine CPP in stressed rats to a level comparable to that observed in mifepristone-treated controls ( Figure 7C and D ) . Therefore , GR activation during stress is required for CPP enhancement .
Repeated stressful experience leads to metaplasticity , i . e . , experience-dependent changes in the capacity of synapses to undergo activity-dependent plasticity ( Abraham , 2008 ) , in different brain areas ( Kim and Diamond , 2002; Joels et al . , 2006; Schwabe et al . , 2012; Chattarji et al . , 2015 ) . The present study demonstrates that repeated social defeat facilitates the induction of LTP of NMDAR-mediated transmission in VTA dopamine neurons while causing no alterations in global NMDAR-mediated excitation or intrinsic firing activity . Importantly , socially defeated animals display enhanced acquisition of cocaine CPP , a form of Pavlovian conditioning that requires NMDAR-dependent bursting in the VTA ( Zweifel et al . , 2009; Wang et al . , 2011; Whitaker et al . , 2013 ) . Repeated social defeat results in increased sensitivity of IP3Rs , which serve as a coincidence detector of presynaptic activity ( causing mGluR-dependent IP3 generation ) and postsynaptic bursting ( driving Ca2+ influx ) during NMDAR LTP induction ( Harnett et al . , 2009 ) . Inhibition of PKA completely reversed the increase in the potency of IP3 , indicating the role of PKA-dependent phosphorylation in stress-induced IP3R sensitization , as has been suggested in previous studies demonstrating similar changes following repeated drug exposure ( Ahn et al . , 2010; Bernier et al . , 2011 ) . It is of note that dopamine neurons in the substantia nigra pars compacta , in contrast to VTA neurons recorded in the present study , display significant PKA-dependent regulation of IP3-induced Ca2+ signaling and NMDAR LTP induction in control rats , which cannot be further enhanced by repeated drug exposure ( Harnett et al . , 2009; Ahn et al . , 2010 ) . It has been reported that repeated social defeat ( 10 days ) in mice leads to long-lasting ( >4 weeks ) alterations in gene expression and behavior ( e . g . , reduced social contact ) , with little recovery unless treated with antidepressants ( Berton et al . , 2006; Tsankova et al . , 2006 ) . In the present study , mGluR/IP3 action on burst-evoked Ca2+ signals remained elevated for 10–30 days following the 5-day defeat paradigm in rats , although displaying gradual decline during the 30-day stress-free period . It remains to be determined how this recovery is affected by different stress paradigms ( e . g . , duration or type/severity ) and treatments following stress experience . Mouse studies have also shown that individual animals display different susceptibility to repeated social defeat when assessed by the degree of social avoidance , which correlates with biochemical and physiological changes in the mesolimbic system ( Krishnan et al . , 2007; Cao et al . , 2010 ) . In particular , these studies observed hyperactivity of VTA dopamine neurons , assessed in vivo or ex vivo , associated with an increase in Ih , after 10-day defeat in susceptible mice . However , these parameters were not affected by 5-day social defeat in the current study conducted in rats . How could repeated stress lead to increased PKA activity regulating IP3Rs in the VTA ? Stress , including social defeat , promotes bursting in a subpopulation of VTA dopamine neurons , causing increased dopamine transients in the nucleus accumbens ( Anstrom et al . , 2009; Brischoux et al . , 2009 ) . Bursting also releases dopamine locally from the soma and dendrites , activating somatodendritic D2 autoreceptors ( Beckstead et al . , 2004 ) . Chronic stimulation of Gi-coupled receptors , such as D2 receptors , is known to upregulate the cAMP-PKA pathway ( Hyman et al . , 2006 ) , which may contribute to stress-induced IP3R sensitization . Interestingly , intra-VTA blockade of NMDARs during each social defeat episode , which would suppress dopamine neuron bursting , has been shown to prevent repeated stress-induced increases in cocaine self-administration ( Covington et al . , 2008 ) . GR signaling during defeat sessions is necessary for the enhancement of IP3R sensitivity . Stress-induced activation of the mesolimbic dopamine system is regulated by glucocorticoids ( Marinelli and Piazza , 2002 ) . Recent evidence implicates GRs expressed in projection areas , not in the VTA , in long-term glucocorticoid regulation of dopamine neuron activity ( Ambroggi et al . , 2009; Butts et al . , 2011; Barik et al . , 2013 ) . Furthermore , glucocorticoids can also enhance synthesis of corticotropin-releasing factor ( CRF ) , a major stress-related neuropeptide , and activation of CRF neurons in brain areas providing major CRF inputs to the VTA ( Makino et al . , 1994; Rodaros et al . , 2007; Kolber et al . , 2008 ) . GR blockade may therefore attenuate CRF-induced excitation of dopamine neurons during stress ( Kalivas et al . , 1987; Ungless et al . , 2003; Wanat et al . , 2008; Holly et al . , 2015 ) . We found that GR activation alone is not sufficient for IP3R sensitization . Thus , the potential GR mechanisms described above may act to amplify glutamatergic input-driven bursting activity during stress episodes , likely further enhanced by stress-induced activation of noradrenergic inputs stimulating dopamine neurons via α1 adrenergic receptors ( Grenhoff et al . , 1995; Paladini et al . , 2001; Morilak et al . , 2005 ) , thereby enabling large local dopamine release in the VTA . In this regard , it is interesting that repeated cocaine treatment was capable of causing similar enhancement of mGluR/IP3 action in a GR-independent manner . Dopamine levels in the VTA caused by cocaine alone are likely sufficient to induce D2-mediated upregulation of the cAMP-PKA pathway . Increased IP3R sensitivity drives the enhancement of NMDAR LTP induction in socially defeated animals . Our recent study demonstrated the involvement of L-type Ca2+ channels ( LTCCs ) in NMDAR LTP ( Degoulet et al . , 2015 ) . Although glucocorticoid-induced upregulation of LTCCs has been reported in the hippocampus and amygdala ( Karst et al . , 2002; Chameau et al . , 2007 ) , pharmacological activation of these channels does not enhance NMDAR LTP in dopamine neurons ( Degoulet et al . , 2015 ) ; thus changes in LTCCs are unlikely to play a role in LTP enhancement . CPP experiments showed that repeated social defeat promoted acquisition of the preference for contextual cues paired with cocaine experience , in accordance with previous studies demonstrating enhanced drug CPP following a period of repeated stress ( Kreibich et al . , 2009; Burke et al . , 2011 ) . Blockade of the critical components regulating NMDAR LTP induction ( i . e . , NMDARs , group I mGluRs , PKA , or LTCCs ) in the VTA during conditioning has been shown to suppress CPP acquisition ( Harnett et al . , 2009; Ahn et al . , 2010; Whitaker et al . , 2013; Degoulet et al . , 2015 ) . In the present study , systemic GR blockade during defeat episodes prevented both the enhancement of the LTP induction mechanism and that of cocaine CPP acquisition , consistent with the potential role of NMDAR plasticity in this form of Pavlovian learning . However , enhanced CPP acquisition observed in defeated rats may well be caused by an increase in the primary rewarding action of cocaine itself . The relative contribution of these two possibilities , i . e . , enhanced learning mechanism vs . enhanced cocaine reward , remains to be determined . It is well known that repeated stress impairs LTP of AMPAR-mediated transmission in the hippocampus ( Foy et al . , 1987; Shors et al . , 1989 ) , an effect that requires GR activation during stress ( Xu et al . , 1998 ) . LTP is similarly impaired in the prefrontal cortex ( Goldwater et al . , 2009 ) . By contrast , repeated stress leads to enhancement of AMPAR LTP in the lateral amygdala , which underlies Pavlovian fear conditioning driven by stressful/aversive stimuli ( Rodriguez Manzanares et al . , 2005; Suvrathan et al . , 2014 ) . Alterations in the function/expression of NMDARs are implicated in these forms of metaplasticity , as NMDARs play a key role in AMPAR LTP induction ( Kim and Diamond , 2002; Chattarji et al . , 2015 ) . Here , we described a distinct form of stress-induced metaplasticity in the VTA , i . e . , enhancement of mGluR/IP3-dependent NMDAR LTP , which may , at least in part , contribute to the enhanced drug reward-based Pavlovian learning . This may illuminate a key mechanism by which stressful experience increases vulnerability to addiction , a chronic relapsing disorder perpetuated by memories of drug-associated stimuli .
Sprague-Dawley rats ( Harlan Laboratories , Houston , Texas ) were housed in groups of 2–3 on a 12 hr light/dark cycle with food and water available ad libitum . All procedures were approved by the University of Texas Institutional Animal Care and Use Committee . Twelve week-old male resident rats were vasectomized and pair-housed with 6 week-old females . Residents ( used for ~8–10 months ) were screened for aggression ( biting or pinning within 1 min ) by introducing a male intruder to the home cage . Intruders and controls were young males ( 4–5 weeks old at the beginning ) housed in groups of 2–3 . For defeat sessions , residents and intruders were taken to a darkened procedure room at the end of the dark cycle . Intruders were introduced to residents’ home cages after removing females . Following 5 min of direct contact , a perforated Plexiglass barrier was inserted for 25 min to allow sensory contact . For repeated defeat , intruders underwent one session daily with a novel resident . Handled controls were taken to a darkened procedure room and placed in novel cages for 30 min . Unhandled controls remained undisturbed in the colony . Intruders and controls were housed separately . All drug and vehicle solutions were administered via i . p . injections ( 1 ml/kg ) . Mifepristone and corticosterone ( both from Tocris Bioscience , Ellisville , Missouri ) were dissolved in 30% propylene glycol plus 1% Tween-20 in 0 . 9% saline . Cocaine-HCl ( Sigma-Aldrich , St . Louis , Missouri ) was dissolved in 0 . 9% saline . Midbrain slices were prepared and recordings were made in the lateral VTA located 50–150 μm from the medial border of the medial terminal nucleus of the accessory optic tract , as in our previous studies ( Ahn et al . , 2010; Whitaker et al . , 2013; Degoulet et al . , 2015 ) . Tyrosine hydroxylase-positive neurons in this area ( i . e . , lateral part of the parabrachial pigmented nucleus ) largely project to the ventrolateral striatum ( Ikemoto , 2007 ) and show little VGluT2 coexpression ( Trudeau et al . , 2014 ) . Putative dopamine neurons in the lateral VTA were identified by spontaneous firing of broad APs ( >1 . 2 ms ) at 1–5 Hz in cell-attached configuration and large Ih currents ( >200 pA; evoked by a 1 . 5 s hyperpolarizing step of 50 mV ) in whole-cell configuration ( Ford et al . , 2006; Lammel et al . , 2008; Margolis et al . , 2008 ) . Cells were voltage-clamped at –62 mV ( corrected for –7 mV liquid junction potential ) . A 2 ms depolarizing pulse of 55 mV was used to elicit an unclamped AP . For bursts , 5 APs were evoked at 20 Hz . The time integral of the outward tail current , termed IK ( Ca ) ( calculated after removing the 20 ms window following each depolarizing pulse; expressed in pC ) , was used as a readout of AP-evoked Ca2+ transients , as it is eliminated by TTX and also by apamin , a blocker of Ca2+-activated SK channels ( Cui et al . , 2007 ) . Cells were loaded with caged IP3 ( 50–400 µM; generous gift from Dr . Kamran Khodakhah ) through the recording pipette . A UV flash ( ~1 ms ) was applied with a xenon arc lamp driven by a photolysis system ( Cairn Research , Faversham , UK ) . The UV flash was focused through a 60× objective onto a ~350 μm area surrounding the recorded neuron . Photolysis of caged compounds is proportional to the UV flash intensity; therefore , the concentration of IP3 was defined as the product of caged IP3 concentration in the pipette ( µM ) and flash intensity ( µJ ) measured at the focal plane of the objective ( expressed in µM·µJ ) . Synaptic stimuli were delivered with a bipolar tungsten electrode placed ~50–100 μm rostral to the recorded neuron . To isolate NMDAR EPSCs , recordings were performed in DNQX ( 10 µM ) , picrotoxin ( 100 µM ) , CGP54626 ( 50 nM ) , and sulpiride ( 100 nM ) to block AMPA/kainate , GABAA , GABAB , and D2 dopamine receptors , and in glycine ( 20 µM ) and low Mg2+ ( 0 . 1 mM ) to enhance NMDAR activation . NMDAR EPSCs were monitored every 20 s . The LTP induction protocol consisted of photolytic application of IP3 ( 250 µM·µJ ) 50 ms prior to the simultaneous delivery of synaptic stimulation ( 20 stimuli at 50 Hz ) and a burst ( 5 APs at 20 Hz ) , repeated 10 times every 20 s . LTP magnitude was determined by comparing the average EPSC amplitude 30 min post-induction with the average EPSC amplitude pre-induction ( each from a 5 min window ) . A CPP box ( Med Associates , St . Albans , Vermont ) consisting of two distinct compartments separated by a small middle chamber was used for conditioning . One compartment had a mesh floor with white walls , while the other had a grid floor with black walls . A discrete cue ( painted ceramic weight ) was placed in the rear corner of each compartment ( black one in the white wall side , white one in the black wall side; Figure 7—figure supplement 3 ) for further differentiation . One day after undergoing repeated stress or control procedures , rats were pretested for initial side preference by exploring the entire CPP box for 15 min . The percentage of time spent in each compartment was determined after excluding the time spent in the middle chamber . Rats with initial side preference >60% were excluded . Starting the next day , rats were subjected to 1-day or 3-day conditioning , in which they were given a saline injection in the morning and confined to one compartment , then in the afternoon given cocaine ( 5 mg/kg ) and confined to the other compartment ( 10 min each ) . Compartment assignment was counterbalanced such that animals had , on average , ~50% initial preference for the cocaine-paired side . A 15 min posttest was performed 1 day after the last conditioning session . The CPP score was determined by subtracting the preference for the cocaine-paired side during pretest from that during posttest . The experimenter performing CPP experiments was blind to animal treatments . Data are expressed as mean ± SEM . Statistical significance was determined by Student's t-test or ANOVA followed by Bonferroni post hoc test . Normality of data distribution was confirmed by Kolmogorov-Smirnov test . The difference was considered significant at p<0 . 05 . | Daily stress increases the likelihood that people who take drugs will become addicted . A very early step in the development of addiction is learning that certain people , places , or paraphernalia are associated with obtaining drugs . These ‘cues’ – drug dealers , bars , cigarette advertisements , etc . – become powerful motivators to seek out drugs and can trigger relapse in recovering addicts . It is thought that learning happens when synapses ( the connections between neurons in the brain ) that relay information about particular cues become stronger . However , it is not clear how stress promotes the learning of cue-drug associations . Stelly et al . investigated whether repeated episodes of stress make it easier to strengthen synapses on dopamine neurons , which are involved in processing rewards and addiction . For the experiments , rats were repeatedly exposed to a stressful situation – an encounter with an unfamiliar aggressive rat – every day for five days . Stelly et al . found that these stressed rats formed stronger associations between the drug cocaine and the place where they were given the drug ( the cue ) . Furthermore , a mechanism that strengthens synapses was more sensitive in the stressed rats than in unstressed rats . These changes persisted for 10-30 days after the stressful situation , suggesting that stress might begin a period of time during which the individual is more vulnerable to addiction . The experiments also show that a hormone called corticosterone – which is released during stressful experiences – is necessary for stress to trigger the changes in the synapses and behavior of the rats . However , corticosterone must work with other factors because giving this hormone to unstressed rats was not sufficient to trigger the changes seen in the stressed rats . Future experiments will investigate what these other stress factors are and how they work together with corticosterone . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Repeated social defeat stress enhances glutamatergic synaptic plasticity in the VTA and cocaine place conditioning |
DNA re-identification is used for a broad suite of applications , ranging from cell line authentication to forensics . However , current re-identification schemes suffer from high latency and limited access . Here , we describe a rapid , inexpensive , and portable strategy to robustly re-identify human DNA called 'MinION sketching' . MinION sketching requires as few as 3 min of sequencing and 60-300 random SNPs to re-identify a sample enabling near real-time applications of DNA re-identification . Our method capitalizes on the rapidly growing availability of genomic reference data for cell lines , tissues in biobanks , and individuals . This empowers the application of MinION sketching in research and clinical settings for periodic cell line and tissue authentication . Importantly , our method enables considerably faster and more robust cell line authentication relative to current practices and could help to minimize the amount of irreproducible research caused by mix-ups and contamination in human cell and tissue cultures .
In order to benchmark our re-identification method for real-life applications , we tested it in a variety of technical scenarios . To start , we constructed two large-scale proof-of-principle reference databases of genomic datasets that would stress the specificity of our method . The first reference database contains 31 , 000 genome-wide ~600K-900K SNP genotyping array files from individuals tested by Direct-to-Consumer ( DTC ) companies such as 23andMe , AncestryDNA , and FamilyTreeDNA ( Figure 2A ) ( Erlich , 2015 ) . The second reference database consists of genome-wide ~700K-800K SNP genotyping array files from 1 , 099 cancer cell lines in the Cancer Cell Line Encyclopedia ( CCLE ) ( Barretina et al . , 2012 ) and can be used for cell line authentication . Next , we created a MinION sketch for DNA samples in multiple technical scenarios ( Supplementary file 1A ) . These testing scenarios included extracting DNA from a spit kit or cell line culture , sequencing with either the R7 chemistry or the newer R9 chemistry , and re-identifying a sample with a genetically unstable background . The genetic reference files for each of these sketched samples were included in our reference databases . We found that the MinION sketching procedure re-identified human DNA with high accuracy after just minutes of operation . After 13 min of sketching using the R7 chemistry , the Bayesian algorithm re-identified the NA12890 sample ( a female CEU individual from the HapMap project ) with a posterior probability greater than 99 . 9% . Despite the high error rate of this relatively old chemistry and the low coverage , the algorithm needed only 195 bi-allelic variants to re-identify the sample ( Figure 3 and Supplementary file 1B ) . This is only ~2 times above the theoretical expectation for re-identifying a person by fingerprinting random markers ( Lin et al . , 2004 ) . To further test the robustness of our method , we re-sketched NA12890’s sequencing data against reference files for her first-degree relative ( NA12877 ) and second-degree relative ( NA12879 ) . Importantly , no exact-matching probability was observed , highlighting the specificity of our method ( Figure 3 ) . Using the 31 , 000-individual reference database ( consisting of genetic profiles from individuals genotyped by DTC companies ) , we repeated the R7 experiment with a sample of a mixed Ashkenazi-Uzbeki male ( YE001 ) . Again , we were able to re-identify this person within 13 min after assessing 110 SNPs ( Figure 2B and Supplementary file 1B ) , further showing that the method produces consistent results across ethnic origins . None of the other 31 , 000 individuals reached this level of matching probability ( Figure 2B ) . Finally , given that the number of reference samples in our database is in the thousands , but the number of people in the world in the billions , we wondered about the impact of the prior probability on identifying individuals . To this end , we tested various prior probabilities of identifying the YE001 sketch . We found that the initial selection of the prior probability had no effect on the matching ability and only slightly increased the time required to achieve a high-confidence match . Even with a prior probability that considers a database around a million times bigger than the world’s population ( 1015 ) , the posterior probability reached 99 . 9% with only 25 min of sketching YE001 ( Figure 2—figure supplement 1 ) , showing that our method returns robust results regardless of the chosen prior . Moving to the newer R9 chemistry provided even faster re-identification results . We sketched samples of a Northern European female ( SZ001 ) and a Northern European-Italian-Ashkenazi male ( JP001 ) using this chemistry . We were able to re-identify these two samples using only 98–134 SNPs , and the fastest identification required fewer than 5 min of MinION sketching ( Figure 2C and D and Supplementary file 1C ) . Again , none of the other 31 , 000 individuals in our database were matched to SZ001 or JP001 using this strategy . The rapid re-identification seems to be linked intimately to the increased speed with which DNA strands pass through the pores with the R9 chemistry versus the R7 chemistry ( 250bases/sec vs 70bases/sec ) . These results suggest that future developments in speeding up the DNA reading time could further reduce the re-identification time . Next , we explored the applicability of MinION sketching for cancer cell line authentication , a longstanding issue in the research community . We used MinION sketching and the R9 chemistry to authenticate THP1 , a monocytic leukemia strain , against the second reference database that consisted of cell lines from the CCLE . To show that more than one sample can be authenticated at the same time , we barcoded the THP1 sample and combined it with an additional , barcoded human sample . From the barcoded THP1 reads that were generated in ~3 min of sequencing , the sketching procedure leveraged 91 SNPs to authenticate the THP1 cell line with a posterior probability of 99 . 9% . None of the other 1098 CCLE reference files reached a probability of 99 . 9% or even exceeded a 10% match probability ( Figure 4A , Supplementary file 1D ) . Thus far , re-identification required an intersection of 91–195 SNPs from the MinION sketch and reference SNP file to reach a match probability of 99 . 9% . Having observed this range in the number of SNPs required , we wished to find the minimum number of intersected SNPs necessary to obtain a 99 . 9% match . This way we can optimize the sequencing time . To determine such a ‘stop sketching’ threshold , we simulated 10 , 000 different sketching runs for the THP1 cell line ( Figure 4B ) and SZ001 ( Figure 4—figure supplement 1A ) . The majority of simulated MinION sketches reached a match with a 99 . 9% probability using only 120–140 intersected SNPs . By 300 intersected SNPs , 99 . 6% of all sketches of the THP1 cell line were matched to its reference file with a probability of 99 . 9% , and for sketches of SZ001 by 240 SNPs that intersected with its reference . As expected , none of our simulation files failed to reach a correct match with a 99 . 9% probability with the correct reference file in the database . Although the number of mismatches per run was strongly correlated with the number of SNPs analyzed ( Figure 4—figure supplement 1B ) , the results from our sequencing runs and simulations suggest that even genetically unstable cancer cell lines can be identified with confidence using no more than 300 SNPs . The minimum sequencing run time necessary to infer a match depends on the yield of the specific run and the chemistry used . In summary , the MinION sketching method relies on the presence of the reference file in the database . If computing the posterior probability for 300 SNPs does not result in a 99 . 9% match , then the reference file for that cell line or individual is almost certainly not present in the reference database and further sketching is highly unlikely to yield any success . Next , we wondered how a severe contamination with cells of another origin would affect successful cell line authentication . Cell line cross-contamination is caused mostly by overgrowth from secondary cell lines with a substantially shorter generation time ( Capes-Davis et al . , 2010; Alston-Roberts et al . , 2010 ) . To start assessing the effects of contamination , we re-analyzed the data from the THP1 experiment but without resolving the barcodes , which essentially reflects a 50% contamination . The algorithm correctly showed a 0% match probability to the THP1 reference file or any other cell line in the database ( Figure 4C ) . We further explored the effect of the fraction of contamination on matching sketches with the THP1 reference file . By sampling from the above data in different proportions , we found that the algorithm correctly rejects a match for samples with contamination levels above 25% ( Figure 5 ) . While it may seem that the algorithm is not as sensitive to contamination as current STR-based methods , periodic testing of a cell culture with our method will reveal the contamination in a more timely fashion ( see Discussion ) . Lastly , we aimed to explore a sample preparation strategy that requires minimal hands-on time . To this end , we utilized a simple protocol to extract DNA using the rapid transposase-mediated fragmentation and adaptor ligation kit provided by ONT . This method generates 1D reads , where only one of the two strands passes through the nanopore , resulting in reads with a higher error rate ( Supplementary file 1E ) . The advantage of this method is the speed and convenience of the preparation protocol . In only 55 min , we were able to extract DNA and produce a ready-to-sequence library ( Figure 6A , example of execution: Figure 6—video 1 ) . The increased error rate resulted in the requirement for more SNPs to reach the re-identification threshold . In our experiment with the rapid sample preparation protocol we needed 239 SNPs to identify SZ001 with >99 . 9% probability ( Figure 6B ) . As such , re-identification of DNA and cell line authentication can still be completed with the same level of accuracy in one afternoon and using only minimal hands-on time by the researcher .
The main cause of cell line mix-ups is suggested to be human error ( Alston-Roberts et al . , 2010; Yu et al . , 2015; Almeida et al . , 2016 ) . It is therefore crucial to have means to monitor these errors rapidly and periodically . While the American Type Culture Collection ( ATCC ) offers an STR-based cell line authentication service , the overall procedure requires shipping consumables and samples back-and-forth and takes 2 weeks to complete . This works sufficiently in situations of cell line contamination that originate from mislabeling of a cell culture ( 100% contamination ) . Yet , a processing time of 2 weeks is suboptimal when caused by the mistaken transfer of cells from one culture to another , which can lead to cases of fitness competition between the cell lines ( Alston-Roberts et al . , 2010; Yu et al . , 2015 ) . It takes only 10 cells from a line with a doubling time that is 2–4 hr shorter than that of the original strain to overgrow an initial culture ( 106 cells ) within 2 weeks ( Figure 5—figure supplement 1 ) , which would currently be the time-point when STR typing results would be returned by the ATCC . Strikingly , once cells are in log-phase it can take as few as 2 days to change the contamination level of a culture from 1% to 80% . Our contamination simulations show that a contamination ≥25% , and often less than that , precludes a true matching result . Although the current STR-based methods can pick up on lower levels of contamination , in practice this does not make much difference considering: ( 1 ) the pace with which a contaminant can invade a cell culture , and ( 2 ) the relatively low identification speed of methods currently employed that precludes the timely return of data on the genomic composition of a cell population over multiple time points . Moreover , STR analysis is typically done using human-specific primers for amplification , and this therefore limits the identification of contaminants to ones of human origin ( Alston-Roberts et al . , 2010 ) . Our method , on the other hand , does have the potential ability to detect DNA from contaminants of non-human origin , such as infectious organisms like mycoplasma . Such contaminants could be identified efficiently when our pipeline is run in combination with metagenomic methods for real-time microbial detection ( as in Quick et al . , 2015 ) . The key to detect cell line contamination with human and non-human cells is periodic testing . The MinION can be part of standard lab equipment and facilitates rapid sample preparation and testing just prior to key experiments . We show that with our MinION sketching method , cell line authentication can be achieved in the lab in one afternoon , either using a hands-on or hands-off protocol . The first protocol involves a hands-on ~3 hr library preparation step ( including DNA extraction ) , but after only ~3 mins of sequencing we were able to identify the THP1 cell line out of 1 , 099 cancer cell lines with a posterior probability of 99 . 9% . The second protocol requires just 55 min for DNA extraction and transposase-mediated adapter ligation , after which sequencing can start . Our MinION sketching method reduces re-identification latency so that research does not have to be paused for long until the DNA profiling results return . Our method relies on randomly sampling SNPs from the genome , instead of a fixed set of small numbers of STRs or SNPs in a panel . This way we can omit a time consuming and biased PCR step in our method , and avoid the loss of statistical power caused by allelic dropout . This is particularly advantageous for cancer cell line authentication where genomic instability is prevalent . Cancerous cell lines commonly undergo loss of heterozygosity or exhibit aneuploidy , which affect STR-based re-identification of DNA samples through the loss of alleles ( Capes-Davis , 2013 ) . Furthermore , cancer cells that are deficient in their mismatch repair ( MMR ) pathways and suffer from microsatellite instability are identified more accurately by SNP-based than by STR-based identification methods ( Castro et al . , 2013; Otto et al . , 2017 ) . Because of these challenges , the current official ASN-0002 standards for STR analysis use an 80% matching threshold to positively match the STR profile in question to a reference file . Using this threshold , cell lines can be identified correctly in 98% of the cases ( Capes-Davis , 2013 ) . In our experiments , we see a clear correlation between the number of mismatches and the SNPs that are used to infer a match . The occurrence of mismatches in matching the DNA from a genetically unstable cancer cell line to its reference file results in a need to collect more SNP evidence for a match . Still , when we simulated matching the THP1 cell line to its reference file 10 , 000 times we found that intersecting a minimum of only 300 SNPs leads to a correct match in 99 . 6% of the cases . Furthermore , we found that based on this simulation , the SZ001 simulation and our experimental data , 300 SNPs could be used as a ‘stop-sketching’ threshold . Importantly , we never observed a false-positive match to any other reference file in the CCLE database . As a SNP-based method MinION sketching improves the precision of re-identifying cancer cell lines compared to the STR-based identification methods . ANSI-approved standards for SNP usage in cancer cell line authentication would be useful to promote the community-wide adoption of SNP-based sample re-identification ( also proposed by Yu et al . , 2015 and Otto et al . , 2017 , among others ) . The start-up cost for the MinION is currently $1000 , and multiplexing 12 DNA samples in one run makes the cost of consumables for sequencing a sample around $100 . This cost per sample is already lower than the ATCC STR-typing service , which is $195 , but higher on a per-run basis than the Geneprint system method . However , the latter method can only be used with access to the Applied Biosystems 3500 platform , and involves a more elaborate protocol that requires hands-on time , therefore incurring higher costs of labor . Although the balance between cost of labor and costs of machine depreciation and consumables poses a trade-off for all methodologies , the requirement of extensive hands-on laboratory work seems a main driver for avoiding authentication with current STR-based tests . Given that MinION sketching requires only minimal hands-on time and provides re-identification within hours instead of days/weeks , it is a very efficient and competitive re-identification method , especially when working with a small number of samples . While the costs of MinION sequencing continue to decrease , MinION sketching is currently not competitive in price for high-throughput testing until sequencing costs will have decreased further . In conclusion , to help solve the long-standing issue of ( cancer ) cell line contamination and to enhance the traceability of tissue samples in biobanks we developed an rapid re-identification method for DNA samples . Our method lowers the barrier for adoption of regular cell line authentication , which is important since only periodic testing will detect contamination and mix-ups efficiently and reduce the costs involved with irreproducible research . MinION sketching can easily be done in laboratories , in the clinic , or in biobanks as a routine sample authentication test .
The matching algorithm uses a Bayesian framework to evaluate the posterior probability of a match . Let xi∈{Y , N} be a random variable that either indicates whether the MinION sketch directly matches a known person ( xi=Y ) , or does not match ( xi=N ) with respect to the i-th individual in the database . Let Dk be the observed MinION data for the k-th bi-allelic marker , with Dk∈{A , B} , where A and B denote the two alleles; and let D= ( D1 , D2 , … , Dn ) denote the observation for n bi-allelic markers . The posterior probability of the matching outcome for the i-th sample is: ( 1 ) p ( xi|D ) =p ( xi ) ⋅p ( D|xi ) p ( D ) where p ( xi ) is the prior probability for the matching status of i-th sample and is specified by the user . The likelihood is approximated using the following equation: ( 2 ) p ( D|xi ) =∏kn{1 , . . , n}p ( Dk|xi ) The likelihood of an exact match given the data of the k-th marker , p ( Dk|xi=Y ) , is given by the following matrix: ( 3 ) M=AB[1−ϵϵ0 . 50 . 5ϵ1−ϵ]AAABBB where the rows denote the genotype of the i-th sample for the k-th marker as observed in the DNA database , the columns correspond to the observed genotype in the MinION data , and ∈ denotes the error rate assuming symmetry in confusing allele A for allele B and vice versa . p ( Dk|xi=Y ) corresponds to a specific row of M based on the observed genotype of a sample in the database . For example , if the genotype of the database sample is AA , then p ( Dk=A|xi=Y ) =1−ϵ and p ( Dk=B|xi=Y ) =ϵ . The likelihood of a mismatch given the data of the k-th marker , p ( Dk|xi=N ) , basically corresponds to observing the allele Dk in a random person from the population . This probability is the sum of two processes: ( i ) the random person has the same allele as Dk and the observation is errorless or ( ii ) the random person does not have the same allele as Dk but a sequencing error flipped the observed allele . Therefore: ( 4 ) p ( Dk|xi=N ) = ( 1−ϵ ) ⋅f ( Dk ) +ϵ⋅[1−f ( Dk ) ] where f ( Dk ) denotes the frequency of the observed allele in the population Finally , the evidence , p ( D ) is given by: ( 5 ) p ( D ) =∑xi∈{Y , N}p ( xi ) ⋅p ( D|xi ) We purchased the genomic DNA sample for the 1 , 000 Genomes individual NA12890 from the Coriell Institute . The THP1 cell line ( ECACC Cat# 88081201 , RRID:CVCL_0006 ) was used from laboratory resources and its authenticity was thoroughly verified in this study . YE001 and SZ001 were derived from the corresponding authors ( Y . E . and S . Z . ) using cheek swabs ( Catch-All Sample Collection Swab Epicentre QEC89100 ) or a saliva collection kit ( Supplementary file 1A ) . JP001 was sampled through saliva collection . Genomic DNA from NA12890 and YE001 ( Supplementary file 1A; exp . 1 , exp . 2 respectively ) were prepared for 2D MinION libraries using the SQK-MAP006 kit ( ONT ) as described by Zaaijer et al . ( 2016 ) . 2D libraries are double-stranded DNA fragments with a ligated hairpin loop and adaptors containing a tether and motor protein necessary for MinION sequencing , and were run on the R7 flow cells . DNA samples from SZ001 , JP001 and the THP1 cell line were prepared using the SQK-NSK007 kit from ONT ( Supplementary file 1A; exp . 3 , exp . 4 , exp . 5 ) and run on R9 flow cells . Samples ( Supplementary file 1A , exp . 6 ) were collected by cheek swab ( Catch-All Sample Collection Swab Epicentre QEC89100 ) scraping ~30 s on both sides of the cheek . Cells were recovered in 200 μl PBS . After addition of 20 μl Proteinase K and 200 μl lysis buffer ( DNeasy blood and tissue kit , Qiagen , #69504 ) , the sample was incubated at 56°C for 10 min . The sample was then applied to the column , spun for 1 min , and washed sequentially with buffers AW1 and AW2 . Next , 20 μl elution buffer was applied and the column was spun for 1 min on a regular benchtop centrifuge at maximum speed . Recovery of the DNA sample in 20 μl of sterile water resulted in an average yield of ~3–5 ng/μl . We used the SQK-RAD001 kit to prepare the DNA library . FRM ( 2 . 5 μl , ONT ) was added to the DNA sample ( 20 μl ) and incubated for 1 min at 30°C . Then , 1 μl RAD ( ONT ) plus 0 . 2 μl ligase was added and the mixture was incubated for 10 min . The R9 flow cell was prepared by applying 500 μl priming mix ( RBF 1x ) twice . The library was then added to the flow cell without a purification step . The barcoding protocol was executed according to manufacturer’s instructions for native barcoding kit I ( EXP-NBD002 , ONT ) in conjunction with the Nanopore Sequencing kit ( SQK-NSK007 , ONT ) with some modifications ( Supplementary file 1A , exp . 4 , exp . 5 ) . In brief; 1 . 5 μg DNA was used as starting material for each sample and vigorously vortexed for 1 min . The DNA sample was end-repaired and dA-tailed using the NEBNext Ultra II End Repair/dA-Tailing Module ( New England Biolabs [NEB] E7546S; 5 min 20°C , and 5 min 65°C ) . After an AMPure purification , the DNA fragments were subject to ligation using Blunt/TA Ligase Master Mix ( NEB M0367S ) for 5 min at 20°C and then 5 min at 65°C . The sample was then purified using AMPure magnetic beads and the DNA was eluted off the beads using 31 μl nuclease-free water . The NB01 and NB02 barcodes were ligated to the fragments of each sample with Blunt/TA Ligase mix ( NEB ) and incubated for 15 min . After an AMPure purification step , the two samples were pooled . Next , we ligated the adaptor ( BAM ) and hairpin ( BHP ) to the barcoded DNA fragments using NEB Quick Ligase ( NEB ) for 20 min at room temperature ( 22°C ) . The HTP ( ONT ) was added and incubated for another 10 min . The 50 μl MyOne C1 beads were prepared in the incubation step , which tethered the hairpin and ligated DNA fragments . The DNA library was eluted off the beads by ELB ( ONT ) at 37°C for 10 min and was applied to the flow cell . YE001 , JP001 ( https://dna . land/consent ) and three HapMap samples ( NA12890 , NA12977 , NA12879 ) are publicly available reference files . The 1 , 099 cancer cell line files were downloaded ( GSE36139 , CCLE ) , base-called using Birdseed and converted into 23andMe file format . The 31 , 000 DTC genomes were available from two sources: ( i ) 1 , 446 DTC genomes were downloaded from the public website OpenSNP . org and ( ii ) 29 , 554 genomes were collected using DNA . Land , an online website ( https://dna . land ) . The website procedures were approved by our IRB . Based on current consent , this set of 29 , 554 genomes cannot be shared . All experiments with this collection were done using an automatic algorithm on a secure server without access to the explicit identifiers of the samples ( e . g . names or contact information ) . To start a MinION run , we primed the flow cell according to the manufacturer’s protocol . We started MinKnow ( protocol ‘MAP_48 Hr_Sequencing_Run_SQK_MAP006’ for R7 and ‘NC_48 hr_Sequencing_Run_FLO-MIN104’ for R9 ) , uploaded the collected reads to Metrichor ( a cloud-based program that base-called the reads ) , and stored them on our computer . We used Poretools ( Loman and Quinlan , 2014 ) to extract the FASTQ data and time stamps from the local files . Only reads with an average base quality greater than nine were used for the downstream analysis . Next , we aligned the files to hg19 using bwa-mem ( v0 . 7 . 14 ) ( Li , 2013 ) using the command ‘bwa mem –V –x ont2d –t 4’ . Reads with multiple alignments were not considered for further analysis . To extract variants , we used a script to retain nucleotides from the MinION output that overlapped known positions of bi-allelic SNPs from dbSNP build-138 with an allele frequency between 1–99% . To minimize the effects of sequencing error , we considered only MinION read bases that matched the common SNP alleles in dbSNP . For example , if at position chr1:10 , 000 the MinION reported ‘A’ and dbSNP reported a variant ‘C/G’ , then we treated this position as a sequencing error . The R7 chemistry run with NA12890 generated 4 , 920 variants after 1 hr of MinION sequencing , of which 7 . 7% were rejected after filtering for common SNPs . Intersecting these with the reference file and analyzing the true error from the matched SNPs resulted in 8 . 9% mismatches . This contrasted with the R9 chemistry , which only resulted in 2% true mismatches ( Supplementary file 1C-E ) . The Bayesian model was integrated in a Python script in order to match between the MinION data and each entry in the database . To accelerate the search , we implemented the following procedure: ( i ) if the posterior probability drops below 10−9 , the script concludes that the database entry does not match and moves to the next entry , and ( ii ) the script uses only up to 1 hr of data to determine the posterior probability of a sample . As a default setting , we used a prior probability of 10−5 for exact matching . The only exception was Figure 2—figure supplement 1 ( YE001 ) , where we employed a range of prior probabilities . As a default setting , we used the computed error rate from each read as the εin our Bayesian algorithm . All code is publicly available on github at https://github . com/TeamErlich/personal-identification-pipeline ( Erlich , 2017 ) . A copy is archived at https://github . com/elifesciences-publications/personal-identification-pipeline . For the simulations , we took reads from exp . 4 and 5 ( Supplementary file 1A ) . The total number of reads was set to 3 , 000 and a random number of reads that represented the required proportion were selected . For example , for 50% contamination , we took 1 , 500 random reads from exp . 4 and 1 , 500 random reads from exp . 5 . These were pooled together and again shuffled to simulate a mixture . This process was repeated five times for each contamination fraction . The resulting pooled file was processed using our pipeline and matched to the reference file of the corresponding MinION sketch ( either THP1 , or JP001 ) . Availability of the data: The code for our method is available on https://github . com/TeamErlich/personal-identification-pipeline ( Erlich , 2017 ) . A copy is archived at https://github . com/elifesciences-publications/personal-identification-pipeline . Replicating the experiments can be done using the THP1 test example ( all data available on Github ) . The MinION reads for THP1 are also available in Zenodo at https://zenodo . org/record/1035914; 10 . 5281/zenodo . 1035913 ) . Building databases for your own MinION queries can be done by using opensnp . org , for cancer cell lines by downloading the CCLE reference files or using your own private SNP array files relevant to your query . The 29 , 554 genomes provided by DNA . land are not available for distribution to ensure genomic privacy of the individuals who donated their genomes to DNA . land ( see Materials and methods section: Reference databases and Table 1 ) . | The human genome represents the complete set of genetic information needed to make a person . DNA sequencing technologies used to study genomes have become much faster , cheaper and more accessible over recent years . This has enabled them to be used more regularly in various fields like precision medicine , in research laboratories and forensics . Even so , there are still fields where optimization is critical . Reproducibility is an important issue in biomedical research; one group of scientists working with human cells may report results that other scientists cannot reproduce . Sometimes this is because the original work was done in the wrong type of cells by mistake . Human cells used in biomedical research are very hard to discriminate from each other using microscopes; however , DNA analysis can be used to ensure the origin of the cells . The MinION device , a USB compatible handheld DNA sequencer , has become available in the last few years . Its size , speed and portability could enable many new uses for DNA sequencing . Technology like this could be used to confirm which cells the scientists are working with before they publish their results . Yet , currently DNA readings from the MinION are not accurate enough to be used to reliably confirm the identity of human cells used in research . Zaaijer et al . have now developed an approach that can accurately identify human cells using the MinION device . The approach involves “DNA re-identification” , which works by comparing an unknown DNA sample to a collection of known DNA profiles . Using their new method , Zaaijer et al . report that , with three minutes of DNA sequencing , they can correctly identify a DNA sample , with 99 . 9% confidence . This is a high enough level of accuracy for the system to tell the difference between one person and another , using only their DNA . This new technology is much faster than current rapid DNA sequencing approaches . Previously , processing DNA samples could take hours or even days and was not particularly portable . The new technology has many applications from finding criminals to diagnosing illnesses and tracking epidemics . It is also an affordable way for laboratories to confirm the identity of cells they are working with . This has the potential to save billions in research funding each year and speed up scientific progress . | [
"Abstract",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"tools",
"and",
"resources",
"genetics",
"and",
"genomics"
] | 2017 | Rapid re-identification of human samples using portable DNA sequencing |
Receptor-ligand pairs are ordinarily thought to interact through a lock and key mechanism , where a unique molecular conformation is formed upon binding . Contrary to this paradigm , cellulosomal cohesin-dockerin ( Coh-Doc ) pairs are believed to interact through redundant dual binding modes consisting of two distinct conformations . Here , we combined site-directed mutagenesis and single-molecule force spectroscopy ( SMFS ) to study the unbinding of Coh:Doc complexes under force . We designed Doc mutations to knock out each binding mode , and compared their single-molecule unfolding patterns as they were dissociated from Coh using an atomic force microscope ( AFM ) cantilever . Although average bulk measurements were unable to resolve the differences in Doc binding modes due to the similarity of the interactions , with a single-molecule method we were able to discriminate the two modes based on distinct differences in their mechanical properties . We conclude that under native conditions wild-type Doc from Clostridium thermocellum exocellulase Cel48S populates both binding modes with similar probabilities . Given the vast number of Doc domains with predicteddual binding modes across multiple bacterial species , our approach opens up newpossibilities for understanding assembly and catalytic properties of a broadrange of multi-enzyme complexes .
Cellulosomes are hierarchically branching protein networks developed by nature for efficient deconstruction of lignocellulosic biomass . These enzyme complexes incorporate catalytic domains , carbohydrate binding modules ( CBMs ) , cohesin:dockerin ( Coh:Doc ) pairs , and other conserved features ( Demain et al . , 2005; Bayer et al . , 2004; Schwarz , 2001; Béguin and Aubert , 1994; Smith and Bayer , 2013; Fontes and Gilbert , 2010 ) . A central attribute of cellulosome assembly is the conserved ~75 amino acid type-I Doc domain typically found at the C-terminus of cellulosomal catalytic domains . The highly conserved consensus Doc sequence from Clostridium thermocellum ( Ct ) is shown in Figure 1A . Dockerins guide attachment of enzymes into the networks by binding strongly to conserved Coh domains organized within non-catalytic poly ( Coh ) scaffolds . In addition to their nanomolar binding affinities , many archetypal Coh:Doc pairs are thought to exhibit dual binding modes ( Carvalho et al . , 2007; Pinheiro et al . , 2008; Currie et al . , 2012 ) . The bound Doc domain can adopt two possible orientations that differ by ~180° rotation on the Coh surface , as shown in Figure 1B . The two binding modes originate from duplicated F-hand sequence motifs , a conserved structural feature found among type-I dockerins ( Pagès et al . , 1997 ) . The duplicated F-hand motifs resemble EF-hands found in eukaryotic calcium binding proteins ( e . g . , calmodulin ) , and provide internal sequence and structural symmetry to Doc domains . Rotating Doc by ~180° with respect to Coh ( Figure 1B , C ) results in an alternatively bound complex with similarly high affinity involving the same residues on Coh recognizing mirrored residues within Doc . The dual binding mode is thought to increase the conformational space available to densely packed enzymes on protein scaffolds , and to facilitate substrate recognition by catalytic domains within cellulosomal networks ( Bayer et al . , 2004 ) . From an evolutionary perspective , the dual binding mode confers robustness against loss-of-function mutations , while allowing mutations within Doc to explore inter-bacterial species cohesin-binding promiscuity in cellulosome-producing microbial communities . Coh:Doc interactions and dual binding modes are therefore important in the context of cellulose degradation by cellulosome-producing anaerobic bacterial communities . 10 . 7554/eLife . 10319 . 003Figure 1 . Cohesin:Dockerin dual binding modes . ( A ) Secondary structure and consensus sequence logo ( Crooks , 2004 ) assembled from 65 putative Ct type-I Doc variants . Dots above the amino acid codes indicate residues involved in: Ca2+ coordination ( yellow ) , mode A binding ( black ) , and mode B binding ( gray ) . Letter colors represent chemical properties: Green , polar; purple , neutral; blue , basic; red , acidic; black , hydrophobic . Crucial Coh-binding residues are located at positions 11 , 12 , 18 , 19 , 22 , and 23 in each F-hand motif . ( B ) Coh:Doc complex crystal structures showing overlaid Doc domains in the two binding modes . Images were generated by aligning the Coh domain ( gray ) from PDB 2CCL ( green , binding mode ( A ) and 1OHZ ( red , binding mode ( B ) using the VMD plugin MultiSeq ( Humphrey et al . , 1996; Roberts et al . , 2006 ) . ( C ) View of the Doc binding interface for each mode from the perspective of Coh . The conserved binding residues at positions 11 , 12 , 18 , and 19 in the F-hand motif relevant for binding in the corresponding mode are depicted as stick models ( yellow ) . ( D ) Close-up view of the interface for each binding mode with arrows indicating the location and direction of applied force . Binding residues 11 , 12 , 18 , and 19 for binding mode A and 45 , 46 , 52 , and 53 for binding mode B are shown as blue stick models . The Coh domain is oriented the exact same way in both views . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 003 However , direct experimental observation of the dual binding modes for wild-type Doc has thus far proven challenging . Ensemble average bulk biochemical assays ( e . g . , surface plasmon resonance , calorimetry , enzyme-linked immunosorbent assays ) are of limited use in resolving binding mode populations , particularly when the binding modes are of equal thermodynamic affinity . Crystallography is challenging because the complex does not adopt a unique molecular conformation , but rather exhibits a mixture of two conformations thereby hindering crystal growth . Structural data on the dual binding mode have typically been collected using a mutagenesis approach , where one of the binding modes was destabilized by mutating key recognition elements ( Carvalho et al . , 2007; Pinheiro et al . , 2008 ) . This approach , however , while resolving the structures of each bound complex , cannot determine if one binding mode is dominant for wild-type Doc , or if that dominance is species or sequence dependent . Coarse grained molecular dynamics has also predicted dual modes of interaction between Coh and Doc ( Hall and Sansom , 2009 ) , but direct experimental evidence of both binding modes for wild-type Doc has remained elusive . Improved fundamental understanding of the dual binding mode could shed light onto the molecular mechanisms by which these multi-enzyme complexes self-assemble and achieve synergistic conformations , as well as provide a new approach to designing systems for protein nanoassembly ( Kufer et al . , 2009; 2008 ) . Here , we used SMFS ( Li and Cao , 2010; Engel and Müller , 2000; Woodside and Block , 2014 ) to study wild-type and mutant Doc from exocellulase Cel48S of C . thermocellum ( Ct-DocS ) . We demonstrate that specific unfolding/unbinding trajectories of individually bound Coh:Doc complexes are characteristic of the binding modes . To validate our approach , we produced Doc mutants that exhibited a preferred binding mode . We performed single-molecule pulling experiments on bound Coh:mutant Doc complexes and observed a strong bias in the probability of two clearly distinguishable unfolding patterns , termed ‘single’ and ‘double’ rupture types for each binding mode mutant . We further probed the unbinding mechanism of the double rupture events using poly ( Gly-Ser ) inserts to add amino acid sequence length to specific sections of Doc as a means to identify which portions of Doc unfolded . Finally , we used the inherent differences in mechanical stability of each binding mode , and the effects these differences had on the unfolding force distributions of an adjacent domain , to directly observe and quantify binding mode populations for wild-type Doc .
The wild-type and mutant Doc sequences used in this work were aligned ( Beitz , 2000 ) and are presented in Figure 2 . Among Ct-Doc domains , a Ser-Thr pair located at positions 11 and 12 of F-hand motif 1 ( N-terminal helix 1 ) is highly conserved ( Figure 1A ) . This Ser-Thr pair is H-bonded to Coh in binding mode A ( Figure 1A , black dots ) . Analogously , binding mode B refers to the configuration where the Ser-Thr pair from helix 3 dominates the H-bonding to Coh ( Figure 1A , gray dots ) . Binding mode B was previously crystallized for a homologous Ct-Doc ( Carvalho et al . , 2003 ) . Mutation of the Ser-Thr pair in helix 3 to Ala-Ala was used to bias binding and thereby crystallize binding mode A for the same Doc ( Carvalho et al . , 2007 ) . A similar targeted mutagenesis approach was also used to obtain crystal structures of a Clostridium cellulolyticum Doc in each binding mode ( Pinheiro et al . , 2008 ) . 10 . 7554/eLife . 10319 . 004Figure 2 . Doc sequences used in this study ( N- to C-terminus ) . Doc_wt: wild-type sequence; hydrophobicity and charge graphs are displayed for the wild-type-Doc ( red: positively charged , blue: negatively charged ) ; ( GS ) x8_insert: A ( Gly-Ser ) 8 linker was incorporated between helix 1 and helix 2; Q1_mutant: Quadruple mutant in helix 1 . Four point mutations ( DE/AA ) were incorporated into Doc helix 1 to knock out binding mode A; Q3_mutant: Quadruple mutant in helix 3 . Four point mutations ( DE/AA ) were incorporated into Doc helix 3 to knock out binding mode B; QQ_mutant: Non-binding control with both binding modes knocked out . Numbers below indicate amino acid number of the fusion protein construct starting from the xylanase N-terminus . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 004 To preferentially select for a specific binding mode ( A or B ) , we prepared Doc sequences that incorporated 4 amino acid point mutations , referred to as quadruple mutants ( ‘Q’ ) . To design quadruple mutants , we noted that recent structural work reported a set of Ct-Doc domains that differ from the canonical duplicated Ser-Thr sequences . These non-canonical Docs were found to exhibit only a single binding mode ( Brás et al . , 2012; Pinheiro et al . , 2009 ) . In one of these non-canonical Doc domains , an Asp-Glu pair was found in place of Ser-Thr . Since the Coh surface is negatively charged , we postulated that including Asp-Glu in place of Ser-Thr within one of the F-hands could be used to effectively knock out a given binding mode for our canonical Doc . Additionally , we incorporated double alanine mutations to replace the conserved Lys-18 Arg-19 residues of a given F-hand motif , further destabilizing a targeted binding mode . Q1 refers to a quadruple mutant where helix 1 has been modified at four positions ( i . e . S11D-T12E-K18A-R19A ) . Q3 refers to the quadruple mutant where helix 3 has been modified at four positions ( i . e . S43D-T44E-K50A-R51A ) . As a negative control , we prepared a mutant referred to as ‘QQ’ that incorporated quadruple mutations into both helices 1 and 3 . Doc domains were expressed as fusion domains attached to the C-terminal end of xylanaseT6 ( Xyn ) from Geobacillus stearothermophilus to improve solubility and expression levels as previously reported ( Stahl et al . , 2012 ) . The Xyn domain also acts as a so-called fingerprint in AFM force extension traces to provide a means for screening datasets and searching for known contour length increments . We use the term ‘contour length’ to refer to the maximum length of a stretched ( unfolded ) polypeptide chain . Our screening process identified single-molecule interactions and ensured correct pulling geometry . For the Coh domain , we chose cohesin 2 from Ct-CipA expressed as a C-terminal fusion domain with the family 3a carbohydrate binding module ( CBM ) from Ct-CipA . In order to exclude artifacts arising from fingerprint domains , protein immobilization or pulling geometry , a second set of fusion proteins was cloned , expressed and probed in complementary experiments using a flavoprotein domain from the plant blue light receptor phototropin ( iLOV ) ( Chapman et al . , 2008 ) . All protein sequences are provided in the ‘Materials and methods’ section . The pulling configuration for single-molecule AFM experiments is shown in Figure 3A . CBM-Coh was site-specifically and covalently attached to an AFM cantilever tip and brought into contact with a glass surface modified with Xyn-Doc . The mechanical strength of protein domains and complexes will strongly depend on the pulling points ( i . e . sites at which the molecule is attached to cantilever/surface ) . The site-specific attachment chemistry used here was precisely defined by the chosen residue of immobilization , ensuring the same loading geometry was used on the complex for each and every data trace . After formation of the Coh:Doc complex , the cantilever was retracted at a constant speed that ranged from 200 to 3200 nm/s while the force was monitored by optical cantilever deflection . The resulting force-distance traces were characteristic of the series of energy barriers crossed by the protein complex along the unfolding/unbinding pathway . A sawtooth pattern was consistently observed when molecular ligand-receptor complexes had formed . Sorting the data using contour length transformation ( Puchner et al . , 2008 ) and identifying traces that contained a Xyn contour length increment ( ~89 nm ) allowed us to screen for single-molecule interactions ( Stahl et al . , 2012 ) , as described in our prior work on Coh:Doc dissociation under force ( Stahl et al . , 2012; Schoeler et al . , 2014; Jobst et al . , 2013; Otten et al . , 2014; Schoeler et al . , 2015 ) . 10 . 7554/eLife . 10319 . 005Figure 3 . Overview of the experimental configuration and recorded single-molecule unfolding and unbinding traces . ( A ) Schematic depiction showing the pulling geometry with CBM-Coh on the AFM Cantilever and Xyn-Doc on the glass substrate . Each fusion protein is site-specifically and covalently immobilized on a PEG-coated surface . ( B-C ) Each force vs . extension trace shows PEG linker stretching ( black ) , xylanase unfolding and subsequent stretching ( blue ) , and Coh:Doc complex rupture . The Coh:Doc complex rupture occurred in two distinct event types: single ( B ) and double ( C ) ruptures . The 8-nm contour length increment separating the double peaks was assigned to Doc unfolding ( C , green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 00510 . 7554/eLife . 10319 . 006Figure 3—figure supplement 1 . Representative sample of force traces . Traces showing xylanase unfolding and ( A ) single-event rupture , ( B ) double-event rupture , ( C ) CBM unfolding prior to unbinding , ( D ) shielded double-event for Q3 mutants , ( E ) nonspecific multiple interactions , and ( F ) no interactions . Traces such as E and F were excluded from the analysis . Multiple traces are offset on the force axis for readability . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 006 Typical single-molecule interaction traces from such an experiment are shown in Figure 3B , C and in Figure 3—figure supplement 1 . Following PEG linker stretching , an initial set of peaks sequentially decreasing in force was assigned to xylanase unfolding and stretching . This domain when unfolded added ~89 nm of free contour length to the system . The final peak ( s ) corresponded to rupture of the Coh:Doc complex , and occurred as either ‘single’ or ‘double’ rupture events . The contour length increment between the two double event peaks was found to be ~8 nm , that is , 8 nm of hidden contour length was added to the biopolymer during a sub-step of Doc unbinding ( see ‘Discussion’ ) . The 8-nm contour length increment was also observed in complementary experiments employing other fusion domains: xylanase was swapped for an sfGFP domain and CBM was swapped out for an iLOV domain . In these new fusions , the 8 nm Doc increment was still observed , indicating it was not caused by a specific fusion domain . As we show below , double and single rupture events were associated with binding modes A and B , respectively . CBM unfolding length increments ( ~57 nm ) were only rarely observed because the Coh:Doc complex only rarely withstood forces sufficiently high to unfold CBM ( Stahl et al . , 2012 ) . Binding experiments were carried out in bulk to evaluate the binding affinity of wild-type , Q1 , Q3 , and QQ Doc sequences to wild-type Coh . Xyn-Doc fusion protein variants were immobilized in a microwell plate and exposed to tag red fluorescent protein ( TagRFP ) ( Merzlyak et al . , 2007 ) fused to Coh ( TagRFP-Coh ) across a range of concentrations , followed by rinsing and subsequent fluorescence readout ( Figure 4A ) . The data clearly showed that Q1 and Q3 Doc sequences , each with a mutated binding mode , maintained high-binding affinity with dissociation constants ( Kd ) in the nM range . These values are in good agreement with previous reports on homologous type-I Doc domains ( Brás et al . , 2012; Sakka et al . , 2011 ) . This suggested that mutant Doc domains with one destabilized binding mode were still able to recognize fluorescent protein fused Coh with strong affinity by relying on the alternative binding mode that was preserved . The QQ double knockout mutant , however , showed no appreciable binding over the concentration range tested . This negative control showed that DEAA quadruple mutations were in fact effective at eliminating binding for the targeted modes . 10 . 7554/eLife . 10319 . 007Figure 4 . Bulk and single-molecule characterization of Doc mutants . ( A ) Fluorescence binding curve showing binding of TagRFP-labelled Coh to wild-type and mutant Doc nonspecifically immobilized in a 96-well plate . Both Q1 and Q3 mutants bound TagRFP-Coh similarly to wild-type with dissociation constants ( KD ) in the low nM range . The negative control QQ mutant showed no binding . Solid lines are 4 parameter logistic nonlinear regression model fits to the data . Error bars represent the standard deviation of three independent samples . ( B ) Event probabilities for single ( opaque colors ) and double ( translucent colors ) Coh:Doc rupture peaks determined for Doc wild-type and DE/AA quadruple mutants . Data originate from 947 , 4959 , and 1998 force-extension traces from wild-type , Q1 and Q3 variants , respectively . Error bars represent 95% Clopper-Pearson confidence intervals based on the beta probability distribution . ( C ) Relative difference in double peak rupture forces for the different Doc variants . Positive values indicate a stronger final peak . Histograms represent concatenated data from various pulling speeds . Drawn lines are kernel density estimates calculated on the raw data . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 00710 . 7554/eLife . 10319 . 008Figure 4—source data 1 . Probability Data . Single/double peak counting statistics displayed in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 008 For each Doc tested , we collected tens of thousands of force-extension traces and selected for further analysis only those traces showing the ~89 nm xylanase contour length increments and no other anomalous behavior , resulting in typically 200–3000 usable single-molecule interaction curves per experiment . We determined the number of Coh:Doc unbinding events that occurred as single or double rupture peaks . The results are shown in Figure 4B . The wild-type Doc showed double rupture events in ~57% of the cases , and single rupture events in ~43% of the cases . The mutant designed to knock out binding mode A ( Q1 ) , showed a single event probability of ~77% , and a double event probability of ~23% . The mutant designed to knock out binding mode B ( Q3 ) showed a single event probability of ~41% , and a double event probability of ~59% . It is clear from these data that the Q1 mutant has a strong bias toward single peaks that is not observed in the wild-type leading to preliminary assignment of single peaks to binding mode B . For all double events , we determined the force difference of the second peak relative to the first ( Figure 4C ) . Q1 and wild-type on average showed second peaks that were ~15–20% higher in force than the first peak . Q3 meanwhile showed clearly different behavior . Although the ratios of single to double peaks were nearly identical between wild-type and Q3 , differences in the relative force between the first and second peaks differentiated wild-type and Q3 ( Figure 4C ) . Double peaks for the Q3 mutant were more likely to show a shielded behavior , where the second peak was lower in force than the first peak by ~10% . Although the Q3 mutant showed the same single vs . double event probability as wild-type , the double events for Q3 were distinguishable from those of the wild-type based on this observed decrease in the rupture force of the second peak . The second barrier of the double events was therefore weaker in Q3 than for wild-type . This weaker 2nd double peak for the Q3 mutant combined with similar single/double peak ratios as wild-type leads us to believe that the number of double peaks is being underestimated systematically for the Q3 mutant . Generally , each binding mode still allows for the occurrence of a single event ( albeit with different likelihood ) , in which the whole Doc domain unbinds without an additional unfolding substep . Since the second and final energy barrier for complex dissociation is weaker than the first for the Q3 mutant , the probability for the molecule to pass both barriers simultaneously is increased , thus resulting in a higher percentage of single events . We sought to identify the molecular origin of the 8 nm contour length increment separating the double event peaks by engineering additional amino acid sequence length into the Doc domain . Amino acid insert sequences have previously been used to probe length increments in AFM force spectroscopy experiments ( Bertz and Rief , 2009 ) ( Carrion-Vazquez et al . , 1999 ) . By adding additional amino acids to the polypeptide chain at a particular location , insert sequences increase the gain in contour length following unfolding of a subdomain in a predictable way . Any change in the observed length increment can be pinpointed to the position in the molecule where the unfolding event occurs . In this case , we engineered flexible ( GS ) 8 insert sequences directly into wild-type Doc between helices 1 and 2 , in a flexible loop that was not expected to interfere with either of the two binding modes . Structural homology models ( Figure 5A ) of the wild-type Doc and ( GS ) 8 insert sequence were calculated using the Phyre server ( Kelley and Sternberg , 2009 ) . If the 8-nm contour length increment was caused by sequential unbinding of Doc helices 1 and 3 in wild-type Doc , then double peaks for the poly ( GS ) constructs should show an increase in the double peak contour length increment . As shown in Figure 5B , C and D , the contour length histogram for ( GS ) 8 Doc was indistinguishable from the wild-type Doc . No additional contour length was gained due to additional amino acids inserted between Doc helices 1 and 2 . Since the Doc was anchored to the glass slide through an N-terminal xylanase domain , this result indicated that the unfolding event responsible for the 8-nm length increment must be located upstream ( i . e . N-terminal ) from the site of the ( GS ) 8-insert . This result suggested that unfolding of calcium binding loop 1 and helix 1 in Doc was the source of the 8-nm length increment . 10 . 7554/eLife . 10319 . 009Figure 5 . Probing the final contour length increment with Poly ( GS ) inserts . ( A ) Structural homology model overlay of wild-type and mutant Doc containing a ( GS ) 8-linker between helix 1 and helix 3 . The wild-type Doc is shown in green . The 16 amino acid long GS-insert is shown in purple ( Kelley and Sternberg , 2009 ) ( remaining Doc domain not shown ) . ( B ) Typical force extension trace with final double rupture event depicted in green ( arrow ) . ( C ) Histogram and kernel density estimate of the transformation of the single force extension trace in panel B into contour length space ( black ) and kernel density estimate of the whole dataset of single molecule Xyn-Doc:Coh-CBM traces bearing xylanase fingerprint and final double rupture ( gray , offset in y-direction for readability ) in contour length space . ( D ) Histograms ( bars , bin width: 1 nm ) , kernel density estimates ( drawn lines , bandwidth: 0 . 75 nm , gaussian kernel ) , and statistical test ( Kolmogorov-Smirnov , ‘KS test’ ) are each calculated on the raw data of the final increments ( peak-to-peak distances ) in contour length space ( x-distance between arrow 1 and 2 in panel ( C ) . Maxima for final double event increments lie at 7 . 75 nm and 7 . 73 nm for iLOV-Coh:Doc ( wild-type ) -sfGFP ( N = 255 ) and Xyn-Doc ( GS ) 8:Coh-CBM ( N = 320 ) final ruptures , respectively ( a two-sample KS test on the raw data indicates no significant difference in the data distributions ( p-value of 21 . 7% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 009 To finally confirm the presence of both bound conformations in wild-type Coh:Doc complexes , we replaced xylanase with sfGFPand CBM with iLOV as the contour length marker or fingerprint domains . iLOV was chosen as a superior unfolding fingerprint domain because it does not show multiple unfolding substeps ( in contrast to xylanase ) , which simplified analysis . Also iLOV has an unfolding force distribution that lies in a similar range as the Coh:Doc complex dissociation single and double peaks , allowing for effective biasing of the iLOV unfolding force distributions by the inherent stability difference between single and double event peaks . Figure 6A shows characteristic single and double event curves containing iLOV unfolding ( 36-nm contour length increment ) followed by Coh:Doc rupture as a single or double event . The rupture force distributions of the single and double event ( second peak ) ruptures are shown in Figure 6B . The most probable rupture force for single events was ~104 pN , while for double events this value was ~140 pN at a pulling speed of 800 nm/s . We next calculated the unfolding force distributions of the iLOV domain for curves that terminated with single events or double events . If the Coh:Doc complex ruptured before iLOV unfolding was observed , the curve was eliminated from the dataset because it lacked a fingerprint domain length increment . This criterion for inclusion in the dataset results in a biasing of the iLOV unfolding forces , since the maximum of the fingerprint unfolding force distribution that can be observed must lie below that of the Coh:Doc complex . The fact that we observed a downward shift in the iLOV unfolding forces ( Figure 6C ) for curves that terminated in the less mechanically stable single rupture event is confirmation that the single- and double-event peaks arise from separate bound conformations . Each mode has a distinct mechanical stability and energy landscape that is set at the time of receptor-ligand binding , that is once bound , the conformation of the complex does not change . If single- and double-event unbinding patterns were simply two competing pathways out of the same bound state , then the downward shift in rupture force distribution would not be observed for the iLOV unfolding forces . Although this shift in rupture force distributions is comparatively subtle , it can be observed accurately with high statistical significance . We note that the datasets for both binding modes were measured with the same cantilever throughout the runtime of the whole experiment . Calibration and drift issues therefore did not interfere with the required accuracy . 10 . 7554/eLife . 10319 . 010Figure 6 . Biasing of unfolding force distributions by dual binding mode . ( A ) Typical force traces showing iLOV unfolding with final single ( green ) and double ( purple ) complex ruptures . The curve terminating in a double peak is offset in the y-direction for clarity . ( B ) Final complex rupture force distribution for single and double events . Double events are more mechanically stable . ( C ) iLOV domain unfolding forces for final single ( green ) and double ( red ) events at a pulling velocity of 800 nm/s . Histograms ( bars ) , kernel density estimates ( lines ) , and statistical tests are each obtained from the raw data . Maxima for iLOV unfolding lie at 96 . 0 pN and 102 . 7 pN for single ( N = 172 ) and double ( N = 277 ) final ruptures , respectively . A two-sample Kolmogorov-Smirnov test showed significant differences in the data distributions ( p-value of 0 . 09% ) . Since the data were all recorded with a single cantilever and both event types were distributed equally throughout the runtime of the measurement , no systematic biasing is expected . Because of the lower force distribution of final single peaks , the iLOV unfolding force distribution is truncated compared to final double peak force traces , supporting the notion that the binding mode is set prior to mechanical loading of the complex . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 010
The relatively small ~8 kDa Doc domains exhibit an internal sequence and structural symmetry that is believed to give rise to a dual mode of binding to Coh , as shown in Figure 1 . In order to study this remarkable plasticity in molecular recognition in greater detail , we prepared a series of mutants ( Figure 2 ) designed to either knock out a specific binding mode or add length to the molecule at a specific position . Bulk experiments showed that Doc mutants Q1 and Q3 , originally designed to suppress one of the binding modes , were still able to bind Coh with high affinity , while the double knockout did not bind ( Figure 4A ) . The equilibrium affinities of Coh binding to Q1 , Q3 , or wild-type were all similarly high with KDs in the low nM range , in good agreement with literature values ( Sakka et al . , 2011 ) , suggesting the two binding modes are thermodynamically equivalent and rendering them indistinguishable with conventional methods such as ELISA or calorimetry . Techniques like surface plasmon resonance could possibly show differing values for on- and off-rates for the mutants , but would still not be able to resolve the binding modes within a wild-type population . Force spectroscopy with the AFM interrogates individual molecules , and measures their mechanical response to applied force . Since the technique is able to probe individual members of an ensemble , it provided a means to quantify binding mode configurations by assigning unfolding/unbinding patterns to the binding mode adopted by the individual complexes . Site-directed Q1 and Q3 mutations supported the assignment of binding mode A to a characteristic double rupture peak dissociation pathway . Single events were assigned to binding mode B and showed no Doc unfolding sub-step prior to complex rupture . We consistently observed 8 nm of added contour length that separated the Doc double peaks . Since force is applied to Doc from the N-terminus , we analyzed the Doc sequence starting at the N-terminus and searched for reasonable portions of Doc that could unfold in a coordinated fashion to provide 8 nm of contour length . The results from the GS-insert experiments ( Figure 5 ) indicated no change in the double-event contour length increment , regardless of the added GS-insert length located between helix 1 and 3 in Doc . This result is consistent with the 8 nm length increment being located N-terminally from the GS-insert site , implicating unfolding of Doc calcium binding loop 1 and helix 1 as the source of the 8 nm . This length accurately matches the estimated length increment for unfolding calculated from the crystal structure ( Figure 1D ) . Although this result could also be consistent with the 8 nm increment being located somewhere outside the Doc domain in the polyprotein , we deem this scenario highly unlikely . The 8 nm increment cannot be located in the Xyn or CBM domains because we have accounted for Xyn and CBM lengths in their entirety based on the observed 89 nm and rare 57 nm length increments here and in a previous study ( Stahl et al . , 2012 ) , and for confirmation swapped out those domains for different proteins completely ( i . e . iLOV and GFP ) . The remaining possibility that the 8 nm is located within the Coh domain is also not likely since the barrel-like structure of the Coh is known to be mechanically highly stable ( Valbuena et al . , 2009; Hoffmann et al . , 2013 ) . Also , if the 8-nm length increment were due to partial Coh unfolding , the Q1 and Q3 mutants would not be expected to affect the single/double peak ratio or force differences between the double event peaks as was observed ( Figure 4B , C ) . The GS-insert data suggest the 8-nm length increment is located within Doc , upstream ( N-terminal ) from the GS-insert site implicating calcium loop 1 and helix 1 in this unfolding event . Finally , we observed that an inherent difference in the mechanical stability of single and double event rupture peaks ( Figure 6B ) could be used as a feature by which to discriminate the binding modes . Our analysis algorithm accepted only the force curves that first showed iLOV fingerprint domain unfolding followed by either a single- or double-rupture peak . By observing a small but significant downward shift in the iLOV unfolding force distribution when analyzing curves that terminated in the less stable single-event peak , we confirmed the single-event peaks originate from a unique conformation that is ‘set’ at the time of complex formation . Taken together , we propose an unbinding mechanism where the first barrier of the double peaks represents unfolding of the N-terminal calcium binding loop and unraveling of alpha helix 1 up to the Lys-Arg pair at sequence positions 18 and 19 in the wild-type structure in binding mode A . Based on a length per stretched amino acid of 0 . 4 nm , the expected contour length for unfolding the Doc domain up to this position would be 7 . 6 nm , in good agreement with the measured value of 8 nm within experimental error . A portion of the N-terminal calcium binding loop ( i . e . residues S11-T12 ) is involved in binding to D39 in Coh . The first peak of the double events is attributed to breakage of this interaction and simultaneous unfolding of calcium loop 1 and alpha helix 1 up to the Lys-Arg pair at sequence positions 18 and 19 . Another contributing factor is the intramolecular clasp that has been identified as a stabilizing mechanism among similar type-I Doc domains ( Slutzki et al . , 2013 ) . A recent NMR structural study ( Chen et al . , 2014 ) on the same wild-type Doc used in this work confirmed a hydrophobic ring-stacking interaction between Tyr-5 and Pro-66 . Confirmation of this clasp motif by NMR means the head and tail of the Doc are bound together , additionally stabilizing the barrier that is overcome in the first of the double event peaks . In this scenario , subsequent to breaking the interactions between the calcium binding loop and Coh , disrupting the intramolecular clasp and unfolding the N-terminal loop-helix motif , the remaining bound residues including Lys-18 , Arg-19 , Lys-50 , Leu-54 , and Lys-55 stay bound to Coh and are able to withstand substantial force on their own , eventually breaking in the second and final of the double rupture peaks . Prior work further supports this unbinding mechanism , revealing that a progressive N-terminal truncation of Doc did not affect the interaction largely , unless the truncation reached the Lys-18 and Arg-19 residues ( Karpol et al . , 2009 ) . This corroborates the idea of the C-terminal end of helix 1 being a crucial part of the binding site within the complex . Single rupture peaks were thus observed when the wild-type complex was bound in binding mode B , and no unfolding of Ca-binding loop 1 or helix 1 occurred . Force was propagated directly to bound residues Lys-18 , Leu-22 , and Arg-23 which when broken resulted in complete complex dissociation . Given the fingerprint biasing phenomenon ( Figure 6C ) , we finally sought to correct the single/double peak counting statistics ( Figure 4B ) in order to correct for undercounting of single peaks due solely to their failure to reach sufficiently high forces to unfold the fingerprint domain . Only traces showing a fingerprint were analyzed to ensure defined unfolding geometry . Using the rupture force distributions of singles , doubles , iLOV , and xylanase domains , we calculated the probability of occurrence of fingerprint unfolding at a force higher than the single-event ruptures . This overlap probability was found to be 0 . 85 for iLOV and 0 . 40 for xylanase . When the single/double peak ratios for were corrected for this effect , the final binding mode ratios ( binding mode A/binding mode B , i . e . , doubles/singles ) were found to be 0 . 95 and 0 . 87 for xylanase-Doc and iLOV , respectively . These ratios are close to 1 indicating comparable probability of each binding mode after accounting for biasing the single/double peak counting statistics due to fingerprint domain stability . We note that these numbers are also slightly lower than unity due to the exclusion of double peaks that occurred before unfolding of the fingerprint domains . Further details on rupture force distributions and overlap statistics are shown in Figure 7 . As the magnitude of biasing changes with the unfolding force distributions of each fingerprint domain , overlaps in the probability distributions allow for normalizing single/double event ratios of experimental data sets with different fingerprinting domains . For the Coh:Doc complex unbinding event , biasing ( undercounting ) is more pronounced for the mechanically weaker single ruptures . This normalization procedure shows the relative difference of biasing between single and double events , as double events are less biased than single events . 10 . 7554/eLife . 10319 . 011Figure 7 . Fingerprint unfolding and complex unbinding forces . ( A ) Rupture force distribution of final complex ruptures for single ( green ) , first ( purple ) and second ( red ) double unbinding events . ( B ) Overlap area ( purple ) of iLOV domain unfolding force distribution ( red ) ( iLOV-doubles curve class ) with the rupture force distribution ( green ) for single-event complex ruptures . ( C ) Overlap area ( purple ) of Xyn domain unfolding force distribution ( red ) ( Xyn-doubles curve class ) with the rupture force distribution ( green ) for single-event complex ruptures . Overlaps in probability distributions allow normalizing single-event counts to double events to account for different biasing caused by the different unfolding forces of the fingerprint domain . Biasing occurs , because for overlapping force distributions of fingerprint unfolding and complex ruptures , unbinding events are more likely to take place without fingerprint unfolding if the two distributions are closer together . For the Coh:Doc unbinding , this effect is more pronounced for the weaker single ruptures . Because double events are also biased , this still does not give a true quantification , but only compensates for the differences of biasing . The non-bell-evans-like shape of the single rupture peaks , especially in the region of the 1st double event peak ( A ) suggests that this class of curves does not contain a single type of unbinding mechanism , but rather a superposition of different event types . DOI: http://dx . doi . org/10 . 7554/eLife . 10319 . 011 The biological significance of Coh-Doc interactions in the context of cellulosome assembly and catalysis cannot be overstated . Their high affinity and specificity , along with their modularity , thermostability , and their ultrastable mechanical properties all make Coh-Doc unique from a biophysics perspective , and attractive from an engineering standpoint . Dual binding mode Doc domains are broadly predicted among many cellulosome producing bacteria ( e . g . C . thermocellum , C . cellulolyticum , R . flavefaciens ) , however relatively few have been confirmed experimentally ( Carvalho et al . , 2007; Pinheiro et al . , 2008; Brás et al . , 2012 ) . In fact , the direct effect of single vs . dual binding modes on the ability of cellulosomes to convert substrate into sugars is currently unknown . It is therefore unclear whether or not dual binding modes affect , for example , the catalytic properties of native or engineered synthetic cellulosomes . However , it is important to note that cellulosome producing bacteria invariably live among communities with other microorganisms , which may be producing cellulases and cellulosomes of their own . In such an environment , a dual binding mode could enable organisms to produce enzymes that are able to bind to a neighboring species’ scaffoldins , yet still retain high-affinity interactions with host scaffoldins . They would be able to combine resources with neighboring cells in a mixed microbial consortium . The dual binding mode could therefore allow genetic drift to explore interspecies protein binding . Indeed , cross-species reactivity between Coh and Doc has been reported ( Haimovitz et al . , 2008 ) . Cellulosome-producing microbes may therefore be pursuing a middle ground between protein synthesis strictly for selfish vs . communal usage . By distinguishing the presence of each binding mode for wild-type Doc domains , the single-molecule biophysical approach presented here based on differences in mechanical hierarchies will facilitate further study into the significance of the dual binding mode . In summary , the dual binding mode of Coh:Doc domains has so far proven resistant to explicit experimental characterization . Crystallography combined with mutagenesis has provided snapshots of the two modes , but resolving each of the modes for wild-type Doc under near native conditions has up until now not been possible . We have demonstrated the advantages of a single-molecule approach in resolving these subtle differences in molecular conformations of bound complexes . Despite having equal thermodynamic binding affinity , when mechanically dissociated by pulling from the N-terminus of Doc , binding mode A was more mechanically stable with an additional energy barrier . This mechanical difference was exploited to probe the two binding modes independently from one another , providing direct observation of this unique mechanism in molecular recognition . In the future , harnessing control over binding modes could offer new approaches to designing molecular assembly systems that achieve defined protein orientations .
A pET28a vector containing the previously cloned xylanaseT6 from Geobacillus stearothermophilus ( Salama-Alber et al . , 2013 ) and DocS dockerin from Clostridium thermocellum Cel48S were subjected to QuikChange mutagenesis ( Wang and Malcolm , 1999 ) to install the following mutations: Q1 , Q3 , and QQ in the dockerin and T129C in the xylanase , respectively . For insertion of the ( GS ) 4 and ( GS ) 8 linkers into the Doc domain , exponential amplification with primers bearing coding sequences for the inserts at their 5’-ends was performed with a Phusion High-Fidelity DNA polymerase ( New England Biolabs , MA ) . PCR products were then blunt end ligated using KLD Enzyme Mix and KLD Reaction Buffer from the Q5 site directed mutagenesis kit ( New England Biolabs , MA ) . The modified DNA constructs were used to transform Escherichia coli DH5-alpha cells , grown on kanamycin-containing agar plates and subsequently screened . All mutagenesis products were confirmed by DNA sequencing analysis . Primers used for inserting the ( GS ) 8 linker into the Doc domain: Fw 5’-ggttctggctccggttctggctccagcatcaacactgacaat-3’ Rev 5’-agaaccggagccagagccggaacctatacctgatctcaaaacatatct-3’ Fusion proteins HIS-CBM A2C-Coh2 ( C . t . ) were expressed in E . coli BL21 ( DE3 ) RIPL cells in kanamycin-containing media supplemented with 2mM calcium chloride overnight at 16°C . After harvesting , cells were lysed by sonication , and the lysate was subjected to heat treatment at 60°C for 30 min to precipitate the bulk of the host bacterial proteins , leaving the expressed thermophilic proteins in solution . The lysate was then pelleted , and the supernatant fluids were applied to a beaded cellulose column and incubated at 4°C for 1 hr . The column was then washed with 50 mM Tris buffer ( pH 7 . 4 ) containing 1 . 15 M NaCl , and the protein was eluted using a 1% ( vl/v ) triethylamine aqueous solution . Tris buffer was added to the eluent and the solution was neutralized with HCl . Fusion proteins HIS-Xyn T129C-DocS ( C . t . ) wild-type , Q1 , and Q3 mutants were expressed as described above . Following heat treatment , the supernatant fluids were applied to a Ni-NTA column and washed with TBS buffer containing 20mM imidazole and 2mM calcium chloride . The bound protein was eluted using TBS buffer containing 250 mM imidazole and 2 mM calcium chloride . The solution was then dialyzed to remove the imidazole . Fusion proteins ybbR-HIS-CBM A2C-Coh2 ( C . t . ) , ybbR-HIS-Xyn T129C-DocS ( C . t . ) wild-type and QQ mutants and ybbR-HIS-Xyn T129C-DocS ( C . t . ) ( GS ) 4 insert were expressed in E . coli BL21 ( DE3 ) RIPL cells; ybbR-HIS-Xyn T129C-DocS ( C . t . ) ( GS ) 8 insert fusion protein variants were expressed in E . coli NiCo21 ( DE3 ) RIPL cells . Cultivation and expression was done in ZYM-5052 autoinduction media ( Studier , 2005 ) containing kanamycin ( and chloramphenicol , in case of the NiCo21 ( DE3 ) RIPL cells ) overnight at 22°C , overall 24 hr . After harvesting , cells were lysed using sonication . The lysate was then pelleted by centrifugation at 39 , 000 rcf , the supernatant fluids were applied to Ni-NTA columns and washed with TBS buffer . The bound protein was eluted using TBS buffer containing 200 mM imidazole . Imidazole was removed with polyacrylamide gravity flow columns or with polyacrylamide spin desalting columns . All protein solutions were concentrated with Amicon centrifugal filter devices and stored in 50% ( v/v ) glycerol at -20°C ( ybbR-free constructs ) or -80°C ( ybbR-bearing constructs ) . The concentrations of the protein stock solutions were determined to be in the order of 1–15 mg/mL by absorption spectrophotometry at a wavelength of 280 nm . 1 µM of Xyn-Doc fusion proteins ( wild-type Q1 , Q3 , QQ Doc fusions ) bearing either wild-type or mutant Doc domains were adsorbed onto surfaces of the wells of a 96-well nunc maxi sorp plate ( Thermo Scientific , Pittsburgh , PA ) . After blocking ( 2% ( w/v ) BSA , 0 . 05% Tween 20 in TBS buffer ) and several rinsing steps , a red fluorescent protein-cohesin ( StrepII-TagRFP-Coh2 ( C . t . ) , Addgene ID 58 , 710 ( Otten et al . , 2014 ) ) fusion construct was incubated to the unspecifically immobilized Doc fusion proteins over a range of concentrations . After further rinsing , the fluorescence of the TagRFP domain was measured with a multi-well fluorescence plate reader ( M1000 PRO , Tecan Group Ltd . , Männedorf , Switzerland ) . Fluorescence values were plotted against their corresponding concentration values for each protein variant , and 4 parameter logistic nonlinear regression model functions were fitted to the data to determine the transition point of the curve . The Xyn domain had a cysteine point mutation at position 129 ( Xyn T129C ) to facilitate covalent attachment to a glass surface via Polyethylene glycol ( PEG ) -maleimide linkers . There were no other cysteines within the Xyn or Doc domains , which ensured site-specific immobilization of the molecule and defined mechanical loading of Doc from the N-terminus for the AFM experiments . The CBM domain likewise contained an A2C cysteine point mutation for covalent attachment to the cantilever tip via PEG-maleimide linkers . The second set of fusion proteins sfGFP-Doc and iLOV-Coh was covalently attached to coenzyme A bearing PEG linkers by their terminal ybbR tags . For AFM measurements , silicon nitride cantilevers ( Biolever mini , BL-AC40TS-C2 , Olympus Corporation nominal spring constant: 100 pN/nm; 25 kHz resonance frequency in water ) , and glass coverslips ( Menzel Gläser , Braunschweig , Germany; diameter 22mm ) were used . 3-Aminopropyl dimethyl ethoxysilane ( APDMES , ABCR GmbH , Karlsruhe , Germany ) , ɑ-Maleinimidohexanoic-ω-NHS PEG ( NHS-PEG-Mal , Rapp Polymere , Tübingen , Germany; PEG-MW: 5 kDa ) , immobilized tris ( 2-carboxylethyl ) phosphine ( TCEP ) disulfide reducing gel ( Thermo Scientific , Pittsburgh , PA ) , tris ( hydroxymethyl ) aminomethane ( TRIS , >99% p . a . , Carl Roth , Karlsruhe , Germany ) , CaCl2 ( >99% p . a . , Carl Roth , Karlsruhe , Germany ) , sodium borate ( >99 . 8% p . a . , Carl Roth , Karlsruhe , Germany ) , NaCl ( >99 . 5% p . a . , Carl Roth , Karlsruhe , Germany ) , ethanol ( >99% p . a . ) , toluene ( >99 . 5% p . a . , Carl Roth , Karlsruhe , Germany ) were used as received . Sodium borate buffer was 150 mM , pH 8 . 5 . Measurement buffer for AFM-SMFS was tris-buffered saline supplemented with 1 mM CaCl2 ( TBS , 25 mM TRIS , 75 mM NaCl , 1 mM CaCl2 pH 7 . 2 ) . All buffers were filtered through a sterile 0 . 2 µm polyethersulfone membrane filter ( Nalgene , Rochester , NY ) prior to use . Force spectroscopy measurement samples , measurements and data analysis were prepared and performed according to previously published protocols ( Jobst et al . , 2013;Otten et al . , 2014 ) . In brief , NHS-PEG-Maleimide linkers were covalently attached to cleaned and amino-silanized silicon nitride AFM cantilevers and cover glasses . The respective protein constructs were covalently linked either via engineered cysteine residues to the maleimide groups of the surface on the sample directly , or via Sfp phosphopantetheinyl transferase-mediated attachment of a terminal ybbR tag to coenzyme A , which was previously attached to the maleimide groups of the surface . AFM data were recorded in 25 mM TRIS pH 7 . 2 , 75 mM NaCl and 1mM CaCl2 buffer solution ( TBS ) . Retraction velocities for constant speed force spectroscopy measurements varied between 0 . 2 and 3 . 2 µm/s . Cantilever spring constants were calibrated utilizing the thermal method applying the equipartition theorem to the one dimensionally oscillating lever ( Hutter and Bechhoefer , 1993; Cook et al . , 2006 ) . Measurements were performed on custom built instruments , deploying an Asylum Research ( Santa Barbara , CA , USA ) MFP-3D AFM controller and Physik Instrumente ( Karlsruhe , Germany ) or attocube ( Munich , Germany ) piezo nanopositioners ( Gumpp et al . , 2009 ) . After each measurement , the xy-stage was actuated by 100 nm to probe a new spot on the surface and measure new individual Xyn-Doc fusion molecules . Instrument control software was programmed in Igor Pro 6 . 3 ( Wavemetrics ) . The retraction speed was controlled with a closed-loop feedback system running internally on the AFM controller field-programmable gate array ( FPGA ) . Data analysis and plotting was performed in Python ( Python Software Foundation . Python Language Reference , version 2 . 7 . Available at http://www . python . org ) utilizing the libraries NumPy and SciPy ( van der Walt et al . , 2011 ) and Matplotlib ( Hunter , 2007 ) . Measured raw data were analyzed by determining the zero force value with the baseline position and applying a cantilever bending correction to the z-position . The resulting force distance traces were coarsely screened for peaks as sudden drops in force and curves with less than three peaks ( such as in Figure 3—figure supplement 1 , panel F ) were excluded , as they contain no clearly identifiable signal . Force-distance traces were transformed into contour length space with the inverse worm-like-chain model ( Jobst et al . , 2013 ) , assuming a fixed persistence length of 0 . 4 nm . Screening for the 89 nm xylanase , the 36nm iLOV and the final 8 nm final double rupture increment was performed by finding their corresponding local maxima in a kernel density estimate with bandwidth b = 1 nm . Thresholds in force , distance , and peak counts were applied to sort out nonspecific and multiple interactions . All curves were ultimately selected for the xylanase or iLOV fingerprint and checked manually . For the counting statistics , double peaks were detected as an increment of 8 +- 4 nm in contour length for final rupture peaks in the contour length plot , given that the curve showed one of the fingerprints . If a double peak was detected , the force difference was determined as the percentual difference between the first and the final rupture peak force . Barrier position diagrams were assembled using optimal alignment through cross-correlation ( Puchner et al . , 2008; Otten et al . , 2014 ) . The numbers of points included in fitted histograms are provided in the figure captions , along with the statistical tests and significance values obtained . | Some bacteria use cellulose , the main component of plant cell walls , as a food source . The enzymes that break down cellulose are anchored onto a protein scaffold in a structure called the cellulosome on the bacteria’s surface . This anchoring occurs through an interaction between receptor proteins known as ‘cohesin’ domains on the scaffold proteins and ‘dockerin’ ligands on the enzymes . Most receptor-ligand interactions only allow the two proteins to bind in a single , fixed orientation . However , cohesins and dockerins are suspected to bind in two different configurations . It has been difficult to investigate the populations of these different configurations because most experimental techniques investigating protein binding take average measurements from many molecules at once . As the binding modes are extremely similar , these methods have been unable to distinguish between the two cohesin-dockerin binding configurations without introducing mutations , in part because these configurations are very similar to each other . Jobst et al . used a technique called single-molecule force spectroscopy to investigate cohesin-dockerin interactions between individual molecules . This technique applies a force that separates , or ‘unbinds’ , cohesin and dockerin , by pulling individual complexes of the two binding partners apart with a nanoscale probe . In the experiments , E . coli bacteria were made to produce mutant versions of dockerin that can only bind to cohesin in one orientation . This allowed each binding configuration to be studied individually . The results of these experiments revealed the mechanical unbinding patterns of each cohesin-dockerin configuration , and showed that it is possible to use these patterns to distinguish between the two configurations . A complimentary set of experiments revealed that wild-type ( non-mutated ) cohesin-dockerin complexes occupy both configurations in approximately equal amounts , and do not switch modes once bound . Further single-molecule experiments together with computer simulations will provide a more detailed picture of how cohesin and dockerin fit together in the two configurations . Such experiments could also reveal how cohesin and dockerin contribute to the break down of cellulose inside living cells and how they could be used for the precise assembly of single proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Resolving dual binding conformations of cellulosome cohesin-dockerin complexes using single-molecule force spectroscopy |
We report the discovery of a simple environmental sensing mechanism for biofilm formation in the bacterium Bacillus subtilis that operates without the involvement of a dedicated RNA or protein . Certain serine codons , the four TCN codons , in the gene for the biofilm repressor SinR caused a lowering of SinR levels under biofilm-inducing conditions . Synonymous substitutions of these TCN codons with AGC or AGT impaired biofilm formation and gene expression . Conversely , switching AGC or AGT to TCN codons upregulated biofilm formation . Genome-wide ribosome profiling showed that ribosome density was higher at UCN codons than at AGC or AGU during biofilm formation . Serine starvation recapitulated the effect of biofilm-inducing conditions on ribosome occupancy and SinR production . As serine is one of the first amino acids to be exhausted at the end of exponential phase growth , reduced translation speed at serine codons may be exploited by other microbes in adapting to stationary phase .
Bacteria constantly monitor their environment and internal physiological state so that they can adapt to changing conditions . A wide variety of sensing mechanisms are deployed for this purpose , including dedicated protein sensors , such as histidine kinases , which mediate changes in gene expression by controlling the phosphorylation of cognate response regulators in response to environmental cues ( West and Stock , 2001 ) . Bacteria also sense changes in their environment and physiology by means of dedicated RNAs , such as the highly structured , leader RNA for the tryptophan operon , which controls the transcription of downstream genes in the operon by a mechanism involving ribosome stalling at tryptophan codons ( Henkin and Yanofsky , 2002 ) . Here we report the discovery of an unusually simple mechanism of environmental sensing involved in the process of biofilm formation by the bacterium B . subtilis that does not require a dedicated RNA or protein . Biofilm formation involves a switch from planktonic growth as individual cells to the formation of complex , multicellular communities in response to environmental cues ( Kolter and Greenberg , 2006 ) . In B . subtilis , these communities are embedded in a self-produced matrix consisting of polysaccharide and an amyloid-like protein , which are specified by the epsA-O and the tapA-sipW-tasA operons , respectively ( Branda et al . , 2001; Kearns et al . , 2005 ) . The transition to multicellularity is governed in part by four histidine kinases ( KinA , KinB , KinC and KinD ) that control the phosphorylation of the response regulator , Spo0A , a master regulator of post-exponential phase gene expression ( Figure 1A ) ( Jiang et al . , 2000; Vlamakis et al . , 2013 ) . Recent studies suggest that KinA and KinB respond to impaired respiration ( Kolodkin-Gal et al . , 2013 ) , whereas KinC responds to membrane perturbations and KinD to unknown chemical signals ( López et al . , 2009; Shemesh et al . , 2010; Chen et al . , 2012; Beauregard et al . , 2013 ) . Once phosphorylated , Spo0A turns on sinI , a gene encoding a small protein antagonist of the biofilm-specific regulatory protein SinR ( Molle et al . , 2003; Kearns et al . , 2005 ) . SinR , which is produced constitutively , is a repressor of the matrix operons , epsA-O and the tapA-sipW-tasA , as well as other biofilm-related genes ( Kearns et al . , 2005; Chu et al . , 2006; Chai et al . , 2009 ) . SinR is also a repressor of the gene for SlrR ( Chu et al . , 2008 ) , which together with SinR sets up a self-reinforcing , double-negative feedback loop for matrix gene expression ( Figure 1A ) ( Chai et al . , 2010; Norman et al . , 2013 ) . A special feature of SinR of relevance to this investigation is that the expression of matrix genes is hypersensitive to small perturbations in the level of the protein ( Chai et al . , 2011 ) . This hypersensitivity is attributed to molecular titration of SinR by SinI and cooperativity among SinR molecules bound to tandem target sequences at regulatory sites for the matrix operons ( Chai et al . , 2011; Chai et al . , 2008 ) . 10 . 7554/eLife . 01501 . 003Figure 1 . Switching synonymous serine codons in sinR affects biofilm formation . ( A ) Regulatory circuit controlling biofilm formation in B . subtilis . ( B ) Top: Serine codon usage in the sinR coding sequence . Number within parenthesis indicates the frequency of the corresponding codon in sinR . Bottom: Average serine codon usage across 4153 protein-coding sequences in the B . subtilis genome . Number within parenthesis indicates the relative frequency of each codon in the genome . ( C ) Colony morphology for the wild-type strain and the indicated sinR synonymous variants grown on solid biofilm-inducing medium . Three TCA codons in the wild-type sequence of sinR were switched to each of the other five serine codons . The wild-type ( WT ) sinR sequence was replaced by the sinR synonymous mutant at the native sinR locus of the strain 3610 . ( D ) SinR protein level during entry into biofilm formation ( OD600 = 2 ) measured using an anti-SinR antibody that also cross-reacts with SlrR , a protein that is 85% identical to SinR . Western blot against the RNA polymerase subunit SigA was used as the loading control . Whole cell lysates were loaded at different dilutions ( indicated as X , X/2 , and X/3 ) . ( E ) Densitometry of SinR bands in ( D ) after normalization by SigA . ( F ) Top panel: Western blot against SinR and SlrR using anti-SinR antibody . Bottom panel: Densitometry ratio of the SlrR and SinR bands in the top panel . Error bars represent standard error over three replicate Western blots . The SlrR/SinR ratio for each blot was normalized such that the wild-type strain had a ratio of 1 . ( G ) Matrix gene expression monitored using a PepsA–lacZ transcriptional reporter inserted at the chromosomal amyE locus . β-galactosidase activity was measured at OD600 = 2 in liquid biofilm-inducing medium . Error bars represent standard error of three measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 00310 . 7554/eLife . 01501 . 004Figure 1—figure supplement 1 . sinR coding sequence . The three TCA codons ( switched in Figure 1 ) are highlighted in red . The three TCC codons and the two AGC/AGT codons ( switched in Figure 1—figure supplement 2 ) are highlighted in green and blue respectively . The remaining serine codons are shown in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 00410 . 7554/eLife . 01501 . 005Figure 1—figure supplement 2 . Effect of TCC and AGC/AGT synonymous substitutions in the sinR gene on colony morphology and biofilm reporter activity . ( A ) Colony morphology for the indicated sinR synonymous variants grown on solid biofilm-inducing medium . Either three TCC codons or two AGC/AGT ( AGY ) codons in the wild-type sequence of sinR were switched to remaining serine synonymous codons . The wild-type ( WT ) sinR sequence was replaced by the sinR synonymous variant at the native sinR locus of the strain 3610 . Colony morphology of the wild-type strain is shown in Figure 1 . ( B and C ) Matrix gene expression monitored using a PepsA–lacZ transcriptional reporter inserted at the chromosomal amyE locus . Strains were grown in liquid biofilm-inducing medium and β-galactosidase activity was measured at an OD600 = 2 . Error bars represent standard error of three measurements . The synonymous variants highlighted in red do not follow the hierarchy between TCN and AGC/AGT codons seen for the six TCA synonymous variants in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 005 In the current study , we present evidence for the existence of a novel cellular sensing mechanism controlling biofilm formation . Rather than relying on regulation by a dedicated RNA or protein , the translation speed of ribosomes decreases at certain serine codons , resulting in lower SinR levels , which as a consequence , contributes to derepression of SinR-controlled genes . We propose that specific serine codons in the sinR mRNA act as a simple sensor for monitoring , and triggering a response to , serine depletion under biofilm-inducing conditions .
In a genetic screen to identify suppressor mutations that rescued the biofilm-defective phenotype of a B . subtilis mutant ( ‘Materials and methods’ ) , we unexpectedly recovered a variant that contained a ‘silent’ mutation that resulted in a switch from one serine codon to a synonymous codon in sinR . This observation prompted us to ask whether switching other serine codons might also influence biofilm formation . Serine is specified by six codons: AGC , AGT , TCA , TCC , TCG and TCT ( where T is U in the mRNA ) . We noticed that the sinR coding sequence has a slightly higher frequency of the four TCN serine codons as compared to the average frequency of TCN codons in the B . subtilis genome ( Figure 1B , p=0 . 22 , N = 12 ) . To test whether this bias towards TCN codons has an effect on biofilm formation , we systematically replaced the three TCA codons in sinR ( Figure 1—figure supplement 1 ) with each of the other five serine codons . Replacing the TCA codons with AGT or AGC resulted in flat , featureless colonies on solid biofilm-inducing medium , indicating severely impaired biofilm formation ( Figure 1C ) . In contrast , replacing the TCA codons with either TCC or TCT had little or no effect on colony morphology whereas switching to TCG increased the wrinkled appearance of the colonies ( Figure 1C ) , which is indicative of robust biofilm formation . Next , we asked whether switching serine codons was altering the level of SinR protein in liquid biofilm-inducing medium . Immunoblot analysis with anti-SinR antibodies revealed slightly yet consistently higher SinR levels in the strain with the AGT variant of sinR when compared to either the wild-type strain with three TCA codons or the TCG variant of sinR ( Figure 1D , E ) . SinR is highly similar ( 85% identity ) to SlrR , which also plays a critical role in biofilm formation and whose gene ( slrR ) is under the direct negative control of SinR ( Chu et al . , 2008 ) . Because SlrR cross reacts with the anti-SinR antibodies , we were also able to detect SlrR in our immunoblot analysis . Strikingly , the levels of SlrR were almost perfectly anti-correlated with those of SinR , with the differences in the SlrR protein levels among the sinR synonymous variants being much higher than the corresponding differences in SinR protein levels ( Figure 1F ) . Because repression by SinR is ultrasensitive to SinR levels ( Chai et al . , 2011 ) , small differences in SinR protein levels among sinR synonymous variants might be sufficient to cause large differences in the levels of expression of SinR-repressed genes such as slrR . Consistent with this idea , an eps-lacZ transcriptional fusion reporter for the SinR-repressed epsA-O matrix operon showed that the four TCN sinR variants had 3- to 19-fold higher β-galactosidase activity than the AGT and AGC variants ( Figure 1G ) . To test the generality of the observed hierarchy between the synonymous variants of sinR , we generated an additional set of eleven sinR synonymous variants in which we replaced either three TCC codons or two AGC/AGT codons ( Figure 1—figure supplement 1 ) with their synonymous counterparts . Eight of these variants conformed to the hierarchy described above , namely , the four TCN variants behaved oppositely to the two AGC/AGT variants in colony morphology and in eps-lacZ reporter expression ( Figure 1—figure supplement 2 ) . The three variants that did not conform to the hierarchy could potentially reflect alterations to the mRNA sequence context near the mutation rather than the effect of a synonymous substitution per se . Taken together , the above results suggest that serine synonymous codons in the sinR coding sequence have a stereotypical effect on biofilm formation that is primarily determined by the differential usage of the four TCN and the two AGC/AGT codons . What is the mechanism by which serine codon usage affects SinR protein levels and biofilm formation ? Synonymous codon changes can alter the synthesis of the encoded protein through changes in the translation initiation rate , mRNA levels or the ribosome elongation rate ( Plotkin and Kudla , 2010 ) . However , the effect of synonymous codon usage on the initiation rate and mRNA levels is context-specific; only codons near the AUG start site affect translation initiation ( Kudla et al . , 2009 ) , whereas only codons that are located in certain regions of secondary structure or at ribonuclease cleavage sites affect mRNA levels ( Bernstein et al . , 2002 ) . Our observation that synonymous codon replacements at multiple locations along sinR have a stereotypical effect on biofilm formation argues against such context-specific mechanisms ( except for the three exceptional cases noted above ) . To test the alternative hypothesis that serine codon usage might alter the ribosome elongation rate , and given that the ribosome elongation rate at a codon varies inversely with the average ribosome density at that codon , we measured ribosome density on mRNAs at single-codon resolution using the ribosome profiling method ( Ingolia et al . , 2009; Oh et al . , 2011; Ingolia et al . , 2012 ) . We grew B . subtilis in liquid biofilm-inducing medium , harvested cells either during exponential phase growth ( OD600 = 0 . 6 ) or during stationary phase when biofilm formation is induced ( OD600 = 1 . 4 ) , and performed deep-sequencing of ribosome protected mRNA fragments and size-matched total mRNA fragments . Ribosome profiling yielded 3 . 75 and 2 . 55 million sequencing reads aligning to annotated protein-coding sequences for the exponential phase sample and the biofilm entry sample , respectively . The number of reads aligning to a single codon on individual mRNAs was too low for accurate quantification of ribosome density , and was not sufficient to directly detect increased ribosome density on the sinR transcript . However , we reasoned that the global pattern of ribosome density at codons across all mRNAs should reflect the ribosome density on individual transcripts such as sinR . Therefore , we calculated the median ribosome density at each of the 61 sense codons across the 1556 protein-coding sequences in the exponential phase sample and the 1148 sequences in the biofilm entry sample that had an average coverage greater than one sequencing read per codon ( Figure 2—figure supplement 1 , ‘Materials and methods’ ) . This analysis reproduced the previous observation ( Li et al . , 2012 ) of increased ribosome density 8 to 11 nt downstream of Shine-Dalgarno-like trinucleotide sequences both during exponential phase and during biofilm entry ( Figure 2—figure supplement 2 ) . During exponential phase , median ribosome density varied over a twofold range , with no systematic difference between serine codons and the remaining codons ( Figure 2A ) . By contrast , during biofilm entry , median ribosome density was significantly higher at serine and cysteine codons as compared to the remaining codons ( Figure 2B ) , suggesting that the translation speed of ribosomes is selectively reduced at these codons during biofilm entry . Notably , the ribosome density at serine codons was not uniform: the four UCN codons had 1 . 9 to 2 . 1-fold higher ribosome density whereas the AGC and AGU codons had only 1 . 1 and 1 . 3-fold higher ribosome density respectively , relative to the median value across 61 sense codons . Further , this difference in ribosome density between UCN codons and the AGC/AGU codons was essentially identical when computed separately for codons located in the first half or in the second half of each gene ( Figure 2—figure supplement 3 ) , a finding that underscores the statistical robustness and the context independence of the observed difference . 10 . 7554/eLife . 01501 . 006Figure 2 . Entry into biofilm formation is accompanied by codon-specific increase in ribosome density . Genome-wide median ribosome density and total mRNA density at 61 sense codons ( excluding start and stop codons ) ( A ) during exponential phase growth ( OD600 = 0 . 6 ) , and ( B ) during stationary phase when biofilm formation is induced ( OD600 = 1 . 4 ) . The six serine ( red ) and two cysteine ( green ) codons are highlighted . Genome-wide ribosome density and total mRNA density were measured by deep-sequencing of ribosome-protected mRNA fragments and total mRNA fragments respectively , of a B . subtilis 3610 derivative ( ΔepsH ) grown in liquid biofilm-inducing medium . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 00610 . 7554/eLife . 01501 . 007Figure 2—figure supplement 1 . Computational workflow for deep-sequencing data analysis . All steps outlined here were performed in Bash and Python programming languages . For further details on individual steps , see ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 00710 . 7554/eLife . 01501 . 008Figure 2—figure supplement 2 . Increase in ribosome density downstream of Shine-Dalgarno-like trinucleotide sequences . Median ribosome density across all protein coding sequences was computed for the 60 nt region around each of six Shine-Dalgarno-like trinucleotide sequences ( Li et al . , 2012 ) for the exponential phase sample ( left-hand panel ) and the biofilm entry sample ( right-hand panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 00810 . 7554/eLife . 01501 . 009Figure 2—figure supplement 3 . Context independence of ribosome and mRNA densities during biofilm formation . Each gene was conceptually divided into two equal halves and the ribosome density and mRNA density was computed separately for codons located either in the first half ( left-hand panel ) or in the second half ( right-hand panel ) of each gene . All other analysis steps were identical to those in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 009 Because ribosomes density increased at serine and cysteine codons only during biofilm formation and not during exponential phase growth , we hypothesized that the increased ribosome density was caused by the depletion of intracellular pools of these two amino acids during biofilm entry rather than by any intrinsic feature of the mRNA ( Li et al . , 2012 ) or the nascent polypeptide ( Charneski and Hurst , 2013 ) . Our hypothesis is also supported by the previous observation that synonymous codon usage can have a starvation-specific effect on protein levels ( Subramaniam et al . , 2013 ) . Further , serine is a precursor metabolite for the biosynthesis of cysteine ( Gagnon et al . , 1994 ) ; hence cysteine depletion was likely the result of a decrease in intracellular serine concentration . Consistent with this hypothesis , we also observed an increase in ribosome density at both serine and cysteine codons during serine starvation of a B . subtilis serine-auxotrophic mutant ( Figure 3C ) . Importantly , serine starvation resulted in an increase in ribosome density at only the four UCN serine codons but not at the AGC and AGU codons , matching the hierarchy seen during biofilm entry ( Figure 2B ) . Serine starvation also resulted in differential levels of production of SinR-YFP protein fusions bearing different synonymous serine codons , whereas serine-rich growth resulted in identical levels of fusion protein production from these variants ( Figure 3A , B ) . Finally , the addition of excess serine or cysteine ( but not any of the other 18 amino acids ) blocked biofilm formation in wild type cells as judged after 48 hr of growth in biofilm-inducing medium ( Figure 3—figure supplement 1 , and data not shown ) . 10 . 7554/eLife . 01501 . 010Figure 3 . Serine starvation reduces translation speed and inhibits SinR synthesis in a codon-specific manner . ( A and B ) Three sinR synonymous variants were synthesized with 10 serine codons switched to AGC , TCA or TCG . The variants were expressed as SinR-YFP fusions from the amyE locus under the control of a lac promoter in a 3610-ΔserA serine auxotroph strain growing in serine-limited medium . Black arrow around 300 min indicates the onset of serine starvation caused by depletion of exogenously-added serine in the growth medium . Cell density ( A ) and the corresponding SinR-YFP protein level ( B ) were monitored using a 96-well spectrophotometer . ( C ) Genome-wide median ribosome density for 61 sense codons ( excluding start and stop codons ) during serine starvation ( vertical axis ) and serine-rich growth ( horizontal axis ) of a serine auxotrophic strain . ( D ) Fold-change in average ribosome density for individual genes upon biofilm entry ( vertical axis ) or serine starvation ( horizontal axis ) . Genes that were preferentially up-regulated at least 10-fold upon biofilm entry in comparison to serine starvation are highlighted in red ( 68 genes , Table 1 ) . Only genes with a minimum of 100 ribosome profiling reads in at least one of the samples were included in this analysis ( 1724 genes ) and the reported log2 fold-changes are median-subtracted values across this gene set . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 01010 . 7554/eLife . 01501 . 011Figure 3—figure supplement 1 . Addition of excess serine or cysteine blocks pellicle formation by B . subtilis . Single amino acids were added at 300 µg ml−1 to liquid MSgg medium . Biofilm formation of 3610 was assayed visually by pellicle formation at the air-liquid interface 48 hr after inoculation . Serine and cysteine were found to block pellicle formation out of all 20 amino acids tested . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 011 Given that both biofilm entry and serine starvation resulted in increased ribosome density at serine and cysteine codons , we asked whether these two apparently different conditions invoke the same gene expression program in B . subtilis . The fold-change in average ribosome density of B . subtilis genes was positively correlated between biofilm entry and serine starvation ( Figure 3D , R2 = 0 . 27 , p=10−5 , 1724 genes ) . However , a subset of 68 genes was induced at least 10-fold higher upon biofilm entry than during serine starvation ( indicated by red markers in Figure 3D , Table 1 ) . This subset included genes for anaerobic metabolism such as lctEP , nasDE , and cydAB , and is consistent with the recently proposed role of impaired respiration in biofilm formation ( Kolodkin-Gal et al . , 2013 ) . We observed that sulfur metabolism genes were also enriched in this subset , possibly indicating a stronger response to cysteine depletion during biofilm entry than during serine starvation . 10 . 7554/eLife . 01501 . 012Table 1 . B . subtilis genes that have greater than 10-fold difference in expression ratio between biofilm formation and serine starvationDOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 012GeneABCDEFFunctionalbA9 . 35−0 . 693210129175144antilisterial bacteriocin ( subtilosin ) production proteinalsD6 . 380 . 9534138453135alpha-acetolactate decarboxylasealsS6 . 961 . 7976461386396acetolactate synthasecah2 . 08−2 . 74253152541486295S-deacylasectc3 . 78−0 . 73327220632225750S ribosomal protein L25cydA5 . 36−3 . 18234472040860cytochrome bd ubiquinol oxidase subunit IcydB9 . 6−1 . 889337412946cytochrome bd ubiquinol oxidase subunit IIcysK1 . 52−2 . 21157602212089912580cysteine synthasegcvPA0 . 97−2 . 38150014401001255glycine dehydrogenase subunit 1gcvPB1 . 05−2 . 28202920571224335glycine dehydrogenase subunit 2gspA4 . 54−0 . 751111261135106glycosyl transferase ( general stress protein ) iseA2 . 24−1 . 141117257925211516inhibitor of cell-separation enzymeslctP6 . 62−2 . 962300924744L-lactate permeaseldh3 . 47−3 . 88281015 , 2541732157L-lactate dehydrogenasemaeN4 . 08−1 . 611481226347151Na+/malate symportermccA3 . 23−2 . 72759349438678cystathionine beta-synthasemetE2 . 72−2 . 1333 , 02710647332 , 23197605-methyltetrahydropteroyltriglutamate/homocysteine S-methyltransferasemgsR4 . 72−1 . 53163209920293stress transcriptional regulatormtlA3 . 92−0 . 22126934104119PTS system mannitol-specific transporter subunit IICBmtnA1 . 84−2 . 79252244253501671methylthioribose-1-phosphate isomerasemtnD1 . 72−1 . 634095660031821361acireductone dioxygenasemtnK2 . 54−2 . 64429012 , 19658661250methylthioribose kinasenasD6 . 21−0 . 652177872635536assimilatory nitrite reductase subunitnasE6 . 050 . 3834110663109assimilatory nitrite reductase subunitrbfK3 . 74−2 . 571329868143497RNA-binding riboflavin kinasesboA7 . 770 . 6412112 , 895256530subtilosin AsboX8 . 260 . 1921312876115bacteriocin-like productssuA2 . 09−2 . 336917682877236aliphatic sulfonate ABC transporter binding lipoproteinssuB2 . 57−2 . 0924167004779242aliphatic sulfonate ABC transporter ATP-binding proteinssuC2 . 17−2 . 1637848362692206aliphatic sulfonate ABC transporter permeasessuD2 . 2−2 . 1912 , 96129 , 1352812817alkanesulfonate monooxygenasetcyJ3 . 25−3 . 261046485642759sulfur-containing amino acid ABC transporter binding lipoproteintcyK3 . 79−3 . 55281519 , 0871095124sulfur-containing amino acid ABC transporter binding lipoproteintcyL3 . 33−3 . 05855422338762sulfur-containing amino acid ABC transporter permeasetcyM3 . 54−2 . 97185910 , 55658199sulfur-containing amino acid ABC transporter permeasetcyN3 . 38−2 . 74336317 , 1491119223sulfur-containing amino acid ABC transporter ATP-binding proteinureA4 . 360 . 78124125277176urease subunit gammaycgL0 . 99−3 . 0151250027846hypothetical proteinycgM2 . 46−1 . 353910517893proline oxidaseycgN2 . 45−1 . 6928247512675561-pyrroline-5-carboxylate dehydrogenaseycnJ0 . 75−2 . 6316813812126copper import proteinydaG4 . 140 . 497060480150general stress proteinydbL3 . 07−0 . 312941210130139hypothetical proteinyeaA1 . 34−2 . 5911213813931hypothetical proteinyezD2 . 52−4 . 1714540634025hypothetical proteinyitJ3 . 01−2 . 43278510 , 99046871153bifunctional homocysteine S-methyltransferase/5 , 10-methylenetetrahydrofolate reductaseyjbC3 . 67−0 . 57104647248222thiol oxidation management factor; acetyltransferaseyjnA1 . 5−2 . 819721342790149hypothetical proteinyoaB2 . 28−2 . 26220652442862792negatively charged metabolite transporteryoaC2 . 92−1 . 89120044591436513hydroxylated metabolite kinaseyrhB2 . 92−2 . 93380614 , 0961904333cystathionine beta-lyaseyrrT2 . 97−3 . 12546208944368AdoMet-dependent methyltransferaseytlI1 . 66−3 . 220631814120LysR family transcriptional regulatorytmI3 . 34−3 . 27345217 , 1731640226N-acetyltransferaseytmO3 . 4−2 . 88386620 , 0071179213monooxygenaseytnI3−2 . 6352213 , 770867189redoxinytnJ3 . 14−2 . 8610 , 64545 , 9972495456monooxygenaseytnL3 . 56−2 . 451281737135486aminohydrolaseytnM3 . 5−2 . 6454225 , 2021264277transporteryuaF1 . 89−1 . 748715715863membrane integrity integral inner membrane proteinyvzB0 . 95−2 . 4812511816840flagellinyxaL3 . 54−0 . 4110696077442442membrane associated protein kinaseyxbB3 . 7−0 . 01108685118155S-adenosylmethionine-dependent methyltransferaseyxeK0 . 86−2 . 72270224061073216monooxygenaseyxeL1 . 29−2 . 9443752520235acetyltransferaseyxeM0 . 87−2 . 57300326921047233ABC transporter binding lipoproteinyxeP1 . 75−2 . 2425774246736207amidohydrolaseyxjH2 . 02−1 . 824162824338111432methyl-tetrahydrofolate methyltransferaseA—median-subtracted log2 fold-change: biofilm/exponential-phase , B—median-subtracted log2 fold-change: serine starvation/serine rich , C—raw counts: biofilm entry , D—raw counts: exponential phase , E—raw counts: serine rich , F—raw counts: serine starvation . In toto , the results with the serine auxotroph support the inference that ribosome stalling observed during biofilm formation is due to a drop in intracellular serine levels . Our efforts to measure intracellular serine levels directly during growth in minimal , biofilm-inducing medium ( MSgg ) have been unsuccessful . Hence , we cannot rule out the less likely possibilities that biofilm entry causes serine and cysteine to be sequestered away from protein synthesis or that aminoacylation rate of the corresponding tRNAs decreases without changes in the intracellular serine and cysteine pools . Here we found that the translation speed of ribosomes decreases at UCN serine codons and thereby modulates production of SinR and entry into biofilm formation . Based on our genome-wide measurements of ribosome density , we expect that the translation speed of ribosomes should decrease during biofilm entry not only on the sinR mRNA , but also on other mRNAs that are enriched for any of the four UCN codons . For example , we found that the four TCN codons are over-represented in two nucleotide biosynthesis genes , pyrAA and purB ( Figure 4A ) . These two genes are also transcriptionally down regulated during biofilm entry ( Figure 4B ) . A high frequency of TCN codons in these genes might serve to reinforce their transcriptional down regulation by reducing translation speed . Conversely , TCN codons are under-represented in the genes encoding lactate dehydrogenase , ldh ( Cruz Ramos et al . , 2000 ) and a master regulator of post-exponential phase gene expression , spo0A ( Molle et al . , 2003 ) ( Figure 4A ) . The two genes are transcriptionally up regulated upon biofilm entry ( Figure 4B ) . The low frequency of TCN codons in these genes might represent a mechanism for optimizing the production of their protein products by minimizing the slowing down of translation during biofilm entry . Consistent with this idea , replacement of AGC/AGT codons by TCN codons in spo0A , whose protein product positively regulates biofilm entry ( by turning on the synthesis of the SinR antagonist SinI ) , resulted in defective biofilm formation ( Figure 4C ) in contrast to the stimulatory effect of replacing AGC/AGT codons with TCN codons in sinR ( Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 01501 . 013Figure 4 . Serine codon bias of biofilm-regulated genes reflects their expression under serine starvation . ( A ) Relative serine codon fraction in genes for nucleotide biosynthesis ( pyrAA , purB ) , lactate dehydrogenase ( ldh ) and a sporulation regulator ( spo0A ) . Numbers in parentheses indicate the number of serine codons in each gene . Relative fraction of serine codons across the B . subtilis genome is shown for comparison . ( B ) Fold-change ( expressed in log2 units ) in average ribosome density upon biofilm entry for the four genes shown in A . ( C ) Colony morphology of a wild-type strain and two spo0A synonymous variants grown on solid biofilm-inducing medium . Seven AGC/AGT codons in wild-type spo0A were replaced by either 7 TCC codons or 3 TCC and 4 TCG codons and inserted at the chromosomal spo0A locus . Both the wild-type spo0A and the synonymous spo0A variants were inserted with a downstream selection marker . ( D ) Left: Codon Adaptation Index ( CAI ) for the four genes shown in A . Right: Distribution of CAI values for 4153 protein-coding sequences of B . subtilis . DOI: http://dx . doi . org/10 . 7554/eLife . 01501 . 013
Together , our results implicate serine depletion as an environmental cue that contributes to promoting biofilm formation in B . subtilis together with other cues that are sensed by the histidine kinases KinA–D . Serine depletion is sensed through a remarkably simple mechanism based on reduced translation speed at UCN serine codons in the mRNA for a regulatory protein , SinR , whose repressive effects are highly sensitive to small changes in the level of the protein . We presume that UCN codons lower SinR levels simply by slowing the rate of ribosome movement along the mRNA ( elongation ) . However , it is possible that the reduced translation speed at UCN codons during biofilm entry could be followed by downstream events such as ribosome rescue ( Keiler et al . , 1996 ) or mRNA decay ( Hayes and Sauer , 2003 ) that might also contribute to lowering the levels of the SinR protein . The serine sensing mechanism uncovered here operates through over-representation of the TCN serine codons in the sinR gene without the necessity for any other dedicated protein or RNA for sensing serine depletion . By contrast , transcriptional attenuation , a widespread mechanism among bacteria for sensing amino acids that also relies on changes in translation speed , involves a translation-transcription coupling mechanism that is mediated by highly structured mRNAs and leader peptides ( Henkin and Yanofsky , 2002 ) . We note that the biased usage of the four TCN serine codons , which act as starvation sensors during biofilm formation , is not evident from widely-used phenomenological measures of codon bias such as the codon adaptation index ( Figure 4D ) , which primarily reflects codon preferences during exponential growth ( Sharp and Li , 1987; Andersson and Kurland , 1990 ) . The difference in translation speed between the four UCN codons and the two AGC/AGU codons under biofilm-inducing conditions is likely mediated by differences in concentration of the corresponding aminoacylated tRNAs ( Elf et al . , 2003; Dittmar et al . , 2005 ) , as was recently observed in serine-starved E . coli ( Subramaniam et al . , 2013 ) . Interestingly , the hierarchy between UCN codons and AGC/AGU codons in B . subtilis during serine starvation is similar to the one in E . coli even though copy numbers of the corresponding tRNA genes have diverged significantly between these two organisms ( Lowe and Eddy , 1997 ) . Despite different tRNA gene copy numbers , it is possible that the relative abundances of the serine tRNA isoacceptors are similar between the two organisms or that their relative abundances might be regulated in the same manner in response to nutrient deprivation ( Doi et al . , 1966 ) . Serine is one of the first amino acids to be completely consumed from the culture medium when either B . subtilis or E . coli cells are grown in complex rich medium ( Liebs et al . , 1988; Prüss et al . , 1994; Sezanov et al . , 2007 ) . Indeed , increased ribosome density has been observed at serine codons during growth of E . coli in Luria-Bertani broth ( Li et al . , 2012 ) . Thus the role of synonymous serine codons as starvation sensors discovered here in the specific context of biofilm formation might be a general regulatory strategy in microbes for adapting to nutrient depletion at the end of exponential phase growth . It is noteworthy that depletion of specific amino acids affects developmental transitions in several eukaryotic cells ( Marin , 1976; Sundrud et al . , 2009; Wang et al . , 2009 ) . It will be interesting to test whether a codon-based sensing mechanism , similar to the one found here in bacterial biofilm development , also plays a role in eukaryotic cells during amino acid depletion .
For ribosome profiling during biofilm formation , a 3610-ΔepsH strain ( RL3852 ) was used to ensure dispersed growth in liquid media ( Kearns et al . , 2005 ) . For serine starvation experiments , a serine-auxotrophic 3610-ΔserA strain ( YC865 ) was used . A list of strains , plasmids , and oligonucleotides used in this work are summarized in Supplementary file 1 . For general purposes , B . subtilis strains PY79 , 3610 , and their derivatives were grown in Luria-Bertani ( LB ) medium ( 10 g tryptone , 5 g yeast extract , and 5 g NaCl per liter broth ) at 37°C . Antibiotics were added to the media at the following concentrations for B . subtilis strains: 10 µg ml−1 of tetracycline , 100 µg ml−1 of spectinomycin , 10 µg ml−1 of kanamycin , 5 µg ml−1 of chloramphenicol , and 1 µg ml−1 of erythromycin . Minimal MSgg medium was used as the liquid biofilm-inducing medium . The same medium with 1 . 5% Bacto-agar ( Difco , Franklin Lakes , NJ ) was used as the solid biofilm-inducing medium . MSgg medium composition: 5 mM potassium phosphate ( pH 7 ) , 100 mM MOPS ( pH 7 ) , 2 mM MgCl2 , 700 µM CaCl2 , 50 µM MnCl2 , 50 µM FeCl3 , 1 µM ZnCl2 , 2 µM thiamine , 0 . 5% glycerol , 0 . 5% glutamate , 50 µg ml−1 tryptophan , 50 µg ml−1 phenylalanine and 50 µg ml−1 threonine . For overnight growth of serine auxotrophic 3610 strains , MSgg medium was supplemented with serine to a final concentration of 5 mM . For serine starvation experiments in which YFP fluorescence was measured ( Figure 3A , B ) , MSgg medium was supplemented with 800 µM serine and 400 µM serine methyl-ester ( Sigma , St . Louis , MO ) . Serine methyl-ester was added to ensure slow growth under serine starvation conditions . General methods for molecular cloning followed published protocols ( Sambrook 2001 ) . SPP1 phage-mediated transduction was used to transfer antibiotic-marked DNA fragments between different strains ( Kearns et al . , 2005 ) . Long-flanking PCR mutagenesis was applied to generate insertional deletion mutations ( Wach 1996 ) . Synonymous switches in sinR and spo0A were generated by using synthetic DNA fragments ( Genewiz , South Plainfield , NJ ) or by applying site-directed mutagenesis ( Roche , Switzerland ) . Sequences of the primers used in constructing mutant sinR alleles are described in Supplementary file 1 . Incorporation of synonymous substitutions into the sinR or spo0A gene at the native locus was done by allele exchange and followed a method described previously ( Chai et al . , 2011 ) . B . subtilis cells were first grown in LB broth at 37°C to mid-exponential phase . For formation of biofilm colonies , 2 µl of the cells was spotted onto MSgg medium solidified with 1 . 5% agar . Plates were incubated at 23°C for 3–4 days before analysis . All images were taken using either a Nikon CoolPix 950 digital camera or using a SPOT camera ( Diagnostic Instruments , Sterling Heights , MI ) . Assays for the β-galactosidase activities were described previously ( Kearns et al . , 2005 ) . Following a previously-published protocol ( Chai et al . , 2010 ) , we set up a genetic screen to search for spontaneous mutations that suppressed the defective biofilm phenotype of a B . subtilis ΔslrR mutant ( YC131 ) . The defective biofilm phenotype of the ΔslrR mutant is manifested as an inability to form robust floating pellicles ( Chu et al . , 2008; Chai et al . , 2010 ) . Briefly , the ΔslrR strain was inoculated into liquid MSgg medium in 6-well plates and incubated at 30°C . After 48 hr , pellicle formation was examined visually . In some wells , robust pellicles appeared possibly due to a second , suppressor mutation elsewhere in the genome . Cells from those wells were picked and streaked out on fresh LB agar plates to isolate single colonies . Cells from the single colonies were then tested for altered colony morphology on solid MSgg medium to confirm the suppressor phenotype . Similar genetic screens in previous studies ( Kearns et al . , 2005; Chai et al . , 2010 ) had established that mutations in the sinR gene could suppress the defective biofilm phenotype of the ΔslrR mutant . Therefore , we isolated genomic DNA from the putative suppressor mutants , amplified the sinR gene by PCR , and then sequenced the sinR locus . Once a mutation in the sinR gene was confirmed by sequencing , the same mutation was reconstituted in the wild type background ( 3610 ) following a previous protocol ( Chai et al . , 2010 ) , and assayed for alteration in colony morphology on solid MSgg medium . For ribosome profiling during biofilm formation , fresh colonies were inoculated into 8 ml of MSgg liquid medium and grown for 12 hr at 30°C , 200 rpm . Saturated cultures were diluted 1:1000 into 200 ml aliquots of fresh MSgg medium and shaken in a 1L flask at 30°C , 200 rpm . For exponential-phase ribosome profiling ( Figure 2A ) , cultures were harvested at OD600 = 0 . 6 . For ribosome profiling during biofilm entry ( Figure 2B ) , cultures were harvested at OD600 = 1 . 4 . For the serine repletion experiment ( Figure 4 ) , serine was added to a final concentration of 2 . 5 mM at OD600 = 1 . 4 and harvested after 30 min at 30°C , 200 rpm . For the serine starvation experiment ( Figure 3B , C ) , pre-cultures were grown in MSgg medium supplemented with 5 mM serine and then diluted into 200 ml of the same medium . At an OD600 = 0 . 6 , the cultures were filtered and re-suspended either in MSgg medium ( starvation ) or in MSgg medium with 5 mM serine ( control ) , and harvested after 60 min at 30°C , 200 rpm ( Figure 3D ) . Cultures were grown with shaking at 37°C , and 14 ml culture aliquots were harvested at an OD600 between 2 . 0 and 2 . 5 . Cell pellets were collected by centrifugation and washed once with 10 ml of lysis buffer ( 20 mM Tris-HCl , 200 mM NaCl , 1 mM EDTA pH 7 . 4 ) . Pellets were resuspended in 1 . 2 ml lysis buffer and incubated with 20 µg/ml of lysozyme ( Sigma ) for 1 hr on ice . Cells were further lysed by sonication . Cell debris was removed by centrifugation ( 14 , 000 rpm , 30 min , 4°C ) . The concentration of total protein in the lysates was determined by a Bradford assay ( Bio-Rad , Hercules , CA ) . Samples for SDS-PAGE were prepared in Laemmli buffer normalized to equal protein concentration . Samples were ran on an NuPAGE 12% gel ( Invitrogen , Carlsbad , CA , 1 . 0 mm , Bis/Tris , 200 V , ∼50 min ) and transferred to a PVDF membrane ( Millipore , Billerica , MA ) at 100 V for 1 hr . The membrane was blocked in 5% milk-TPBS for 1 hr , and then incubated with anti-SinR antibody ( 1:2500 , polyclonal ) and anti-SigA ( 1:100 , 000 , polyclonal ) overnight . The membranes were washed 3 times in TPBS for 5 min each . Blots were incubated with goat anti-rabbit secondary antibody conjugated to Horseradish peroxidase ( 1:10 , 000 , Bio-Rad ) . Blots were washed three times in TPBS , and developed with SuperSignal West Dura chemiluminescent substrate ( Thermo , Waltham , MA ) and imaged on a gel-doc ( Bio-Rad ) . Densitometry analysis of Western blot images was performed using ImageJ software ( NIH , http://rsbweb . nih . gov/ij/ ) . Rectangles were drawn around each distinct band and the average pixel intensity in this rectangle was calculated , followed by subtraction of background pixel intensity from an identical rectangle drawn in a region without any band . These background subtracted values were used for calculating SinR/SigA , SlrR/SigA and SinR / SlrR ratios in Figure 1 . A previously published measurement protocol was used ( Subramaniam et al . , 2013 ) . Fresh colonies were inoculated into 1 ml of MSgg liquid medium with 5 mM serine and grown overnight in deep 96-well plates at 30°C , 1400 rpm . In the morning , saturated cultures were diluted 1:100 into 1 ml of MSgg medium with 800 µM serine and 400 µM serine methyl-ester and shaken at 30°C , 1350 rpm for 3 hr . Three aliquots of 150 µl from each culture was pipetted into three wells of three 96-well plates ( 3799 , Costar , Corning , NY ) . Wallac Victor2 plate reader ( PerkinElmer , Waltham , MA ) was used to monitor cell density ( absorbance at 600 nm ) and YFP expression ( fluorescence , excitation 504 nm and emission 540 nm ) . Each plate was read every 15 min using a robotic system ( Caliper Life Sciences , Hopkinton , MA ) and shaken at 1000 rpm in between readings ( Variomag Teleshake shaker , Daytona Beach , FL ) . 30°C and 60% relative humidity was maintained throughout the experiment . Ribosome profiling protocol was adapted from published literature ( Ingolia et al . , 2009; Oh et al . , 2011; Ingolia et al . , 2012 ) with minor modifications as described below . Briefly , 200 ml of bacterial culture was harvested by filtration . The filter was immediately inserted into a 50 ml conical tube , flash frozen in liquid nitrogen , and stored at −80°C until further processing . Frozen cells were re-suspended in 8 ml of polysome resuspension buffer ( 20 mM Tris pH 8 . 0 , 10 mM MgCl2 , 100 mM NH4Cl , and 100 µg ml−1 Chloramphenicol ) . Re-suspended cells were pelleted by centrifugation ( 3000 g , 4°C , 5 min ) and the supernatant was discarded . The cell pellet was re-suspended in 500 µl of polysome lysis buffer ( 1X polysome resuspension buffer , 5 mM CaCl2 , 0 . 4% TritonX-100 , 0 . 1% NP-40 , and 100 U/ml RNase-free DNase [04716728001; Roche] ) , and transferred to an ice-cold 1 . 5 ml tube containing 500 µl of 0 . 2–0 . 3 µm acid-washed glass beads ( G1277; Sigma ) . Cells were lysed by vortexing at maximum speed on a vortexer in a 4°C room ( Vortex Genie 2 , 10 × 30 s with 1 min cooling on ice in between ) . The lysate was clarified by centrifugation ( 20 , 000 g , 4°C , 10 min ) and the supernatant was transferred to a fresh 1 . 5 ml tube . 500 µg of total RNA ( A260 units ) was digested ( 25°C , 1400 rpm , 60 min , 150 µl vol ) with 2 U/µg of Micrococcal nuclease ( LS004797; Worthington , Lakewood , NJ ) . The digestion was quenched with 1 . 5 µl of 0 . 5 M EGTA , loaded on top of a 10–50% sucrose gradient and ultra-centrifuged in a SW41 rotor ( 35000 rpm , 4°C , 150 min ) . Monosomes collected by gradient fractionation ( Biocomp Instruments , Canada ) . For total RNA extraction , 100 µl of polysome lysate was mixed with 400 µl of RNA extraction buffer ( 0 . 3 M sodium acetate , 10 mM EDTA , pH 4 . 5 ) and the aqueous phase was extracted twice with phenol-chloroform and once with chloroform . RNA was precipitated with an equal volume of isopropanol . The pellet was washed with 70% ice-cold ethanol and re-suspended in 100 µl of 10 mM Tris pH 7 . 0 . 10 µg of total RNA was DNase-treated and mRNA enriched using the Microbe Express kit ( Invitrogen ) . mRNA was fragmented by heating at 95°C with a bicarbonate buffer ( Ingolia et al . , 2009 ) for 20 min . Collected monosome fractions were purified using the same phenol-chloroform method as used for total RNA extraction above . Monosomes and fragmented mRNA were then used for small RNA sequencing library preparation . Size selection , dephosphorylation , polyadenylation , reverse-transcription , circularization and PCR amplification were performed using the same protocol as in ( Ingolia et al . , 2009 ) . An rRNA subtraction step was carried out between the circularization and PCR amplification steps using the same protocol as in ( Oh et al . , 2011 ) . Typically , several samples were multiplexed for sequencing in an Illumina HiSeq sequencer such that at least 1 million reads were obtained for each sample . Deep-sequencing data analysis was carried out in Bash and Python programming languages , and performed on the Harvard research computing Odyssey cluster . Main steps are summarized below and shown schematically in Figure 2—figure supplement 1 . Each 50 nt single-end read was polyA-trimmed by identifying 10 or greater number of adjacent adenines and discarding all nucleotides starting from −1nt of the polyA run . The first 5′ nt of the read was also discarded since its identity was ambiguous in several reads . PolyA-trimmed reads were first aligned against all non-coding RNAs in the B . subtilis genome using bowtie aligner ( ver . 0 . 12 . 7 , Langmead et al . , 2009 ) . The non-coding RNA sequences were downloaded from NCBI ( NC_000964 . frn ) . Reads that did not align to non-coding RNAs were aligned against the whole B . subtilis genome using bowtie aligner . The B . subtilis genome was downloaded from NCBI ( NC_000964 . fna ) . Only reads that had less than three mismatches with the reference genome were considered for further analysis . Reads that aligned to the B . subtilis genome were further trimmed by 8 nt from each end to approximate the ribosome A-site coordinate . The remaining sequence was normalized by its length and assigned to the corresponding genomic coordinate , and this value was designated as the read density at this genomic coordinate during further downstream analyses . Average ribosome and mRNA density for a single gene was calculated by summing the read density between the start and stop codon , and then normalizing by the length of the gene . Fold-change in ribosome density for a single gene between two samples was calculated by taking the log2 of the ratio of average ribosome density between the two samples for that gene . The median value of this log2 fold-change across all genes that received a minimum of 100 reads in at least one of the two samples was then subtracted from the fold-change value for each gene . This median-subtracted log2 fold change in reported throughout this work . We note that the average ribosome density on any gene is directly proportional to the corresponding mRNA level in the absence of specific translational regulation . Hence fold-changes in ribosome density ( such as the one shown in Figure 3D ) primarily reflect fold-changes in mRNA level . To calculate the ribosome and mRNA density at individual codons , The start codon was treated as a single separate codon irrespective of its identity . Only genes with average ribosome density of at least 1 read per codon were considered . The ribosome density at the first nucleotide of each codon was assigned as the ribosome density at that codon . For each of the 64 codons and the start codon , read density at the first nucleotide of the codon was averaged across all occurrences of that codon in a single gene and then normalized by the average ribosome density for that gene . Hence a codon without over- or under- representation of ribosome or mRNA density will have a density value equal to 1 . Genome-wide ribosome and mRNA density was calculated as the median of the individual gene read density from Step ( 3 ) across all genes that pass the threshold of Step ( 2 ) . Ribosome profiling measurement resulted in a high ribosome density at start and stop codons . It is unclear whether this increase is a true biological signal or caused by the measurement protocol ( Ingolia et al . , 2011 ) . Hence these codons were excluded from the ribosome density plots shown in Figure 2A , B and 3C . However , the increased ribosome density at these codons resulted in a concomitant decrease in ribosome density at the remaining codons due to normalization by the average ribosome density for each gene ( which included the start and stop codons ) . Codon Adpatiaton Index ( CAI ) was calculated according to the original prescription of Sharp and Li ( Sharp and Li , 1987 ) . For this calculation , 68 genes that had an annotation as ‘ribosomal’ were used as the reference set of highly expressed genes . The annotations file used for this analysis was downloaded from NCBI ( NC_000964 . ptt ) . The p value for the higher frequency of TCN codons in sinR was calculated by assuming a binomial distribution of TCN and AGC/AGT codons . The p value represents the binomial probability that there are 10 or more TCN codons in the sinR gene ( 12 serine codons total ) given the genome-wide frequency of serine codons ( 0 . 66 for TCN codons and 0 . 34 for AGC/AGT codons ) . | Bacteria use several different mechanisms to recognize and respond to changes in their environment . Protein sensors , for example , relay signals from the cell surface to target molecules within the cell . Additionally , RNA sensors can respond to internal levels of chemicals by regulating the expression of genes . Now , Subramaniam et al . have discovered a sensing mechanism in bacteria that does not rely on either protein sensors or RNA sensors . Bacteria normally live as free swimming cells in water . But sometimes , in response to certain environmental conditions , they form a multicellular community called a biofilm . In this biofilm , bacteria encase themselves with layers of carbohydrates and proteins , which then protect the bacteria from adverse chemicals such as antibiotics . A protein known as SinR plays a key role during biofilm formation in the model bacterium Bacillus subtilis . SinR normally prevents the formation of the carbohydrates and proteins that make up the biofilm . Upon the decision to form a biofilm , B . subtilis counters the effect of SinR by producing an anti-SinR protein called SinI . Now Subramaniam et al . have found that as well as producing the SinI protein , B . subtilis use an additional mechanism to promote biofilm formation . This mechanism relies on codons , the elements within genes that correspond to specific amino acids . Six different codons correspond to the amino acid serine , and the gene for the SinR protein contains above average numbers of four of them . These four codons are highly sensitive to serine levels , and they decrease the levels of the SinR protein when there is less serine in the environment , as happens to be the case in biofilms . And since SinR prevents the production of the carbohydrates and proteins that make up the biofilms , a decrease in the levels of SinR leads to an increase in the production of biofilms . Since the serine codons at the heart of the sensing mechanism discovered by Subramaniam et al . are present in all forms of life , from viruses to humans , it is possible that similar sensing mechanisms might be found in contexts other than bacterial biofilms , such as in viral infection and cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2013 | A serine sensor for multicellularity in a bacterium |
Mammalian oocytes are arrested in the dictyate stage of meiotic prophase I for long periods of time , during which the high concentration of the p53 family member TAp63α sensitizes them to DNA damage-induced apoptosis . TAp63α is kept in an inactive and exclusively dimeric state but undergoes rapid phosphorylation-induced tetramerization and concomitant activation upon detection of DNA damage . Here we show that the TAp63α dimer is a kinetically trapped state . Activation follows a spring-loaded mechanism not requiring further translation of other cellular factors in oocytes and is associated with unfolding of the inhibitory structure that blocks the tetramerization interface . Using a combination of biophysical methods as well as cell and ovary culture experiments we explain how TAp63α is kept inactive in the absence of DNA damage but causes rapid oocyte elimination in response to a few DNA double strand breaks thereby acting as the key quality control factor in maternal reproduction .
The p53 protein family with its three members p53 , p63 and p73 plays very important roles in the surveillance of genetic and cellular stability ( Levine et al . , 2011 ) . Probably the most ancient function of this family is the maintenance of genetic quality in germ cells since even short lived eukaryotic animals express a p63-like protein in their germ cells ( Ollmann et al . , 2000; Derry et al . , 2001; Brodsky et al . , 2000; Suh et al . , 2006; Ou et al . , 2007 ) . In mammals , up to 10 diverse p63 isoforms exist with the longest one , TAp63α , being highly expressed in primary oocytes that are arrested in prophase of meiosis I . After homologous recombination , oocytes are kept in this dictyate arrest phase until they are recruited for ovulation , a period that can take decades in humans . Once oocytes reenter the cell cycle , expression of TAp63α is lost ( Suh et al . , 2006 ) . Since p63 can initiate apoptosis the high expression level of TAp63α in oocytes requires that its activity is tightly regulated . Recently we could show that TAp63α assembles into a closed and only dimeric conformation in which the protein is inactive ( Deutsch et al . , 2011 ) . Detection of DNA damage leads to activation of p63 triggered by phosphorylation ( Suh et al . , 2006; Bolcun-Filas et al . , 2014 ) that results in the formation of open tetramers with a twentyfold higher DNA binding affinity and the induction of apoptosis . This p63-based quality control is unique to oocytes , making them very sensitive to DNA damage . Irradiation with 0 . 45 Gy is sufficient to eliminate all p63-expressing oocytes in mice while all surrounding cells of the ovaries survive . To understand the mechanism of inhibition and activation we have started to characterize the structural requirements for the formation of the closed and dimeric state of TAp63α . In previous experiments we have shown that the very C-terminus contains a transactivation inhibitory domain ( TID ) that is of central importance for creating the closed dimeric state ( Serber et al . , 2002; Straub et al . , 2010 ) . We have suggested a model in which both the C-terminal TID and the N-terminal transactivation domain ( TAD ) interact with the central tetramerization domain ( TD ) thereby preventing the formation of tetramers . This central TD is a dimer of dimers suggesting that blocking the interface by which two dimers form a tetramer is the most likely mechanism of inhibition . In the past we have identified mutations in all three domains – TAD , TD and TID – that break the inhibitory mechanism , establishing that at least these three domains are involved in this process . In the absence of a high resolution structure we have now used systematic alanine scanning and charge swap mutagenesis in combination with SAXS ( small angle X-ray scattering ) experiments to build a model of the closed and dimeric complex . In addition , we show that the inhibited conformation is a kinetically trapped state and that the oocyte contains all factors necessary to activate p63 without requirement of further protein expression . Together our data show that activation of TAp63α follows a spring-loaded mechanism and explains why oocytes are far more sensitive to DNA damage than the surrounding follicular cells .
TAp63α contains three folded domains , the DNA binding domain ( DBD ) , the tetramerization domain ( TD ) and the SAM domain that are linked by unstructured regions . NMR experiments with a tetrameric construct containing all three folded domains showed that these domains behave independently as pearls on a string ( Figure 1—figure supplement 1 ) . All sequences outside of these folded domains are not structured in isolation but may be folded when interacting with other segments of the protein as part of the inhibitory mechanism . To identify the exact sequence elements required to form the closed state , we systematically deleted sequences in these linker regions . Deletion of sequences crucial for the formation of the closed state results in the formation of an open conformation . Previously we have shown that the open state can be detected by a conformation sensitive pull-down experiment: tetrameric mutants with an intact TAD can be pulled down with a GST-TID construct ( 569–616 ) ( Straub et al . , 2010 ) . Thus , mutants that cannot be pulled down are assumed to adopt the closed dimeric state . After several rounds of deletion mutagenesis , a minimal dimeric construct was obtained . Size exclusion chromatography combined with multi angle light scattering ( SEC-MALS ) confirmed that this minimal construct ( TAp63αmin ) comprising deletions Δ ( 1–9; 64–119; 417–453; 460–505; 571–593; 615–641 ) is a stable dimer in solution ( Figure 1A and Figure 1—figure supplement 2B ) . In addition , deletion of amino acids 322–342 between DBD and TD does not disrupt the dimeric state ( Figure 1—figure supplement 3 ) , but results in quite low expression levels in E . coli . For the experiments described below we have , therefore , used either TAp63αmin , wild type TAp63α or a slightly shortened version TAp63α ( 10–614 ) lacking unstructured sequences in the N- and C-terminus ( Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 13909 . 003Figure 1 . Mapping of structurally important regions within dimeric TAp63α . ( A ) Domain organization of TAp63α: transactivation domain ( TAD ) , DNA binding domain ( DBD ) , tetramerization domain ( TD ) , sterile alpha motif ( SAM ) domain , transactivation inhibitory domain ( TID ) . The minimal construct of TAp63α ( TAp63αmin ) lacks the first 9 and the last 27 amino acids as well as linker regions between TAD and DBD ( 64–119 ) , TD and SAM ( 417–453; 460–505 ) and SAM and TID ( 571–593 ) . Residues 454–459 were used as a linker between TD and SAM . ( B ) WB and corresponding bar diagram of pull-down experiments with constructs lacking either the DBD or the SAM domain using immobilized TID . Ratio of pull-down ( P ) and input ( I ) is shown relative to TAp63α ( 10–614 ) ( set to 1 ) . Pull-downs were performed in technical triplicates and error bars denote standard deviation . ( C , D , F , H ) TAp63α ( 10–614 ) constructs were expressed in rabbit reticulocyte lysate ( RRL ) and subjected to size exclusion chromatography ( SEC ) . SEC profiles were obtained by WB ( using an anti-myc antibody ) . ( C , D ) SEC profiles of TAp63α ( 10–614 ) ΔSAM ( C; pink ) and TAp63α ( 10–614 ) R ( DBD; sfGFP ) ( D; green ) compared with wild type ( TAp63α ( 10–614 ) , grey ) . R ( DBD; sfGFP ) indicates the replacement of the DBD by sfGFP . ( E ) Secondary structure prediction and mapping of structural motifs that stabilize the dimeric TAp63α . Cylinders and arrows represent α-helices and β-strands , respectively . Mutations ( color-coded and indicated by filled circles ) were introduced into TAp63α ( 10–614 ) on different faces of predicted secondary structure elements . The TAD is subdivided into TA1 ( residues 10–26 ) , TA2A ( 33–41 ) and TA2B ( 46–61 ) . The TA1 forms an α-helix and the F16/W20/L23 motif constitutes the single interaction motif of the TA1 . See Figure 1—figure supplement 5 for a thorough mapping of the TA1 . ( F ) The two faces of the β-stranded TA2B were mutated ( residues i , i+2 , i+4 to alanine ) . SEC profiles of I50A I52A M54A ( orange ) and K49A E51A S53A ( blue ) . See Figure 1—figure supplement 6 for a thorough mapping of the TA2 . SEC of I50A I52A M54A was performed in technical triplicates and error bars denote standard deviation . ( G ) Transcriptional activities of TAp63α TD mutants on the p21 promoter in SAOS2 cells . Triple and double alanine mutations were introduced on the central hydrophobic interface of the TD . Bar diagrams show n-fold induction relative to the activity of the empty vector . Experiments were performed in biological triplicates and error bars denote standard deviation . ( H ) Mutations were introduced on the two faces of the TID β-strand . SEC profile of R598A I600A ( red ) , E597A V599A D601A ( blue ) , V603A F605 L607A ( green ) and R604A R608A ( purple ) , Q609A I611A F613A ( green ) and R604A R608A ( purple ) . See Figure 1—figure supplement 7 for SEC profiles of other mutants . ( I ) Central hydrophobic interface of the dimeric TD , showing the important I378 L382 M385 motif . ( J ) Transactivation assay of TAp63α ( 10–614 ) mutants that appeared tetrameric in previous experiments ( see F , H and Figure 1—figure supplement 7 ) . Transcriptional activities on the p21 promoter in SAOS2 cells were normalized to the protein level ( determined by WB and referenced on GAPDH level ) . Experiments were performed in biological triplicates and error bars denote standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00310 . 7554/eLife . 13909 . 004Figure 1—figure supplement 1 . Domains behave as pearls on a string in tetrameric p63 . [15N , 1H]-TROSY spectra of 15N-labeled DBD-TD-SAM and individual domains at 303 K . The construct ranging from DBD to SAM is used to investigate the behavior of tetrameric p63 proteins , specifically referring to ΔNp63α and activated TAp63α . Despite its high molecular weight of 200 kDa a well-resolved spectrum of 15N-labeled DBD-TD-SAM was obtained . The spectra of DBD and SAM overlay well with the spectrum of DBD-TD-SAM . The spectrum of the TD can be recognized with lower confidence , likely owing to unfavorable relaxation properties in the center of the protein . The ability to obtain such a spectrum already proofs that the domains do not form a globular structure but that they tumble independent of each other in solution . Titrations of the individual domains ( DBD , TD and SAM ) to each other also did not show any interaction ( data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00410 . 7554/eLife . 13909 . 005Figure 1—figure supplement 2 . SEC-MALS proves the dimeric nature of TAp63αmin . ( A ) Domain organization of TAp63α: transactivation domain ( TAD ) , DNA binding domain ( DBD ) , tetramerization domain ( TD ) , sterile alpha motif ( SAM ) , transactivation inhibitory domain ( TID ) . TAp63α ( 10–614 ) lacks the first 9 and the last 27 amino acids . In addition to these N- and C-terminal truncations the minimal construct of TAp63α ( TAp63αmin ) lacks linker regions between TAD and DBD ( 64–119 ) , TD and SAM ( 417–453; 460–505 ) and SAM and TID ( 571–593 ) . Residues 454–459 were used as a linker between TD and SAM . Identical to Figure 1A . ( B ) SEC-MALS of TAp63αmin . Change of molecular weight ( Mw ) is shown in red . Marked area in green was used to calculate the Mw . ( C , D ) SEC profiles of RRL ( rabbit reticulocyte lysate ) expressed TAp63α and TAp63α ( 10–614 ) , obtained by western blots ( using an anti-myc antibody ) of eluted fractions and subsequent signal integration , are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00510 . 7554/eLife . 13909 . 006Figure 1—figure supplement 3 . Deletion of 322–342 does not disrupt the dimeric state . SEC profiles of RRL expressed TAp63α ( 10–614 ) constructs Δ ( K322-N342 ) and Δ ( K322-N352 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00610 . 7554/eLife . 13909 . 007Figure 1—figure supplement 4 . DBD is not essential to retain the dimeric state . ( A ) Constructs were designed based on TAp63α ( 10–614 ) . R ( DBD; sfGFP ) indicates the replacement of the DBD by sfGFP . All constructs were expressed in rabbit reticulocyte lysate ( RRL ) and subjected to size exclusion chromatography ( SEC ) on a Superose 6 3 . 2/300 column . SEC profiles were obtained by western blots ( using an anti-myc antibody ) of eluted fractions and subsequent signal integration . ( B ) SEC profile of TAp63α ( 10–614 ) R ( DBD; sfGFP ) ( green ) and wild type ( TAp63α ( 10–614 ) , grey ) . R ( DBD; sfGFP ) indicates the replacement of the DBD by sfGFP . Identical to Figure 1D . ( C ) SEC profile of TAp63α ( 10–614 ) R ( DBD; sfGFP ) F16A W20A L23A ( green ) and TAp63α ( 10–614 ) F16A W20A L23A ( grey ) . ( D ) SEC profile of TAp63α ( 10–614 ) R ( DBD; sfGFP ) I50A I52A M54A ( green ) and TAp63α ( 10–614 ) I50A I52A M54A ( grey ) . ( E ) SEC profile of TAp63α ( 10–614 ) R ( DBD; sfGFP ) F605A T606A L607A ( green ) and TAp63α ( 10–614 ) F605A T606A L607A ( grey ) . ( F ) SEC profiles of TAp63α R ( DBD; sfGFP ) ( green ) and TAp63α R ( DBD; MBP ) ( dark blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00710 . 7554/eLife . 13909 . 008Figure 1—figure supplement 5 . The TA1 forms an α-helix . ( A ) Secondary structure prediction and mapping of structural motifs in the TAD that stabilize the dimeric TAp63α . Cylinders and arrows represent α-helices and β-strands , respectively . Mutations ( color-coded and indicated by filled circles ) were introduced into TAp63α ( 10–614 ) on different faces of predicted secondary structure elements . The TAD is subdivided into TA1 ( residues 10–26 ) , TA2A ( 33–41 ) and TA2B ( 46–61 ) . ( B ) The four faces of the α-helical TA1 were mutated ( residues i , i+4 , i+7 to alanine ) . SEC profiles of E14A H18A D21A ( blue ) , V15A I19A F22A ( red ) , F16A W20A L23A ( green ) and Q17A D21A E24A ( purple ) . Only the F16A W20A L23A mutation disrupts the dimeric state . Therefore , the F16 W20 L23 motif constitutes the single interaction motif of the helical TA1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00810 . 7554/eLife . 13909 . 009Figure 1—figure supplement 6 . Mapping of structural motifs in the TA2 . ( A ) Secondary structure prediction and mapping of structural motifs in the TAD that stabilize the dimeric TAp63α . Cylinders and arrows represent α-helices and β-strands , respectively . Mutations ( color-coded and indicated by filled circles ) were introduced into TAp63α ( 10–614 ) on different faces of predicted secondary structure elements . The TAD is subdivided into TA1 ( residues 10–26 ) , TA2A ( 33–41 ) and TA2B ( 46–61 ) . ( B ) The two faces of the β-stranded TA2A were mutated ( residues i , i+2 , i+4 to alanine ) . SEC profiles of I33A L35A F37A ( yellow ) and D34A N36A V38A ( brown ) . SEC of I33A L35A F37A was performed in technical triplicates and error bars denote standard deviation . ( C , D ) The two faces of the β-stranded TA2B were mutated ( residues i , i+2 , i+4 to alanine ) . ( C ) SEC profiles of K49A E51A S53A ( blue ) and I50A I52A M54A ( orange ) . SEC of I50A I52A M54A was performed in technical triplicates and error bars denote standard deviation . Identical to Figure 1F . ( D ) SEC profiles of C56A R58A Q60A ( green ) and I57A M59A D61A ( purple ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 00910 . 7554/eLife . 13909 . 010Figure 1—figure supplement 7 . Mapping of structural motifs in the TID . ( A ) Secondary structure prediction and mapping of structural motifs in the TID that stabilize the dimeric TAp63α . The TID is predicted to form a β-strand . Mutations ( color-coded and indicated by filled circles ) were introduced into TAp63α ( 10–614 ) on different faces of the β-strand . Mutations were performed to evaluate the contribution of the single amino acid mutants to the effect shown for the double mutations R598A I600A and R604A R608A ( Figure 1H ) . In addition , the C-terminal part of the TID is mapped . ( B ) SEC profile of I600A ( red ) and R598A ( blue ) . ( C ) SEC profile of R604A ( green ) and R608A ( purple ) . ( D ) SEC profile of Q609A I611A F613A ( black ) and T610A S612A ( cyan ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01010 . 7554/eLife . 13909 . 011Figure 1—figure supplement 8 . Mapping of structural motifs in the TD by measurement of transcriptional activities . ( A ) Transcriptional activities of TAp63α TD mutants on the p21 promoter in SAOS2 cells . Triple and double alanine mutations were introduced on the surface of the two helices . Experiments were performed in triplicates . Bar diagrams show n-fold p21 promoter induction relative to the activity of the empty vector control . Mutations M374A I378A L382A and L382A M385A L388A suggest that the hydrophobic interface starting from the center to the end of the first α-helix is important for the stabilization of dimeric TAp63α . Further detailed experiments are shown in Figure 1G . ( B ) SEC profiles of TAp63α mutants M374A I378A L382A ( green ) I378A L382A M385A ( red ) and L382A M385A L388A ( blue ) are identical to wild type TAp63α although they are transcriptionally active and should therefore exhibit a more open conformation . Since the tetrameric interface is mutated , the mutants cannot form tetramer but only dimers . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01110 . 7554/eLife . 13909 . 012Figure 1—figure supplement 9 . Validation of structural motifs by pull-down with GST-TID . ( A ) Secondary structure prediction and mapping of structural motifs that stabilize the dimeric TAp63α . Cylinders and arrows represent α-helices and β-strands , respectively . Mutations ( color-coded and indicated by filled circles ) were introduced into TAp63α ( 10–614 ) on different faces of predicted secondary structure elements . Transcriptional activities of identical mutations were investigated in a separated experiment ( see Figure 1J ) . ( B , C ) Western blot ( B ) and corresponding bar diagram ( C ) of pull-down experiments ( using immobilized TID ) with TAp63α ( 10–614 ) mutants that appeared tetrameric in previous experiments and the I33 L35A F37A mutant . ( B ) Western blots used for quantification of pull-down with GST-TID . Experiments were performed in technical triplicates . ( C ) Quotient of pull-down ( P ) and input ( I ) is shown relative to TAp63α ( 10–614 ) ( set to 1 ) . Error bars denote standard deviation . All mutants showed a more than 2-fold pull-down compared to TAp63α ( 10–614 ) which indicates that they exist in an open conformation , exposing hydrophobic patches . Surprisingly the I33A L35A F37A mutant exhibited the highest pull-down , indicating that I33 , L35 , and F37 do indeed play a structural role inside TAp63α , likely in forming a beta-strand as predicted . Error bars denote standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01210 . 7554/eLife . 13909 . 013Figure 1—figure supplement 10 . Transcriptional activities of tetrameric TAp63γ mutants . ( A ) Motifs in the TAD of TAp63γ are tested for their importance in transcriptional activation . ( B ) Transcriptional activities of human TAp63γ mutants on the p21 promoter in SAOS2 cells . Bar diagrams show n-fold p21 promoter induction relative to the activity of the empty vector control . Experiments were performed in biological triplicates and error bars denote standard deviation . Means were compared using Student’s t-test . ( C ) TAp63γ forms tetramers ( expected molecular weight: 204 kDa ) . TAp63γ was expressed in rabbit reticulocyte lysate ( RRL ) and subjected to size exclusion chromatography ( SEC ) on a Superose 6 3 . 2/300 column . SEC profile of TAp63γ was obtained by western blot ( using an anti-myc antibody ) of eluted fractions and subsequent signal integration . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 013 In contrast to the TD , an involvement of the SAM domain and DBD in the formation of the closed dimeric state is not immediately obvious . To investigate whether these domains participate in the stabilization of the closed conformation we deleted each domain separately in TAp63α ( 10–614 ) and performed pull-down experiments with GST-TID . Interestingly , deletion of the SAM domain did not show any significant pull-down and size exclusion chromatography confirmed the formation of closed dimers ( Figure 1B and C ) . On the contrary , deletion of the DBD resulted in a strong pull-down signal suggesting an open state ( Figure 1B ) . Initially we expected the DBD to participate in essential domain-domain contacts that stabilize the closed conformation and therefore conducted an extensive mutagenesis screen of surface residues of the DBD ( Supplementary file 1 ) . However , none of the mutants formed tetramers making this hypothesis unlikely . Alternatively , the DBD may be important for geometric reasons , acting as a spacer between TAD and TD . To test this hypothesis , we replaced the DBD by superfolder GFP ( sfGFP ) which is very stable and of similar size as the DBD . SEC analysis of this chimeric protein expressed in rabbit reticulocyte lysate ( RRL ) suggested that it adopts a closed dimeric conformation ( Figure 1D ) . Moreover , mutations F16A W20A L23A within the TAD and F605A T606A L608A within the TID resulted in the formation of a tetrameric state similar to experiments with wild type TAp63α ( Straub et al . , 2010 ) ( Figure 1—figure supplement 4C and E ) . Similarly , replacement of the DBD by MBP enables the formation of a closed dimeric state ( Figure 1—figure supplement 4F ) . These results suggest that the DBD does not participate in essential domain-domain interactions necessary to form the dimeric state and that the closed dimeric state of TAp63α is formed by interaction of the N-terminal TAD , the central TD and the C-terminal TID . Nonetheless , constructs that only contain these three domains did not form dimers but aggregated , suggesting that the DBD or a domain of similar size is necessary for structural reasons or for the folding process . To build a first model of the closed state we used secondary structure prediction programs to identify potential secondary structure elements within the TAD and TID and alanine scanning in combination with SEC analysis to experimentally verify these predictions . The theoretical analysis predicted the existence of an α-helix in the TA1 region , two β-strands in the TA2A and TA2B regions of the TAD and a β-strand in the TID ( Figure 1E ) . Alanine scanning of the TA1 confirmed that only mutations of F16 , W20 and L23 that have previously been identified as crucial for binding of the TA1 to the TD ( Deutsch et al . , 2011 ) , disrupted the closed conformation while mutations on the three remaining faces of the hypothetical helix had no effect ( Figure 1—figure supplement 5 ) . To test the existence of the various β-sheets we mutated all amino acids on one side of each predicted β-strand to alanine ( i , i+2 , i+4 ) . While mutations on both faces of the presumed first beta-strand ( TA2A ) did not affect the oligomeric state ( Figure 1—figure supplement 6B ) , the mutations I50A I52A M54A located on one face of the predicted TA2B β-strand disrupted the dimeric state ( Figure 1F ) . Alanine scanning of the TID showed that mutations on both sides of the presumed β-strand disrupt the dimeric state ( Figure 1H and Figure 1—figure supplement 7B ) . Stabilizing the dimeric state is most likely achieved by blocking the tetramerization interface of the TD and we also used alanine scanning of the TD to identify essential residues ( Figure 1—figure supplement 8 ) . Since mutations in the tetramerization interface that destabilize the dimeric state most likely also inhibit the formation of the tetramer , we did not use SEC analysis . Previously , we have shown that an open dimeric state is transcriptionally more active than the closed dimeric state ( Deutsch et al . , 2011 ) . Mutating the hydrophobic amino acids I378 , L382 and M385 alongside the second half of the α-helix of the TD led to high transcriptional activity as expected for an open conformation ( Figure 1G and I , Figure 1—figure supplement 8 ) . We also used the measurement of the transcriptional activity as well as pull-down experiments with GST-TID to validate the results of our SEC analysis with the different alanine mutants ( Figure 1J and Figure 1—figure supplement 9 ) . As expected , all mutants that behaved like open and tetrameric conformations showed high transcriptional activity . The only exception was the F16A W20A L23A mutant since these mutations compromise the function of the TAD ( Figure 1—figure supplement 9 ) . The experiments described above support the prediction that TA2B and TID form regular secondary structure elements , most likely β-strands . In the closed dimer , two TID and two TA2B sequences must be involved in the stabilization of the closed state . For symmetry reasons , the β-strands probably adopt an antiparallel orientation . Based on the results of the alanine scanning experiments we speculated that the two TID strands form the inner pair since mutations on both faces of the predicted β-sheet show strong effects . Further , we propose that the two TA2B strands form the two outer strands of a four stranded anti-parallel β-sheet which might be further extended by β-strands contributed by the TA2A segment . Such an arrangement would create one hydrophobic surface formed by I50/I52/M54 of TA2B and V603/F605/L607 of TID and a hydrophilic surface with residues E51/D55 of TA2B and R604/R608 of TID . The arrangement shown in Figure 2B brings charged amino acids on neighboring strands in close proximity , making it possible to test this hypothetical model by charge change and charge swap mutagenesis . Exchanging R604 and R608 in the TID to glutamic acids disrupted the dimeric state ( Figure 2C ) . In our model these mutants created in combination with the negative charges on the TA2B strands a cluster of negatively charged amino acids that destabilized the dimer . Additional charge reversal of E51R and D55R in TA2B resulted in the formation of a stable dimer . Similarly , the R595E and R598E mutants are open tetramers and the additional charge reversal of D61R , D63R in TA2B rescued the dimer ( Figure 2D ) . To refine our model and to identify the register of the proposed β-strands we used further pairwise charge swap mutations . The results of these experiments that all support our structural model are summarized in Figure 2—figure supplement 1 . Since the predicted β-sheet has one hydrophobic face and the interface used by the TD to form tetramers is also hydrophobic , we propose that the β-sheet covers the tetramerization interface of the TD , thus inhibiting the formation of tetramers ( Figure 3B and C ) . In addition , the TA1 helix binds to the TD as well , further stabilizing the closed and compact conformation . 10 . 7554/eLife . 13909 . 014Figure 2 . TA2B and TID form an anti-parallel β-sheet with a polar and a hydrophobic face . ( A ) Domain organization of TAp63α and secondary structure elements of TAD and TID . ( B ) Proposed interaction of TA2 and TID through β-sheet formation . This interaction is thought to be stabilized by hydrophobic amino acids clustered on one face of the β-sheet ( bottom ) and electrostatic interactions between charged amino acids on the other face ( top ) . Extensive charge swap experiments ( see Figure 2—figure supplement 1 ) revealed interactions between TA2B and TID . Interactions are depicted in green . ( C , D ) Introduction of negative charges in the TID and charge swaps between TID and TA2B show interaction via β-sheet formation . ( C ) SEC profiles of TAp63α R604E R608E ( orange ) and the charge swap mutant TAp63α E51R D55R R604E R608E ( blue ) . ( D ) SEC profiles of TAp63α R595E R598E ( orange ) and the charge swap mutant TAp63α D61R D63R R595E R598E ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01410 . 7554/eLife . 13909 . 015Figure 2—figure supplement 1 . TA2B and TID form an anti-parallel β-sheet . ( A ) Secondary structure prediction of TAD and TID . TA2B and TID are predicted to form β-strands . ( B ) On validation , charges were swapped between presumably distant amino acids . SEC profiles of TAp63α E51R D55R R595D R598D ( orange ) and TAp63α D61R D63R R604D R608E ( blue ) . ( C ) Charge swaps are used to reveal interactions between charged amino acids across the presumed β-sheet formed by TA2B and TID . A destabilizing mutation ( arginine to aspartate or glutamate ) is introduced into TAp63α . Solely mutation R604D/E leads to the formation of tetramers . Any other arginine in TA2B or TID mutated to aspartate or glutamate does not change the oligomeric state . In order to break the interaction between TA2B and TID , a second destabilizing mutation is introduced that disrupts the dimeric state and leads to the formation of tetramers . Additional compensating mutations in the double charge swap recover the dimeric state . To prove a single interaction between two differently charged amino acids , they are swapped in presence of a destabilizing mutation ( R608E ) resulting in a triple mutant . Similar SEC profiles of the destabilizing and the triple mutation prove the interaction between the swapped amino acids . ( D ) Extensive charge swap experiments ( shown in E-I ) revealed interactions ( shown in green ) between TA2B and TID . ( E ) SEC profiles of a single charge swap between R604 and D55 ( blue ) and the destabilizing mutation R604D show a direct interaction between D55 and R604 . ( F , G , H , I , J ) SEC profiles of double arginine mutants ( top , orange ) and double charge swaps ( top , blue ) . SEC profiles of triple mutants ( bottom , blue ) and the R608E mutant ( bottom , orange ) should be identical to verify the interaction shown in bold ( top ) . For comparison identical western blots / SEC profiles of TAp63α R608E are shown on bottom . ( F ) SEC profiles of TAp63α R58D R608E ( top , orange ) , charge swap TAp63α E51R R58D D601R R608E ( top , blue ) , triple mutant TAp63α R58D D601R R608E ( bottom , blue ) and mutant TAp63α R608E ( bottom , orange ) . ( G ) SEC profiles of TAp63α R598D R608E ( top , orange ) and charge swap TAp63α E51R D61R R598D R608E ( top , blue ) , triple mutant TAp63α D61R R598D R608E ( bottom , blue ) and mutant TAp63α R608E ( bottom , orange ) . ( H ) SEC profiles of TAp63α R595D R608E ( top , orange ) and charge swap TAp63α E51R D61R R595D R608E ( top , blue ) , triple mutant TAp63α D61R R595D R608E ( bottom , blue ) and mutant TAp63α R608E ( bottom , orange ) . ( I ) SEC profiles of TAp63α R58E R608E ( top , orange ) and charge swap TAp63α E51R R58E E597R R608E ( top , blue ) , triple mutant TAp63α R58E E597R R608E ( bottom , blue ) and mutant TAp63α R608E ( bottom , orange ) . ( J ) SEC profiles of TAp63α R598D R608E ( left , orange , identical blot/profile as shown in G ) and charge swap TAp63α E51R D63R R598D R608E ( left , blue ) , triple mutant TAp63α D63R R598D R608E ( right , blue ) and mutant TAp63α R608E ( right , orange ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01510 . 7554/eLife . 13909 . 016Figure 3 . Model of the closed dimeric conformation of TAp63α . ( A ) Domain organization of TAp63α . All domains and structural elements are color coded . ( B ) The TD of p63 forms a dimer of dimers ( colored in dark and light grey ) . Its two tetrameric interfaces ( in light blue and rose ) must be blocked in the inactive dimer to inhibit tetramerization . The TA1 was shown to bind to the upper interface ( in rose ) ( Deutsch et al . , 2011 ) . The I378 L382 M385 motif in the central interface ( in light blue ) must be covered by hydrophobic amino acids . The hydrophobic interface of the proposed 6-stranded β-sheet is expected to cover this central tetrameric interface of the TD . ( C ) Model of the intramolecular interactions between TAD , TD and TID . The angles between structural elements are speculative . The TD was placed on top of the TA2/TID β-sheet so that the hydrophobic amino acids mask each other . The second helix of the TD is not modelled . ( D ) Pair distribution function P ( r ) from inline SEC-SAXS ( small-angle X-ray scattering ) data of TAp63αmin . Derived function transformed smoothly and appears to indicate globular central part with short extensional component . ( E ) Average ab-initio SAXS envelopes of TAp63αmin without ( left ) and with ( right ) P2 symmetry , calculated using DAMMIF ( Franke and Svergun , 2009 ) . The similar shape suggests the presence of C2 symmetry in TAp63αmin . Envelopes were filtered and averaged using DAMFILT and were obtained from inline SEC-SAXS . ( F ) Simulated annealing multiphase model from simultaneous curve fits to wild type TAp63αmin and λ-cro-TAp63αmin ( N-terminal fusion ) . Models constructed using MONSA allowing co-refinement of ab-initio models simultaneously . Blue segments give density differences derivative when refined against the native dataset . ( G ) Localization of the N-terminus . Multiphase fits to data sets , wild type TAp63αmin in green and λ-cro-TAp63αmin in blue . ( H ) WB and corresponding bar diagram of the pull-down experiments with ΔNp63α , TAp63α and ΔNp63α R279H from RRL using either immobilized GST or GST-ASPP2 fusion . WB signal for input ( IP ) and pull-down ( PD ) are shown . The pull-down efficiency of ΔNp63α was set to 100% . Pull-downs were performed in technical triplicates and error bars display the standard deviation . ( I ) Structure of the human p63 DBD alone , bound to DNA and a model of the p63 DBD bound to ASPP2 based on the co-crystal structure between the p53 DBD and ASPP2 . ( J ) TAD , DBD , TD and TID are placed manually inside the P2 calculated average SAXS envelope . The DBDs are likely positioned at the outside of the molecule , leaving the center to be occupied by TAD , TD and TID . The SAM domain is not modelled . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 016 The mutational analysis described above predicted the formation of a compact structure with C2 symmetry . To verify this prediction , we performed SAXS measurements with TAp63αmin . To identify the localization of the N-termini we also collected SAXS data on a construct containing mutated λ-cro ( Q27P , A29S , K32Q ) at its N-terminus . Low resolution models derived from unbiased evaluation of the SAXS data showed indeed a C2 symmetry ( Figure 3E ) with the N-termini located in the center of the molecule ( Figure 3G ) . Based on these results and the volumes of the individual domains we propose that the DBDs are positioned at the outside while the complex formed by the TAD , TD and TID builds the center of the molecule ( Figure 3J ) . In this model the SAM domain is also located in the center where the molecule showed the largest volume . To obtain additional information on the orientation of the DBD we performed binding studies with the Ankyrin Repeat and SH3 domain of the protein ASPP2 . This protein is known to bind to the DNA binding interface of the DBD ( Figure 3I ) . In pull-down experiments we were not able to detect interaction of TAp63α with ASPP2 while the open and tetrameric ΔNp63α isoform showed strong interaction ( Figure 3H ) . This observation suggests that the DNA binding interface of the DBD is not freely accessible but points towards the core of the molecule . Activation of TAp63α entails breaking of the interactions described above to expose the tetramerization interface leading to the formation of active tetramers . In oocytes this transition is triggered by phosphorylation . In principle phosphorylation could provide a new interface contributing interactions that stabilize the tetrameric state , making it thermodynamically more stable while the dimeric state would be thermodynamically favored in the absence of phosphorylation . However , the observation that dephosphorylation of the open tetrameric state using λ-phosphatase does not result in converting TAp63α back to a dimer argues against this model ( Deutsch et al . , 2011 ) . An alternative explanation would be that the tetrameric state is always the thermodynamically most stable one and the dimeric state is a kinetically trapped conformation . Phosphorylation would then function as a trigger to overcome a kinetic barrier and convert p63 into the thermodynamically preferred tetramer . Such spring-loaded mechanisms have been observed for example in the activation of influenza hemagglutinin ( Carr et al . , 1997; Carr and Kim , 1993 ) . Characteristic for this type of activation mechanism is that perturbing the kinetically trapped conformation by moderate amounts of denaturants , changes in pH or an increase in temperature initiates the transition to the thermodynamically more stable conformation even without the natural trigger . Since the stability towards chemical denaturants of the three folded domains of TAp63α is quite high ( Klein et al . , 2001; Sathyamurthy et al . , 2011 ) ( Figure 4—figure supplement 1 ) we hypothesized that using low to moderate amounts of urea might disrupt the inhibitory structure , thus triggering the formation of the tetramer without affecting the folding of the DBD , the SAM or the TD . To investigate if activation of TAp63α follows a spring-loaded mechanism we equilibrated a SEC column with different concentrations of urea , incubated TAp63αmin in buffer containing the same urea concentration and analyzed the percentage of dimer and tetramer . Figure 4A shows that a concentration of 1 . 75 M urea leads to an approximately 1:1 ratio of dimer and tetramer and at concentrations above 3 M no dimer was detected . Higher urea concentrations resulted in further shifts on the SEC column probably representing partially denatured conformations ( Figure 4—figure supplement 2 ) . To validate the data we performed SEC-MALS measurements at concentrations of 2 M and 2 . 5 M urea ( Figure 4F and G ) . The first SEC peak had a mass of 197 . 9 ± 12 . 7 kDa ( at 2 . 5 M urea ) and the second peak a mass of 96 . 3 ± 6 . 4 kDa ( at 2 M urea ) , consistent with the first one representing a tetrameric ( 202 . 8 kDa ) and the second one a dimeric ( 101 . 4 kDa ) conformation . 10 . 7554/eLife . 13909 . 017Figure 4 . The closed dimeric conformation of TAp63α constitutes a kinetically trapped state . ( A ) TAp63αmin samples were incubated for 1 hr at different urea concentrations and subjected to size exclusion chromatography ( SEC ) at corresponding urea concentrations . ( B ) TAp63αmin samples were incubated in 1 . 75 M urea and injected into a Superose 6 3 . 2/300 column equilibrated with 1 . 75 M urea at different time points . ( C ) SEC profiles of TAp63αmin injected after incubation for 50 min in 1 . 75 M urea . Fractions of tetrameric and dimeric protein are highlighted in orange and blue , respectively . ( D , E ) SEC profiles of reinjected tetrameric ( E ) and dimeric ( D ) fractions ( originating from SEC shown in C ) after dialysis to 0 M urea for 13 hr . ( F , G ) SEC-MALS of TAp63αmin at different urea concentrations to proof the tetrameric nature of the early eluting peak in A . a , t and d denote aggregate , tetramer and dimer respectively . Colored areas where used to calculate the mean molecular weight and standard deviation . ( F ) SEC-MALS of TAp63αmin in 2 M urea ( preincubated in 2 M urea for 14 min at RT ) . ( G ) SEC-MALS of TAp63αmin in 2 . 5 M urea ( preincubated in 2 . 5 M urea for 25 min at RT ) . ( H ) WB and corresponding bar diagram of pull-down experiments with ΔNp63α , TAp63α R604E R608E and TAp63α incubated either during or after expression in RRL at 30°C for 1 . 5 hr with His6-tagged p73 TD or a mutant that is not able to form hetero-tetramers ( His6-p73 TDHOMO ) . Pull-down is achieved by hetero-tetramerization of His6-tagged p73 TD with specified p63α constructs . Quotient of pull-down ( P ) and input ( I ) is shown relative to TAp63α incubated after expression with p73 TD ( set to 1 ) . Pulldowns were performed in technical triplicates and error bars denote standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01710 . 7554/eLife . 13909 . 018Figure 4—figure supplement 1 . Urea treatment of p63 structured domains . To prove that moderate concentrations of urea ( up to 3 M ) do not unfold domains inside TAp63αmin , individual domains were incubated for 1 hr at different urea concentrations and subjected to size exclusion chromatography ( SEC ) on a Superdex 75 3 . 2/300 column at corresponding urea concentrations . SEC profiles of TD ( A ) , SAM ( B ) and DBD ( C ) at urea concentrations of 0 M , 1 . 75 M and 3 M urea . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 01810 . 7554/eLife . 13909 . 019Figure 4—figure supplement 2 . Urea unfolding experiments with TAp63αmin . ( A , B ) TAp63αmin samples were incubated for 1 hr at different urea concentrations and subjected to size exclusion chromatography ( SEC ) on a Superose 6 3 . 2/300 column at corresponding urea concentrations . As in Figure 4A but at higher urea concentrations . At a urea concentration of 4 M urea the tetramers seem to unfold as seen in the partial shift to higher elution volumes . ( C ) TAp63αmin was incubated in 1 . 75 M urea for 24 hr and injected onto a Superose 6 3 . 2/300 column equilibrated in 1 . 75 M urea . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 019 If the interpretation of the spring-loaded activation is correct , removal of urea would not allow the formation of a p63 dimer . To test this hypothesis , we separated the dimer and the tetramer fraction at a urea concentration of 1 . 75 M on the SEC column ( Figure 4C ) and dialyzed both fractions against buffer without urea . Re-analysis of these samples by SEC revealed that the dimeric fraction remained dimeric ( Figure 4D ) and the tetrameric fraction tetrameric with a tendency to aggregate ( Figure 4E ) . These experiments strongly suggest that the dimeric state of TAp63α is a kinetically trapped conformation that is activated by a spring-loaded mechanism . A spring-loaded activation requires that the protein is trapped in a high energy state during protein synthesis . From p53 it is known that this protein forms dimers co-translationally ( Nicholls et al . , 2002 ) , which in the case of TAp63α would enable the protein to fold into its closed conformation . To probe this hypothesis , we expressed TAp63α in RRL in the presence or absence of a high concentration ( 20 µM ) of the isolated TD of p73 . The rationale behind this experiment was that a high concentration of a domain that can interact with the TD of TAp63α during the translation would result in the formation of open tetramers . The TD of p73 was used since the isolated p63 and p73 TDs form hetero-tetramers that are thermodynamically even more stable than homo-tetramers ( Coutandin et al . , 2009 ) . As a control we stopped the translation of TAp63α in RRL by adding cycloheximide ( CHX ) and then added the p73 TD to the same concentration as before and incubated for the same amount of time . Interaction between TAp63α and the p73 TD was monitored by pull-down experiments via the His-tag of the p73 TD . As shown in Figure 4H , expression in the presence of the p73 TD resulted in a strong pull-down while incubation post-translationally showed virtually no interaction with the p73 TD , even at elevated temperatures of 37oC . Replacing TAp63α in these experiments with open and tetrameric ΔNp63α or a tetrameric mutant TAp63α R604E R608E resulted in strong pull-downs both in the co-translational as well as in the post-translational setup . Performing the same experiments with a mutated TD that is not capable of forming hetero-tetramers showed no interaction . These results suggested that the kinetically trapped state of TAp63α is formed during or immediately after protein synthesis . Oocytes survive the high concentration of TAp63α only when the inactivation mechanism is very effective . However , thermodynamics predicts that the closed conformation is always in equilibrium with more open conformations in which the inhibitory network of the TAD , TD and TID is at least partially broken . If during this partial unfolding no thermodynamically more stable tetramer is formed the dimer might be able to refold in its closed conformation . To obtain an estimation of the rate of unfolding of the TID and of the TAD we introduced TEV protease cleavage sites either C-terminal to the TAD or N-terminal to the TID . The rationale of this experiment was that after proteolytic cleavage the cleaved peptides ( either the TAD or the TID ) would diffuse away as soon as the p63 adopts an open conformation , therefore not allowing the protein to refold into its compact dimeric state and forcing it to form open tetramers . From this experiment the off rate of the corresponding domain can be estimated and thus the overall stability of the inhibitory lock mechanism . We incubated RRL expressed TAp63α with TEV protease for 15 min at 37°C which was sufficient to obtain close to 100% cleavage ( Figure 5 ) . The cleaved protein was then analyzed either immediately via SEC or further incubated for up to 12 hr at 37oC . Interestingly , cleavage near the TAD leads to the immediate formation of tetramers ( Figure 5G ) . Unlike the TAD , the TID was bound with remarkable stability and cleaved p63 showed no tendency to assemble into tetramers even after long incubation times ( Figure 5I ) . These results demonstrated that the N-terminus is the least stable part involved in keeping TAp63α dimeric and that its off rate determines the overall stability of the inhibited conformation . In addition , this interpretation further supports our model assuming that the TID forms the core of the central β-sheet . 10 . 7554/eLife . 13909 . 020Figure 5 . Unlike TID , secession of TAD induces the transformation of dimeric TAp63α to tetramers . ( A ) A cleavage site is introduced C-terminal to the TAD ( between residues 66 and 67 ) allowing its secession by TEV protease cleavage . For comparison a TAp63α construct is created that lacks the TAD ( TAp63α Δ ( 1–66 ) ) and resembles the cleavage product . ( B ) A cleavage site is introduced N-terminal to the TID ( between residues 591 and 592 ) allowing its secession by TEV protease cleavage . For comparison a TAp63α construct is created that lacks the TID ( TAp63α Δ ( 593–641 ) ) and resembles the cleavage product . ( C ) Schematic depiction of TAp63α ( 66-TEVsite-67 ) and secession of TAD by TEV protease cleavage . ( D ) Schematic depiction of TAp63α ( 591-TEVsite-592 ) and secession of TID by TEV protease cleavage . ( E ) Secession of TAD and TID from TAp63α derivatives using TEV protease . Cycloheximide ( CHX ) and TEV protease were added to the RRL expressed TAp63α derivative at 37°C and samples were taken after indicated time points and analyzed by western blotting . Both constructs are cleaved nearly completely within approximately 10 min . ( F , G , H , I ) TAp63α constructs were expressed in reticulocyte lysate ( RRL ) , treated with CHX and optionally with TEV protease ( G , I ) at 37°C for denoted time , cooled to 4°C and subjected to SEC . SEC profiles were obtained by WB . ( F ) SEC profiles of TAp63α ( 66-TEVsite-67 ) and of TAp63α Δ ( 1–66 ) . ( G ) SEC profiles of TAp63α ( 66-TEVsite-67 ) after treatment with CHX and TEV protease for either 15 min or 1 hr at 37°C . ( H ) SEC profiles of TAp63α ( 591-TEVsite-592 ) and of TAp63α Δ ( 593–641 ) . ( I ) SEC profiles of TAp63α ( 591-TEVsite-592 ) after treatment with CHX and TEV protease for either 4 or 12 hr at 37°C . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 020 The experiments described above have demonstrated that TAp63α exists in a kinetically trapped state , poised to become activated upon the detection of DNA damage . Such a mechanism allows the cell to build an apoptotic switch with a sharp transition between survival and cell death . Indeed , measurements of the dose dependence of oocyte death have shown such a sharp transition with fewer than 10 double strand breaks per cell leading to oocyte death . To make such a system efficient the cell would need to be able to activate TAp63α fast which is best achieved when the activation machinery , i . e . the kinases required are already present and do not have to be expressed first . To investigate if oocytes have established such a pre-existing machinery , we harvested ovaries from eight day old mice and γ-irradiated them with or without prior incubation with cycloheximide . Activation of TAp63α was followed by native geleletrophoresis . Addition of cycloheximide did neither prevent phosphorylation ( Figure 6A ) nor the formation of a tetrameric state ( Figure 6B–D ) , suggesting that the kinases involved in detecting DNA damage and activating TAp63α are already present in resting oocytes . As a control to verify the effectiveness of the translation inhibitor cycloheximide we investigated the level of polyubiquitination ( Figure 6—figure supplement 1A ) . Adding a proteasome inhibitor results in a strong accumulation of polyubiquitinated proteins that is suppressed by the addition of cycloheximide , as previously shown ( Mimnaugh et al . , 2004 ) . 10 . 7554/eLife . 13909 . 021Figure 6 . The cellular machinery for TAp63α activation in murine oocytes is always present and ready to act upon genotoxic insults . ( A ) WB of CHX treatment of nonirradiated ( NIRR ) and γ-irradiated ( IRR ) murine ovary samples . The signals of p63 , the oocyte marker Msy2 and β-actin are displayed for each time point after NIRR/IRR . The asterisk marks phosphorylated p63 . ( B ) WB of SDS-PAGE loaded with the ovary samples of the Native PAGE in ( C ) . The asterisk marks phosphorylated p63 . ( C ) WB of Native PAGE from ( un- ) treated and either NIRR or IRR murine ovaries . The p63 signal in the range from 20 kDa to 1 , 236 kDa is shown . ( D ) Intensity projection of the Native PAGE p63 signal from ( C ) . The molecular weight range of the p63 dimer and tetramer is colored in green and red , respectively . ( E ) Quantitative Real-Time PCR of isolated murine oocytes . The bar diagram shows the fold induction of p21 , Puma , Mdm2 and Msy2 mRNA after γ-irradiation . Error bars show the standard deviation of the biological duplicates . Brackets above the bars display the p-test results showing no significance ( n . s . ) between untreated and CHX treated oocytes for all targets . ( F ) Inhibition of Chk2 suppresses the DNA-damage induced phosphorylation of TAp63α in γ-irradiated ovaries . Chk2 inhibitor II at concentrations of 5 and 25 µM was added 2 hr before irradiation with 1 . 5 Gy . Ovaries were harvested 4 hr after irradiation and analyzed by SDS PAGE and Western Blot . Activated TAp63α gets degraded fast while preventing activation via inhibition of Chk2 preserves the original cellular concentration . ( G ) Native PAGE analysis of the same samples used as in ( F ) . Inhibition of Chk2 prevents tetramerization and keeps TAp63α in a closed and dimeric state . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 02110 . 7554/eLife . 13909 . 022Figure 6—figure supplement 1 . p63 is responsible for inducing apoptosis in oocytes . ( A ) Verification of CHX activity in ovary culture . WB of CHX and MG132 treatment of mouse ovaries . After overnight culture with either DMSO or CHX , ovaries were incubated with DMSO , CHX , MG132 or a combination of the latter two for additional 8 hr . The signal of ubiquitin ( mono- and poly-ubiquitin bands ) and β-actin as a loading control are displayed . ( B ) Immunohistochemistry staining of P8 mouse ovaries either non-irradiated ( NIRR ) or 8h after γ-irradiation ( IRR ) for Msy , p53 , p63 or p73 . Stainings for Msy and p63 were developed with a 30 s exposure time and then stopped due to high signal intensity . Stainings for p53 and p73 were exposed for 5 min . The red arrows indicate primordial follicles , which express high amount of TAp63α and are responsive to a low dosage of γ-irradiation . Scale bar: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 022 Induction of apoptosis requires the transcriptional activity of TAp63α and the translation of pro-apoptotic factors such as PUMA and NOXA ( Kerr et al . , 2012 ) . To test whether the treatment with cycloheximide affects the transcriptional activity of TAp63α we used qPCR to detect mRNA levels of the three p63 targets p21 , Puma and Mdm2 ( Figure 6E ) . As a control we used the oocyte specific marker Msy2 . The data showed that both with and without cycloheximide treatment significant induction of the target genes occurred while the level of Msy2 was unaffected . We could not detect the presence of p53 before or eight hours after irradiation by immunohistochemistry , suggesting that p53 is not involved in the apoptosis of oocytes ( Figure 6—figure supplement 1B ) . This interpretation is also consistent with the observation that oocytes only from the TAp63 but not the p53 knock out mouse are protected from irradiation induced apoptosis ( Suh et al . , 2006 ) . For p73 we could detect a weak , diffuse staining consistent with earlier reports of low levels of cytoplasmic p73 in oocytes ( Livera et al . , 2008 ) . The very low level compared to p63 and the strong induction of target genes such as PUMA or p21 in the presence of the translational inhibitor cycloheximide , however , argue against a significant role of p73 for the irradiation induced cell death of oocytes . Our results suggest that oocytes contain all kinases necessary to initiate tetramerization of TAp63α and all factors essential for p63’s transcriptional function ( Figure 7B ) . One of the kinases that has been identified in the activation process is Chk2 that phosphorylates TAp63α on Ser 582 ( numbering according to the TA-isoform of p63 ) ( Bolcun-Filas et al . , 2014 ) . To investigate if phosphorylation by Chk2 is required for tetramerization we treated mouse ovaries with increasing amounts of the Chk2 inhibitor II BML-277 and irradiated them with a dose of 1 . 5 Gy two hours after adding the inhibitor . At a concentration of 25 µM phosphorylation of TAp63α was almost completely suppressed and almost no tetramer was formed ( Figure 6F ) . These data confirm the essential role of Chk2 in the activation process and demonstrate that phosphorylation by Chk2 is also a prerequisite for the formation of tetramers . Interestingly , these data also show that activation of TAp63α leads to a very significant drop of the intracellular concentration and inhibition of the activation by the Chk2 inhibitor to a preservation of the original level . This effect is due to fast proteasomal degradation of activated TAp63α and is consistent with other observations showing that the cellular concentration of active isoforms of p63 is low while inactive isoforms can accumulate to high concentrations ( Serber et al . , 2002 ) . Interestingly , it has been shown that the N-terminal TAD is involved in this degradation process and that degradation is linked to DNA-binding competent and transactivating p63 isoforms ( Ying et al . , 2005 ) . This observation is also consistent with our model in which the TAD is involved in the formation of the inhibitory lock structure that covers the tetramerizatoin interface and is therefore protected from ubiquitination . After the formation of the open and active state , however , the TAD is accessible , leading to fast degradation . This competition between activation and degradation probably constitutes an intracellular threshold that protect oocytes from apoptosis by low levels of activated TAp63α . 10 . 7554/eLife . 13909 . 023Figure 7 . Spring-loaded activation mechanism of TAp63α on the molecular and cellular level . ( A ) Schematic energy landscape of TAp63α . The kinetically trapped closed dimer is opened by phosphorylation or artificially by moderate concentrations of urea ( Figure 4 ) . The resulting open dimer is less stable and forms tetramers with a dissociation constant of 12 ± 1 nM ( Brandt et al . , 2009 ) . ( B ) Schematic representation of TAp63α activation . Oocytes express high levels of dimeric TAp63α and harbor normally inactive kinases ready to be activated and to phosphorylate TAp63α upon genotoxic stress leading to active tetramers and , consequently , cell death . DOI: http://dx . doi . org/10 . 7554/eLife . 13909 . 023
Oocytes are very special cells that have developed a unique quality control system . In humans the approximately seven million oocytes that are created during embryogenesis are diminished to one to two million at the time of birth ( Tilly , 2001 ) . A large drop in numbers is also seen for mouse oocytes . Of the original roughly 25 , 000 cells only 10 , 000 remain at the time of birth ( Di Giacomo et al . , 2005 ) . During the late embryonic stage , the sensitivity of oocytes to DNA double strand breaks changes dramatically . While oocytes in the leptotene stage of prophase I ( around E14 ) tolerate hundreds of Spo11 induced double strand breaks as part of the process of homologous recombination , postnatal oocytes are killed by fewer than 10 DNA double strand breaks per cell . This dramatic shift in sensitivity is correlated with the expression of TAp63α which starts to get expressed in the diplotene stage beginning around E18 . 5 when chromosomes have been repaired after homologous recombination ( Livera et al . , 2008 ) . Most likely , the p63 system developed as a safeguard to ensure that cells that still contain chromosome damage do not survive . The finding that the p63 expression level is kept high during the long dictyate arrest in mammals , however , shows that p63 is not only used as a short term quality control check point but also as a factor that guarantees the long term genetic stability of germ cells . In particular , this long term quality control function requires a tightly controlled activity of p63 . A basal activity that is too high would lead to premature loss of the oocyte pool and ovary failure while a low activity bears the risk that oocytes acquire a high level of chromosomal defects . Our extensive mutagenesis study and biophysical characterization now provides a first model how interaction of N-terminal and C-terminal sequences blocks the tetramerization interface of the TD and therefore prevents tetramerization . Our biochemical analysis has also revealed that neither the SAM domain nor the DBD are essential for formation of the closed state . However , unlike the SAM domain the DBD cannot be completely removed but must be replaced with a domain of similar size . At the same time , the ASPP2 binding assays in combination with the SAXS analysis predicts that the DBD has a defined orientation within the dimeric structure which makes the DNA binding interface inaccessible . This orientation , however , seems to be stabilized by interactions that are not essential for the formation of the core inhibitory structure consisting of the TAD , TD and TID . At the same time , this finding explains why both in cells as well as in in vitro fluorescence anisotropy measurements the DNA binding affinity of TAp63α is roughly 20-fold lower than the affinity of open and tetrameric isoforms or mutants ( Deutsch et al . , 2011; Suh et al . , 2006 ) . Mutations in the SAM domain as well as in the TID cause the Ankyloblepharon-ectodermal defects-cleft lip/palate syndrome ( AEC ) syndrome ( McGrath et al . , 2001 ) . Two mutations identified in human patients , R598L and D601V ( Rinne et al . , 2009 ) , are located in a region of the TID that is responsible for stabilizing the dimeric state . According to our model both R598 and D601 are involved in charge-charge interactions with the TA2B β-strand and their mutation likely destabilizes the closed dimeric state which might cause in addition to the severe skin phenotype of the patients further ovary related problems . Mutations found in other domains of p63 such as the DBD that cause ectrodactyly–ectodermal dysplasia–cleft ( EEC ) syndrome ( Celli et al . , 1999; Kouwenhoven et al . , 2015 ) might also affect the stability of the closed dimer in oocytes . While effective inhibition is a prerequisite for a stable long term quality control with a minimal protein turnover rate , an effective activation mechanism is also of paramount importance . Our results show that the closed conformation of TAp63α is a metastable state and that activation follows a spring-loaded mechanism ( Figure 7A ) . In oocytes , phosphorylation is used as the natural trigger to initiate the transition from the closed dimeric state to the thermodynamically more stable tetrameric state . Once the active tetramer is formed , the phosphate groups can be removed without affecting the oligomeric state of the protein ( Deutsch et al . , 2011 ) . Spring loaded activation mechanisms are known from other proteins as well . One prominent example is the Influenza virus hemagglutinin A ( HA ) . This membrane protein is trapped in a metastable native pre-fusion state in which the fusion peptide is buried inside the trimeric structure ( Carr and Kim , 1993 ) . Following endocytosis of the virus and a pH drop in the endosome , the protein changes its conformation resulting in the exposure of the fusion peptides that are subsequently inserted into the host membrane ( Lin et al . , 2014 ) . While the drop in pH is the natural trigger , activation can also be initiated by high temperatures or urea ( Carr et al . , 1997 ) . Another example is α-lytic protease , a secreted serine protease that is expressed with an N-terminal pro-region that catalyzes folding from a stable molten globule-like intermediate . Proteolytic degradation of the pro-region results in release of the native and active protease , which is thermodynamically less stable than the partially unfolded state but remains folded due to a large barrier to unfolding ( Sohl et al . , 1998; Baker , 1998 ) . The kinetically trapped state of dimeric TAp63α raises the question how and when this state is formed during protein synthesis . Interestingly , it was shown that p53 forms dimers co-translationally and tetramers post-translationally ( Nicholls et al . , 2002 ) . Our expression experiments in the presence of high concentrations of the p73 TD in principle support a co-translational folding of TAp63α . However , our deletion mutagenesis also implicates that the last amino acid of TAp63αmin , P614 , has to emerge from the ribosomal exit tunnel before the closed dimeric state can be formed . As a model we propose that open dimers form co-translationally via the TD that acts as the interaction platform for the TAD and the TID to fold into the trapped conformation after completion of translation . The exact mechanism of folding and a potential role for chaperones remains to be investigated . Not only the metastable state of TAp63α sensitizes oocytes for DNA damage induced cell death , the entire machinery that detects DNA damage and activates TAp63α is present in resting oocytes without the need for further protein expression . So far , ATM/ATR as upstream kinases and Chk2 as a direct phosphorylating kinase have been shown to be involved in this process ( Bolcun-Filas et al . , 2014; Kim et al . , 2011 ) . Other factors might contribute as well ( Gonfloni et al . , 2009 ) in stabilizing the tetrameric state and forming active transcriptional complexes on promotor sites . The special metabolic state that oocytes reside in during dictyate arrest requires them to express a limited number of genes , essential for keeping the cells stable . Proteins involved in the surveillance of DNA damage as well as transmitting the signal to the central integrator , p63 , are part of this cellular repertoire . Quality control in oocytes by TAp63α is therefore based on a spring-loaded activation mechanism on the molecular and the cellular level .
TAp63α was codon-optimized for expression in E . coli and ordered from Genscript ( Piscataway , NJ , USA ) . Deletions were introduced using the QuikChange II Site-Directed Mutagenesis Kit ( Agilent Technologies ) . TAp63αmin comprising deletions Δ ( 1–9; 64–119; 417–453; 460–505; 571–593; 615–641 ) was cloned into pNIC28-Bsa4 ( SGC Oxford ) by ligation independent cloning ( Gileadi et al . , 2008 ) . The protein , bearing a N-terminal His6-tag and a TEV ( tobacco etch virus ) protease cleavage site was expressed in BL-21 ( DE3 ) -R3-Rosetta ( SGC Oxford ) and initially purified using Ni-Sepharose Fast Flow and HiTrap Q HP ( GE Healthcare ) according to standard protocols . After His6-tag removal using TEV protease the protein was further purified using a HiTrap Q HP and a HiLoad 16/600 Superdex 200 prep grade column . TAp63αmin was stored concentrated ( 100 mg/mL ) at -80°C . ASPP2 ( 891–1128 ) was cloned into pGEX 6p2 ( GE Healthcare ) with an additional C-Terminal His6-tag . The resulting GST-fusion of ASPP2 was expressed in BL-21 ( DE3 ) -R3-Rosetta ( SGC Oxford ) and purified by Ni-Sepharose Fast Flow and Gluthation-Sepharose Fast Flow ( GE Healthcare ) using standard protocols followed by size-exclusion chromatography with a HiLoad 16/600 Superdex 200 prep grade column . SEC-MALS experiments were performed at room temperature using a Superose 6 3 . 2/300 column ( GE Healthcare ) in phosphate buffer containing 0 , 2 or 2 . 5 M urea on an Agilent 1200 Series HPLC system at a flow rate of 0 . 05 ml/min . Prior to injection the protein was incubated in phosphate buffer containing 2 M urea for 14 min or 2 . 5 M urea for 25 min . Elution of 10 μL of purified proteins of 6 . 4 mg/ml concentration was detected using Dawn Heleos II ( 11 angles were used ) and an Optilab rEX Refractive Index Detector at a Laser wavelength of 658 nm ( Wyatt Technology ) to determine the weight average molar mass MW of peak locations . Data were processed using ASTRA software package 6 . 1 . 2 . 84 ( Wyatt Technology ) . For Native PAGE analysis of the oligomeric state of p63 two ovaries per indicated condition were harvested in 20 µl of ice-cold lysis buffer A ( 50 mM Tris pH 8 . 0 , 100 mM NaCl , 1 mM DTT , 2 mM MgCl2 , supplemented with 1x cOmplete and PhosSTOP ( Roche ) ) . Lysis was performed by mechanical force using a pestle , pipetting and two cycles of freeze and thaw . After addition of 20 µl lysis buffer B ( lysis buffer A containing 40 mM CHAPS ) and 1 µl benzonase , samples were incubated for 1 hr on ice and subsequently centrifuged for 10 min at 4°C and 13 . 2 krpm to remove cell debris . 20 µl of supernatant were supplemented with 5 µl of 5x Native PAGE sample buffer ( 60% glycerol , 25 mM coomassie G250 ) for Native PAGE analysis . The remaining lysate was used for analysis of p63 level and phosphorylation-induced mobility shift via SDS-PAGE . The separation of ovary lysate by Native PAGE followed by detection of p63 via subsequent Western Blot analysis was performed with the Native PAGE Novex 3–12% Bis-Tris protein gel system ( Life Technology ) according to the manufacturer’s instructions . The cathode buffer was supplemented with 0 . 002% coomassie G250 and the separation was performed at 4°C for 60 min at 150 V and 90 min at 250 V . For NMR spectroscopy [u-15N]-labeled human p63 DBD-TD-SAM , DBD , TD and SAM were measured at concentrations between 0 . 1–0 . 3 mM in a total volume of 350 μL in shigemi NMR tubes . Complete Protease Inhibitor ( Roche ) and 6% of a D2O/DSS ( 3 mM DSS ) solution was added . NMR-Experiments were performed on a Bruker Avance spectrometer equipped with 1H triple resonance , z-gradient cryogenic probes at a proton frequency of 900 MHz . All experiments were performed at 303 K . DSS ( 4 , 4-dimethyl-4-silapentane-1-sulphonate ) was used as an internal chemical shift reference . Spectra were processed with Bruker Topspin 2 . 1 and analyzed with UCSF SPARKY 3 . 114 ( Kneller and Kuntz , 1993 ) . Animal care and handling were performed according to the guidelines set by the World Health Organization ( Geneva , Switzerland ) . Eight-day-old ( P8 ) female CD-1 mice were purchased from Charles River Laboratories . Ovaries were harvested , transferred in sterile flat-bottom 96-well plates with 100 µl MEM ( + L-Glu , Gibco ) supplemented with 5% FBS , 0 , 4% BSA ( w/v ) , Pen/Strep and 70 µM Br-cAMP and cultured in an incubator at 37°C with 5% CO2 . Ovaries were treated overnight with either DMSO or CHX ( 50 µg/mL ) prior following experiments . IRR ovaries were exposed to 1 . 5 Gy of γ-irradiation on a rotating turntable in a 137Cs irradiator , at a dose rate of 2 . 387 Gy/min . For inhibition of Chk2 in ovary culture the inhibitor BLM-277 ( Merck Millipore , 220486 ) was used 2 hr prior γ-irradiation in indicated concentrations . The following antibodies were used for detection of endogenous protein of ovary samples by Western Blotting: Msy2 ( Santa Cruz , N-13 ) , Ubiquitin ( Santa Cruz , P4D1 ) , p63 ( Santa Cruz , H-129 ) and β-Actin ( Santa Cruz , C4 ) . Dissected ovaries were cultured overnight and subsequently treated with γ-irradiation as indicated . Ovaries were fixed in formalin , embedded in paraffin and sectioned into 6 µm thickness ( Morphisto GmbH , Frankfurt , Germany ) . For 3 , 3'-Diaminobenzidine ( DAB ) IHC staining sections were deparaffinised and rehydrated followed by 30 min antigen retrival in boiling 0 . 1 M citrate buffer . Sections were blocked for 1 hr at room temperature in 5% donkey normal serum ( Santa Cruz , sc-2044 ) in TBS and incubated with primary antibody raised either against the oocyte marker Msy ( Santa Cruz , N-13 , 1:200 ) , p53 ( Santa Cruz , DO-1 , 1:100 ) , p63 ( Santa Cruz , H-129 , 1:200 ) or p73 ( Merck Millipore , ER-15 , 1:100 ) in 1% BSA in TBS overnight . Sections were developed after incubation with biotin-conjugated secondary antibodies for 1 hr at room temperature in 1% BSA in TBS ( Vector Labs ) with the ABC DAB Peroxidase System ( Vector Labs ) . Nuclei were stained for 5 min in Mayer’s hematoxylin followed by dehydration and mounting of the stained sections . N-terminally myc-tagged human TAp63α , TAp63α ( 10–614 ) , TAp63γ , ΔNp63α and all mutants that base on these constructs were expressed from pcDNA3 . 1 vector in RRL as described ( Straub et al . , 2010 ) . Proteins were used for SEC analysis and pull-down experiments . GST pull-down experiments were performed with RRL expressed proteins and immobilized GST-TID ( aa 569–616 ) as described previously ( Straub et al . , 2010 ) . For His6 pull-down experiments , ΔNp63α , TAp63α and TAp63α R604E R608E were expressed in presence or absence of His6-tagged p73 TD ( 20 μM ) in 50 μL RRL for 90 min at 30°C . In the latter case , cycloheximide ( 50 μg/mL final ) and His6-tagged p73 TD ( 20 μM final ) were added after expression and incubated for another 90 min at 30°C . Afterwards 5 μL samples were removed as input controls ( I ) . For each pull-down 50 μL Ni-IDA beads were washed inside an Ultrafree centrifugal filter unit ( Durapore PVDF 0 . 65 μm , Millipore ) with binding buffer ( 500 mM NaCl , 50 mM Tris pH 7 . 8 , 5 mM imidazole , 5% ( v/v ) glycerol ) . The remaining 45 μL of the RRL expression was added to the beads and incubated for 1 hr at 4°C . Subsequently the beads were washed 5 times with ice-cold wash buffer ( 500 mM NaCl , 50 mM Tris pH 7 . 8 , 30 mM imidazole , 5% ( v/v ) glycerol ) and the proteins were eluted with 40 μL of 80°C hot SDS-PAGE buffer ( P ) . After SDS-PAGE and western blotting the quotient of pull-down ( P ) and input ( I ) band intensity was normalized to TAp63α incubated after expression with His6-tagged p73 TD ( set to 1 ) . Real-time quantitative PCR was performed with two independent sets of samples . For each condition per set four dissected ovaries were pooled . Oocytes were isolated by trypsin-digestion and multiple centrifugation steps . Total RNA was extracted applying the PicoPure RNA Isolation Kit ( Applied Biosystems ) with on-column DNAseI ( Qiagen ) digestion and subsequently subjected to reverse transcription with random primers using the RETROscript Kit ( Ambion ) followed by cDNA amplification with the TaqMan PreAmp Kit ( ThermoFisher Scientific ) . Real-time quantitative PCR to determine the fold-induction of p63 target genes was performed with the TaqMan Gene Expression System ( ThermoFisher Scientific ) using a LightCycler 480 ( Roche ) . For one biological set , each sample and TaqMan assay probe combination was measured in duplicates . All Kits were used according to the manufacturer’s instruction . The following TaqMan assays ( ThermoFisher Scientific ) were purchased for the preamplification step and the gene expression analysis: TBP ( Mm00446971_m1 ) , Msy ( Mm01250826_g ) , p21 ( Mm04205640_g1 ) , PUMA ( Mm00519268_m1 ) and Mdm2 ( Mm01233136_m1 ) . Target gene signals were referenced to the house keeping gene TBP and mean fold-induction upon irradiation was calculated for the biological duplicates including error propagation . The significance levels were determined by the student’s t-test . Permission for the experiments with mouse ovaries was obtained from the “Tierschutzbeauftragte” of the Goethe University . Analytical SEC was performed in phosphate buffer ( 50 mM sodium phosphate pH 7 . 8 , 100 mM NaCl ) at 4°C using a Superose 3 . 2/300 column ( GE Healthcare ) ( injection volume 50 μL; flow rate 50 μL/min; fraction size 50 μL ) . SEC fractions were quantified by western blotting . Analytical SEC of TAp63αmin in urea was performed as described detailed in Supplemental Experimental Procedures . SEC experiments were performed on an ÄKTApurifier system at 4°C using a Superpose 6 3 . 2/300 column ( GE Healthcare ) , monitoring absorption at 280 nm . The column was equilibrated in a phosphate buffer containing urea at a variable concentration X . 5 μL of TAp63αmin ( 102 mg/mL ) were diluted with 75 μL of buffer X ( to a final concentration of 6 . 4 mg/mL ) and incubated for one hour at 4°C before being injected on the column . This experiment was performed at different urea concentrations X [M]: 1 , 1 . 25 , 1 . 5 , 1 . 75 , 2 , 2 . 25 , 2 . 5 , 3 , 3 . 5 , 4 , 4 . 5 , 5 , 6 , and 7 . The column was equilibrated in a phosphate buffer containing 1 . 75 M urea . TAp63αmin was diluted to a final concentration of 6 . 4 mg/mL in a buffer with a final urea concentration of 1 . 75 M ( first TAp63αmin was diluted with x μL of buffer X and then with additional y μL of buffer Y , whereby cy = cx + 1 M , so that the final concentration was exactly 1 . 75 M ) . Injections were performed at different time points [hours:minutes]: 0:01 , 0:53 , 1:45 , 2:43 , 3:36 , 4:30 , 5:29 , 6:20 , 7:17 , 8:14 , 9:06 , 24 hr . The column was equilibrated in a phosphate buffer containing 1 . 75 M urea . TAp63αmin ( 32 mg/mL final concentration ) was incubated in phosphate buffer with 1 . 75 M urea for one hour and then injected on the column . The dimer and tetramer peak ( two fractions each ) was dialyzed back to 0 M urea using D-tube Dialyzer Mini ( MWCO 12–14 kDa ) in a 50 mL falcon filled with phosphate buffer under continuous stirring . After 13 hr of dialysis the samples were reinjected on the column equilibrated with phosphate buffer . We used the crystal structure of the p63 tetramerization domain ( PDB: 4A9Z ) ( Muniz et al . , 2011 ) to highlight interactions relevant in context of dimeric TAp63α . The crystal structure of the p63 DNA binding domain ( DBD ) in complex with DNA ( PDB: 3QYN ) ( Chen et al . , 2011; Chen and Herzberg , 2011 ) was used to model the interaction with ASPP2 by structural alignment with the p53-ASPP2 complex ( PDB: 1YCS ) ( Gorina and Pavletich , 1996; 1997 ) . All structures and models were illustrated using PyMOL 1 . 7 . 6 . 6 . To obtain a qualitative measure of TAD and TID dissociation , constructs with a TEVsite ( ENLYFQGS ) between residues 66 and 67 ( 591 and 592 ) and with a C-terminal ( N-terminal ) myc-tag were created . After RRL expression cycloheximide ( 50 μg/mL final ) and TEV protease ( 10 μg ) were added . The sample was incubated for either 15 min , 1 hr , 4 hr or 12 hr at 37°C before being cooled to 4°C and subsequently analyzed by SEC . Transcriptional activities of TAp63α and TAp63α ( 10–614 ) mutants were measured in triplicates as described previously ( Luh et al . , 2013 ) . Western blot ( WB ) analysis was performed as described previously ( Straub et al . , 2010 ) . In-line size exclusion chromatography small-angle X-ray scattering of TAp63αmin was performed at bending magnet beamline B21 at Diamond Light Source ( Harwell , UK ) . The output from an Agilent HPLC was connected to an in-vacuum quartz flow cell . The SAXS detector was triggered by the 280 nm UV sensor in the Agilent HPLC , and allowed the collection of data in 1 s time bins across the peak of interest . A Shodex KW404 column was utilised for these experiments . At the end of each experimental run , SAXS data were integrated using beamline software and the background subtracted using running buffer . The integration procedure ensured that only SAXS data from the peak of interest were abstracted and subjected to further analysis . Data were inspected for radiation damage and aggregation by inspection of Guinier plots . This method ensured that SAXS data were unperturbed by any other oligomers which may have formed or been present in the analysis solution . The beamline was also used to collect data in batch mode , whereby protein and corresponding buffer solutions were exposed to the beam using an Arinax ( Grenoble , France ) BioSAXS automated sample changer robot , consisting of temperature controlled storage and exposure units . The exposure unit contained a 1 . 6 mm diameter quartz capillary in which the samples were illuminated with the x-ray beam; the exposure unit temperature was set to 15°C . The sample capillary was held in vacuum and subjected to a cleaning cycle between each measurement . Samples were stored in 96 well plates at 5°C . A Pilatus 2M two-dimensional detector was used to collect 10 frame exposures of 10 s from each sample and the corresponding buffer . The detector was placed at 3 . 9 m from the sample , giving a useful q-range of 0 . 008 Å-1 < 0 . 4 Å-1 , where q = 4π sin θ / λ , 2θ is the scattering angle and λ is the wavelength , which was set to 1 Å . Two dimensional data reduction consisted of normalization for beam current and sample transmission , radial sector integration , background buffer subtraction and averaging . Each frame was inspected for the presence of radiation induced protein damage; if this was found to be the case , the frames were not reduced and processed . Further data analysis , such as scaling , merging and Guinier analysis were performed in Scatter ( Forster et al . , 2010 ) . Three concentrations were measured of each mutant with each experimental data frame being inspected for signs of radiation damage . Frames which appeared to demonstrate radiation damage were excluded from averaging . Ab-initio shape reconstruction of the wild type was performed by averaging and filtering 13 runs of DAMMIF ( Franke and Svergun , 2009 ) , with a final refinement in DAMMIN ( Svergun , 1999 ) , utilizing slow mode . The wild type was found to have Rg of 38 . 6 Å , with Dmax of 132 Å . λ-cro-TAp63αmin was analyzed using MONSA , allowing a simultaneous bead modelling from the wild type and the N-terminal fusion . A relative volume difference for MONSA was derived from Porod analysis of the wild type and derivative scattering curves . Secondary structure and disorder were predicted with Phyre2 ( Kelley et al . , 2015 ) and the Protein Crystal Structure Propensity Prediction Server ( Price II et al . , 2009 ) which uses PredictProtein ( Rost et al . , 2004 ) . | The irradiation and chemotherapy drugs that are used to destroy cancer cells also damage healthy cells . Germ cells – from which egg cells and sperm cells develop – are particularly vulnerable as they contain sensitive quality control mechanisms that kill any cell that contain damaged DNA . Consequently , after surviving cancer many patients are confronted with infertility . A protein called p63 , which is closely related to another protein that suppresses the formation of tumors , plays an essential role in detecting and responding to DNA damage . In immature egg cells ( also known as oocytes ) , p63 mostly exists in an inactive form . The protein then switches to an active form when DNA damage is detected to trigger the process of cell self-destruction . Now , Coutandin , Osterburg et al . have performed a range of biochemical , biophysical and cell culture experiments to study how p63 is kept in its inactive form in the oocytes of mice . The experiments showed that in the inactive form , the two ends of the protein form a sheet that closes a key site on the protein and prevents it from changing into its active form . However , this closed form can be thought of as being like a spring-loaded trap – it doesn’t take much energy to spring the trap and open the protein into its active form . Once this change has occurred , it is irreversible . Coutandin , Osterburg et al . also found that the oocytes of mice already contain all the proteins necessary to activate p63 . This means that once the switch to the active form is triggered there is no delay waiting for other proteins to be made , which makes oocytes extremely sensitive to DNA damage . Further work is now needed to investigate the exact molecular mechanisms behind the activation of p63 . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Quality control in oocytes by p63 is based on a spring-loaded activation mechanism on the molecular and cellular level |
Both pathogen- and tissue damage-associated molecular patterns induce inflammation through toll-like receptors ( TLRs ) , while sialic acid-binding immunoglobulin superfamily lectin receptors ( Siglecs ) provide negative regulation . Here we report extensive and direct interactions between these pattern recognition receptors . The promiscuous TLR binders were human SIGLEC-5/9 and mouse Siglec-3/E/F . Mouse Siglec-G did not show appreciable binding to any TLRs tested . Correspondingly , Siglece deletion enhanced dendritic cell responses to all microbial TLR ligands tested , while Siglecg deletion did not affect the responses to these ligands . TLR4 activation triggers Neu1 translocation to cell surface to disrupt TLR4:Siglec-E interaction . Conversely , sialidase inhibitor Neu5Gc2en prevented TLR4 ligand-induced disruption of TLR4:Siglec E/F interactions . Absence of Neu1 in hematopoietic cells or systematic treatment with sialidase inhibitor Neu5Gc2en protected mice against endotoxemia . Our data raised an intriguing possibility of a broad repression of TLR function by Siglecs and a sialidase-mediated de-repression that allows positive feedback of TLR activation during infection .
Nearly 25 years ago , Janeway proposed that the innate immune system discriminates infectious nonself from non-infectious self through non-clonally distributed pattern recognition receptors ( PRRs ) ( Janeway , 1989 , 1992 ) . This concept was supported by the observations of microbial induction of costimulatory activity on antigen-presenting cells ( Liu and Janeway , 1991 , 1992; Wu and Liu , 1994 ) and bolstered as a major pillar in immunology by the identification of TLRs ( Medzhitov et al . , 1997 ) and its ligands ( Poltorak et al . , 1998 ) . Over the following decades , TLRs have emerged as a family of PRRs that sense a variety of pathogen-associated molecular patterns ( PAMPs ) , ranging from microbial glycan , bacterial glycolipids , flagellin , and viral and bacterial nucleic acids ( Kawai and Akira , 2010 ) . Surprisingly , accumulating data demonstrated that TLRs also sense cellular components released after cellular injuries , such as heat-shock proteins ( Millar et al . , 2003 ) , HMGB1 ( Apetoh et al . , 2007; Ivanov et al . , 2007; Tian et al . , 2007 ) and S100 ( Hiratsuka et al . , 2008; Tsai et al . , 2014 ) . These components were collectively called DAMPs for danger-associated molecular patterns , based on Matzinger's ‘danger theory’ ( Matzinger , 1994 ) . In addition to TLRs , Nod-like receptors ( NLRs ) have also been shown to respond to both microbial components ( Franchi et al . , 2012 ) and cellular injuries ( Ting et al . , 2008; Tschopp and Schroder , 2010 ) . Since these PRRs respond to both infectious and noninfectious inflammatory stimuli , additional regulatory mechanism was deemed needed if pattern recognition receptors were to be the key components to discriminate the two ( Liu et al . , 2009 ) . We have reported that CD24-Siglec-G/10 interactions selectively inhibit host responses to DAMPs without affecting responses to PAMPs ( Chen et al . , 2009 ) . These data suggest a new mechanism for limiting inflammation to autologous molecules , and a framework for integrating the self-nonself and danger theories of immunity ( Liu et al . , 2009 ) . Siglecs are membrane-bound lectins that constitute the sialic acid-binding immunoglobulin super family with distinct cellular distribution and glycan specificities ( Crocker et al . , 2007 ) . Most Siglecs have intracellular domains capable of inhibitory signaling . With a few notable exceptions ( Stamenkovic et al . , 1991; Chen et al . , 2009; Bandala-Sanchez et al . , 2013; McMillan et al . , 2013 ) , the natural ligands for most Siglecs remain to be determined . The selective effect of Siglec-G in regulating host response to DAMPs raised two interesting questions: First , do all Siglecs perform similar functions in discriminating self from non-self in innate immunity ? Second , do Siglec family members directly interact with other families of pattern recognition receptors and play a broad function in regulating innate immunity ? Here we addressed these two issues by revealing a broad and direct interaction between TLRs and many Siglecs , and by highlighting Siglec-G for its lack of direct interactions with TLR . A standard test to validate Siglec-mediated interactions is to measure their susceptibility to sialidase treatment ( Crocker et al . , 2007 ) . The implication of this cardinal feature in immune recognition has not been widely explored . Our previous studies have demonstrated that CD24-Siglec-G interactions are disrupted by bacterial sialidase and that such disruptions exacerbates sepsis ( Chen et al . , 2011 ) . Since most pathogens do not express sialidases , the implication of this observation on immune regulation during infection and autoimmune diseases is less clear . On the other hand , mammals express at least four sialidases , Neu1-4 ( Miyagi and Yamaguchi , 2012 ) . Several studies have suggested that activation of TLR can be attenuated by sialidase inhibitors ( Amith et al . , 2010; Abdulkhalek et al . , 2011 ) . Both Neu1 and Neu3 have been suggested as targets of the inhibitors ( Amith et al . , 2010; Abdulkhalek et al . , 2011 ) . Neu1 has been shown to cause TLR4 dimerization , which was suggested to be important for its regulation of TLR4 activation ( Amith et al . , 2010 ) . Here we report a critical role for Neu1 in regulating Siglec-TLR interaction and endotoxemia . Together , our data reveal an overlooked network of interactions among Siglecs , host sialidases and TLRs and , with more potent Neu1 inhibitors , propose host sialidases as therapeutic targets for lethal endotoxemia .
To test whether TLRs directly interact with Siglecs , we first tested a panel of recombinant human SIGLEC fusion proteins for their interaction with human TLRs synthesized by the THP1 cell line using a sandwich capture assay . Plates were coated with recombinant SIGLEC-Fc fusion proteins or control human Fc . Lysates from the human myeloid cell line THP1 was used as the source of cellular TLRs , as this line expresses transcripts of all TLRs at significant levels ( Figure 1A ) . SIGLEC-Fc-bound TLRs were detected with anti-TLR antibodies . This assay revealed extensive interaction between the two families of pattern recognition receptors ( Figure 1B ) . Among them , Siglec 5 and 9 cross-reacted with virtually all TLR tested . In contrast , very little interaction was observed between TLRs and SIGLECS-1 , 2 , 7 , 10 and 11 . 10 . 7554/eLife . 04066 . 003Figure 1 . Extensive direct interactions between Siglecs and TLRs . ( A ) Evaluation of TLR expression using TLR-primer set . Data shown are means and SEM of triplicate % of GAPDH levels . ( B ) and ( C ) Interactions between human ( B ) or mouse ( C ) SIGLEC-Fc fusion proteins and TLRs from THP-1 cells ( B ) or murine splenocyte lysates ( C ) . Recombinant SIGLEC-Fc or control IgG Fc were coated on 96 well plates to capture TLR in the cell lysates . The associated TLRs were detected with biotinylated anti-TLR antibodies that cross-react with both mouse and human TLR . Data shown are the log2 ratios between Siglec-Fc and IgG Fc in triplicate and were repeated three times . ( D ) Recombinant Siglec-E binds to endogenous Tlr4 . Lysates from C57BL/6 mouse splenocytes were incubated with either Fc control or SIglec E-Fc . After precipitation with protein A beads , the precipitates were analyzed by Western blot , using antibodies against Siglec-E , Tlr4 , and Fc . ( E ) Interaction between endogenous Tlr4 and Siglec-E in D2SC dendritic cells . Lysates from D2SC cells were immunoprecipitated with either anti-Siglec-E ( top ) or Tlr4 ( bottom ) antibodies . The precipitates were analyzed by Western blot , using antibodies against Siglec-E or Tlr4 . Similar results were obtained when performed with WT mouse splenocyte lysates ( data not shown ) . ( F ) Direct interaction between human Siglecs and ectodomain of TLR4 . As in ( A ) , except the cell lysates were replaced with recombinant TLR4 . ( G ) Direct interaction between mouse Siglecs and ectodomain of TLR4 . As in ( B ) , except the cell lysates were replaced with recombinant TLR4 . Data presented in this figure have been reproduced at least three times . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 003 Since the anti-TLR antibodies used also cross-react with their mouse homologs , we carried out a similar analysis between mouse Siglecs and Tlrs . As shown in Figure 1C , Siglec-3 , E , F and H were broadly cross-reactive , while Siglec-G displayed no interaction with any murine Tlr . We validated the interaction between Siglec-E and Tlr4 using both pull down and bi-directional immunoprecipitation assays . Siglec-E-Fc , but not control IgG , pulled down Tlr4 in the spleen cell lysates ( Figure 1D ) . Anti-Siglec-E monoclonal antibodies co-precipitated Tlr4 , while anti-Tlr4 co-precipitated Siglec-E ( Figure 1E ) . To determine whether Siglecs directly interact with TLR , we coated the 96-well plates with recombinant Siglecs or control IgG . After blocking with bovine serum albumin , recombinant TLR4 extracellular domain was added , and the bound TLR4 were determined using biotinylated anti-TLR4 and horseradish peroxidase-conjugated streptavidin . The data revealed that TLR4 directly interact strongly with SIGLEC-5 , 6 , 9 , 11 and modestly with SIGLEC-7 ( Figure 1F ) . In contrast , no interaction was observed between SIGLEC-1 , 2 , 3 , 10 and TLR4 . Likewise , mouse Siglec-E and F bound strongly to TLR4 , with Siglec-3 and H showing modest binding to TLR4 , and no binding to TLR4 and Siglec-1 , 2 and G ( Figure 1G ) . Comparisons between Figure 1A , E , and between Figure 1C , G , show that , with the modest exception of Siglec-7 , recombinant TLR4 recapitulated the specificity of endogenous TLR4 . Taken together , data in Figure 1 demonstrate broad and direct interaction between TLR and Siglec families of PRRs . Since Siglec-E contains intracellular immunoreceptor tyrosine inhibitory motif ( ITIM ) and ITIM-like domains and associates with molecules known to negatively regulate production of inflammatory cytokines , we compared bone marrow-derived dendritic cells ( DC ) from WT and Siglece−/− mice for their responses to prototypic TLR ligands . After overnight stimulation with TLR ligands , we analyzed the production of IL-6 and TNFα by DC . As shown in Figure 2A , Siglece−/− DC were 1000-fold more responsive to LPS stimulation than Siglece+/+ DC . Likewise , Siglece−/− DC produced at least 10-fold more cytokines in response to CpG ( Figure 2B ) . Given the multiple interactions of Siglec-E with other TLR ( Figure 1C ) , we tested if endogenous Siglec-E negatively regulates production of inflammatory cytokines to other TLR ligands . As shown in Figure 2C , in addition to an enhanced response to TLR4 and TLR9 ligands , Siglece−/− DC also produced significantly more IL-6 in response to synthetic triacylated lipoprotein Pam3CSK4 ( TLR1/2 agonist ) , heat-killed Listeria monocytogenes ( HKLM , Tlr2 agonist ) , poly ( I:C ) , ( TLR3 agonist ) , Salmonella typhimurium flagellin ( ST-FLA , TLR5 agonist ) , synthetic lipoprotein derived from Mycoplasma salivarium ( FSL-1 , TLR2/6 agonist ) and ssRNA40 , a 20-mer phosphorothioate-protected single-stranded RNA oligonucleotide containing a GU-rich sequence ( TLR8 agonist ) . Since Siglec-E negatively regulates responses to all TLR ligands tested , we suggest that the physical interactions between Siglec-E and TLRs are biologically significant . Except for a modest elevated response to ssRNA40 , Siglecg−/− and Siglecg+/+ DC exhibited similar responses to all TLR ligands tested ( Figure 2C ) . This , along with data from our previous report indicating that Siglec-G does not inhibit inflammatory response to LPS and poly ( I:C ) , ( Chen et al . , 2009 ) , is consistent with the lack of physical interactions between Siglec-G and TLRs ( Figure 1C , G ) . 10 . 7554/eLife . 04066 . 004Figure 2 . Siglec-E negatively regulates production of inflammatory cytokines by DC in response to TLR ligands . ( A ) and ( B ) Siglec-E inhibits production of IL-6 and TNFα by bone marrow derived DC . DC cultured from WT or Siglece−/− bone marrow were stimulated with indicated concentrations of LPS ( A ) , or poly ( I:C ) ( B ) for 16 hr , and supernatant cytokine concentrations were analyzed with cytokine bead array . ( C ) . Targeted mutation of Siglec-E , but not Siglec-G , enhances production of IL-6 to multiple TLR ligands . The TLR agonists used are: synthetic triacylated lipoprotein Pam3CSK4 ( TLR1/2 agonist ) , heat-killed Listeria monocytogenes ( HKLM , Tlr2 agonist ) , poly ( I:C ) , ( TLR3 agonist ) , Salmonella typhimurium flagellin ( ST-FLA , TLR5 agonist ) , synthetic lipoprotein derived from Mycoplasma salivarium ( FSL-1 , TLR2/6 agonist ) and ssRNA40 , a 20-mer phosphorothioate-protected single-stranded RNA oligonucleotide containing a GU-rich sequence ( TLR8 agonist ) . All agonist were used at 100 ng/ml . Data represent the mean ± SD for three independent cultures of DCs in each genotype and were repeated at least three times . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 004 We next tested the impact of LPS stimulation on Siglec-E-TLR4 interaction . As shown in Figure 3A , co-precipitation between endogenous TLR4 and Siglec-E was substantially reduced after LPS stimulation . These data demonstrate a dynamic regulation of TLR4-Siglec-E association on DC . Since interaction of Siglecs with their ligands is dependent on sialic acid and can be disrupted by sialidase ( Crocker et al . , 2007 ) , and since LPS is devoid of sialidase activity , we evaluated the contribution of endogenous sialidase , using the D2SC DC cell line ( Bachmann et al . , 1996 ) . Real-time PCR analysis indicated that D2SC cells express Neu1 and to a lesser extent Neu3 , but not Neu2 or Neu4 ( Figure 3B ) . Since TLR4 is a cell surface glycoprotein while Neu1 is primarily lysosomal , we evaluated whether Neu1 translocates to the cell surface following LPS stimulation . Fluorescent microscopy revealed a robust translocation of Neu1 to cell surface where it co-localized with TLR4 ( Figure 3C ) , which is similar to a previous report on macrophages ( Liang et al . , 2006 ) . Flow cytometry confirmed a time-dependent translocation of Neu1 between 6–18 hr following LPS stimulation ( Figure 3D ) . To test if Neu1 and TLR4 interact with each other , live cells were cross-linked with dithiobis[succinimidyl propionate ] ( DSP ) to stabilize transient enzyme–substrate interactions . After cross-linking , the LPS-treated and untreated D2SC cells were lysed for co-immunoprecipitation . As shown in Figure 3E , a specific Neu1-TLR4 association was observed only after LPS stimulation . 10 . 7554/eLife . 04066 . 005Figure 3 . A critical role for Neu1 in Tlr4 activation . ( A ) Siglec-E-Tlr4 association is disrupted by LPS stimulation . D2SC cells were cultured in the presence or absence of LPS overnight . Immunoprecipitation was used to test Siglec-E-Tlr4 association as detailed in the Figure 1D legend . ( B ) Expression of Neu1-4 mRNA in D2SC dendritic cells was determined by RT-PCR . Data shown are mean ± SD transcript levels , expressed as % of the housekeeping gene HPRT . ( C ) Translocation of Neu1 and its co-localization with TLR4 in D2SC cells . D2SC were cultured in the presence or absence of LPS ( 100 ng/ml ) for 18 hr and co-stained with anti-TLR4 and anti-Neu1 antibodies . ( D ) Increased cell surface expression of Neu1 on D2SC cells after stimulation with LPS as revealed by flow cytometry . D2SC cells were treated with LPS ( 100 ng/ml ) or vehicle for 6 hr ( upper panel ) or 18 hr ( lower panel ) . The expression of cell-surface Neu1 was determined by FACS . ( E ) Physical association between Neu1 and TLR4 . D2SC 2 cell lines were stimulated with 2 μg/ml LPS or vehicle for 16 hr were crossed linked with 1 mM DSP at room temperature for 30 min . The lysates were immunoprecipitated with anti-Neu1 and then probed with anti-Neu1 or anti-TLR4 . ( F ) Silencing Neu1 by lentivirus shRNA increased the cell surface Siglec-E ligands . Histograms shown on top panels are FACS profiles . The bar graphs in the bottom panels represent geometric means ± SD of fluorescence intensity ( n = 3 ) . ( G ) Neu1 disrupts Tlr4-Siglec-E association in DC . D2SC dendritic cells were transduced with lentiviral vector carrying scrambled shRNA , three independent Neu1 shRNAs or Neu3 shRNA . After LPS stimulation for 24 hr , the lysates were used for immunoprecipitation . ( H ) ShRNA silencing of Neu1 affect cell surface Neu1 levels . Data shown are histogram of flow cytometry data depicting cell surface expression of Neu1 in LPS-stimulated D2SC clones . ( I ) An essential role for Neu1 in production of TNFα by D2SC cells . Aliquots of 2 × 105 D2SC transfectants were stimulated with LPS ( 100 ng/ml ) for 12 hr . The culture supernatants were subsequently collected and analyzed for TNF-α . Scramble , Neu1sh1 , 2 , 3 represent stable clones expressing three independent Neu1 ShRNAs . Experiments depicted in this figure have been reproduced two to three times . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 005 To determine if Neu1 regulates cell surface Siglec-E ligand levels , we compared binding of Siglec-E-Fc to either scrambled or Neu1 shRNA-transduced D2SC cells . As shown in Figure 3F , silencing of Neu1 increased binding of Siglec-E-Fc to DC . To determine whether Neu1 contributes to LPS-induced disassociation between Siglec-E and TLR4 , we tested the impact of Neu1 silencing on Siglec-E-TLR4 interaction , as measured by co-immunoprecipitation . As a control , we also tested the impact of silencing Neu3 , which is a constitutive cell surface sialidase . As shown in Figure 3G , shRNA silencing of Neu1 abrogated the LPS-induced disassociation of Siglec-E-TLR4 complexes . In contrast , shRNA silencing of Neu3 had no effect on Siglec-E-TLR4 interaction . To determine whether Neu1 regulates production of inflammatory cytokines , we tested three independent Neu1 shRNAs ( Neu1sh1-3 ) -silenced D2SC cell lines for their responses to LPS . As shown in Figure 3H , the three shRNAs reduced cell surface Neu1 in LPS-stimulated D2SC cells with different efficiencies: Neu1 was completely silenced by Sh1 and Sh2 , while only partially suppressed by Sh3 . Corresponding with the silencing efficiencies , LPS-induced cytokine production was more significantly reduced by Sh1 and Sh2 than by Sh3 , which had partial effect when compared to scramble control ( Figure 3I ) . Taken together , the data presented in this section demonstrate dynamic regulation of the TLR-Siglec-E interaction by Neu1 and its impact on DC response to LPS , the prototypic TLR4 ligand . To confirm the impact of LPS stimulation on Neu1 translocation and sialylation of primary leukocytes , we injected intraperitoneally ( i . p . ) sublethal doses of LPS ( 200 μg/mouse ) or PBS control into C57BL/6 mice , and analyzed spleen cells 16 hr later for cell surface expression of Neu1 . As shown in Figure 4A , LPS stimulation led to upregulation of cell surface Neu1 on DCs , macrophages , and neutrophils . Correspondingly , Siglec-E-Fc showed reduced binding to DCs after LPS stimulation , and this effect was rescued by incubation with a sialidase inhibitor , indicating that Siglec-E binding to its ligands was disrupted by sialidase activity ( Figure 4B ) . 10 . 7554/eLife . 04066 . 006Figure 4 . A critical role for hematopoietic cell-expressed Neu1 in endotoxic shock . ( A ) LPS stimulation in vivo increased cell surface Neu1 on DC , macrophage , and neutrophil . Data are representative of those from two independent experiments involving two mice per group . Splenocytes were collected from mice 16 hr after they received an injection of LPS ( i . p . 200 μg/mouse ) . ( B ) The cell surface Siglec-E ligands are down-regulated on DC following LPS stimulation . Data shown are means ± SD ( n = 3 ) and have been reproduced twice . ( C–H ) Lethally irradiated CD45 . 1 congenic B6 mice were transplanted with WT or Neu1−/− BM ( 5 × 106 cells/mouse ) . 21 weeks later , the mice were bled to analyze hematopoiesis . ( C ) Comparable reconstitution of donor-derived hematopoietic cells , based on the frequencies of donor-derived CD45 . 2+ cells . ( n = 5 ) . ( D ) Normal CD45 . 2+ leukocyte composition of PBL as determined by flow cytometry . Populations were defined as: B cells , B220+; monocytes , NK1 . 1−CD11b+; CD4+ T cells , CD4+CD3+; CD8+ T cells , CD3+CD8+; NK cells , NK1 . 1+ . ( E ) Increase of α2 , 3 and α2 , 6 sialylation and Siglec-E ligand on Neu1−/− DC . ( Gated CD11c+ as DCs ) . Splenocytes from Neu1+/+ and Neu1−/− mice were stained with FITC-conjugated SNA , MAA , or unconjugated Siglec-E-Fc ( detected with PE-anti-mouse IgG Fc ) in conjunction with APC-conjugated anti-CD11c mAb . Data shown are histrograms depicting the binding of SNA , MAA and Siglec-E-Fc respectively . ( F ) Siglec-E-Tlr4 association is increased on Neu1−/− splenocytes , as measured by captured Tlr4 in plates coated with Siglec-E-Fc . ( G ) Neu1 deficiency increases resistance to LPS challenge . Data shown are Kaplan Meier survival curves of mice that received the indicated doses of LPS ( i . p . , n = 5 in all groups ) . The mice were observed for 2 weeks , though all death occurred within 72 hr . ( H ) Cytokine production in blood measured 16 hr after LPS treatment . Data represent the mean ± SD . All data in this figure are representative of two to three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 006 Since we showed Neu1 to be a key regulator of Siglec-E-TLR4 interaction and of the TLR4 function in vitro , we sought a genetic model to definitively address the potential role of Neu1 in the host response to endotoxemia . However , mice with homozygous deletion of the Neu1 locus develop clinical abnormalities reminiscent of early-onset sialidosis in children , including severe nephropathy , progressive edema , splenomegaly , kyphosis and urinary excretion of sialylated oligosaccharides ( de Geest et al . , 2002 ) . To avoid these developmental defects , we produced bone marrow chimeric mice by reconstituting WT congenic mice with WT or Neu1−/− bone marrow cells , and compared their survival and cytokine responses following LPS challenge . Bone marrow from CD45 . 2+ Neu1−/− donors was as competent as WT bone marrow in reconstituting lethally irradiated CD45 . 1 hosts ( Figure 4C ) , and the major hematopoietic cellular components were not affected by Neu1 deficiency ( Figure 4D ) . Animals reconstituted with Neu1−/− bone marrow had increased sialylation and Siglec-E binding to DC ( Figure 4E ) , and more specifically , Siglec-E-TLR4 association ( Figure 4F ) . To test the role for Neu1 in susceptibility to LPS , we challenged chimeric mice with 100–400 μg LPS/mouse and monitored their survival over a 2 week period . As shown in Figure 4G , 100% of mice reconstituted with WT bone marrow cells succumbed to all doses of LPS tested . In contrast , the majority of mice reconstituted with Neu1−/− bone marrow were resistant to 100 and 200 μg doses of LPS . Corresponding with the increased resistance , the Neu1−/− >WT chimera mice produced significantly less IL-6 and TNFα following LPs stimulation ( Figure 4H ) . These data conclusively demonstrate a critical role for Neu1 in host response to endotoxemia . The significant reduction of endotoxemia in mice with Neu1-deficient hematopoietic cells suggests that endogenous sialidase may be a valuable therapeutic target . To test this concept , we injected 450 µg LPS i . p . into C57BL/6 mice . Immediately after LPS injection , the mice were treated with either PBS or a mixture of two sialidase inhibitors , Neu5Ac2en and Neu5Gc2en , that we previously found to protect mice against polymicrobial sepsis ( Chen et al . , 2011 ) . As shown in Figure 5A , while 80% of the vehicle-treated mice succumbed to LPS challenge , all mice that received the inhibitors survived throughout the observation period . Interestingly , the inhibitors had no significant effect on the first cytokine storm , but drastically inhibited IL-6 and TNF-α production 24 and 48 hr post LPS administration ( Figure 5B , C ) . Of the two inhibitors used , Neu5Gc2en accounted for most , if not all , of the therapeutic effect ( Figure 5D ) . 10 . 7554/eLife . 04066 . 007Figure 5 . Sialidase inhibitors protect mice against endotoxic shock and preserve Siglec-TLR interactions . ( A ) and ( D ) Survival analyses of mice that were treated with 450 µg/mouse of LPS ( i . p . , Escherichia coli 0111:B4 ) . The mice received NeuAc2en and/or NeuGc2en ( 100 µg/mouse/injection ) immediately after LPS administration and every 24 hr thereafter . ( n = 8–10 for a and n = 5–6 mice for ( D ) , 6–8 week old male mice were used ) . ( B ) and ( C ) Sialidase inhibitors reduce the levels of IL-6 ( B ) and TNF-α ( C ) , measured at indicated time after LPS treatment . Data shown are means ± SD . ( n = 8–10 , 6–8 week old male mice ) . ( E ) Therapeutic effect of sialidase inhibitors . C57BL/6 mice received 250 µg/mouse of LPS ( i . p . , E . coli 0111:B4 ) . The mice received NeuGc2en ( 100 µg/mouse/injection ) 6 hr after LPS administration and every 24 hr thereafter . ( n = 8–10 , 6–8 week old male mice ) . ( F ) Neu5Gc2en protects mice against lethal E . coli ( strain 25 , 922 , 107 CFU ) infection . ( G ) Sialidase inhibitor prevents LPS-induced disruption of the Siglec-TLR4 interaction . 1 × 107 splenocytes were cultured in RPMI supplemented with 10% FBS , stimulated with or without 2 μg/ml LPS in the presence of 100 μg/ml NeuGc2en or vehicle for 16 hr , and then the cells were lysed . The lysates were added to wells precoated with Siglece-Fc , Siglecf-Fc or hIgG-fc . The amounts of TLR4 captured were measured using anti-TLR4 mAb . All data in this figure are representative of two to three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 007 The fact that sialidase inhibitors have no effect on the first major wave of inflammatory cytokines suggests that the drug can be used therapeutically . Indeed , as shown in Figure 5E , Neu5Gc2en conferred complete protection even when it was administered 6 hr after LPS injection . Furthermore , since clinical manifestations are usually caused by a combination of bacteremia and enodoxemia in a more chronic manner ( Hurley , 1995 ) , we tested the protective effect of NEU1 inhibitor for infection with endotoxin-producing Escherichia coli . As shown in Figure 5F , Neu5Gc2en protected mice against lethal E . coli ( strain 25922 ) infection . As shown in Figure 1B , F , TLR4 interacts strongly with both Siglec-E and F . It is therefore of interest whether the sialidase inhibitor preserved the Tlr4-Siglec-E and/or Tlr4-Siglec-F interactions . We tested whether Neu5Gc2en preserved Tlr4-Siglec interaction using the in vitro capture assay . As shown in Figure 5G , LPS treatment reduced both Siglec-E:Tlr4 and Siglec-F:Tlr4 interactions , while treatment with Neu5Gc2en preserved both interactions . Functional redundancy of the two Siglecs explains lack of strong phenotype of mice with Siglec-E deletion in endotoxemia ( data not shown ) . As the first step to identify the therapeutic targets of Neu5Gc2en , we compared the inhibitory effect of Neu5Ac2en and Neu5Gc2en against Neu1-4 . We transfected 293T cells with cDNA encoding mouse Neu1-4 and use the lysates of the transfectants as the source of sialidases and determined the inhibition of their function by the sialidase inhibitors Neu5Ac2en and Neu5Gc2en . As shown in Figure 6A , while both Neu1 and Neu2 were efficiently inhibited by Neu5Ac2en and Neu5Gc2en , Neu3 and Neu4 are largely resistant . Importantly , while Neu2 was equally sensitive to both inhibitors , Neu1 was 10-fold more sensitive to Neu5Gc2en than to Neu5Ac2en . Since Neu5Gc2en but not Neu5Gc2en conferred a therapeutic effect ( Figure 5D ) , we hypothesized that Neu1 is likely the relevant target . To test this notion , we compared DC cultured from WT and Neu1−/− bone marrow for their cytokine response to LPS in the presence or absence of Neu5Gc2en . As shown in Figure 6B , targeted mutation of Neu1 reduced production of both IL-6 and TNFα by more than 70% . Importantly , while WT DC responded to Neu5Gc2en inhibition , Neu1−/− DC were largely resistant , confirming that Neu5Gc2en inhibits cytokine responses by targeting Neu1 . 10 . 7554/eLife . 04066 . 008Figure 6 . NeuGc2en targets Neu1 to inhibit inflammation and confer protection against endotoxemia . ( A ) Comparison of Neu5Ac2en and Neu5Gc2en for inhibitory activity against mouse Neu1-4 . Lysates from 293T cells transiently transfected with murine Neu1-4 cDNA were assayed for sialidase activity in the presence of indicated concentrations of inhibitors . Data shown are means ± SD of % inhibition of the activity of each sialidase by indicated concentration of inhibitors . ( B ) Neu5Gc2en targets Neu1 to inhibit production of inflammatory cytokine by DC in response to LPS . TNFα ( upper panel ) and IL-6 ( lower panel ) production by DC cultured from Neu1+/+ or Neu1−/− bone marrow . The DC were stimulated with LPS ( 100 ng/ml ) in the presence or absence of indicated doses of Neu5Gc2en . Data shown are means ± SD ( n = 3 ) and have been reproduced twice . ( C ) and ( D ) Chimeras consisting of either WT or Neu1−/− bone marrow were challenged with lethal doses of LPS ( 200 μg/mouse ) for WT ( C ) , and 400 μg/mouse for mutant chimeras ( D ) . Sialidase inhibitor was injected at 6 hr after LPS challenge . Data shown are Kaplan–Meier survival curves . Data are representative of those obtained from two to four independent experiments . Statistical significance of survival analysis was determined using log-rank tests , while the pairwise comparison was performed with Student's t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 008 To ascertain if NEU1 is the therapeutic target of Neu5Gc2en in endotoxic shock , we challenged bone marrow chimeric mice with a lethal dose of LPS ( 200 μg for WT chimera and 400 μg for Neu1−/− chimera , see Figure 4G ) and determined the effect of Neu5Gc2en , administrated at 6 hr after LPS challenge . As shown in Figure 6C , D , Neu5Gc2en was therapeutic only in WT , but not Neu1−/− chimeric mice . Therefore , Neu5Gc2en protects mice against endotoxemia by targeting Neu1 . While Neu1 normally resides in the lysosome , it has been shown to translocate to the plasma membrane after LPS stimulation ( Liang et al . , 2006 ) . To test whether the Neu5Gc2en inhibits cell surface or intracellular Neu1 , we first incubated either unstimulated or LPS-stimulated D2SC cells with either Neu5Gc2en or vehicle control for 1 hr and washed away unbound inhibitor . The cells were then lysed to measure sialidase activity . Regardless of LPS stimulation , all sialidase activity was eliminated by shRNA silencing of Neu1 , which indicated that this assay measured primarily Neu1 activity in the D2SC cells ( Figure 7A ) . Upon LPS stimulation , the total Neu1 activity was increased , and all activity was inhibited by Neu5Gc2en ( Figure 7A ) . In contrast to LPS-stimulated D2SC , a depot of Neu1 resistant to extracellularly-administered Neu5Gc2en was present in the unstimulated D2SC . However , this residual sialidase activity was eliminated by adding Neu5Gc2en to the lysates ( Figure 7A ) . Since unstimulated DC also have cell surface Neu1 ( Figure 3C ) , we suspect that Neu5Gc2en-sensitive Neu1 resides on the plasma membrane , while the Neu5Gc2en-resistant fraction is intracellular . To test this notion , we fractionated plasma membrane and cytoplasm from D2SC after the live cells were incubated with either vehicle or Neu5Gc2en , and compared their sialidase activities . As shown in Figure 7B , while the plasma membrane sialidase activity was significantly reduced , the cytoplasmic sialidase activity was unaffected . Since Neu5Gc2en selectively inhibited cell surface Neu1 , it is possible to preserve intracellular Neu1 function while targeting cell surface Neu1 with Neu5Gc2en . 10 . 7554/eLife . 04066 . 009Figure 7 . Exogenously added Neu5Gc2en inhibits cell surface but not intracellular Neu1 . ( A ) LPS stimulation increased sensitivity of Neu1 to exogenously added Neu5Gc2en . D2SC were transduced with lentiviral vectors encoding either scrambled or Neu1 shRNA and incubated with or without 1 µg/ml Neu5Gc2en for 30 min . After washing away the inhibitors , the lysates were analyzed for residual sialidase activity . Sialidase activity in the lysates was detected with 4-MU-NANA . To confirm that remaining activity is susceptible to Neu5Gc2en , the lysates were also assayed in the presence or absence of 1 µg/ml Neu5Gc2en . ( B ) When added to intact cells , Neu5Gc2en inhibits cell surface but not intracellular Neu1 . D2SC cells were incubated with 1 μg/ml of Neu5Gc2en for 1 hr . The plasma membrane and cytoplasmic membranes were prepared as described ( Liu and Fagotto , 2011 ) and measured for sialidase activity . The top panel shows Neu1 activity in cell membrane and cytosolic fractions of the D2SC cells ( triplicate data , and mean ± SEM ) , while the lower panel shows commonly used marker proteins to show the purities of the plasma membrane ( caveolin ) or GAPDH ( cytoplasm ) . Data are representative of those obtained from two independent experiments . Pairwise comparison was performed with Student's t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 009
Siglecs are cellular receptors for sialic acid-decorated biomolecular structures . Consistent with its intracellular ITIM- or ITIM-like motifs , engagement of Siglec-G with its natural ligands has been shown to inhibit both innate immune responses and adaptive T cell responses ( Chen et al . , 2009; Bandala-Sanchez et al . , 2013; Toubai et al . , 2014 ) . The function of Siglec 2 ( CD22 ) and Siglec-G in immune tolerance has also been demonstrated using both synthetic ligands and mice with deletions of the Cd22 and Siglecg genes ( Duong et al . , 2010; Jellusova et al . , 2010; Pfrengle et al . , 2013 ) . Targeted mutation of Siglecg revealed its critical role for inflammatory cytokine production to tissue injuries ( Chen et al . , 2009 ) and type I interferon response to RNA viruses ( Chen et al . , 2013 ) . Likewise , mutation of Siglece has been shown to enhance inflammatory cytokine production to bacterial infection ( Chang et al . , 2014 ) and neutrophil recruitment to lung ( McMillan et al . , 2013 ) . A fusion protein consisting of extracellular domain of Siglec-E and IgG-Fc ( Siglec-E-Fc ) is a potent inhibitor of TLR4 response in vitro ( Boyd et al . , 2009 ) . However , the target of Siglec-E-Fc was not identified . Despite these interesting observations , it is unclear how pattern recognition by Siglec family members may relate to other prototypic innate pattern recognition receptors such as TLRs . By showing a broad physical interaction between Siglecs and TLRs , our data provides a missing link between these two PRR families . The biological significance of the interactions is confirmed by the broad impact of Siglece mutation on response of DCs to all TLR ligands tested . Although a number of intracellular negative regulators have been described to regulate TLR function ( Kawai and Akira , 2010 ) , to our knowledge , Siglecs constitute the first class of cell surface pattern recognition receptors that are directly involved in regulation of TLR function . By virtue of the broad and direct interaction between TLRs and Siglecs , one may envision that under steady-state conditions , TLR-induced inflammation may be relatively moderate as TLR functions are either directly ( in case of Siglec-E ) or indirectly ( in case of Siglec-G ) restrained by sialoside-based pattern recognition . In case of tissue injury , a moderate inflammation may be beneficial for tissue remodeling and regeneration ( Takahashi et al . , 2008 ) . Infection can boost inflammation through microbial sialidase , as we have recently reported ( Chen et al . , 2011 ) , or as shown herein , by inducing translocation of host Neu1 to the cell surface to disarm the Siglec-mediated negative regulation of TLR function . Since the Neu1 translocation is induced by microbial TLR ligand , a positive feedback is created to release the Siglec brake for TLR activation . The molecular mechanism for TLR4-induced Neu1 translocation is largely unclear , although Neu1 translocation during differentiation from monocytes to macrophage occurs by a route used by MHC class II ( Liang et al . , 2006 ) . Engagement of TLR4 also activates cell surface Neu1 activity , perhaps by G protein-coupled receptor and matrix metalloproteinase-9-dependent mechanisms ( Amith et al . , 2009; Abdulkhalek et al . , 2011 ) . In addition to regulating Siglec-TLR interaction , Neu1 has also been shown to induce TLR4 dimerization and thus potentially stimulate TLR4 activity ( Amith et al . , 2010 ) . It is also of interest to note that TLR-mediated trans-regulation of cellular glycosylation was described through genetic screen in Drosophila ( Seppo et al . , 2003 ) , although the functional significance was not clear . The positive feedback described herein suggests a physiological function of TLR-induced glycosylation switch . Our data showed that Siglec-E mutation affects the function of both cell surface and endosomal TLRs . It is unclear how Siglec-E may affect endosomal TLRs . However , As a member of CD33-like Siglecs that are known to be readily endocytosed ( Walter et al . , 2008 ) , Siglec-E may form physical complexes with endosomal TLRs . The Siglec family of PRR may help the innate immune system to discriminate between infection and tissue injury by two distinct mechanisms . First , as represented by Siglec-G/10 , which does not directly associate with TLR , Siglecs may selective repress inflammation by binding to a DAMPs-associated natural ligand , such as CD24 , to inhibit host response to DAMPs . Second , as represented by Siglec-E which directly associates with TLR , Siglecs may regulate inflammation to endogenous TLR ligands by default . By either inducing translocation of intracellular sialidase or by producing microbial sialidases , infection can exacerbate inflammatory cytokine production to disrupt direct Siglec-TLR interactions ( Figure 8 ) . 10 . 7554/eLife . 04066 . 010Figure 8 . Sialoside-based pattern recognition and self-nonself discrimination by the innate immune system . TLR signaling is restrained by Siglecs that are either directly ( such as Siglec-E ) or indirectly through Cd24 ( such Siglec-G ) in the case of tissue injuries . Infections cause a positive feedback in TLR signaling . Infections cause translocation of Neu1 to cell surface and/or to production of bacterial/viral sialidases . Both host and cellular to desialylate TLR and/or CD24 . The Siglecs are dissociated from TLR to allow a more robust inflammation . DOI: http://dx . doi . org/10 . 7554/eLife . 04066 . 010 The broad impact of Siglecs in the inflammatory response suggests that the interaction with TLR may be preserved to reduce inflammation in cases of detrimental inflammation , such as sepsis . However , since one Siglec can interact with multiple TLRs , and each TLR with multiple Siglecs , it is unlikely that blocking specific Siglecs may significantly impact the susceptibility to endotoxemia . In contrast , as sialidases may regulate multiple Siglec-TLR interactions , targeting host sialidase may be more effective . Indeed , we showed that inhibition of Neu1 by either gene deletion or sialidase inhibitor administration confers a strong protection against endotoxemia . Therefore , Neu1 may serve as a new therapeutic target for endotoxic shock . Sepsis remains a major challenge despite advances in antibiotics , and many sepsis cases are caused by gram-negative bacteria ( Hurley , 1995; Angus et al . , 2001; Seyrantepe et al . , 2003; Dombrovskiy et al . , 2007 ) . We have recently shown bacterial sialidase as a valuable target for polybacterial sepsis ( Chen et al . , 2011 ) . However , most pathogens do not encode sialidase . With the identification of a critical function of endogenous sialidase , we can now extend the potential of this strategy to more sepsis-causing pathogens . Unlike bacterial and viral sialidases , endogenous sialidases may be easier to target as they are both limited in diversity and unlikely to acquire drug resistance through mutation . However , since mutation of Neu1 cause sialidosis ( Seyrantepe et al . , 2003 ) , it is of interest to consider how potential side-effects associated with Neu1 inactivation may be avoided . Fortunately , our data demonstrate that an effective sialidase inhibitor can confer protection by selectively targeting cell surface Neu1 , perhaps by virtue of poor drug accessibility to the intracellular compartment . Therefore , it is feasible to target Neu1 therapeutically without side effects associated with disruption of the intracellular sialidase activity , the known cause of sialidosis . Taken together , our data establish a missing link between the TLR and Siglec families of pattern recognition receptors . The specificity of individual Siglecs allow them to play distinct roles in innate immunity , while regulation of the interaction by host sialidase suggest a novel approach for treatment of inflammatory diseases , such as sepsis .
Recombinant proteins consisting of human IgG Fc and extracellular domains of human SIGLEC1 , 2 , 3 , 5 , 6 , 7 , 9 , 10 , 11 , and mouse Siglec1 , 2 , E , F were purchased from R&D Systems ( Minneapolis , MN ) . Recombinant mouse Siglec3-hIgGFc was purchased from Sino Biological , Inc . ( Beijing , China ) and anti-mouse TLR4 ( MTS510 ) from Biolegend ( San Diego , CA ) . Anti-human TLR4 and Anti-Siglec-E were from R&D . Anti-mouse CD11c , CD11b , CD4 , CD8 , B220 and Gr1 were purchased from eBioscience ( San Diego , CA ) . Anti-Neu1 ( H-300 ) and anti-Neu3 ( M-50 ) antibodies and Horseradish perioxidase conjugated anti-mouse , anti-goat or anti-rabbit secondary-step reagents , as well as anti-TLRs antibodies that are cross-reactive for mouse and human TLR1 ( H-90 ) , TLR2 ( A-9 ) , TLR3 ( M-300 ) , TLR5 ( M-300 ) , TLR6 ( N-18 ) , TLR7 ( N-20 ) , TLR8 ( H-114 ) , TLR9 ( H-100 ) and TLR10 ( V-20 ) were purchased from Santa Cruz Biotechnology ( Santa Cruz , CA ) and biotinylated for studies used in Figure 1A , B . Lipopolysaccharide ( LPS , from E . coli 0111:B4 ) was from Sigma . E . coli ( strain 25922 ) was obtained from ATCC ( Manassas , VA ) . Neu5Ac2en and Neu5Gc2en were synthesized as described ( Li et al . , 2008 , 2011 ) . D2SC cell line was obtained from Dr Yong-Jun Liu , and maintained in Dulbeco's minimal essential medium supplied with10% fetal calf serum and 1% penicillin and streptomycin . The lentiviral vectors expressing Neu1 shRNAs or Neu3 shRNA were from Thermo Scientific ( San Diego , CA ) . Puromycin was purchased from Sigma . Stable clones were obtained after selection with puromycin ( 2 . 5 μg/ml ) for 3 weeks after infection . All mice were used at 6–8 weeks of age . All animal procedures were approved by the Animal Care and Use Committee of University of Michigan and that of Children's National Medical Center . Mice with targeted mutations of Siglece were derived from 129/Sv ES cells produced by Mutant Mouse Regional Resource Center ( MMRRC ) at UC Davis ( Davis , CA ) . The mice were backcrossed to C57BL/6 for three generations . Bone marrow from Siglece+/+ and Siglece−/− littermates were used to prepare DC for the current studies . Since the 129/Sv has a mutation in the Caspase11 gene which is involved in the LPS response ( Broz et al . , 2012 ) , we also typed the genotype of Caspase11 and excluded mice homozygous Caspase11 mutation from the current study . Bone marrow chimeras were produced as described ( Chen et al . , 2008 ) , using a total of 5 × 106 bone marrow cells from either WT or Neu1−/− mice as donors , and WT mice as recipients . Neu1-4 and TLRs expression measured by real-time polymerase chain reaction . The primers for human TLR were from the TLR-primer set ( InvivoGen , version #09H31-MM , Toulouse , France ) ; those for murine Neu1-4 were:Neu1 Sense ATGTGACCTTCGACCCTGAGNeu1 antiSense TCCTTCTGCCAGGATGTACCNeu2 Sense GCTCTACCTGAAGAAGCAGAAGNeu2 antiSense GACATGGATTCATGGAGCGGTGNeu3 Sense TGCGTGTTCAGTCAAGCCNeu3 antiSense GCAGTAGAGCACAGGGTTACNeu4 Sense TGGTCTGCGGAGCCTGATATTGNeu4 antiSense AGTAACGCAGGCACACGGTAGHprt Sense AGCCTAAGATGAGCGCAAGTHprt antiSense TTACTAGGCAGATGGCCACA Samples were run in triplicate , and the relative expression was determined by normalizing expression of each target to the endogenous reference , hypoxanthine phosphoribosyltransferase ( Hprt ) transcripts or GAPDH . To generate a construct expressing recombinant proteins consisting of human IgG Fc and extracellular domains of Siglec-G or H , the corresponding cDNA fragment was amplified by PCR and subcloned into expression vector pFUSE-hIgG1-Fc ( Invivogen ) . cDNA for Neu1-4 were amplified by RT-PCR and subcloned into expression vector pCDNA6 ( Life technologies , Grand Island , NY ) . All constructs were verified by restriction enzyme digestion and DNA sequencing . For purification of Siglec-G-Fc or Siglec-H-Fc , the corresponding expression vector was co-transfected with a GFP expression vector into 293 T cells , and stable clones were obtained after 2 weeks culture in selection medium containing 2 . 5 μg/ml puromycin and 50 μg/ml Zeocin . The stable clones were amplified and cultured in serum free medium , Siglec-G-Fc or Siglec H-Fc was purified with a protein A column from the cell culture supernatants . Bone marrow cells were isolated from femurs and incubated in RPMI complete medium supplemented with 10% fetal calf serum ( Life Technologies ) and 10 ng/ml recombinant mouse GM-CSF ( Peprotech , Coconut Creek , Florida ) and 1 ng/ml IL-4 ( Peprotech ) for 12 days and then stimulated with different TLR ligands from the Mouse TLR1-9 Agonist Kit , InvivoGen . Cytokines in the supernatant were determined using mouse inflammation CBA kit ( 552364; BD Biosciences , San Diego , CA ) . Spleen cells from untreated WT mice , LPS treated WT mice , Neu1 knockout mice or culture cells were washed in flow staining buffer ( 1× PBS , 2% BSA ) , and incubated for 1 hr on ice with different directly conjugated-antibodies in flow staining buffer . The intensity of cell-bound antibodies was analyzed on a FACSCanto II cytometer ( Becton Dickinson , San Diego , CA ) . For Siglec-E binding assay , 1 µg of SiglecE-Fc incubated with the cells in 100 μl of the flow staining buffer . Unbound fusion proteins were washed and then incubated with PE-anti-mouse IgGFc antibody ( 1:400 ) for another hour on ice . The amounts of cell-bound Siglec-E-Fc was analyzed in Canto II cytometer . D2SC cell lysates were prepared in lysis buffer ( 20 mM Tris–HCl , 150 mM NaCl , 1% Triton X-100 , pH 7 . 6 , including protease inhibitors , 1 μg/ml leupeptin , 1 μg/ml aprotinin and 1 mM phenylmethylsulfonyl fluoride ) , sonicated , centrifuged at 13 , 000 rpm for 5 min and then diluted in IP buffer ( 20 mM Tris–HCl , 150 mM NaCl , pH 7 . 6 , including the protease inhibitors as described above ) . Samples were pre-cleared with 60 µl of protein A-conjugated agarose beads ( Upstate , Lake Placid , NY ) for 2 hr at 4°C or 37°C , and then incubated with corresponding antibodies . Immunoprecipitates were washed four times with IP buffer and re-suspended in SDS sample buffer for Western blot analysis . Coimmunoprecipitation between Neu1 and TLR4 were carried out after the live cells were crosslinked . Briefly , D2SC cell lines were stimulated with 2 μg/ml LPS or vehicle for 16 hr . Live cell suspension were incubated with with 1 mM DSP in the reaction buffer ( pH7 . 5 , 20 mM HEPES , 0 . 1 M phosphate , 0 . 15 M NaCl ) at room temperature for 30 min and then stopped by adding 20 mM pH 7 . 5 , Tris and incubated at room temperature for 15 min prior to lysis . 96-Well plates were coated with either Siglecs or IgGFc in 50 mM carbonate/bicarbonate buffer , pH 9 . 5 , overnight at 4°C . Wells were blocked with ELISA buffer ( 20 mM Tris–HCl , 2% bovine serum albumin , 150 mM NaCl , pH 7 . 6 ) for 1 hr . Spleen or THP-1 cell lysates were prepared in the lysis buffer ( 20 mM Tris–HCl , 1% Triton X-100 , 150 mM NaCl , PH 7 . 6 ) , sonicated , centrifuged at 13 , 000 rpm for 5 min and then diluted in ELISA buffer . 100 µl cell lysates ( 1 µg/µl ) were added to the plate and incubated for 2 hr . Between incubations ( all at 37°C ) , the plates were washed five times with the ELISA buffer . Biotinylated-anti-TLR antibody ( 0 . 05 μg/ml ) was used to detect bound TLRs . The plate-associated biotinylated proteins were detected by horse-radish perioxidase ( HRP ) -conjugated streptavidin ( 1:1000 ) for 1 hr and developed with 100 μl/well p-nitrophenyl phosphate liquid substrate system . Absorbance at 450 nm was recorded . In some experiments , cell lysates were replaced with recombinant TLR4 ectodomain ( 1 μg/ml , R&D ) in order to measure direct interaction between Siglecs and TLR4 . Blood or cell culture supernatants were obtained at indicated times and cytokines in the serum or cell culture supernatants were determined using mouse cytokine bead array designed for inflammatory cytokines ( 552364; BD Biosciences ) . Sialidase activity was measured using 2′- ( 4-methylumbelliferyl ) -α-D-N-acetylneuraminic acid sodium salt hydrate ( 4-MU-NANA ) ( M8639; Sigma ) as the substrate . 293 T cells ( 4 × 106 ) were transiently transfected with 10 µg of Neu1-4 expression vectors as described above or empty vector as control . 48 hr after transfection , cells were harvested and suspended in 100 µl of lysis buffer ( 20 mM Tris–HCl , 1% Triton X-100 , 150 mM NaCl , pH 7 . 6 ) , sonicated and centrifuged at 13 , 000 rpm for 5 min . For one reaction , 5 µl of the supernatant was first mixed with different amounts of sialidase inhibitors and then incubated with 4-MU-NANA ( final concentration , 15 µM ) for 30 min at 37°C in 50 µl reaction buffer ( 50 mM Sodium phosphate , pH 5 . 0 ) . The reaction was terminated by adding 600 µl stop buffer ( 0 . 25 M glycine-NaOH , pH 10 . 4 ) and then fluorescence intensity was measured with a FLUOstar OPTIMA ( BMG LABTECH GMBH , Germany ) ( excitation at 360 nm; emission at 460 nm ) . D2SC cells were seeded in the chamber slides and treated with 100 ng/ml LPS for 18 hr and then fixed with methanol/acetone ( 50:50 Vol ) for 20 min at −20°C . After washing with PBS , cells were blocked by PBS containing 5% bovine serum albumin and stained by rabbit anti-Neu1 polyclonal antibody ( H-300; Santa Cruz Biotech ) for 1 hr at room temperature . After washing with PBS , cells were stained by Alexa Fluor 568-conjucated anti-rabbit IgG ( Invitrogen , San Diego , CA ) . Subsequently cells were stained by rat anti-TLR4 monoclonal antibody ( Sa15-21; BioLegend ) and washed and stained by Alexa Fluor 488-conjucated anti-rat IgG ( Invitrogen ) . DNA was stained with DAPI . Fluorescence images were taken under an Olympus X51 microscope . The differences in cytokine concentrations were analyzed by the Student's t test . The differences in survival rates were analyzed by Kaplan-Meier plot and statistical significance determined using a log-rank test . *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . | Many living things have an immune system that is able to detect invading bacteria , viruses and other pathogens and trigger a response targeted against the threat before it causes lasting damage . Cells employ a number of different receptors that can detect these pathogens or the molecules that they produce . In animals , toll-like receptors ( or TLRs ) are a type of protein that recognizes patterns or structures that are found in many different types of pathogen , known as pathogen-associated molecular patterns ( or PAMPs ) . Injured cells release proteins that are also recognized by toll-like receptors and are called danger associated molecular patterns ( or DAMPs ) . An immune response is triggered when PAMPs and DAMPs are recognized , but the response must be properly controlled . If it goes awry , it can result in an over-activation of the immune cells that can lead to life-threatening conditions , one of which is called sepsis . Siglecs are proteins that bind to a sugar molecule , which is found attached to many other proteins , and are known to inhibit the immune response . However , it remained unclear how Siglecs do this and if they can interact directly with toll-like receptors . Chen et al . now show that most ( although not all ) Siglecs bind to TLRs , and that deleting the gene for a Siglec protein that can bind to multiple TLRs boosted the response of the immune cells to a range of microbial PAMPs . Deleting the gene for another Siglec that did not bind to any TLRs had no effect on the immune response . Chen et al . suggest that the Siglec proteins that interact with toll-like receptors act a bit like a brake that slows down the activation of the receptors . However , when an immune cell detects a foreign molecule through a TLR , an enzyme called Neu1 is relocated from the inside of the cell to the cell's surface , where it removes the sugar molecules from the TLRs . This disrupts the interaction between the TLRs and the Siglecs , thus activating the receptors and triggering an immune response against the invading pathogen or damaged cells . This represents a newly discovered mechanism that can regulate the signaling of TLRs . Chen et al . also show that a chemical compound that stops the function of the Neu1 enzyme prevents the toll-like receptors—and hence the immune cells—from becoming overly activated . Mice treated with this compound are protected against sepsis triggered by the presence of a bacterial PAMP . These results suggest that the Neu1 enzyme may be a promising new target for treating sepsis; further work will now be required to assess the potential side effects caused by inhibiting this enzyme . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2014 | Broad and direct interaction between TLR and Siglec families of pattern recognition receptors and its regulation by Neu1 |
One reported mechanism for morphine activation of dopamine ( DA ) neurons of the ventral tegmental area ( VTA ) is the disinhibition model of VTA-DA neurons . Morphine inhibits GABA inhibitory neurons , which shifts the balance between inhibitory and excitatory input to VTA-DA neurons in favor of excitation and then leads to VTA-DA neuron excitation . However , it is not known whether morphine has an additional strengthening effect on excitatory input . Our results suggest that glutamatergic input to VTA-DA neurons is inhibited by GABAergic interneurons via GABAB receptors and that morphine promotes presynaptic glutamate release by removing this inhibition . We also studied the contribution of the morphine-induced disinhibitory effect on the presynaptic glutamate release to the overall excitatory effect of morphine on VTA-DA neurons and related behavior . Our results suggest that the disinhibitory action of morphine on presynaptic glutamate release might be the main mechanism for morphine-induced increase in VTA-DA neuron firing and related behaviors .
Morphine is a potent analgesic with high addictive potential . Morphine-induced addictive behaviors are strongly dependent on the activation of dopamine ( DA ) neurons of the ventral tegmental area ( VTA ) ( Wise , 1989; Gardner , 2011; Luscher and Malenka , 2011 ) . The hedonic response produced by the activation of VTA-DA neurons by morphine is a primary factor behind drug dependence . One previously reported mechanism for morphine activation of VTA-DA neurons is the VTA-DA neuron disinhibition model ( Johnson and North , 1992; Kalivas , 1993; White , 1996 ) . Morphine first inhibits neighboring GABA inhibitory neurons , which shifts the balance between inhibitory and excitatory input to VTA-DA neurons in favor of excitation and then leads to the promotion of VTA-DA neuron excitation . It was also reported that the release of glutamate from least some of the glutamate terminals that synapsed on VTA-DA neurons was inhibited by opioids ( Manzoni and Williams , 1999; Margolis et al . , 2005 ) . This inhibitory effect of opioids on glutamate release is puzzling ( Chartoff and Connery , 2014 ) because one would expect morphine to produce rapid activation of VTA-DA neurons through both the inhibition of GABAergic input and the excitation of glutamatergic input . In addition , postsynaptic inhibition ( Ford et al . , 2006 ) or excitation ( Margolis et al . , 2014 ) in response to μ opioid receptor activation has also been reported in some VTA-DA neurons . However , the amplitude of the inhibitory outward currents produced by opioids in postsynaptic VTA-DA neurons in the study of Ford et al . was small ( 2 . 1 ± 1 . 5 pA in VTA-DA neurons projecting to the nucleus accumbens and 14 ± 4 pA in those projecting to the basolateral amygdaloid nucleus ) ( Ford et al . , 2006 ) , while only a small population of VTA-DA neurons ( 19% ) showed depolarization or an increase in firing rate in response to opioids in the study of Margolis et al . ( 2014 ) . This low postsynaptic responsiveness to opioids was consistent with physiological and anatomical evidence that only a few VTA-DA neurons had µ receptors ( Ford et al . , 2006 ) . Here , we propose that morphine might have a strengthening , rather than an inhibitory , effect on glutamatergic input to VTA-DA neurons because it was reported that blockade of excitatory glutamatergic signaling in the VTA suppressed morphine-induced VTA-DA neuron activation ( Jalabert et al . , 2011 ) and related behaviors ( Kalivas and Alesdatter , 1993; Harris and Aston-Jones , 2003; Harris et al . , 2004 ) . Moreover , the strengthening influence of morphine on glutamatergic input to VTA-DA neurons is more relevant to its promoting effect on VTA-DA neuron excitation . In this article , we report our observations of the effect of morphine on glutamatergic input to VTA-DA neurons using the whole-cell patch-clamp method , and the results of our study of its mechanism and functional consequences using electrophysiological , biochemical , optogenetic , pharmacological , and behavioral approaches .
We observed the effect of morphine ( 10 µM ) on the spontaneous firing frequency of VTA-DA neurons in rats . We select 10 µM of morphine because this concentration is commonly used in in vitro experiments as it elicits a significant effect useful for further analyses ( Akaishi et al . , 2000 ) , and the effect of morphine at this concentration can be essentially abolished by the opioid receptor antagonist naloxone ( Valentino and Dingledine , 1982 ) . VTA-DA neurons were identified based on Ih currents ( Figure 1A ) and tyrosine hydroxylase ( TH ) staining ( Figure 1B ) . Details of the identification method are given in the ‘Materials and methods’ section . Using original recordings of spontaneous firing ( left panel of Figure 2A ) and the time course of spontaneous firing ( middle panel of Figure 2A ) in VTA-DA neurons for comparison , we could see that morphine ( 10 μM ) increased the frequency of spontaneous firing in VTA-DA neurons . The average frequency of spontaneous firing increased from 1 . 0 ± 0 . 3 Hz before to 1 . 4 ± 0 . 4 Hz for 10–15 min after morphine application ( n = 6 cells from five rats , paired t test , p < 0 . 05 , compared to control before morphine , right panel of Figure 2A ) . In order to determine the role of glutamatergic input in the morphine-induced increase in the spontaneous firing frequency of VTA-DA neurons , we observed the influence of the N-methyl-D-aspartic acid ( NMDA ) receptor antagonist DL-2-amino-5-phosphonovaleric acid ( APV ) ( 50 μM ) and the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid ( AMPA ) receptor antagonist 6 , 7-Dinitroquinoxalie-2 , 3-dione ( DNQX ) ( 10 μM ) on the effect of morphine . In the presence of APV and DNQX , morphine no longer increased spontaneous firing frequency ( Figure 2B ) . The average frequency of spontaneous firing was 0 . 7 ± 0 . 1 Hz before and 0 . 8 ± 0 . 1 Hz for 10–15 min after morphine application in the presence of APV and DNQX ( n = 6 cells from five rats , paired t test , p > 0 . 05 , compared to control with APV and DNQX before morphine , right panel of Figure 2B ) . These results suggest that the morphine-induced increase in the spontaneous firing frequency of VTA-DA neurons requires AMPA and NMDA receptor-mediated glutamatergic input , consistent with a recent report using the NMDA antagonist APV and the AMPA receptor antagonist 6-Cyano-7-nitroquinoxaline-2 , 3-dione ( CNQX ) in in vivo experiments ( Jalabert et al . , 2011 ) . 10 . 7554/eLife . 09275 . 003Figure 1 . Identification of VTA-DA neurons in rats . ( A ) Electrophysiological properties of VTA-DA neurons . Left panel: representative traces showing a large hyperpolarization-activated current ( Ih ) in whole-cell voltage-clamp recording . Holding potential: −70 mV . Right panel: representative traces showing a large voltage ‘sag’ when hyperpolarized in whole-cell current-clamp recording . Holding current: 0 pA . ( B ) Immunohistochemical labeling of identified VTA-DA neurons . Panel 1: images of a Lucifer yellow-labeled neuron from the ventral tegmental area ( VTA ) after whole-cell patch-clamp recording under infrared differential interference contrast and fluorescent microscopy . Panel 2: the same neuron labeled with Lucifer yellow ( green color ) under confocal microscopy . Panel 3: VTA images showing tyrosine hydroxylase ( TH ) -positive neurons after immunostaining . Panel 4: Lucifer yellow-filled neuron co-labeled with TH . Scale bar: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 00310 . 7554/eLife . 09275 . 004Figure 2 . Effect of morphine on spontaneous firing of VTA-DA neurons in rats and the influence of the NMDA receptor antagonist APV and the AMPA receptor antagonist DNQX on the effect of morphine on spontaneous firing of VTA-DA neurons in rats . ( A ) Effect of morphine on spontaneous firing of VTA-DA neurons . Left panel: representative spontaneous firing traces before and after morphine ( 10 μM ) . Middle panel: time course of spontaneous firing before and after morphine ( 10 μM ) ( n = 6 cells from five rats ) . Right panel: average frequency of spontaneous firing before and after morphine ( n = 6 cells from five rats , p < 0 . 05 , compared to control before morphine ) . ( B ) Influence of the NMDA receptor antagonist APV and the AMPA receptor antagonist DNQX on the effect of morphine on spontaneous firing in VTA-DA neurons . Left panel: representative spontaneous firing traces before and after morphine ( 10 μM ) in the presence of APV ( 50 μM ) and DNQX ( 10 μM ) . Middle panel: time course of spontaneous firing before and after morphine in the presence of APV ( 50 μM ) and DNQX ( 10 μM ) ( n = 6 cells from five rats ) . Right panel: average frequency of spontaneous firing before and after morphine in the presence of APV ( 50 μM ) and DNQX ( 10 μM ) ( n = 6 cells from five rats , p = 0 . 34 ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 004 In order to study the effect of morphine on glutamatergic input to VTA-DA neurons , we examined the effect of morphine on the frequency of spontaneous excitatory postsynaptic currents ( sEPSCs ) in VTA-DA neurons in rats . First , we observed the effect of morphine on the frequency of sEPSCs when the GABAA receptor antagonist picrotoxin ( PTX ) was added to a bath solution to remove spontaneous inhibitory postsynaptic currents ( sIPSCs ) . Consistent with earlier reports ( Manzoni and Williams , 1999; Margolis et al . , 2005 ) , in the presence of extracellularly applied PTX ( 100 μM ) , morphine ( 10 μM ) decreased the frequency of sEPSCs ( Figure 3A ) . The average frequency of sEPSCs decreased from 4 . 2 ± 0 . 7 Hz before to 3 . 5 ± 0 . 7 Hz for 10–15 min after morphine application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before morphine , right panel of Figure 3A ) . However , bath application of GABAA receptor antagonists can lead to a wide blocking effect on GABAA receptors present on different types of neurons , including DA-neurons and GABA neurons , in VTA slices . To circumvent this wide influence and target the inhibition of GABAA receptors to VTA-DA neurons , we added the GABAA receptor antagonist PTX ( 100 µM ) to the internal solution of microelectrodes as described by Akaike et al . ( 1985 ) when recording sEPSCs . To demonstrate the effectiveness of the blockade of GABAA receptors by intracellularly applied PTX , we first examined the effect of intracellularly applied PTX ( 100 μM ) on inhibitory postsynaptic currents ( IPSC ) in VTA-DA neurons . The results showed that under this experimental paradigm , IPSCs disappeared ( top of panel 1 in Figure 3B , n = 5 cells from two rats ) , demonstrating that GABAA receptors in the recorded VTA-DA neurons were blocked by intracellularly applied PTX . In addition , to further confirm that the spontaneous events we measured in the presence of intracellularly applied PTX were in fact sEPSCs , we observed the effect of the AMPA receptor antagonist DNQX on spontaneous events in the presence of intracellularly applied PTX . The results showed that the spontaneous events were completely blocked by DNQX ( 10 μM ) ( bottom of panel 1 in Figure 3B ) . We repeated the experiment in three cells from different slices and obtained similar results . On this basis , we observed the effect of morphine on the sEPSCs in the presence of intracellularly applied PTX . Raw current traces ( panel 2 of Figure 3B ) and the time course of sEPSCs ( panel 3 of Figure 3B ) before and after morphine application in the presence of intracellularly applied PTX showed that morphine ( 10 μM ) increased the frequency of sEPSCs . The average frequency of sEPSCs increased from 4 . 8 ± 0 . 6 Hz before to 5 . 5 ± 0 . 6 Hz for 10–15 min after morphine application ( n = 8 cells from five rats , paired t test , p < 0 . 05 , compared to control before morphine , panel 4 of Figure 3B ) . However , morphine ( 10 μM ) had no significant effect on the amplitude of sEPSCs . The average amplitude of sEPSCs was 16 . 6 ± 1 . 3 pA before and 15 . 3 ± 1 . 3 pA for 10–15 min after morphine application ( n = 8 cells from five rats , paired t test , p > 0 . 05 , compared to control before morphine , panel 5 of Figure 3B ) . To confirm that the increased spontaneous events we measured in the presence of intracellularly applied PTX after morphine were also in fact sEPSCs , we observed the effect of the AMPA receptor antagonist DNQX on spontaneous events after morphine application in the presence of intracellularly applied PTX . The results showed that the spontaneous events after morphine application were completely blocked by adding DNQX ( 10 μM ) ( panel 6 of Figure 3B ) . We repeated the experiment in three cells from different slices and obtained similar results . We also used the first excitatory postsynaptic current ( EPSC ) of paired pulse facilitation ( PPF ) as an index of EPSC ( Maejima et al . , 2001 ) , and the PPF of paired EPSC as an indicator of presynaptic glutamate release ( Zucker and Regehr , 2002 ) to confirm the effect of morphine on presynaptic glutamate release in VTA-DA neurons in the presence of intracellularly applied PTX . As shown in the left panel of Figure 3C , morphine ( 10 μM ) increased the amplitude of the first EPSC , which was accompanied by a clear change in the presynaptic parameter PPF . The average amplitude of the first EPSCs was 124 . 1 ± 9 . 0 pA before and 161 . 9 ± 10 . 8 pA for 10–15 min after morphine application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before morphine , middle panel of Figure 3C ) . The average PPF was decreased from 1 . 5 ± 0 . 2 before to 1 . 2 ± 0 . 1 for 10–15 min after morphine application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before morphine , right panel of Figure 3C ) . These results supported the suggestion that morphine increased presynaptic glutamate release in VTA-DA neurons in the presence of intracellularly applied PTX . More importantly , when we clamped the membrane potential of VTA-DA neurons at the reversal potential of Cl− channels to remove sIPSCs , as an alternative to the application of a GABAA antagonist , either by the intracellular or bath approach , morphine still exerted a promoting effect on presynaptic glutamate release in VTA-DA neurons ( Figure 3D ) . The average frequency of sEPSCs in this condition increased from 4 . 4 ± 0 . 3 Hz before to 5 . 5 ± 0 . 4 Hz for 10–15 min after morphine application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before morphine , right panel of Figure 3D ) . 10 . 7554/eLife . 09275 . 005Figure 3 . Effects of morphine on the frequency of spontaneous excitatory postsynaptic currents ( sEPSCs ) and paired pulse facilitation ( PPF ) of VTA-DA neurons in rats . ( A ) Effects of morphine on the frequency of sEPSCs in the presence of extracellularly applied picrotoxin ( PTX ) in VTA-DA neurons . Left panel: typical current traces of sEPSCs before and after morphine ( 10 μM ) in the presence of extracellularly applied PTX . Middle panel: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of extracellularly applied PTX . Right panel: average frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of extracellularly applied PTX ( n = 6 cells from four rats , p < 0 . 05 , compared to control before morphine ) . ( B ) Effects of morphine on sEPSCs in the presence of intracellularly applied PTX in VTA-DA neurons . Top of panel 1: inhibitory postsynaptic currents ( IPSCs ) in the normal intracellular recording solution and the presence of intracellularly applied PTX . Bottom of panel 1: typical current traces of sEPSCs before and after DNQX ( 10 μM ) in the presence of intracellularly applied PTX . Panel 2: typical current traces of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX . Panel 3: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX . Panel 4: average frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX ( n = 8 cells from five rats , p < 0 . 05 , compared to control before morphine ) . Panel 5: average amplitude of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX ( n = 8 cells from five rats , p = 0 . 24 , compared to control before morphine ) . Panel 6: typical current traces of sEPSCs before and after DNQX ( 10 μM ) in the presence of intracellularly applied PTX and morphine . ( C ) Effects of morphine on the PPF in VTA-DA neurons . Left panel: representative traces of the PPF before and after morphine ( 10 μM ) , and superimposition of the two traces normalized to the first excitatory postsynaptic current ( EPSC ) before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX . Middle panel: average amplitude of the first EPSC in control and morphine ( 10 μM ) in the presence of intracellularly applied PTX ( n = 6 cells from four rats , p < 0 . 05 , compared to control before morphine ) . Right panel: average PPF before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX ( n = 6 cells from four rats , p < 0 . 05 , compared to control before morphine ) . ( D ) Effects of morphine on the frequency of sEPSCs when VTA-DA neurons were clamped the membrane potential at the reversal potential of Cl− channels in VTA-DA neurons . Left panel: typical current traces of sEPSCs before and after morphine ( 10 μM ) when DA neurons was clamped the membrane potential at the reversal potential of Cl− channels . Middle panel: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) when DA neurons was clamped the membrane potential at the reversal potential of Cl− channels . Right panel: average frequency of sEPSCs before and after morphine ( 10 μM ) when DA neurons was clamped the membrane potential at the reversal potential of Cl− channels ( n = 6 cells from four rats , p < 0 . 05 , compared to control before morphine ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 005 The mechanism underlying the promoting effect of morphine on presynaptic glutamate release in VTA-DA neurons may involve different processes . One suggestion is that morphine may directly act at glutamatergic terminals to promote glutamate release . To test this hypothesis , we studied the effect of morphine on glutamate release from the VTA synaptosomes of rats , which are sealed particles containing vesicles , viable mitochondria , and all components necessary to store , release , and retain neurotransmitters ( Breukel et al . , 1997 ) , using on-line fluorometry . However , we did not find that morphine ( 10 μM ) had a direct effect at glutamatergic terminals to promote glutamate release . The average concentration of glutamate was 8 . 1 ± 0 . 1 nmol/mg before and 8 . 5 ± 0 . 7 nmol/mg after morphine application ( n = 6 samples from eight rats , paired t test , p > 0 . 05 , Figure 4A ) . We also studied the effect of morphine on the frequency of sEPSCs from mechanically dissociated single VTA-DA neurons in rats , which retained functional synaptic terminals ( Akaike and Moorhouse , 2003; Ye et al . , 2004; Deng et al . , 2009 ) . The left top image of panel 1 of Figure 4B shows mechanically dissociated neurons from the VTA . These VTA-DA neurons were identified by only using Ih currents ( the graph on the right of panel 1 of Figure 4B ) without TH staining because in dissociated DA neurons , it was difficult to fix the cell after recording for TH staining . The results showed that morphine ( 10 µM ) had no significant effect on the frequency of sEPSCs ( panels 2 and 3 of Figure 4B ) . The average frequency of sEPSCs was 2 . 1 ± 0 . 3 Hz before and 2 . 0 ± 0 . 2 Hz for 10–15 min after morphine application ( 10 µM ) ( n = 6 cells from six rats , paired t test , p > 0 . 05 , compared to control before morphine , panel 4 of Figure 4B ) . These results suggest that morphine may promote presynaptic glutamate release in VTA-DA neurons in an indirect way . 10 . 7554/eLife . 09275 . 006Figure 4 . Effects of morphine on glutamate release from the ventral tegmental area ( VTA ) synaptosomes of rats and the frequency of sEPSCs in mechanically dissociated VTA-DA neurons from rats . ( A ) Effects of morphine on glutamate release from VTA synaptosomes . Average concentration of glutamate release before and after application of morphine ( 10 μM ) from VTA synaptosomes ( n = 6 samples from eight rats , p = 0 . 46 ) . ( B ) Effects of morphine on the frequency of sEPSCs in mechanically dissociated VTA-DA neurons . Panel 1 , left: images of an acutely dissociated single neuron from the VTA under phase contrast microscopy . Scale bar: 10 μm . Panel 1 , right: representative current traces showing a large hyperpolarization-activated current ( Ih ) in whole-cell voltage-clamp recording . Holding potential: −70 mV . Panel 2: typical current traces of sEPSC before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX . Panel 3: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX . Panel 4: Average frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of intracellularly applied PTX ( n = 6 cells from six rats , p = 0 . 65 ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 006 To explore how morphine promoted presynaptic glutamate release in VTA-DA neurons , we hypothesized that glutamatergic input to VTA-DA neurons was inhibited by GABAergic interneurons and morphine disinhibited glutamatergic input by removing this inhibition , thus promoting glutamate release . To test this hypothesis , we first studied whether GABA could inhibit presynaptic glutamate release in VTA-DA neurons by examining the effect of exogenous applied of GABA on the frequency of sEPSCs of VTA-DA neurons in rats . From raw current traces ( left panel of Figure 5A ) and the time course of sEPSCs ( middle panel of Figure 5A ) , we could see that GABA ( 10 μM ) apparently decreased the frequency of sEPSCs . The average frequency of sEPSCs decreased from 6 . 7 ± 0 . 7 Hz before to 4 . 7 ± 0 . 6 Hz for 10–15 min after GABA application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before GABA , right panel of Figure 5A ) . Then , we observed whether the activation of intrinsic GABAergic neurons could inhibit presynaptic glutamate release of VTA-DA neurons in mice . To do this , AAV virus expressing a double floxed-stopped channelrhodopsin-2 ( ChR2 ) -mCherry was stereotaxically injected into the VTA of mice expressing Cre recombinase in GABA neurons . 2 weeks after infection , expression of ChR2–mCherry was observed in the VTA ( panel 1 of Figure 5B ) . We then performed whole-cell patch-clamp recording in GABA neurons of the VTA and observed the light-induced action potentials in GABA neurons in order to confirm that ChR2 was indeed expressed in the GABA neurons of the VTA in mice . The results showed that 470 nm light stimulation ( 20 Hz ) elicited action potentials in GABA neurons ( panel 2 of Figure 5B ) . On this basis , whole-cell patch-clamp recording was performed in VTA-DA neurons to observe the effect of 470 nm light stimulation on the frequency of sEPSCs in the presence of intracellularly applied PTX ( 100 µM ) . Following 5–10 min of baseline recording of sEPSCs , twenty 470 nm light pulses of 5 ms at 20 Hz were delivered every 4 s for 15 min . Raw current traces ( panel 3 of Figure 5B ) and the time course of sEPSCs ( panel 4 of Figure 5B ) before and after light stimulation showed that the light stimulation apparently decreased the frequency of sEPSCs . The average frequency of sEPSCs decreased from 4 . 1 ± 0 . 7 Hz before to 2 . 7 ± 0 . 3 Hz for 10–15 min after light stimulation ( n = 6 cells from five mice , paired t test , p < 0 . 05 , compared to control before light stimulation , panel 5 of Figure 5B ) . 10 . 7554/eLife . 09275 . 007Figure 5 . Effect of exogenous application of GABA and 470 nm light stimulation on the frequency of sEPSCs in the presence of intracellularly applied PTX in VTA-DA neurons . ( A ) Effect of exogenous application of GABA on the frequency of sEPSCs of VTA-DA neurons in rats . Left panel: typical current traces of sEPSCs before and after GABA ( 10 μM ) in the presence of intracellularly applied PTX . Middle panel: typical time course of the frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of intracellularly applied PTX . Right panel: average frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of intracellularly applied PTX ( n = 6 cells from four rats , p < 0 . 05 , compared to control before GABA ) . ( B ) Effect of 470 nm light stimulation on the frequency of sEPSCs of VTA-DA neurons in mice . Panel 1: coronal image showing the expression of ChR2-mCherry ( red ) following injection of the viral construct bilaterally into the ventral tegmental area ( VTA ) of GADcre+ mice . Scale bar: 500 µm . Panel 2: 470 nm light ( 20 Hz ) -induced firing of VTA GABA neurons in current-clamp mode . Panel 3: typical current traces of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of intracellularly applied PTX . Panel 4: typical time course of the frequency of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of intracellularly applied PTX . Panel 5: average frequency of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of intracellularly applied PTX ( n = 6 cells from five mice , p < 0 . 05 , compared to control before light stimulation ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 007 We also studied which kinds of GABA receptors ( GABAA or GABAB receptors ) mediated the decreasing effect of GABA on presynaptic glutamate release in VTA-DA neurons by examining the influence of the GABAA or GABAB receptor antagonist on the effect of GABA on the frequency of sEPSCs in rats . The results showed that the GABAA receptor antagonist PTX ( 100 µM ) had no significant influence on the effect of GABA ( Figure 6A ) . The average frequency of sEPSCs still decreased from 4 . 0 ± 0 . 3 Hz before to 3 . 0 ± 0 . 2 Hz for 10–15 min after GABA application in the presence of PTX ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to PTX before GABA , right panel of Figure 6A ) . The percentage of GABA-produced response in the presence of PTX ( −25 . 8 ± 3 . 8% ) was not statistically significant ( n = 6 , independent t test , p > 0 . 05 ) compared to that without PTX ( −29 . 2 ± 6 . 0% ) . However , in the presence of the GABAB receptor antagonist CGP54626 , the effect of GABA on the frequency of sEPSCs disappeared ( Figure 6B ) . The average frequency of sEPSCs was 3 . 8 ± 0 . 4 Hz before and 3 . 8 ± 0 . 4 Hz for 10–15 min after GABA application in the presence of CGP54626 ( 2 µM ) ( n = 6 cells from four rats , paired t test , p > 0 . 05 , compared to CGP54626 before GABA , right panel of Figure 6B ) . These results suggest that it is GABAB receptors , rather than GABAA receptors , that mediate the decreasing effect of GABA on presynaptic glutamate release in VTA-DA neurons . 10 . 7554/eLife . 09275 . 008Figure 6 . Influence of the GABAA receptor antagonist PTX and the GABAB receptor antagonist CGP54626 on the effect of exogenous application of GABA as well as the influence of the GABAB receptor antagonist CGP54626 on 470 nm light-induced inhibition of the frequency of sEPSCs in VTA-DA neurons . ( A ) Influence of the GABAA receptor antagonist PTX on the effect of exogenous application of GABA of VTA-DA neurons in rats . Left panel: typical current traces of sEPSC before and after GABA ( 10 μM ) in the presence of PTX ( 100 μM ) . Middle panel: typical time course of the frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of PTX ( 100 μM ) . Right panel: average frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of PTX ( 100 μM ) ( n = 6 cells from four rats , p < 0 . 05 , compared to PTX before GABA ) . ( B ) Influence of the GABAB receptor antagonist CGP54626 on the effect of exogenous application of GABA of VTA-DA neurons in rats . Left panel: typical current traces of sEPSC before and after GABA ( 10 μM ) in the presence of CGP54626 ( 2 μM ) . Middle panel: typical time course of the frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of CGP54626 ( 2 μM ) . Right panel: average frequency of sEPSCs before and after GABA ( 10 μM ) in the presence of CGP54626 ( 2 μM ) ( n = 6 cells from four rats , p = 0 . 87 ) . ( C ) Influence of the GABAB receptor antagonist CGP54626 on 470 nm light-induced inhibition of the frequency of sEPSCs of VTA-DA neurons in mice . Left panel: typical current traces of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of CGP54626 ( 2 μM ) . Middle panel: typical time course of the frequency of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of CGP54626 ( 2 μM ) . Right panel: average frequency of sEPSCs before and after blue light ( 470 nm ) stimulation in the presence of CGP54626 ( 2 μM ) ( n = 6 cells from five mice , p = 0 . 21 ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 008 To study the role of GABAB receptors in intrinsic GABA-induced inhibition of presynaptic glutamate release in VTA-DA neurons , we observed the effect of the GABAB receptor antagonist CGP54626 on the 470 nm light-induced inhibition of the frequency of sEPSCs of VTA-DA neurons in mice . From raw current traces ( left panel of Figure 6C ) and the time course of sEPSCs ( middle panel of Figure 6C ) before and after the 470 nm light stimulation in the presence of CGP54626 ( 2 µM ) , we could see that the inhibitory effect of the 470 nm light stimulation on the frequency of sEPSCs disappeared in the presence of CGP54626 . The average frequency of sEPSCs was 4 . 2 ± 0 . 3 Hz before and 4 . 1 ± 0 . 2 Hz for 10–15 min after 470 nm light stimulation in the presence of CGP54626 ( n = 6 cells from five mice , paired t test , p > 0 . 05 , compared to control before 470 nm light stimulation in the presence of CGP54626 , right panel of Figure 6C ) . We examined the expression of GABAB receptors in the presynaptic terminals of the VTA in rats . Western blotting indicated that GABAB receptors were present in synaptosomes from the VTA ( Figure 7A ) . Immunohistochemistry results for GABAB receptors , VGLUT2 ( a glutamatergic terminal marker ) and TH ( a DA neuron marker ) showed that GABAB receptors ( green color , panel 1 of Figure 7B ) , VGLUT2 ( red color , panel 2 of Figure 7B ) and TH-positive neurons ( blue color , panel 3 of Figure 7B ) were present in VTA slices and the coexpression of GABAB receptors and VGLUT2 ( yellow color , panel 4 of Figure 7B ) indicated that GABAB receptors were present in glutamatergic terminals of the VTA in rats . Moreover , this coexpression was close to TH-positive neurons ( panel 4 of Figure 7B ) . 10 . 7554/eLife . 09275 . 009Figure 7 . The presence of GABAB receptors in the presynaptic glutamatergic terminals of the ventral tegmental area ( VTA ) in rats . ( A ) GABAB receptor expression in the synaptosomes from the VTA . A representative Western blot shows GABAB receptor expression in the synaptosomes from the VTA . ( B ) GABAB receptor expression in the VTA shown using the triple-immunofluorescence method . Panel 1: GABAB receptor immunolabeling ( green-colored ) in the VTA . Panel 2: VGLUT2-labeled axon terminals ( red-colored ) in the VTA . Panel 3: DA neuron labeled with TH ( blue-colored ) in the VTA . Panel 4: Merged image ( yellow-colored ) of GABAB receptors and VGLUT2 in the VTA . The insets marked with small white rectangles in panels 1–4 are magnified views . Scale bars: 5 µm ( in panels 1–4 ) ; 1 µm ( in the insets in panels 1–4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 009 In addition , to determine whether there is basal GABAergic inhibitory control through GABAB receptors on presynaptic glutamate release in VTA-DA neurons , we observed the effect of the GABAB receptor antagonist CGP54626 on the frequency of sEPSCs in rats . The results showed that after the application of CGP54626 , the frequency of sEPSCs significantly increased ( Figure 8A ) . The average frequency of sEPSCs was 3 . 6 ± 0 . 3 Hz before and 4 . 5 ± 0 . 3 Hz for 10–15 min after CGP54626 application ( 2 µM ) ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before CGP54626 , right panel of Figure 8A ) . However , the GABAA receptor antagonist PTX had no significant effect on the frequency of sEPSCs ( Figure 8B ) . The average frequency of sEPSCs was 5 . 8 ± 1 . 0 Hz before and 5 . 7 ± 0 . 9 Hz for 10–15 min after PTX ( 100 µM ) application ( n = 6 cells from four rats , paired t test , p > 0 . 05 , compared to control before PTX , right panel of Figure 8B ) . This is consistent with the above result showing that GABA-mediated inhibition of presynaptic glutamate release in VTA-DA neurons is via GABAB receptors rather than GABAA receptors . 10 . 7554/eLife . 09275 . 010Figure 8 . Effect of the GABAB receptor antagonist CGP54626 and the GABAA receptor antagonist PTX on the frequency of sEPSCs of VTA-DA neurons in rats . ( A ) Effect of the GABAB receptor antagonist CGP54626 on the frequency of sEPSCs in VTA-DA neurons . Left panel: typical current traces of sEPSCs before and after CGP54626 ( 2 μM ) . Middle panel: typical time course of the frequency of sEPSCs in DA neurons before and after CGP54626 ( 2 μM ) . Right panel: average frequency of sEPSCs before and after CGP54626 ( 2 μM ) ( n = 6 cells from four rats , p < 0 . 05 , compared to control before CGP54626 ) . ( B ) Effect of the GABAA receptor antagonist PTX on the frequency of sEPSCs in VTA-DA neurons . Left panel: typical current traces of sEPSCs before and after PTX ( 100 μM ) . Middle panel: typical time course of the frequency of sEPSCs before and after PTX ( 100 μM ) . Right panel: average frequency of sEPSCs before and after PTX ( 100 μM ) ( n = 6 cells from four rats , p = 0 . 29 ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 010 To study whether the morphine-induced increase in presynaptic glutamate release in VTA-DA neurons was via presynaptic disinhibition , we observed the effect of ‘closing’ local GABAergic interneurons on the morphine-induced increase in glutamate release in VTA-DA neurons using optogenetic methods in mice . AAV virus expressing a double floxed-stopped eNpHR3 . 0-EYFP was stereotaxically injected into the VTA of mice expressing Cre recombinase in GABA neurons . 2 weeks after infection , we performed whole-cell patch-clamp recording in GABA neurons of the VTA and observed light-induced inhibition of firing of action potentials in GABA neurons . The results showed that yellow light stimulation ( 590 nm ) could reliably inhibit the current injection-induced firing of action potentials in GABA neurons ( left panel of Figure 9A ) . On this basis , whole-cell patch-clamp recording was performed in VTA-DA neurons to observe the influence of the light-induced disinhibition of glutamatergic input on the effect of morphine . First , we observed whether this inhibition by light stimulation of GABA neurons affected the frequency of sEPSCs in VTA-DA neurons . As expected , light stimulation increased the frequency of sEPSCs in VTA-DA neurons ( middle panel of Figure 9A ) . The average frequency of sEPSCs was 3 . 9 ± 0 . 4 Hz before and 4 . 5 ± 0 . 4 Hz for 10–15 min after light stimulation ( n = 6 cells from five mice , paired t test , p < 0 . 05 , compared to control before light stimulation , right panel of Figure 9A ) . Then , we observed the influence of the light-induced disinhibition of glutamatergic input on the effect of morphine . From raw current traces ( left panel of Figure 9B ) and the time course of sEPSCs ( middle panel of Figure 9B ) , we could see that the effect of morphine on the frequency of sEPSCs disappeared in the presence of the light stimulation . The average frequency of sEPSCs was 4 . 5 ± 0 . 4 Hz before and 4 . 4 ± 0 . 5 Hz for 10–15 min after morphine application in the presence of the light stimulation ( n = 6 cells from five mice , paired t test , p > 0 . 05 , compared to control before morphine with light stimulation , right panel of Figure 9B ) . These results suggest that prior removal of the inhibition of GABAergic input on presynaptic glutamate release leads to disappearance of the effect of morphine on presynaptic glutamate release of VTA-DA neurons , indicating that morphine promotes presynaptic glutamate release of VTA-DA neurons via presynaptic disinhibition . 10 . 7554/eLife . 09275 . 011Figure 9 . Effect of 590 nm light stimulation on the frequency of sEPSCs and influence of light-induced disinhibition of glutamatergic input and the GABAB receptor antagonist CGP54626 on the effect of morphine on the frequency of sEPSCs in VTA-DA neurons . ( A ) Effect of 590 nm light stimulation on the frequency of sEPSCs of VTA-DA neurons in mice . Left panel: 590 nm light inhibits current injection-induced firing of action potentials in VTA GABA neurons . Middle panel: typical time course of the frequency of sEPSCs before and after yellow light ( 590 nm ) stimulation . Right panel: average frequency of sEPSCs before and after yellow light ( 590 nm ) stimulation ( n = 6 cells from five mice , p < 0 . 05 , compared to control before light stimulation ) . ( B ) Influence of light-induced disinhibition of glutamatergic input on the effect of morphine on the frequency of sEPSCs of VTA-DA neurons in mice . Left panel: typical current traces of sEPSC before and after morphine ( 10 μM ) in the presence of yellow light ( 590 nm ) stimulation . Middle panel: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of yellow light ( 590 nm ) stimulation . Right panel: average frequency of sEPSCs before and after morphine in the presence of yellow light ( 590 nm ) stimulation ( n = 6 cells from five mice , p = 0 . 78 ) . ( C ) Influence of the GABAB receptor antagonist CGP54626 on the effect of morphine on the frequency of sEPSCs of VTA-DA neurons in rats . Left panel: typical current traces of sEPSCs before and after morphine ( 10 μM ) in the presence of CGP54626 ( 2 μM ) . Middle panel: typical time course of the frequency of sEPSCs before and after morphine ( 10 μM ) in the presence of CGP54626 ( 2 μM ) . Right panel: plots of the average frequency of sEPSCs in control , in CGP54626 ( 2 μM ) , and in morphine ( 10 μM ) ( n = 6 cells from four rats; p < 0 . 05 , CGP54626 compared to control , p < 0 . 05 , morphine compared to control ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 , #p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 011 In addition , we used the GABAB receptor antagonist CGP54626 to remove inhibition of presynaptic glutamate release by GABAergic input and then observed its influence on the effect of morphine on glutamate release in rats . The results showed that after application of the GABAB receptor antagonist CGP54626 ( 2 µM ) , the frequency of sEPSCs significantly increased ( Figure 9C ) . The average frequency of sEPSCs was 2 . 5 ± 0 . 4 Hz before and 3 . 0 ± 0 . 4 Hz for 5–10 min after CGP54626 application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before CGP54626 , right panel of Figure 9C ) , suggesting that CGP54626 induced a disinhibitory effect on presynaptic glutamate release . On this basis , morphine ( 10 µM ) was applied in the same cell , but did not further increase the frequency of sEPSCs ( Figure 9C ) . The average frequency of sEPSCs was 3 . 0 ± 0 . 4 Hz before and 2 . 8 ± 0 . 4 Hz for 5–10 min after morphine application in the presence of CGP54626 ( n = 6 cells from four rats , paired t test , p > 0 . 05 , compared to CGP54626 before morphine , right panel of Figure 9C ) . We studied the influence of the selective presynaptic GABAB receptor antagonist CGP36216 on the effect of morphine on the frequency of spontaneous firing of VTA-DA neurons in rats . First , we observed the effect of CGP36216 on the frequency of spontaneous firing in VTA-DA neurons . The results showed that CGP36216 ( 100 µM ) could increase the frequency of spontaneous firing in VTA-DA neurons ( Figure 10A ) . The average frequency of spontaneous firing increased from 1 . 8 ± 0 . 3 Hz before to 2 . 1 ± 0 . 3 Hz after CGP36216 application ( n = 6 cells from four rats , paired t test , p < 0 . 05 , compared to control before CGP36216 , right panel of Figure 10A ) . Then , we observed the influence of CGP36216 on the effect of morphine on the frequency of spontaneous firing in VTA-DA neurons . As shown by raw spontaneous firing traces ( left panel of Figure 10B ) and the time course of spontaneous firing ( middle panel of Figure 10B ) in VTA-DA neurons , the effect of morphine ( 10 µM ) disappeared in the presence of CGP36216 ( 100 µM ) . The average frequency of spontaneous firing in VTA-DA neurons was 1 . 3 ± 0 . 2 Hz before and 1 . 2 ± 0 . 2 Hz for 10–15 min after morphine application in the presence of CGP36216 ( n = 6 cells from four rats , paired t test , p > 0 . 05 , compared to CGP36216 before morphine , right panel of Figure 10B ) . 10 . 7554/eLife . 09275 . 012Figure 10 . Effect of the selective presynaptic GABAB receptor antagonist CGP36216 on spontaneous firing and the influence of the selective presynaptic GABAB receptor antagonist CGP36216 on the effect of morphine on spontaneous firing of VTA-DA neurons in rats . ( A ) Effect of the selective presynaptic GABAB receptor antagonist CGP36216 on spontaneous firing in VTA-DA neurons . Left panel: representative spontaneous firing traces before and after CGP36216 ( 100 μM ) . Middle panel: time course of spontaneous firing before and after CGP36216 ( 100 μM ) ( n = 6 cells from four rats ) . Right panel: average frequency of spontaneous firing before and after CGP36216 ( 100 μM ) ( n = 6 cells from four rats , p < 0 . 05 , compared to control before CGP36216 ) . ( B ) Influence of the selective presynaptic GABAB receptor antagonist CGP36216 on the effect of morphine on spontaneous firing in VTA-DA neurons . Left panel: representative spontaneous firing traces before and after morphine ( 10 μM ) in the presence of CGP36216 ( 100 μM ) . Middle panel: time course of spontaneous firing before and after morphine ( 10 μM ) in the presence of CGP36216 ( 100 μM ) ( n = 6 cells from four rats ) . Right panel: average frequency of spontaneous firing before and after morphine ( 10 μM ) in the presence of CGP36216 ( 100 μM ) ( n = 6 cells from four rats , p = 0 . 35 , compared to CGP36216 before morphine ) . Data are shown as the mean ±s . e . m . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 012 Since enhanced DA function in the VTA has most often been assessed as increased locomotor activity at the behavioral level ( Vezina , 2004 ) , we also studied the contribution of the morphine-induced disinhibition of glutamatergic input in the VTA to behavioral changes produced by morphine . In behavioral experiments , to avoid morphine acting on other brain areas sensitive to morphine , we injected morphine locally into the VTA in rats . Injection sites were verified under light microscope ( left panel of Figure 11A ) ; the injection sites used for data analysis are shown in the middle and right panels of Figure 11A . We first observed the effect of CGP36216 on locomotor activity . The results showed that CGP36216 ( 20 μg/rat ) increased locomotor activity . The distance traveled by rats was 6 . 3 ± 1 . 4 m in the control group compared with 12 . 2 ± 2 . 0 m in the CGP36216 group ( n = 6 rats , independent t test , p < 0 . 05 ) . Then we observed the influence of CGP36216 on the effect of morphine on locomotor activity in rats . As shown in Figure 11B , locomotor activity was significantly increased following intra-VTA injection of morphine ( 1 µg/rat ) . The distance traveled by rats was 6 . 3 ± 1 . 4 m in the control group compared with 37 . 6 ± 9 . 6 m in the morphine alone group ( n = 6 rats , independent t test , p < 0 . 05 , right panel of Figure 11B ) . However , in animals with intra-VTA injection of CGP36216 ( 20 µg/rat ) , the morphine-induced increase in locomotor activity disappeared ( Figure 11B ) . The distance traveled by rats was 37 . 6 ± 9 . 6 m in the morphine alone group compared with 7 . 5 ± 1 . 4 m in the morphine plus intra-VTA CGP36216 group ( n = 6 rats , independent t test , p < 0 . 05 , right panel of Figure 11B ) . In addition , to evaluate the role of VTA GABAB receptors in the overall effect of intraperitoneal ( i . p . ) morphine , we observed the effect of intra-VTA injected CGP36216 on i . p . morphine-induced increase in locomotor activity . As shown in Figure 11C , locomotor activity was significantly increased by i . p . injection of morphine ( 10 mg/kg ) plus intra-VTA injection of saline . The distance traveled by rats was 5 . 4 ± 0 . 8 m in the control group compared with 28 . 3 ± 5 . 7 m in the i . p . morphine ( 10 mg/kg ) plus intra-VTA saline group ( n = 6 rats , independent t test , p < 0 . 05 , right panel of Figure 11C ) . However , in animals with intra-VTA injection of CGP36216 ( 20 µg/rat ) , the i . p . morphine-induced increase in locomotor activity disappeared ( Figure 11C ) . The distance traveled by rats was 28 . 3 ± 5 . 7 m in the i . p . morphine ( 10 mg/kg ) plus intra-VTA saline group compared with 6 . 6 ± 2 . 3 m in the i . p . morphine plus intra-VTA CGP36216 group ( n = 6 rats , independent t test , p < 0 . 05 , right panel of Figure 11C ) . 10 . 7554/eLife . 09275 . 013Figure 11 . Influence of intra-ventral tegmental area ( VTA ) injection of the presynaptic GABAB receptor antagonist CGP36216 on morphine-induced increase in locomotor activity in rats . ( A ) Injection sites were verified under light microscope . Left panel: Representative Nissl-stained photomicrograph of cannula tracts terminating in the VTA . Right panel: location of the injection cannula tips in the VTA of rats used in data analyses . Numbers indicate coordinates relative to bregma . ( B ) Influence of intra-VTA injection of the presynaptic GABAB receptor antagonist CGP36216 on intra-VTA injection of morphine inducing an increase in locomotor activity . Left panel: time course of locomotor activity before and after intra-VTA injection of saline or morphine ( 1 μg/rat ) , or morphine ( 1 μg/rat ) with CGP36216 ( 20 µg/rat ) ( n = 6 rats ) . Right panel: average distance traveled by rats during 120 min after treatment with an intra-VTA injection of saline or morphine ( 1 μg/rat ) , or morphine ( 1 μg/rat ) with CGP36216 ( 20 µg/rat ) ( n = 6 rats; *p < 0 . 05 , compared with intra-VTA injection of saline , #p < 0 . 05 , compared with intra-VTA injection of morphine . ( C ) Influence of intra-VTA injection of the presynaptic GABAB receptor antagonist CGP36216 on intraperitoneal ( i . p . ) morphine-induced increase in locomotor activity . Left panel: time course of locomotor activity before and after intra-VTA injection of saline co-administered with either i . p . saline ( 1 ml/kg ) or morphine ( 10 mg/kg ) , and i . p . morphine ( 10 mg/kg ) with intra-VTA injection of CGP36216 ( 20 μg/rat ) ( n = 6 rats ) . Right panel: average distance traveled by rats during 120 min after treatment with intra-VTA injection of saline co-administered with either i . p . saline ( 1 ml/kg ) or morphine ( 10 mg/kg ) , and i . p . morphine ( 10 mg/kg ) with intra-VTA injection of CGP36216 ( 20 μg/rat ) ( n = 6; *p < 0 . 05 , compared with intra-VTA injection of saline co-administered with i . p . saline , #p < 0 . 05 , compared with intra-VTA injection of saline co-administered with i . p . morphine ) . Data are shown as the mean ±s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 013 We also observed the effect of the presynaptic GABAB receptor antagonist CGP36216 on morphine-induced conditioned place preference ( CPP ) in rats . Morphine ( 1 µg/rat ) and CGP36216 ( 20 µg/rat ) were locally injected into the VTA . Injection sites for data analysis are shown in Figure 12A . A schematic of the experimental design for CPP and drug application is shown in Figure 12B . As shown in Figure 12C , two-way ANOVAs conducted on the CPP score using treatment with different drugs as the between-subjects factors and test condition ( preconditioning and postconditioning ) as the within-subjects factor , revealed that there was a significant interaction of treatment ( F ( 3 , 40 ) = 4 . 45; p = 0 . 008 ) and test condition ( F ( 1 , 40 ) = 19 . 79; p < 0 . 001 ) . Post-hoc analysis showed that after CPP training , the morphine group ( n = 6 rats ) exhibited greater CPP compared with the control group ( n = 6 rats , p < 0 . 05 ) but in groups receiving CGP36216 ( n = 6 rats ) , morphine-induced CPP was absent ( p < 0 . 05 ) . These results suggest that intra-VTA injection of a presynaptic GABAB receptor antagonist abolishes CPP induced by intra-VTA injection of morphine . 10 . 7554/eLife . 09275 . 014Figure 12 . Influence of intra-ventral tegmental area ( VTA ) injection of the presynaptic GABAB receptor antagonist CGP36216 on intra-VTA injected morphine-induced conditioned place preference ( CPP ) in rats . ( A ) Schematic representations of injection cannula tips in the VTA of rats used in data analyses . Numbers indicate coordinates relative to bregma . ( B ) A schematic of the experimental design for CPP and administration of drugs . ( C ) Influence of the presynaptic GABAB receptor antagonist CGP36216 on morphine-induced CPP in rats . Averaged CPP score of preconditioning and postconditioning in different groups ( n = 6 rats in the every group; *p < 0 . 05 , intra-VTA injection of morphine group compared with intra-VTA injection of saline group , #p < 0 . 05 , intra-VTA injection of morphine plus CGP36216 group compared with intra-VTA injection of morphine group ) . Data are shown as the mean ±s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 09275 . 014
Previous studies reported that µ receptor agonists decreased EPSCs ( Bonci and Malenka , 1999; Margolis et al . , 2005 ) . In agreement , as shown by the results above , if a GABAA receptor antagonist was applied using a similar bath approach , morphine inhibited the frequency of sEPSCs . However , if the GABAA receptor antagonist was applied using an intracellular approach , morphine had a promoting rather than an inhibitory effect . Why did morphine produce opposite effects on the frequency of sEPSCs under different GABAA receptor antagonist application conditions ? We believe that under the condition of bath application of a GABAA receptor antagonist , the antagonist can produce a wide inhibitory effect on GABAA receptors present on various neuron subtypes , such as DA neurons and GABA neurons , in VTA slices . Maybe this wide effect shifts the effect of morphine on presynaptic glutamate release in VTA-DA neurons toward an inhibitory effect . However , the intracellular application of a GABAA receptor antagonist only blocks GABAA receptors in the single VTA-DA neuron being recorded . In this case , a clear promoting effect on presynaptic glutamate release in VTA-DA neurons was observed . In addition , to confirm this promoting effect under intracellularly applied PTX , we used the PPF of EPSC as another indicator of presynaptic glutamate release in order to observe the effect of morphine; the result substantiated our conclusion that morphine promoted presynaptic glutamate release in VTA-DA neurons . More importantly , when we clamped the membrane potential of VTA-DA neurons in the reversal potential of Cl− channels to remove sIPSCs , instead of using the intracellular or bath application of a GABAA antagonist , morphine still promoted presynaptic glutamate release in VTA-DA neurons . These results strongly suggest that morphine can promote presynaptic glutamate release in VTA-DA neurons . Postsynaptic inhibition in response to µ opioid receptor activation has been reported in some VTA-DA neurons ( Ford et al . , 2006 ) . Surprisingly , we did not observe this effect . The amplitude of the outward currents produced by opioids in the study by Ford et al . was small ( 2 . 1 ± 1 . 5 pA in VTA-DA neurons projecting to the nucleus accumbens and 14 ± 4 pA in those projecting to the basolateral amygdaloid nucleus ) and thus might not be reflected by the change in action potential firing in our study . In addition , a recent report demonstrated postsynaptic excitation of some VTA-DA neurons by opioids ( Margolis et al . , 2014 ) , an event that we also did not observe . Only a small population of VTA-DA neurons ( 19% ) showed depolarization or an increase in firing rate in response to opioids in this paper , which is consistent with physiological and anatomical evidence that µ receptors were present on a small population of VTA-DA neurons ( Ford et al . , 2006 ) . Consequently , we believe that we did not observe this effect perhaps because the percentage of the cells responding by excitation was low . Similar to our study , Jalabert et al . did not observe an excitatory effect of firing of VTA-DA neurons by opioids in the presence of glutamate receptor antagonists ( Jalabert et al . , 2011 ) . The mechanism underlying the promoting effect of morphine on presynaptic glutamate release in VTA-DA neurons may involve different processes . One suggestion is that morphine may directly act at glutamatergic terminals to promote glutamate release . To test this hypothesis , we observed the effect of morphine on glutamate release in isolated nerve terminals—synaptosomes ( Breukel et al . , 1997 ) from the VTA . We found that morphine had no significant effect on glutamate release . In addition , using mechanically dissociated single VTA-DA neurons , we still did not observe an effect of morphine on the frequency of sEPSCs . These results suggest that morphine does not exert a direct promoting effect on presynaptic glutamate release in VTA-DA neurons . Another possible mechanism may be through a local neuronal circuit that promotes glutamate release . It has been proposed that µ receptors located in GABAergic interneurons are the primary site of action of opiates in the VTA ( Hyman et al . , 2006 ) . This is consistent with morphological and functional data showing that µ receptors are primarily located on non-dopaminergic neurons in the VTA ( Garzon and Pickel , 2001 ) and that activation of µ receptors can inhibit GABAergic interneurons by hyperpolarizing the cells ( Johnson and North , 1992 ) . In addition , Chefer et al . studied the inter-relationship of µ receptors , GABA , glutamate , and DA in the VTA of freely moving animals ( Chefer et al . , 2009 ) . They found that there was a positive correlation between basal dopamine levels and the glutamate/GABA ratio in the VTA of wild-type animals . Moreover , in µ receptor knockout mice , the GABAergic inhibitory tone in the VTA was significantly increased , but the glutamatergic excitatory tone was decreased . These results suggest that activation of µ receptors may increase presynaptic glutamate release via inhibition of GABAergic interneurons in VTA-DA neurons . This is supported by our results indicating that the µ receptor antagonist CTOP can block the promoting effect of morphine on the frequency of sEPSCs ( the average frequency of sEPSCs was 2 . 95 ± 0 . 35 Hz before and 2 . 97 ± 0 . 40 after morphine application ( 10 µM ) in the presence of CTOP ( 1 µM ) ; n = 6 cells from four rats , p > 0 . 05 ) . Therefore , we hypothesize that morphine may act at μ receptors located in GABA neurons , inhibit the function of GABA neurons , and thus disinhibit the presynaptic glutamate release of VTA-DA neurons , leading to increased glutamate release . To test whether morphine has a disinhibitory effect on the presynaptic glutamate release of VTA-DA neurons , the presence of an inhibitory circuit from neighboring GABA neurons to the presynaptic glutamatergic terminals of VTA-DA neurons must first be demonstrated . Therefore , we designed a series of experiments to determine if this inhibitory circuit existed . Our results showed that ( 1 ) GABA could inhibit the presynaptic glutamate release of VTA-DA neurons and this inhibition was mediated through GABAB receptors , rather than GABAA receptors; ( 2 ) GABAB receptors were present in synaptosomes from the VTA as shown by Western blotting , and in glutamatergic terminals of VTA-DA neurons as shown by the triple immunofluorescence staining method; and ( 3 ) selective stimulation of GABA neurons of the VTA with an optogenetic technique could inhibit the presynaptic glutamate release of VTA-DA neurons and the GABAB receptor antagonist CGP54626 could remove this inhibitory effect . These results strongly support the presence of an inhibitory circuit from neighboring GABA neurons to the presynaptic glutamatergic terminals of VTA-DA neurons . Moreover , they suggest that GABAB receptors mediate this inhibitory effect of GABA . On this basis , we checked whether morphine-promoted presynaptic glutamate release of VTA-DA neurons through the disinhibitory action on presynaptic glutamate release . To test this hypothesis , we used two strategies to remove inhibition by GABAergic input of presynaptic glutamate release and then observed their influence on the effect of morphine on glutamate release . The results showed that ‘closing’ local GABAergic interneurons both with optogenetic methods and with the GABAB receptor antagonist CGP54626 could abolish the promoting effect of morphine on glutamate release in VTA-DA neurons . These results confirm our hypothesis that morphine promotes presynaptic glutamate release in VTA-DA neurons through its disinhibitory action on presynaptic glutamate release . It was previously reported that μ-opioid agonists hyperpolarized the GABAergic interneurons of the VTA , reduced the frequency of sIPSCs , and thus produced a disinhibitory action on VTA-DA neurons ( Johnson and North , 1992 ) . In the present study , we demonstrated that morphine-disinhibited presynaptic glutamate release in VTA-DA neurons by a second mechanism . These two kinds of disinhibitory actions may both contribute to the morphine-induced excitatory effect on VTA-DA neurons . However , there is no direct evidence supporting the suggestion that the disinhibitory action on VTA-DA neurons causes the morphine-induced increase in DA neuron firing . In the present study , we investigated the contribution of the disinhibitory action of morphine on presynaptic glutamate release in VTA-DA neurons to the overall excitatory effect of morphine on VTA-DA neurons by employing the selective presynaptic GABAB receptor antagonist CGP36216 . Several studies have indicated that pre- and postsynaptic GABAB receptors in central neurons may be pharmacologically distinct ( Dutar and Nicoll , 1988; Deisz et al . , 1993 , 1997 ) . CGP36216 is a useful lead compound to differentiate between pre- and postsynaptic GABAB receptors . Up to 1 mM of CGP36216 was ineffective in antagonizing baclofen-induced hyperpolarization mediated through postsynaptic GABAB receptors , but the responses mediated through presynaptic GABAB receptors were reversibly antagonized by CGP36216 with an IC50 of 43 µM ( Ong et al . , 2001 ) . Therefore , in the present study , we evaluated the contribution of the disinhibitory action of morphine on presynaptic glutamate release in VTA-DA neurons to the overall excitatory effect of morphine on VTA-DA neurons by examining the effect of CGP36216 on the morphine-induced increase in VTA-DA neuron firing and related behaviors . Our results showed that if GABAergic inhibitory control of presynaptic glutamate release in VTA-DA neurons was removed , the effect of morphine on the firing of VTA-DA neurons disappeared , indicating that the morphine-induced increase in VTA-DA neuron activity might mainly depend on its disinhibitory action on presynaptic glutamate release . We also observe the effect of removing GABA inhibitory control of presynaptic glutamate release with the presynaptic GABAB receptor antagonist CGP36216 on morphine-induced increase in locomotor activity , which is often assessed as increased VTA-DA neuron activity at the behavioral level ( Vezina , 2004 ) . The results showed that intra-VTA injection of a presynaptic GABAB receptor antagonist could abolish the increase in rat locomotor activity induced by both i . p . and intra-VTA-injected morphine . In addition , we observed the effect of removing GABA inhibitory control of presynaptic glutamate release with the presynaptic GABAB receptor antagonist CGP36216 on morphine-induced CPP , which is initiated by morphine-induced activation of VTA-DA neurons ( Tsai et al . , 2009 ) . The results showed that intra-VTA injection of CGP36216 could abolish morphine-induced CPP in rats . These data suggest that the disinhibitory action of morphine on presynaptic glutamate release may be the main mechanism in the morphine-induced increase in VTA-DA neuron firing and related behaviors .
Male Sprague–Dawley rats ( 14–16 days old ) or mice ( 4–5 weeks old ) were anesthetized with chloral hydrate ( 400 mg/kg , i . p . ) . All experimental procedures conformed to Fudan University as well as international guidelines on the ethical use of animals . All efforts were made to minimize animal suffering and reduce the number of animals used . VTA slices were prepared according to procedures described previously ( Hopf et al . , 2007 ) . The brain was removed rapidly from the skull and placed in modified ACSF containing 75 mM sucrose , 88 mM NaCl , 2 . 5 mM KCl , 1 . 25 mM NaH2PO4 , 7 mM MgCl2 , 0 . 5 mM CaCl2 , 25 mM NaHCO3 , and saturated with 95% O2 and 5% CO2 at ∼0°C . Horizontal 250 μm midbrain slices containing VTA were cut on a vibratome ( VT-1200 , Leica , Wetzlar , Germany ) and transferred to normal ACSF containing 126 mM NaCl , 2 . 5 mM KCl , 1 . 25 mM NaH2PO4 , 2 mM MgSO4 , 2 . 5 mM CaCl2 , 25 mM NaHCO3 , and 10 mM glucose at 32°C . Slices were incubated for at least 60 min before patch-clamp recording . The medial terminal nucleus of the accessory optic tract ( MT ) was used as the anatomical structure to define the VTA ( Hopf et al . , 2007 ) . VTA neurons were visualized on an upright microscope ( BX50WI , Olympus , Tokyo , Japan ) using infrared differential interference contrast or fluorescent optics . Whole-cell current- and voltage-clamp recordings were made using an EPC10 amplifier and PatchMaster 2 . 54 software ( HEKA , Lambrecht , Germany ) . Electrodes had a resistance of 3–4 MΩ when filled with the patch pipette solution . The internal pipette solution contained 130 mM K-gluconate , 8 mM NaCl , 0 . 1 mM CaCl2 , 0 . 6 mM EGTA , 2 mM Mg-ATP , 0 . 1 mM Na3-GTP , and 10 mM HEPES ( pH 7 . 4 ) . Lucifer yellow ( 2 mM ) was added to the internal pipette solution for labeling the recorded neuron . Cells were held at 0 pA under a current-clamp mode to record spontaneous firing . Only neurons with regular spontaneous firing and an action potential amplitude greater than 60 mV were used for electrophysiological analyses . Cells were held at −70 mV under a voltage-clamp mode to record sEPSCs or evoked EPSC . A concentric stimulating electrode ( FHC , Bowdoin , USA ) was placed near the recorded cell . To observe PPF , two synaptic responses were evoked by a pair of stimulating pulses given at short intervals ( 50 ms ) at 0 . 1 Hz . The series resistance ( Rs ) was monitored by measuring the instantaneous current in response to a 5 mV voltage step command . Rs compensation was not used , but cells where Rs changed by >15% were discarded . After forming whole-cell recording mode , we first identified DA neurons based on electrophysiological characteristics , which included both a spontaneous pacemaker-like firing and expression of a hyperpolarization-induced current ( Ih ) in the voltage-clamp configuration , by 1 s hyperpolarizing voltage steps ( −70 mV to −150 mV ) ( Grace and Onn , 1989; Margolis et al . , 2006; Zhang et al . , 2010; Chieng et al . , 2011 ) ( Figure 1A ) . Next , DA neurons were retrospectively confirmed by labeling the recorded neuron with Lucifer yellow ( 2 mM ) in the internal pipette solution and subsequent TH staining of the recorded cell with Lucifer yellow . Slices were fixed immediately after electrophysiological recording in 4% formaldehyde for 2 hr and then washed with 0 . 01 M PBS solution . The slices were incubated for 2 hr at 4°C in a blocking solution containing 10% normal goat serum and 0 . 2% Triton X-100 in PBS and then incubated overnight at 4°C with primary rabbit anti-TH antibody ( 1:1000; Abcam , Cambridge , UK ) . The slices were washed thoroughly in PBS before being incubated for 2 hr at 4°C with goat anti-rabbit-Cy3 secondary antibody ( 1:200; Jackson ImmunoResearch Laboratories , West Grove , USA ) . The recorded cells were identified as DA neurons based on the co-labeling by Lucifer yellow and TH ( Figure 1B ) . Male Sprague–Dawley rats ( 100–150 g ) were anesthetized with chloral hydrate ( 400 mg/kg , i . p . ) . Synaptosomes were prepared as described previously ( Dong et al . , 2005 ) . The VTA was dissected and homogenized in 0 . 32 M sucrose solution at 4°C using the Art-Miccra D-8 tissue grinder with a motor-driven pestle rotating at 900 rpm . The homogenate was centrifuged at 3000×g for 3 min at 4°C . The supernatant ( S1 ) was centrifuged at 14 , 500×g for 12 min at 4°C . The pellet ( P2 ) was resuspended and loaded onto Percoll gradients consisting three steps of 23% , 10% , and 3% Percoll in 0 . 32 M sucrose additionally containing 1 mM EDTA and 250 μM DTT . The gradients were centrifuged at 32 , 500×g for 6 . 5 min at 4°C . Synaptosomes were harvested from the interface between the 23% and 10% Percoll layers and washed in Hanks' balanced salt solution ( HBSS ) containing 140 mM NaCl , 5 mM KCl , 5 mM NaHCO3 , 1 mM MgCl2 , 1 . 2 mM Na2HPO4 , 10 mM glucose , and 20 mM HEPES , pH 7 . 4 . Washed synaptosomes were centrifuged at 27 , 000×g for 15 min at 4°C . Glutamate release from synaptosomes was assayed by on-line fluorimetry as described previously ( Nicholls et al . , 1987 ) . Synaptosomal pellets were suspended again by adding 1 . 5 ml of incubation medium: 122 mM NaCl , 3 . 1 mM KCI , 0 . 4 mM KH2PO4 , 5 mM NaHCO3 , 20 mM Na-TES , 1 . 2 mM MgSO4 , 16 pM bovine serum albumin ( fatty acid free , type V ) , and 10 mM D-glucose , pH 7 . 4 . Additionally , NADP+ ( 2 mM ) , glutamate dehydrogenase ( 50 units/ml ) , and CaCl2 ( 1 mM ) were added after 3 min . After a 5-min incubation period , morphine ( 10 μM ) was added to observe its effect on glutamate release . Oxidative decarboxylation of released glutamate , leading to reduction in NADP+ , was monitored by measuring NADPH fluorescence at excitation and emission wavelengths of 340 nm and 460 nm , respectively . Data points were obtained at 60 s intervals . An exogenous glutamate standard ( 5 nmol ) was added at the end of each experiment . The fluorescence change produced by the addition of exogenous glutamate ( 5 nmol ) was used to calculate the released glutamate in nmol/mg . VTA-DA neurons with functional presynaptic terminals attached were obtained by a mechanical dissociation method as previously described ( Akaike and Moorhouse , 2003; Ye et al . , 2004; Deng et al . , 2009 ) . First , VTA slices were prepared as described above . Then the slice was transferred to a 35-mm culture dish ( Sigma , USA ) and held down with a flat U-shaped wire . The dish was filled with standard external solution containing 140 mM NaCl , 5 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , 10 mM HEPES , and 10 mM glucose ( pH 7 . 4 ) . The MT was used as the anatomical structure to define the VTA ( Hopf et al . , 2007 ) under a stereomicroscope ( SZ61 , Olympus ) . A fire-polished glass pipette , lightly touching the surface of the VTA , was vibrated horizontally at 50–60 Hz for 1–2 min by a home-made device . The slice was then removed . The isolated neurons adhered to the bottom of the dish within 20 min and were then ready for electrophysiological experiments . Immunoblot analysis of GABAB R1 receptors was performed on the synaptosomes obtained from the VTA . The synaptosomal pellets were homogenized in a buffer containing 100 mM Tris-HCl ( pH 6 . 7 ) , 1% SDS , 143 mM 2-mercaptoethanol , and 1% protease inhibitor . The lysate was centrifuged at 12 , 000 rpm for 10 min at 4°C . The samples were treated with the SDS sample buffer at 95°C for 5 min , loaded on a 10% SDS polyacrylamide gel , and blotted to a PVDF membrane . Each blot was incubated with a rabbit anti-GABAB R1 ( 1:100; Alomone Labs , Jerusalem , Israel ) . Following extensive washing , membranes were incubated with IRDye 800CW goat anti-rabbit secondary antibodies ( 1:20 , 000; LI-COR , Lincoln , USA ) for 1 hr , and images were acquired on a LI-COR Odyssey system . Sprague–Dawley rats ( 15–20 days old ) were perfused transcardially with 0 . 01 M PBS followed by 4% paraformaldehyde . After perfusion , brains were removed and postfixed in 4% paraformaldehyde overnight . Horizontal brain slices ( 40-μm thick ) containing VTA were prepared using a vibratome ( VT-1000S , Leica ) . Slices were pre-blocked for 2 hr at 4°C in a blocking solution containing 10% horse serum and 0 . 2% Triton X-100 in PBS and again washed three times in PBS for 5 min . Slices were incubated overnight at 4°C with primary antibody mouse anti-TH ( 1:1000; Millipore , Billerica , USA ) , rabbit anti-GABAB R1 ( 1:100; Alomone Labs ) , and guinea pig anti-VGLUT2 ( 1:3000; Millipore ) dissolved in the blocking solution . Afterwards , slices were washed three times in PBS for 5 min and then incubated with the following secondary antibodies: horse anti-rabbit-Alexa 488 ( 1:200; Jackson ImmunoResearch Laboratories ) , horse anti-mouse-Cy5 ( 1:200; Jackson ImmunoResearch Laboratories ) , and horse anti-guinea pig-Cy3 ( 1:200; Jackson ImmunoResearch Laboratories ) for 2 hr at room temperature in 2% horse serum and 0 . 2% Triton X-100 in PBS . Subsequently , slices were washed three times in PBS for 5 min , and sections were mounted on glass slides using aqua-mount mounting medium ( Thermo Fisher Scientific , Waltham , USA ) . Confocal images of sections were obtained using confocal microscopy ( FV1000 , Olympus ) with a 60× oil-immersion lens . The optogenetic approach was based on procedures described previously ( van Zessen et al . , 2012 ) . Bilateral injections of purified and concentrated pAAV-EF1a-double floxed-hChR2 ( H134R ) -mCherry or pAAV-EF1a-double floxed-eNpHR 3 . 0-EYFP virus ( 2 . 05 × 1012 vector genomes/ml; Neuron Biotech Company , Shanghai , China ) were stereotaxically performed in 3-week-old male Gad2-IRES-Cre mice ( B6N . Cg-Gad2tm2 ( cre ) Zjh/J; Jackson Laboratories , USA ) . Each side of the VTA ( final coordinates: AP , −3 . 1 mm; ML , ±0 . 4 mm; DV , −4 . 2 mm from the skull surface ) was injected with 0 . 5 µl AAV for 10 min followed by an additional 10 min to allow diffusion of viral particles away from the injection site . About 10–14 days after AAV virus injection , VTA slices were prepared according to the procedures described above . GABA neurons were stimulated by a 5 ms , 470 nm light or a constant 590 nm light delivered via an optical fiber ( core diameter 200 µm , NA 0 . 39; ThorLabs , USA ) coupled to an LED light source ( Mightex , California , USA ) . The end of the fiber optic cannula ( stainless ferrule , core diameter 200 µm , length 20 mm; ThorLabs , New Jersey , USA ) connected to the optical fiber through a mating sleeve ( ThorLabs ) was placed 500 µm above the recording cell . Male Sprague–Dawley rats ( 270–300 g ) were used . Animals were housed singly under a 12 hr light/dark cycle ( lights on 7:00 AM to 7:00 PM ) in a temperature- and humidity-controlled environment with food and water freely available . The rats were anesthetized with chloral hydrate ( 400 mg/kg , i . p . ) and placed in stereotaxic instruments ( Stoelting , Wood Dale , USA ) . Two 24-gauge stainless steel guide cannulae were implanted bilaterally 2 mm above the VTA . The coordinates were: AP , −5 . 8 mm; ML , ±2 . 5 mm; DV , −8 . 0 mm from the skull surface with a 14° lateral angle . The cannulae were secured to the skull with two anchoring screws and dental cement . To prevent occlusion , 30-gauge wire plugs were inserted into the cannulae . After the surgery , the animals were housed individually and were allowed to recover for over a week . The locomotor activity test was conducted as described previously with some modifications ( Borgland et al . , 2006 ) . The locomotor activity of animals was monitored with a near infrared video camera within the operant chambers ( Med Associates , St . Albans , USA ) . Distance traveled was measured using Open Field Activity Software ( Med Associates ) and analyzed locomotion estimates based on movement over a given distance and resting delays ( movement in a given period of time ) . All animals were habituated to the test room for 2 hr prior to the start of the experiment . Rats were further habituated to the operant chambers for 30 min prior to the 120 min testing session . For bilateral intra-VTA microinjection , 30-gauge injection needles were inserted into the cannulae . The injection needles were connected to a 1-μl microsyringe ( Hamilton , Reno , USA ) by a polyethylene tube and controlled by a syringe pump ( Harvard Apparatus , Holliston , USA ) . After injection , the needles were left in place for another 1 min . Rats were bilaterally injected with 0 . 2 μl morphine ( 0 . 5 μg/side ) or CGP36216 ( 10 μg/side ) for 1 min . After drug administration , the rats were placed in the chambers for the 120 min testing session . On days 1 , 2 , and 3 , all rats were only given saline ( bilateral intra-VTA microinjections or intraperitoneal injections ) to habituate them to the test protocol . On day 4 , rats were given drugs according to experimental group . After the behavioral tests , all rats were anesthetized with an overdose of chloral hydrate and perfused with 0 . 9% saline . The brain was removed and fixed in 4% paraformaldehyde for 24 hr . Coronal sections ( 80 μm ) were cut by a vibratome and stained with cresyl violet . Injection sites were verified under light microscope . Animals where the injection site was outside the VTA were discarded . The CPP test was conducted using a three-compartment place conditioning apparatus ( Med Associates ) with distinct tactile environments to maximize contextual differences . The procedure for CPP testing was similar to that described previously ( Phillips and LePiane , 1980; Bardo and Neisewander , 1986; Wang et al . , 2008 ) , with some modifications . On day 1 , rats underwent a preconditioning test: they were placed in the middle neutral area and were allowed to freely access both sides of the apparatus for 15 min . Rats with a strong preference ( 60% ) for any compartment were discarded . Conditioning was performed using an unbiased , balanced protocol . On day 2 , rats were microinjected with saline , morphine , CGP36216 , or CGP36216 plus morphine 30 min before they were confined to the conditioning chambers for 45 min ( drug-paired ) . On day 3 , rats received a microinjection of saline and were then confined to the other chamber for 45 min . The day after conditioning , rats were tested for drug-induced ( postconditioning test ) CPP in the same conditions as for the preconditioning test . The place preference score ( CPP score ) was defined as the time spent in the drug-paired chamber minus that spent in the drug-unpaired ( saline-paired ) chamber . Injection sites were verified under light microscope as described above . Methyl sulfoxide ( DMSO ) , PTX , K-gluconate , Lucifer yellow , K2-ATP , Na3-GTP , 4- ( 2-hydroxyethyl ) piperazine-1-ethanesulfonic acid ( HEPES ) , N-[tris ( hydroxymethyl ) methyl]-2-aminoethanesulfonic acid ( Na-TES ) , nicotinamide adenine dinucleotide phosphate ( NADP+ ) , glutamate dehydrogenase , sodium dodecyl sulfate ( SDS ) , ethyleneglycol-bis ( β-aminoethyl ether ) -N , N , N′ , N′-tetraacetic acid ( EGTA ) , Triton X-100 , and 0 . 01 M PBS were purchased from Sigma . [S- ( R* , R* ) ]-[3-[[1- ( 3 , 4-Dichlorophenyl ) ethyl]amino]-2-hydroxypropyl] ( cyclohexylmet-hyl ) phosphinic acid ( CGP54626 ) and ( 3-aminopropyl ) ethylphosphinic acid hydrochloride ( CGP36216 ) were purchased from Tocris . Morphine was from Shenyang No . 1 Pharmaceutical Factory , China . Percoll was purchased from Amersham Biosciences Corporation . Other AR grade reagents were from Shanghai Chemical Plant . PTX or CGP54626 was dissolved in DMSO and others were dissolved in ddH2O . When DMSO was used as the vehicle , drugs were initially dissolved in 100% DMSO and then diluted into ASFC at a final DMSO concentration of less than 0 . 5% , which had no detectable effects on the parameters we observed . Numerical data were expressed as the mean ±s . e . m . Off-line data analysis was performed using the Mini Analysis Program ( Synaptosoft , Fort Lee , USA ) , Clampfit ( Axon Instruments , Sunnyvale , USA ) , SigmaPlot ( Systat Software , San Jose , USA ) , and Origin ( Microcal Software , Northampton , USA ) . Spontaneous firing was analyzed using the Event Detection function of Clampfit . sEPSCs were analyzed using the Mini Analysis Program . Detection criteria were set at >8 pA , <1 ms rise-time , and <3 ms decay-time for sEPSCs ( Borgland et al . , 2006 ) . Statistical significance was determined using Student's t-test for comparisons between two groups or ANOVAs followed by the Student–Newman–Keuls test for comparisons among three or more groups . In the patch-clamp studies , n refers to the number of cells . Every cell was from a different slice , and a group of cells in each experiment was from at least four animals . Two-way ANOVAs were performed on the data from CPP with the between-subjects factors treatment ( different drugs ) . All post-hoc comparisons were made using Tukey's test . Results with p < 0 . 05 were accepted as being statistically significant . | Morphine is one of the most commonly used drugs for the treatment of severe pain . It is derived from opium , which is extracted from poppies , and binds to the same receptors in the brain as the body's own naturally produced painkillers . As well as providing pain relief , morphine can act directly on the brain's reward system to trigger a state of euphoria , and can therefore be highly addictive . One of the key components of the brain's reward circuit that morphine affects is called the ventral tegmental area ( VTA ) . The activity of the VTA is regulated by the combined efforts of two groups of cells: excitatory glutamatergic neurons that increase VTA activity and inhibitory interneuronsthat reduce the activity of the VTA . Morphine inhibits the interneurons , thereby allowing the glutamatergic neurons to activate the VTA . But does morphine also strengthen this excitatory input directly ? By examining the effects of morphine on individual VTA neurons , Chen et al . show that the drug does indeed enhance the activity of the glutamatergic neurons . However , it does so indirectly by inhibiting another group of interneurons that would otherwise silence the glutamatergic neurons . This effect of morphine is dependent on the drug acting on a specific receptor type on the interneurons . Chen et al . show that injecting a drug that blocks these receptors straight into the VTA of rats prevents morphine from increasing the animals' activity levels . It also prevents the animals from developing a preference for being in locations where they have previously received morphine . This suggests that morphine could primarily exert its pleasurable effects by preventing the glutamatergic neurons from being inhibited , and thus allowing them to activate the VTA neurons . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Morphine disinhibits glutamatergic input to VTA dopamine neurons and promotes dopamine neuron excitation |
Chronic critical illness is a global clinical issue affecting millions of sepsis survivors annually . Survivors report chronic skeletal muscle weakness and development of new functional limitations that persist for years . To delineate mechanisms of sepsis-induced chronic weakness , we first surpassed a critical barrier by establishing a murine model of sepsis with ICU-like interventions that allows for the study of survivors . We show that sepsis survivors have profound weakness for at least 1 month , even after recovery of muscle mass . Abnormal mitochondrial ultrastructure , impaired respiration and electron transport chain activities , and persistent protein oxidative damage were evident in the muscle of survivors . Our data suggest that sustained mitochondrial dysfunction , rather than atrophy alone , underlies chronic sepsis-induced muscle weakness . This study emphasizes that conventional efforts that aim to recover muscle quantity will likely remain ineffective for regaining strength and improving quality of life after sepsis until deficiencies in muscle quality are addressed .
Sepsis is a common life-threatening condition caused by a deregulated host response to infection . This syndrome is characterized by profound systemic inflammation and disseminated intravascular coagulation , which often lead to multiple organ failure ( MOF ) and subsequent mortality ( Singer et al . , 2016 ) . The incidence of sepsis has risen by 9% to 13% annually , largely due to an expansion of the elderly population , more frequent invasive surgical procedures , and increased antibiotic resistance ( Martin et al . , 2003; Gaieski et al . , 2013; Angus and Wax , 2001; Starr and Saito , 2014 ) . However , advances in critical care medicine and campaigns promoting early identification and treatment of sepsis have led to improved survival rates ( Angus and Wax , 2001 ) . Consequently , nearly 1 . 5 million sepsis survivors are discharged annually from U . S . hospitals , approximately 14 million globally ( Elixhauser et al . , 2011; Prescott and Angus , 2018; Fleischmann et al . , 2016 ) . As the population of survivors grows , post-sepsis physical dysfunction , which exceeds that of intensive care unit acquired weakness ( ICUAW ) alone , has become a clear clinical problem ( Prescott and Angus , 2018; Callahan and Supinski , 2009; Iwashyna et al . , 2010; Schefold et al . , 2010; Contrin et al . , 2013 ) . Sepsis survivors rarely return to baseline functional status after discharge from the ICU . Post-sepsis muscle weakness causes nearly half of previously functionally independent individuals to be discharged either to nursing care facilities or home with home care ( Odden et al . , 2013 ) . Nearly a third of patients which were previously independent and had no prior comorbidities had problems with mobility one year after discharge ( Yende et al . , 2016 ) . Further , survivors continue to develop functional limitations for at least 5 years following discharge , a trend that appears to be sepsis-specific ( Iwashyna et al . , 2010 ) . Moreover , less than half of previously employed individuals are able to return to work within one year after discharge ( Poulsen et al . , 2009; Pettilä et al . , 2000 ) . Together these studies illustrate the dramatic impact of chronic post-sepsis weakness . Although post-sepsis muscle weakness is now widely recognized as a serious medical issue , the lack of an appropriate animal model has greatly impeded the identification of mechanisms that contribute to long-term dysfunction . Current animal models of sepsis are either too severe , causing early death of most animals without recovery from sepsis , or too mild thus not triggering long-term chronic dysfunction . To overcome this issue , we recently refined a non-surgical murine model of polymicrobial sepsis whereby infection is initiated by injection of cecal slurry ( CS ) ( Starr et al . , 2014; Starr et al . , 2016 ) . Therapeutic intervention with a broad-spectrum antibiotic and fluids is provided , but initiated after bacteremia is evident ( Steele et al . , 2017 ) . This delayed ICU-like resuscitation protocol allows for the development of sepsis with organ damage , yet rescues the majority of mice from an otherwise completely lethal condition , thereby allowing the study of survivors . To further optimize our animal model for the current study , careful attention was also given to age , as the large majority of sepsis patients are late middle-age and older , and aging is an established risk factor for sepsis incidence , severity , and mortality ( Angus and Wax , 2001; Starr and Saito , 2014; Elixhauser et al . , 2011; Martin et al . , 2006; Dombrovskiy et al . , 2007 ) . The purpose of the present study was to establish that chronic muscle weakness , similar to the clinical condition among sepsis survivors , can be modeled in age-appropriate mice using our CS protocol with delayed ICU-like intervention . We then aimed to delineate underlying mechanisms responsible for post-sepsis muscle dysfunction . We show that sepsis survivors have significant skeletal muscle weakness for at least one month which cannot be attributed to muscle atrophy , but rather is associated with impaired mitochondrial activity and persistent protein oxidative damage .
We adapted our recently reported ICU-like model of sepsis to late middle-aged C57BL/6 mice ( Steele et al . , 2017 ) ( 16 months; equivalent to ~50-year-old human [Flurkey K et al . , 2007] ) . Sepsis was induced by bolus injection of cecal slurry ( CS ) and therapeutic resuscitation with antibiotics and fluids was initiated at 12h and continued twice daily for five days ( schematic provided in Figure 1A ) . This protocol rescued 74 . 1% of middle-aged males from otherwise completely lethal ( LD100 ) sepsis ( Figure 1B , p<0 . 0001 ) . No further mortality was observed after day 14 . Assessment of bacteremia showed that resuscitation decreased bacterial load by day 2 ( p=0 . 009 ) , and resolved the systemic infection by day 4 ( Figure 1C ) . Similar data were obtained using middle-aged female mice: 72 . 7% survival was achieved compared to 16% survival without therapeutic intervention ( Figure 1D , p=0 . 035 ) , and bacteremia was rapidly resolved ( Figure 1E ) . Next , we evaluated skeletal muscle strength in the murine sepsis survivors using ex vivo specific force analysis ( muscle force normalized to physiological cross section ) of the extensor digitorum longus hind limb muscle . Compared to non-sepsis controls ( NSC ) , male sepsis survivors had a 24 . 6% reduction in maximal specific force at 2 weeks ( p<0 . 0001 ) and 17 . 4% reduction at 1 month ( p=0 . 002 , Figure 1F ) . Similarly , female sepsis survivors were 19 . 8% weaker compared to controls at 2 weeks ( p=0 . 035 , Figure 1G ) . Sepsis-induced chronic muscle weakness could not be explained by atrophy since muscle force measurements are normalized to muscle size ( physiological cross section ) . To exclude the possibility that reduced muscle function in sepsis survivors was due to repeated antibiotic administration , force was also measured in non-sepsis animals which received resuscitation with antibiotics and fluids in parallel to sepsis mice . We found that specific force was unchanged by resuscitation procedures ( maximal force: 19 . 0 ± 0 . 8 N/cm2 in NSC vs 19 . 5 ± 0 . 8 N/cm2 in NSC + resuscitation , p=0 . 359 ) . To determine if chronic muscle weakness was due in part to ongoing inflammation in the sepsis survivors , cytokine concentration in the plasma and gene expression in muscle were measured . Despite resolution of bacteremia , plasma IL-6 concentration was elevated at day 4 ( p<0 . 0001 ) , but comparable to NSC by 2 weeks in both male and female sepsis survivors ( p=0 . 999 and p=0 . 667 , respectively , Figure 2A ) . Plasma TNFα ( Figure 2B ) and IL-10 ( Figure 2C ) followed a similar trend . Likewise , gene expression of these cytokines in gastrocnemius muscle showed that pro-inflammatory IL-6 ( Figure 2D ) and TNFα ( Figure 2E ) were comparable among NSC and sepsis survivors at 2 weeks ( p=0 . 622 and p=0 . 565 , respectively ) . These data indicate that muscle weakness is not associated with ongoing systemic or local muscle inflammation . Interestingly , IL-10 , an anti-inflammatory cytokine , was 2 . 5-fold higher in sepsis survivors compared to NSC ( p=0 . 011 , Figure 2F ) . In addition , we observed splenomegaly in many sepsis animals on day 4 ( males , 76 . 4% larger , p=0 . 0003 compared to NSC ) which remained elevated at 2 weeks ( males , 49 . 7% larger , p=0 . 004; females 20 . 2% larger , p=0 . 034 ) , but recovered by 1 month ( males p=0 . 905 , Figure 2G , H ) which may be indicative of high antigen clearance during recovery from sepsis ( Bronte and Pittet , 2013 ) . Taken together , these results demonstrate that middle-aged sepsis survivors exhibit chronic muscle weakness , even after bacteremia and systemic inflammation are resolved . Animals with sepsis lost significant body weight over time ( males p=0 . 003 compared to NSC , Figure 3A; females p=0 . 001 , Figure 3—figure supplement 1 ) . Analyses of body composition by EchoMRI revealed a significant loss of fat mass after sepsis induction ( p=0 . 004 compared to NSC , Figure 3B ) . On the other hand , loss of lean mass was observed during the first 5 days ( p=0 . 005 compared to NSC ) which steadily recovered thereafter and became comparable to non-sepsis controls by two weeks ( p=0 . 821 , Figure 3C ) . In a selected experiment in which sepsis survivors were kept 1 month post-sepsis , body weight remained significantly lower than baseline ( 6 . 12% lower , p<0 . 0001 , data not shown ) . These results suggest that sustained weight loss in murine sepsis survivors is attributable to loss of fat mass , not lean mass . EchoMRI analysis allowed repeated measures of whole body composition without need to euthanize the animals; however , analysis of lean mass is not specific to skeletal muscle . Thus , we compared wet weight of hind limb skeletal muscles from male mice at day 4 ( the time at which EchoMRI data showed the most profound loss of lean mass ) , 2 weeks , and 1 month , alongside non-sepsis controls ( Figure 3D ) . The wet weight of the predominantly fast-glycolytic muscles tibialis anterior ( TA ) and extensor digitorum longus ( EDL ) , and the mixed fiber-type gastrocnemius ( GA ) were reduced at day 4 compared to non-sepsis controls ( p=0 . 054 , p<0 . 0001 , and p=0 . 001 , respectively ) . The wet weight of these muscles was lower at 2 weeks and was comparable to non-sepsis controls at 1 month ( p=0 . 918 for EDL and p=0 . 242 for GA ) . The weight of the predominantly oxidative soleus muscle was similarly reduced at day 4 , but did not achieve statistical significance ( p=0 . 055 across groups ) . While these data are indicative of muscle loss , it is important to note that changes in muscle size were proportional to loss of body weight as evidenced by skeletal muscle weight to body weight ratio ( Figure 3E ) . Such trends in body weight , muscle wet weight , and muscle weight to body weight ratio were similarly observed in female sepsis survivors ( Figure 3—figure supplement 1 ) . To more robustly assess atrophy , and to evaluate potential fiber-type differences in the post-sepsis condition , we performed myofiber-specific cross-sectional area ( CSA ) analysis on the medial head of the mixed fiber-type gastrocnemius ( Figure 3F , G ) muscle . The mean CSA of fast-twitch ( type II ) fibers showed sepsis-mediated atrophy on day 4 ( IIa p=0 . 023 , IIb p=0 . 044 , and IIx p=0 . 0034 ) which was largely recovered by 2 weeks , whereas little , if any , atrophy was observed in slow ( type I ) fibers ( p=0 . 626 ) . Analysis of the predominantly slow-oxidative soleus muscle did not reveal sepsis-induced atrophy ( type I fibers p=0 . 521 , IIa p=0 . 799 , Figure 3—figure supplement 2 ) which is consistent with the wet weight ( Figure 3D , right panel ) . These data are consistent with clinical reports which show that preferential atrophy of fast-twitch fibers occurs during critical illness ( Gutmann et al . , 1996; Bierbrauer et al . , 2012 ) . Collectively , these data provide evidence that chronic muscle weakness in murine sepsis survivors cannot be attributed to muscle wasting alone . Therefore , a long-term reduction in muscle function of murine sepsis survivors must be largely attributable to impaired quality , rather than quantity . In sepsis patients , skeletal muscle mitochondrial dysfunction is evident during their stay in the ICU ( Fredriksson et al . , 2006; Fredriksson et al . , 2008 ) , but it is unknown whether such mitochondrial derangements persist after recovery from sepsis . Therefore , we assessed mitochondrial integrity in the EDL of sepsis survivors 2 weeks after CS injection . Transmission electron microscopy was used to study intermyofibrillar ( IMF ) mitochondria ( Figure 4A ) and subsarcolemmal ( SS ) mitochondria ( Figure 4B ) . Normal mitochondrial morphology was evident in non-sepsis controls characterized by highly organized lamellar cristae ( i . e . inner mitochondrial membrane arranged in parallel stacks ) . On the other hand , striking morphological alterations were observed frequently in both IMF mitochondria and SS mitochondria in muscle from sepsis survivors . Even observation at low magnification ( 5 , 000X ) revealed numerous mitochondria with fragmented cristae and enlarged matrix space in the muscles from sepsis survivors . These features are better appreciated at higher magnification ( 15 , 000X ) by which we also observed some mitochondria nearly devoid of cristae , ‘onion-like’ concentric swirling of cristae ( Jiang et al . , 2017; Walker and Benzer , 2004 ) , as well as compartmentalization in vesicle-like structures ( Sun et al . , 2007; Vincent et al . , 2016 ) . Further , observation of mitochondrial ultrastructure in TA muscles of the sepsis survivors ( Figure 4—figure supplement 1 ) showed abnormalities consistent with those found in the EDL . In an effort to better understand these marked ultrastructural abnormalities in the muscles of sepsis surviving mice , we performed morphometric analysis . Five representative images were acquired for each mitochondrial population ( IMF and SS ) . On average , 49 IMF and 69 SS mitochondria were observed and categorized in muscles from three animals per group . Normal mitochondria were defined as having intact cristae which occupied ≥80% of the mitochondrial area and had no observable compartmentalization; aberrant cristae were defined as cristae which occupied 20–80% of the mitochondrial space and/or had indications of swirled cristae or compartmentalization; ‘empty’ mitochondria were defined as exceptionally damaged whereby < 20% of the mitochondrial space was devoid of cristae and/or the outer mitochondrial membrane was seemingly ruptured and accompanied by enlargement ( representative examples are provided in Figure 4C ) . Normal mitochondria made up a significantly smaller proportion of the total IMF ( Figure 4D ) and SS ( Figure 4E ) mitochondria in sepsis survivors ( p=0 . 002 and p=0 . 048 , respectively ) . IMF mitochondria were seemingly more affected in the post-sepsis condition , where aberrant and empty mitochondria represented a significantly larger proportion of the mitochondria ( p=0 . 029 and 0 . 049 , respectively ) compared to controls . This trend was similar in the SS mitochondrial population , but did not achieve statistical significance . Since we observed altered mitochondrial morphology post-sepsis , we hypothesized that energy metabolism would be impaired . Thus , mitochondria were isolated from the TA of mice 2 weeks after sepsis and oxygen consumption rates ( OCR ) were measured to assess mitochondrial bioenergetics in the post-sepsis condition . These results showed that the maximal ADP phosphorylation rate ( State III ) and Complex I-driven electron transport ( State V-CI ) were significantly lower ( 27 . 3% p=0 . 003 , and 26 . 6% p=0 . 017 , respectively ) in sepsis survivors compared to that of controls ( Figure 5 ) . On the other hand , Complex II-driven maximum electron transport ( State V-CII ) showed a non-significant ( p=0 . 099 ) trend of lower OCR in sepsis survivors ( Figure 5 ) . These results show that skeletal muscle mitochondria in sepsis survivors have significantly reduced mitochondrial bioenergetics . Since acute sepsis affects food consumption , we confirmed that the observed long-term changes in mitochondrial bioenergetics was not influenced by reduced food intake . After measuring baseline food consumption for 5 days , cecal slurry was injected and food consumption was monitored daily for 2 weeks ( Figure 5—figure supplement 1A ) . On the first day of sepsis , animals ate only 11 . 08% of their baseline ( 0 . 52 g compared to 5 . 19 g; p<0 . 001 ) , which steadily increased to 75 . 51% by day 4 ( 3 . 53 g , p=0 . 042 ) and returned to near-baseline thereafter . To mimic this reduced food consumption , a pair-feeding paradigm was followed by which the daily food intake of a set of non-sepsis animals was restricted to match the daily food consumption by the sepsis animals . Ad libitum ( freely-fed ) non-sepsis mice were included as controls . Unsurprisingly , pair feeding resulted in significant reductions in body weight over time ( p<0 . 001; Figure 5—figure supplement 1B ) . The wet weight of hindlimb skeletal muscles was lower in the pair-fed mice than freely-fed controls ( GA p<0 . 001 , TA p=0 . 004; Figure 5—figure supplement 1C ) , however the reduction was proportional to overall loss of body weight ( GA p=0 . 151 , TA p=0 . 813; Figure 5—figure supplement 1D ) , similar to the trend observed in the sepsis survivors ( Figure 3 ) . Oxygen consumption rate ( OCR ) was measured in isolated mitochondria from the TA which showed similar mitochondrial respiration capacity among pair-fed and freely-fed mice ( State III p=0 . 944 , State V-CI p=0 . 737 , State V-CII p=0 . 185; Figure 5—figure supplement 1E ) . These data confirm that the reduced food intake during sepsis does not contribute to chronic mitochondrial impairment in sepsis surviving mice . Although OCR of isolated mitochondria is widely used to assess functional capacity , the severity of the phenotype is potentially masked because severely damaged mitochondria are lost during isolation . In addition , OCR reflects the maximum capacity in the presence of excess amounts of substrates . In contrast , the efficiency of substrate utilization , which does not influence OCR , is also a major determinant of mitochondrial activity . Therefore , we also performed a series of histochemical stains on tibialis anterior muscle sections as an alternative method to evaluate specific mitochondrial complex enzyme activity in whole tissue ( Figure 6A ) . Quantification of stain intensities was conducted ( Figure 6B ) which showed that complex I activity ( measured by nicotinamide adenine dinucleotide dehydrogenase ( NADH ) , Figure 6A , top row ) was 22 . 2% lower at day 4% and 40 . 8% lower at 2 weeks compared to NSC , although statistical significance was not achieved . Complex II activity ( measured by succinate dehydrogenase ( SDH ) , Figure 6A , middle row ) was 19 . 8% lower at day 4% and 38 . 1% lower at 2 weeks ( p=0 . 015 ) . Overall mitochondrial respiration activity was assessed by cytochrome c oxidase staining ( COX; complex IV; Figure 6A , bottom row ) which was 43 . 7% lower at day 4 ( p<0 . 001 ) and 48 . 6% lower at 2 weeks ( p<0 . 001 ) . Interestingly , 2 weeks after sepsis some myofibers had areas devoid of staining which was consistent among serial sections for the different mitochondrial enzyme activities ( Figure 6 , right column , yellow arrows ) . Using hematoxylin and eosin stained serial sections we confirmed that the areas devoid of mitochondrial enzyme activity were not devoid of tissue . We hypothesize that the areas devoid of histochemical staining in the myofibers of sepsis survivors ( Figure 6A , right column ) are likely the areas that house ‘empty’ or highly ‘aberrant’ mitochondria as depicted in Figure 4 by TEM analysis . Taken together , these data demonstrate that mitochondrial complex enzyme activities are impaired long after sepsis itself is resolved . Complex I-driven respiration seems more affected than Complex II-driven activity as evidenced by respiration analysis in isolated mitochondria ( Figure 5 ) . However , histochemical analysis , which reflects mitochondrial bioenergetics in whole tissue , indicated that CII-driven activity is also significantly affected in the post-sepsis condition ( Figure 6 ) . Significant production of reactive oxygen species ( ROS ) and reactive nitrogen species ( RNS ) is triggered during sepsis due to inflammation , tissue hypoxia ( Ueda et al . , 2008; Starr et al . , 2011 ) , and mitochondrial dysfunction ( Fredriksson et al . , 2006; Zolfaghari et al . , 2015; Singer , 2014; Castello et al . , 2006 ) . To investigate oxidative damage as a potential contributor to muscle weakness in the post-sepsis condition , whole protein was extracted from the tibialis anterior of control and post-sepsis animals ( both at 2 weeks and 1 month ) , and markers of irreversible oxidative damage were assessed by Western blot . Carbonylation was detected in proteins ranging in size from ~37 to 250 kDa , and showed ~2 fold higher levels in sepsis survivors at 2 weeks , as quantified by densitometric analysis of the entire lane ( p=0 . 002 , Figure 7A ) . This trend was no longer evident at 1 month ( p=0 . 832 , Figure 7B ) . Multiple bands were also detected with the oxidative marker 3-nitrotyrosine , which was ~2 . 5 fold higher in muscle from 2 week sepsis survivors ( p=0 . 016 , Figure 7C ) which remained modestly elevated at 1 month ( p=0 . 099 , Figure 7D ) . These results demonstrate that sepsis survivors have significant oxidative and nitro-oxidative damage even long after the acute sepsis condition has passed , which may indicate that ( 1 ) damage is not successfully cleared and therefore persists in the post-sepsis condition , and/or ( 2 ) damage is potentially propagated continuously by impaired mitochondrial activity .
Clinically , severe persistent physical impairment in sepsis survivors is a well-recognized phenomenon; however , progress towards understanding the underlying mechanisms have been limited largely due to the lack of an appropriate animal model with which to study long-term effects . Here , we describe our refinements to a murine model of sepsis , which enabled us to induce severe pathogenesis ( i . e . with multi-organ damage and high risk of mortality ) , yet rescue the majority of mice to obtain survivors in which long-term muscle function could be assessed . To the best of our knowledge , this is the first study to demonstrate skeletal muscle dysfunction long after recovery from experimental sepsis . We show that murine sepsis survivors have prolonged muscle weakness , independent of muscle atrophy , that is associated with ongoing mitochondrial dysfunction and oxidative damage . In the refinement of our model , we addressed four specific areas that were overlooked in prior research: ( Singer et al . , 2016 ) age of animals , ( Martin et al . , 2003 ) sepsis severity , ( Gaieski et al . , 2013 ) inclusion of muscle strength analysis , and ( Angus and Wax , 2001 ) avoiding artificial effects of sepsis model . First , in addition to being more likely to develop sepsis ( Starr and Saito , 2014; Elixhauser et al . , 2011; Dombrovskiy et al . , 2007 ) , patients at late middle-age more frequently suffer from post-sepsis dysfunction due to confounding issues such as preexisting comorbidities , reduced muscle mass , and lower protein intake ( Contrin et al . , 2013; Rahman et al . , 2014 ) . Further , these patients are most vulnerable to economic repercussions since post-sepsis functional limitations often prevent them from returning to work ( Iwashyna et al . , 2012; Hofhuis et al . , 2008 ) . Second , in order to observe long-term muscle weakness , considerably severe sepsis needed to be induced . Battle and colleagues ( Battle et al . , 2014 ) reported that patients who survived septic shock had reduced physical functioning and general health compared to patients who survived uncomplicated sepsis or sterile ( i . e . non-infectious ) systemic inflammatory response syndrome . Likewise , this suggests that animal models of sepsis with mild or modest severity ( i . e . no to little mortality in the absence of resuscitation ) have little clinical relevance . In order to model the morbidities with which sepsis survivors present clinically , experimental sepsis should be severe enough to cause chronic dysfunction , as we demonstrate in the present study . Third , the limited research that has been conducted at later time-points after experimental sepsis did not include muscle function assessment in parallel with molecular analyses , thus actual muscle weakness was not confirmed . One exception , Rocheteau et al . ( 2015 ) assessed muscle force ( specific tension ) in mice 21 days after injury induced by a combination of myotoxin injection and sepsis ( cecal ligation and puncture; CLP ) ; however , it is unclear whether sepsis itself had any effect on muscle force and no comparison was made with non-sepsis controls . Fourth , many sepsis studies use the CLP animal model; however , in this model surgical complications may influence recovery . Further , differences in cecum shape and size among older mice lead to inconsistency in ligation site ( Starr et al . , 2014 ) , and in mice kept long-term , the ligated cecum may result in unresolved necrosis , and sustained inflammation . In an effort to better model the clinical course of sepsis by addressing the above four factors , we induced sepsis by CS injection in late middle-aged mice , and used our ICU-like severe model of sepsis ( Steele et al . , 2017 ) which resulted in ~75% survival after an otherwise lethal insult . In the current study , using ex vivo force analysis , we found that middle-aged sepsis surviving mice exhibit a ~ 20% reduction of muscle force compared to non-sepsis controls up to 1 month after sepsis . This ex vivo muscle strength assessment circumvents potential confounding variables , such as behavioral , sensory , and operational factors , which are introduced by the commonly-used grip strength measurement ( Maurissen et al . , 2003 ) . We then provide clear evidence that long-term muscle weakness cannot be explained by loss of muscle mass , but rather reduced muscle quality . We show that specific force , which is normalized to unit area ( i . e . muscle size ) , is significantly reduced in murine sepsis survivors . These data imply that differences in muscle size among control and sepsis surviving animals did not account for differences in muscle strength . We additionally provide body composition analysis , wet weights of skeletal muscles , and myofiber cross-sectional area which collectively demonstrate that animals indeed have atrophy during sepsis , but that muscle mass largely recovers . Thus , this model mimics the clinical situation since aged subjects are most at risk for sepsis incidence , morbidity , and mortality . Further , it models non-surgical sepsis , and development of severe infection and inflammation prior to therapeutic intervention , similar to the typical clinical scenario . Additionally , this long-term model replicates the clinical presentation of post sepsis syndrome , or the broader issue of post-intensive care syndrome ( PICS ) , in which the patient survives but remains physically weak , even in the absence of muscle wasting . This differentiation between quality vs quantity is highly important , as prevention of muscle wasting is a large clinical focus ( Cohen et al . , 2015 ) . Muscle wasting is a major concern for sepsis patients in the ICU . Numerous factors , most notably muscle unloading during bedrest , insufficient nutrition , pharmacologic exposures , and imbalanced protein catabolism and synthesis , have been shown to contribute to sepsis-mediated atrophy ( Schefold et al . , 2010; Schreiber et al . , 2018; Eikermann et al . , 2006; Spranger et al . , 2003; Dres et al . , 2017; Yu et al . , 2018; Lang et al . , 2007; Yang et al . , 2018; Vary and Kimball , 1992; Puthucheary et al . , 2013; Bouglé et al . , 2016 ) . However , clinical interventions with early mobilization ( Denehy et al . , 2013; Morris et al . , 2016; Moss et al . , 2016; Tipping et al . , 2017 ) and nutritional support ( Goossens et al . , 2017; Hermans et al . , 2013; Ogilvie and Larsson , 2014 ) have not improved long-term outcomes , and some studies even show negative effects of high protein delivery on muscle wasting ( Puthucheary et al . , 2013 ) . Beyond this key issue , muscle weakness is not always accompanied by muscle wasting: they are separate entities that should not be considered synonymously ( Schefold et al . , 2010; Reid and Moylan , 2011 ) . However , in future studies it would be of interest to expand our experimental model to include other factors that are common for patients in the ICU , such as corticosteroid treatment , immobilization , and mechanical ventilation , which may further impact recovery and potentially reproduce the phenomenon of long-term muscle wasting post-sepsis . Nonetheless , taken together our data suggest therapeutic targets beyond muscle wasting are more likely to improve post-sepsis muscle weakness . In this study , we demonstrate that mitochondrial damage may be a key driver of chronic muscle weakness in murine sepsis survivors . Healthy mitochondria are critical for efficient muscle function , an ATP-dependent process . Skeletal muscles are mitochondrial dense and heavy oxygen consumers and thus greatly susceptible to damage during hypoxic conditions , including sepsis ( Zhou et al . , 2014 ) . Using transmission electron microscopy , we demonstrate that mitochondria in sepsis survivors have striking morphological abnormalities , including absent or fragmented cristae , enlarged matrix spaces , vacuolar/compartmentalized structures , and concentric ‘onion-like’ swirling cristae . Notably , the IMF mitochondria , which are primarily responsible for ATP production for muscle contraction ( Hood , 2001; Timpani et al . , 2015 ) , are more significantly damaged than SS mitochondria . As has been described ( Vincent et al . , 2016 ) , mitochondrial structure and function are tightly linked; thus , derangements are associated with profound physiological implications . The loss of inner mitochondrial membrane ( where oxidative phosphorylation occurs ) , as well as the abnormal presence of concentric/swirling cristae likely cause reduced bioenergetics in the post-sepsis condition . Indeed we demonstrate that mitochondrial bioenergetics are impaired in murine sepsis survivors long after sepsis itself is resolved as shown using respiration analysis on isolated mitochondria and complex enzyme activity analyses on whole tissue . We also unexpectedly found that in some myofibers there are areas devoid of any functional mitochondrial enzyme activity , where mitochondria are not capable of converting energy substrates to detectable levels . Measuring glycolytic rates and TCA cycle flux may increase our understanding of mitochondrial bioenergetics in the post-sepsis condition which may be assessed using a metabolomics approach in future studies . As muscle contraction is dependent on ATP for cross-bridge cycling , one can conclude that such mitochondrial impairments would contribute to reduced muscle contraction . It is important to note that accumulation of such grossly damaged and dysfunctional mitochondria in middle-aged sepsis survivors is likely perpetuated by the fact that mitophagy decreases during aging ( Sun et al . , 2007; Jang et al . , 2018 ) . We are currently investigating mitophagy and mitochondrial turnover in murine sepsis survivors to observe if reduced mitophagy and/or biogenesis contribute to mitochondrial dysfunction . Damaged mitochondria not only result in reduced ATP synthesis , but also in increased oxidative stress . As a byproduct of oxidative phosphorylation , mitochondria generate superoxide anions and nitric oxide . Mitochondrial damage , especially to complex I ( Castello et al . , 2006 ) , triggers persistent production of reactive oxygen species ( ROS ) and reactive nitrogen species ( RNS ) ( Balaban et al . , 2005 ) . Previous studies have shown that sarcomeric proteins , including tropomyosin , actin , and creatine kinase , scavenge free radicals during oxidative stress which causes post-translational modifications ( Fedorova et al . , 2009 ) . We observed significant oxidative and nitro-oxidative damage in the muscle homogenates from sepsis survivors . Importantly , these oxidative modifications are irreversible ( Cai and Yan , 2013 ) and require protein turnover for elimination . As oxidative stress has been shown to directly cause muscle fatigue ( Shindoh et al . , 1990; Reid , 2008; Stasko et al . , 2013 ) , our data suggest that persistent oxidative stress in the post-sepsis condition likely contributes to chronic muscle weakness . Moreover , the ROS and RNS secreted as a result of sepsis-induced mitochondrial damage then likely propagate mitochondrial dysfunction and energy failure through inhibition of the electron transport chain , thus acting as a vicious cycle of damage and functional decline ( Harman , 1972; Brandt et al . , 2017 ) . Future proteomic analyses would allow us to further elucidate mechanisms of mitochondrial and sarcomeric protein damage , as well as uncover many other pathways of interest that may underlie chronic muscle weakness after critical illness . This is likely multiphasic due to different mechanisms during the acute and chronic state , and influenced by age , and thus requires careful investigation . In summary , the present study demonstrates that profound weakness is present in the skeletal muscle of murine sepsis survivors and that such muscle weakness is not attributed to loss of muscle quantity alone , but rather is characterized by impaired quality on the mitochondrial and myofibrillar levels . Significant mitochondrial damage and dysfunction , as well as marked oxidative damage to skeletal muscle proteins , together likely contribute to chronic muscle weakness in sepsis survivors . We propose that mitochondrial dysfunction is central to other altered muscle physiology in the post-sepsis state , such as redox balance , calcium homeostasis , repressed autophagy , and mitochondrial permeability transition pore opening , altogether largely contributing to post-sepsis muscle weakness . This work provides strong evidence that therapeutic strategies aimed at restoring mitochondrial health , in addition to restoring muscle mass , will likely allow survivors to regain strength and improve quality of life after sepsis .
Late-middle-aged adult C57BL/6 mice were acquired from the National Institute on Aging , and all experiments were initiated when animals were 16 months old ( average male body weight ~34 grams , female body weight ~28 grams ) . Animals were acclimated in the Division of Laboratory Animal Resources at the University of Kentucky for at least 10 days before experimental procedures were performed . The mice were housed in pressurized intraventilated ( PIV ) cages with ad libitum access to drinking water and chow ( Teklad Global No . 2918 , Madison WI ) . Temperature ( 21–23°C ) , humidity ( 30–70% ) , and lighting ( 14/10 hr light/dark cycle ) were controlled . All experimental procedures were approved by the Institutional Animal Care and Use Committee . All animal handling techniques were in accordance with the National Institutes of Health guidelines for ethical treatment . Experimental groups were allocated using restricted randomization to evenly distribute particularly large or small animals among control and sepsis groups . Profound polymicrobial sepsis was initiated by intraperitoneal ( i . p . ) injection of cecal slurry ( CS; prepared as previously described in detail in Starr and Saito , 2014 , with minor refinements as noted in Steele et al . ( 2017 ) , at a dose which was 100% lethal when administered without subsequent therapeutic resuscitation . Animals which did not develop severe hypothermia ( ≤30°C at 12h ) were excluded from the study . Antibiotics ( imipenem , IPM; 1 . 5 mg/mouse , i . p . ) and fluid resuscitation ( sterile physiological saline 0 . 9% ) were administered beginning 12h following CS injection and continued twice daily . Antibiotic therapy was continued for at least 5 days , and fluid resuscitation ( 700 µL s . c . ) was continued until body temperature recovered to ≥35 . 0°C . Survival , body weight and body temperature were monitored regularly . Fat and lean mass were also monitored using EchoMRI Body Composition Analyzer ( EchoMRI LLC , Houston , TX , USA ) in a select group . Sepsis survivors were euthanized as previously described on days 4 , 2 weeks ( days 14–15 ) , and 1 month ( days 28–31 ) following CS injection along with non-sepsis controls as follows . Animals were anesthetized by inhalation of 5% isoflurane , and were maintained under 2 . 5% as a laparotomy was performed and blood was collected from the inferior vena cava ( IVC ) with a syringe containing 10% vol of 0 . 1M sodium citrate , and the IVC was cut to ensure exsanguination . Mice were singly housed and baseline food consumption was monitored daily at the same time of day for 5 days , and for 2 weeks after cecal slurry injection . Baseline food consumption was calculated as the average food consumed by all mice ( n = 5 sepsis survivors ) over the 5 day period , and daily averages were calculated after sepsis was induced . In a separate experiment , the food of non-sepsis animals was restricted according to the daily average food consumed by animals with sepsis ( i . e . pair-fed mice ) ; this experiment was conducted alongside freely-fed ( ad libitum ) controls ( n = 5 per group ) . Bacteremia was regularly assessed by culture of small blood samples taken from the tail vein as previously described ( Steele et al . , 2017 ) . In another experiment , systemic inflammation was assessed by measuring IL-6 in plasma samples obtained from animals sacrificed on day 4 and 2 weeks alongside non-sepsis controls using a high sensitivity ELISA kit ( eBIOSCIENCE , Vienna , Austria ) . TNFα and IL-10 were assessed in non-sepsis controls and animals with sepsis on day 4 and 2 weeks ( in the same mice ) using Meso Scale Discovery ( Rockville , Maryland ) customized V-PLEX multiplex assay . These assays were conducted in singlet due to limited sample volumes . Frozen muscles ( gastrocnemius ) were homogenized with TRIzol reagent ( Invitrogen , Carlsbad , CA ) , and total cellular RNA was purified using PureLink RNA Mini Kit ( Life Technologies , Grand Island , NY ) . The concentration of the RNA was determined by measuring the absorbance at 260 nm using a NanoDrop ONE microvolume spectrophotometer ( NanoDrop Technologies , Wilmington , DE ) , and integrity was confirmed through visualization of 18S and 28S RNA bands using an Agilent 2100 Bioanalyzer ( Agilent Technologies , Santa Clara , CA ) . Equivalent amounts of RNA were reverse transcribed into complementary DNA using SuperScript III First-Strand Synthesis SuperMix ( Life Technologies ) according to the manufacturer's protocol . TaqMan assays were purchased from ThermoFisher Scientific and the quantitative reverse transcriptase-polymerase chain reaction was performed on a QuantStudio 3 ( Applied Biosystems , Foster City , CA ) . Target gene expression was normalized to hypoxanthine-guanine phosphoribosyl transferase ( HPRT ) expression as an endogenous control , and fold change was calculated as 2- ( ΔΔCT ) , using the mean ΔCT of the Control group as a calibrator . TaqMan assays used were Mm00446190_m1 , Mm00443258_m1 , Mm01288386_m1 , and Mm03024075_m1 for IL-6 . TNFα , IL-10 , and HPRT , respectively . The right hind limb was skinned and immediately placed in oxygenated ( 95% O2-5% CO2 ) Krebs-Ringer solution ( 118 mM NaCl , 4 . 4 mM KCl , 1 . 2 mM MgSO4 , 1 . 3 mM NaH2PO4 , 2 . 5 mM CaCl2 , 25 mM NaHCO3 , and 10 mM D-Glucose; pH 7 . 4 ) . The muscle bath was continuously oxygenated while the extensor digitorum longus ( EDL ) was dissected using a microscope , and tethers were placed on the proximal and distal tendons using braided silk suture ( 4-0 ) . The muscle was freed from the leg , and mounted by attaching the tether at the distal end to a fixed hook and the proximal end to the lever arm of an ASI 300C‐LR Aurora Scientific transducer system ( Aurora , Ontario , Canada ) . The muscle was positioned between platinum electrodes and suspended in a temperature-controlled ( 25°C ) chamber containing the Krebs-Ringer solution which was continually oxygenated , and allowed to acclimate for 5 min . Using an Aurora stimulator ( model 701C ) , the muscle was subjected to electrical field stimulation , and resulting force output was recorded using ASI 610A Dynamic Muscle Control software . The optimum length ( Lo ) was found by adjusting the EDL length to maximum twitch force ( 1 Hz stimulation ) , which was measured using digital calipers . Maintaining the muscle at Lo , the force-frequency relationship was elucidated using stimulus frequencies of 1 , 15 , 30 , 50 , 80 , 150 , and 250 Hz ( note that 250 Hz stimulations were most appropriate for maximum force production for the middle aged mice used in these studies , whereas 300 Hz stimulations are commonly used when using young animals ) . When the protocol was complete , the muscle was transferred from the apparatus to a muscle bath where the suture tethers where carefully removed , then the muscle was weighed . The tissue length and weight were then used to calculate the physiological cross-sectional area ( Brooks and Faulkner , 1988 ) which was used to normalize force outputs ( specific force ) . Upon euthanasia , hind-limb muscles were carefully dissected , embedded in a thin layer of optimal cutting temperature ( OCT ) medium , pinned at resting length to cork board , and immersed in liquid nitrogen-cooled isopentane . After freezing , the cork board was transferred to a bed of dry ice where the muscle was removed and quickly transferred to pre-chilled cryovials and then stored at −80°C until sectioning ( 8 µm ) . The morphology of the mitochondria was evaluated by transmission electron microscopy as previously described ( Nakazawa et al . , 2017 ) with minor modifications . Briefly , extensor digitorum longus and tibialis anterior samples were collected and were immersed in fixation buffer comprised of 2 . 5% glutaraldehyde and 2 . 0% paraformaldehyde in 0 . 1 M sodium cacodylate buffer ( pH 7 . 4 ) overnight . The tissue samples were post-fixed for 1 hr ( 1% osmiumtetroxide , 1 . 5% potassiumferrocyanide , stained for 1 hr ( 1% uranyl acetate ) , and dehydrated . Samples were infiltrated overnight in a 1:1 mixture of propylene oxide and TAAB Epon ( Marivac Ltd . , St . Laurent , Canada ) , and viewed and imaged under the Philips Technai BioTwin Spirit Electron Microscope ( FEI , Hillsboro , OR , USA ) at the Harvard Medical School Electron Microscopy facility ( Nakazawa et al . , 2017 ) . At least five fields of view of subsarcolemmal and intermyofibrillar mitochondria populations were captured for each sample using an AMT 2 k CCD camera in a blinded manner . ImageJ software ( version 1 . 46 r , NIH ) was used to perform morphometric analysis using the cell counter analysis plugin . Muscle mitochondria were isolated as described previously ( Patel et al . , 2009; Gollihue et al . , 2018 ) . Briefly , the tibialis anterior was quickly dissected and placed in ice cold isolation buffer ( 215 mM mannitol , 75 mM sucrose , 0 . 1% BSA , 20 mM HEPES , 1 mM EGTA; pH 7 . 2 ) . The muscle was minced in trypsin ( 0 . 25 mg/ml ) before homogenizing on ice in 3 rounds of 5 s intervals ( total of 15 s; motor-driven Potter-Elvehjem homogenizer ) , and a protease inhibitor cocktail was added to the tissue homogenates . Mitochondrial pellets were obtained through differential centrifugation steps at 4°C . The pellet was suspended in isolation buffer and subjected to protein estimation was performed ( BCA protein assay kit , Thermo Scientific ) . It was previously shown that this method yields pure mitochondria being that the resulting fraction does not express transaminase ( Patel et al . , 2009 ) . Mitochondrial respiration was assessed in terms of mitochondrial oxygen consumption rate ( OCR ) using Seahorse Bioscience XFe24 extracellular flux analyzer , as previously reported ( Gollihue et al . , 2018; Patel et al . , 2014; Sauerbeck et al . , 2011 ) . Briefly , the 24-well dual-analyzer sensor cartridges ( Agilent Technologies , Santa Clara , CA , USA ) were placed in a carbon dioxide-free incubator at 37°C the day prior to the experiment . Once the mitochondria were isolated the day of the experiment , the Seahorse Flux Pak cartridges were filled in the following manner: ( A ) pyruvate , malate , and ADP ( to yield final concentrations of 5 mM , 2 . 5 mM , and 1 mM , respectively ) , ( B ) oligomycin ( 1 µg/mL ) , ( C ) carbonilcyanide p-triflouromethoxyphenylhydrazone ( FCCP; 3 µM ) , and ( D ) rotenone plus succinate ( 100 nM and 10 mM , respectively ) . The Seahorse XF24 Flux Analyzer was calibrated as previously described ( Patel et al . , 2014 ) using mitochondrial protein and respiration buffer ( 125 mM potassium chloride , 2 mM magnesium chloride , 2 . 5 mM potassium phosphate monobasic , 20 mM HEPES , and 0 . 1% BSA , adjusted to pH 7 . 2 ) . The experimental plates contained both non-sepsis control and post-sepsis samples in triplicate , and were subjected to OCR analysis through the series of dispensing solutions containing substrates and inhibitors that were added to injector ports A-D . Rates were generated using the AKOS oxygen consumption rate calibration algorithm; the average rate was determined for each sample well , and the replicates ( three technical replicates per sample ) were averaged . Upon euthanasia , skeletal muscles were dissected , placed in cryovials , snap frozen in liquid nitrogen , and stored at −80°C . Protein was extracted using the method described by Feng et al . ( 2012 ) with slight adaptations . Protein isolation buffer was comprised of 2% SDS , 10% glycerol , 2% 2-mercaptoethanol , in 50 mM Tris base , and a protease inhibitor cocktail was added ( Sigma ) . The muscle samples were homogenized using dounce homogenizers in approximately 20 volumes ( w/v ) of isolation buffer . The homogenates were transferred to Eppendorf tubes , and heated ( 80°C water bath for five minutes ) , centrifuged ( 12 , 000 g for 10 min at room temperature ) , and the resulting supernatant was collected . The protein concentration of the samples was evaluated using the Bio-Rad RC DC assay ( Hercules , CA , USA ) according to the manufacturer’s protocol . The control and post-sepsis skeletal muscle protein isolates ( 20 µg loading protein ) were resolved by SDS-PAGE electrophoresis ( Bio-Rad Mini Protean Tetra Cell system ) using TGX stain-free gradient ( 4–20% ) gels , total protein was visualized using stain-free technology ( ChemiDoc MP imaging system ) , and proteins were electrophoretically transferred ( Bio-Rad Trans-blot Turbo Transfer System ) to polyvinylidene difluoride ( pvdf ) membranes . Protein carbonyls were evaluated using the OxyBlot kit ( EMD Millipore Corp , Billerica , MA , USA ) by following the standard protocol with the minor adjustment of incubating in the primary antibody overnight ( 4°C ) . For 3-nitrotyrosine , the membranes were blocked for 1 hr in 5% milk at room temperature , and incubated in 1:3 , 000 3-nitrotyrosine primary antibody ( Abcam #ab61392 in 5% milk ) overnight at 4°C . Secondary antibody incubation ( 1:10 , 000 Santa Cruz #SC2005 ) was conducted for 1 hr at room temperature , and detected by chemiluminesce ( Bio-Rad Clarity Western ECL Substrate ) . Densitometry analysis was performed on the resulting blots using Image Lab software ( 2017 ) , and normalized to total protein analysis . Survival data were analyzed using Kaplan-Meier curves , with a Log Rank test to confirm significant differences between groups . In this case , and for all other data in this study , males and females were analyzed separately . For outcome variables measured over time or with multiple observations taken from the same subject ( such as specific force ) , a full-factorial repeated-measures ANOVA was performed , first analyzing overall differences across the various treatment groups . Likelihood ratio testing and Akaike Information Criterion ( AIC ) were used to select appropriate covariance structures in each case . When the interaction between treatment group and time was significant , relevant pairwise differences were calculated and reported with their respective F-statistics and p-values along with the full ANOVA table . A Kenward-Roger or Mancl-DeRouen adjustment was used , as appropriate , to correct for negative bias in the standard errors and degrees of freedom calculations induced by small samples . The remaining outcomes of interest were measured for each subject only once . Thus , for these quantitative variables , we performed a one-way ANOVA model or unpaired t-test , as appropriate , to analyze differences in the response across the groups studied . For outcomes whose overall ANOVA p-value was significant ( less than 0 . 05 ) , relevant pairwise differences were calculated and reported with their respective F-statistics and p-values along with the full ANOVA table . No outlier removal was performed , thus all data are presented . Sample sizes were estimated based on preliminary data comparing muscle strength of sepsis survivors and non-sepsis controls by ex vivo specific force analyses . With alpha = 0 . 05 , and conservative estimation of standard deviation for treatment ( i . e . sepsis ) and control at 20 , n = 6 per group had 80% power to detect a difference in the group means . For mitochondrial respiration assessments , previous experiments determined that n = 6 per group was sufficient to detect a 20% change from controls as statistically significant with 80% power ( Patel et al . , 2014 ) . With these analyses , we aimed for groups sizes of at least n = 6–7 for other experiments , anticipating consistent variation in the groups . This was not always achieved due to variation in sepsis survival rates . Due to feasibility , n = 3 per group were used for transmission electron microscopy observation . Although underpowered , differences were still observed after post-hoc and small sample size corrections were performed , where appropriate , as detailed above . Throughout the study , a p-value of less than 0 . 05 was considered significant . All analyses were completed in SAS 9 . 4 ( SAS Institute Inc; Cary , NC , USA ) . Statistical analyses were performed by statisticians ( AJS and GSH ) in the University of Kentucky Department of Statistics . | Sepsis is a life-threatening condition that occurs when a local infection spreads to the bloodstream and the body responds in such an exaggerated way that organs become damaged . Patients often require longs stays in intensive care units , and upon discharge experience chronic physical weakness and fatigue for several years . However , it was difficult to understand how sepsis can create these long-term problems because there was no way to study these issues in animals . To fill this knowledge gap , Owen et al . developed a protocol where they triggered sepsis in adult mice and then used therapeutic treatments similar to the ones found in intensive care units; as a result , most of the animals survived , with many then exhibiting chronic muscle weakness . Further observations in surviving mice revealed that muscle mass recovered after sepsis , so this weakness was not due to a drop in muscle mass: instead , the quality of the muscle fibers had worsened . More specifically , there were striking abnormalities in mitochondria , structures whose role is to power cells . The muscles also showed signs of persistent oxidative damage , a process in which toxic molecules produced by life processes accumulate and end up harming cells . Overall , these data suggest that reduced muscle quality contributes to chronic weakness after sepsis . While current programs for sepsis survivors aim to increase muscle quantity , the results by Owen et al . suggest that improving muscle quality , for example using antioxidant therapies , could be a new avenue of treatment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"immunology",
"and",
"inflammation"
] | 2019 | Chronic muscle weakness and mitochondrial dysfunction in the absence of sustained atrophy in a preclinical sepsis model |
Strong kinetochore-microtubule attachments are essential for faithful segregation of sister chromatids during mitosis . The Dam1 and Ndc80 complexes are the main microtubule binding components of the Saccharomyces cerevisiae kinetochore . Cooperation between these two complexes enhances kinetochore-microtubule coupling and is regulated by Aurora B kinase . We show that the Ndc80 complex can simultaneously bind and bridge across two Dam1 complex rings through a tripartite interaction , each component of which is regulated by Aurora B kinase . Mutations in any one of the Ndc80p interaction regions abrogates the Ndc80 complex’s ability to bind two Dam1 rings in vitro , and results in kinetochore biorientation and microtubule attachment defects in vivo . We also show that an extra-long Ndc80 complex , engineered to space the two Dam1 rings further apart , does not support growth . Taken together , our work suggests that each kinetochore in vivo contains two Dam1 rings and that proper spacing between the rings is vital .
Kinetochores link replicated chromosomes to spindle microtubules . They form attachments flexible and strong enough to stay attached to microtubules during assembly and disassembly . Estimates of the strength of the attachments suggest that they are much stronger than required just to pull a chromosome through the cellular milieu ( Fisher et al . , 2009; Nicklas , 1965 ) . Yet , incorrect attachments are weak enough to be detached and corrected . The Dam1 and Ndc80 complexes are the main microtubule-binding components in the budding yeast kinetochore . The Ndc80 complex connects the Dam1 complex to the rest of the kinetochore . The two complexes interact on microtubules , significantly increasing the strength of Ndc80 complex attachment to microtubules ( Lampert et al . , 2010 , 2013; Tien et al . , 2010 ) . How these two complexes interact on microtubules remains uncertain with conflicting results reported in the literature ( Lampert et al . , 2013; Maure et al . , 2011 ) . In addition to providing strong attachments between chromosomes and spindle microtubules , kinetochores also serve as regulatory hubs . The Ndc80 complex and the Dam1 complex are at the center of this regulation . The Ndc80 complex is a scaffold for the spindle assembly checkpoint ( Dou et al . , 2015; Hiruma et al . , 2015; Ji et al . , 2015 ) . Both the Dam1 and Ndc80 complexes are targets of Aurora B kinase , which corrects aberrant kinetochore-microtubule attachments to achieve bioriented attachments ( Reviewed in Sarangapani and Asbury , 2014 ) . Aurora B kinase phosphorylation of the Dam1 complex components Dam1p , Ask1p , and Spc34p together disrupts the interaction between the Dam1 and Ndc80 complexes ( Lampert et al . , 2013; Tien et al . , 2010 ) ; however , which phospho-protein is responsible for disrupting the interaction has yet to be deciphered . The spontaneous assembly in vitro of the Dam1 complex into microtubule-encircling rings suggests a compelling mechanism for kinetochore-microtubule coupling . The ring might be pushed by the curled depolymerizing microtubule protofilaments and prevented from falling off of flared assembling ends ( Asbury et al . , 2011; McIntosh et al . , 2008 ) . We don’t know the organization of the Dam1 complex in the cell , although oligomerization is essential for formation of strong kinetochore-microtubule attachments ( McIntosh et al . , 2013; Umbreit et al . , 2014 ) . Estimates of the number of Dam1 complexes at the kinetochore in vivo have suggested numbers as low as 10 , which is not enough to form a single 16-membered ring , and as high as 32 , which is enough for two rings ( Aravamudhan et al . , 2013; Joglekar et al . , 2006; Lawrimore et al . , 2011 ) . Despite this uncertainty , previous budding yeast kinetochore models have assumed one Dam1 complex ring ( Aravamudhan et al . , 2015; Cheeseman and Desai , 2008; Joglekar et al . , 2009; Tanaka et al . , 2007 ) . In this study , we provide evidence that the kinetochore requires the Ndc80 complex to bind and bridge two Dam1 complex rings in vivo and that Aurora B kinase regulates the interactions at both rings . Our work suggests that the specific ring-ring distance , defined by the Ndc80 complex bridge , is important for supporting growth . The presence of two Dam1 rings at the kinetochore , in a specific orientation bound to the Ndc80 complex , would have implications for how attachment strength is established and modulated by tension .
We identified where the Dam1 and Ndc80 complexes interact using protein cross-linking and mass spectrometry analysis ( Figure 1A and Figure 1—figure supplement 1 ) following protocols developed previously ( Hoopmann et al . , 2015; Kudalkar et al . , 2015; Tien et al . , 2014; Zelter et al . , 2015 ) . The full decameric Dam1 complex and tetrameric Ndc80 complex were cross-linked in the presence of taxol-stabilized microtubules . The C-terminal regions of Dam1p , Ask1p , and Spc34p formed cross-links to three distinct regions of Ndc80p ( Figure 1A; Figure 1—figure supplement 1 ) . The corresponding regions of Nuf2p also formed cross-links to these same regions of the Dam1 complex ( Figure 1—figure supplement 2 ) . These three C-terminal regions of the Dam1 complex include five sites phosphorylated by Aurora B kinase . Phosphorylation of all five sites fully disrupts the interaction between the Dam1 and Ndc80 complexes ( Lampert et al . , 2010; Tien et al . , 2010 ) . Thus , these sites likely represent primary interaction sites through which the Dam1 complex binds the Ndc80 complex . 10 . 7554/eLife . 21069 . 003Figure 1 . Dam1p , Ask1p , and Spc34p form cross-links to three distinct Ndc80p regions . ( A ) Cross-links between the Dam1p , Ask1p , and Spc34p of the Dam1 complex and Ndc80p of the Ndc80 complex . Dam1 and Ndc80 complexes were cross-linked in the presence of microtubules . Horizontal black bars represent proteins and the six vertical yellow lines indicate Aurora B kinase phosphorylation sites on the Dam1 complex . Red and blue lines show cross-links formed with DSS and EDC cross-linkers , respectively . For clarity , only the cross-links between the Dam1 complex proteins and Ndc80p are displayed . Data are shown for peptides with Percolator ( Käll et al . , 2007 ) assigned q-values ≤ 0 . 05 . Red bars on Ndc80p indicate regions where clusters of lethal mutations mapped ( from Tien et al . , 2013 ) to cross-linked regions . Grey bars on Ndc80p indicate clusters of lethal insertions outside of cross-linked regions . Green , blue , and orange trapezoids represent putative interactions ( A , B , and C ) between the Dam1 and Ndc80 complexes . ( B ) Bar diagram of Ndc80p with structural features . Green , blue and orange arrows indicate the positions of lethal mutations in interaction regions ANdc80p , BNdc80p , and CNdc80p used in this study . ( CH: calponin homology; HP: hairpin ) ( C ) Diagram showing the setup of TIRF microscopy experiments . Single molecule Ndc80-GFP complex binding on microtubules was visualized in the presence or absence of the Dam1 complex . ( D ) Lethal mutation in region ANdc80 , BNdc80 , or CNdc80 partially disrupts the Ndc80 complex’s interaction with Dam1 complex . Average residence time of Ndc80-GFP mutant and wild-type complexes on microtubules in the presence or absence of Dam1 complex . Bars represent average residence time ± error of the mean ( estimated by bootstrapping analysis; see Materials and methods for additional details ) . Ndc80-GFP complex microtubule residence time raw data are included in Figure 1—source data 1 . Refer to Supplementary file 1A for statistical analysis of data in part ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 00310 . 7554/eLife . 21069 . 004Figure 1—source data 1 . Table of Ndc80-GFP microtubule residence times for Figure 1D . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 00410 . 7554/eLife . 21069 . 005Figure 1—figure supplement 1 . Dam1 and Ndc80 complexes robustly react with DSS and EDC cross-linking agents . ( A ) Four types of peptides identified by mass spectrometry after cross-linking and trypsin digestion . Horizontal black bars represent protein sequence and vertical white lines represent residues able to react with the cross-linking agent . Unlinked: peptide that did not react with cross-linking agent; mono-link: only one end of the cross-linking reagent reacted with a residue; loop-link: intra-protein cross-link without a tryptic site in between the two cross-linked residues; self cross-link: intra-protein cross-link with a tryptic site in between the two-cross-linked residues . Dam1 and Ndc80 complexes were cross-linked in the presence of microtubules using DSS ( panels B and C ) or EDC ( panels D and E ) . ( B , D ) Cross-links above and below the black bars represent self-crosslinks and loop-links , respectively . Data are shown for peptides with Percolator ( Käll et al . , 2007 ) assigned q-values ≤ 0 . 05 . Highlighted regions above and below black bars represent sequence coverage from mass spectrometry . ( C , E ) Lines below the black bars represent mono-links . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 00510 . 7554/eLife . 21069 . 006Figure 1—figure supplement 2 . Dam1p , Ask1p , and Spc34p cross-link to Nuf2p , in agreement to the coiled-coil structure of Ndc80p and Nuf2p . ( A ) Cross-links between the Dam1p , Ask1p , and Spc34p of the Dam1 complex and Ndc80p and Nuf2p of the Ndc80 complex . For clarity , cross-links between the Dam1 complex proteins are not shown . Red lines represent cross-links between Ndc80p and Dam1p , Ask1p , or Spc34p . Blue lines represent cross-links between Nuf2p and Dam1p , Ask1p , or Spc34p . Purple lines represent cross-links between Ndc80p and Nuf2p . ( B ) Cross-links above and below the black bar represent self-crosslinks and loop-links , respectively , for Nuf2p . ( C ) Lines below the black bar represent mono-links for Nuf2p . Data from cross-linking experiments using DSS and EDC cross-linking agents are both shown . Data are shown for peptides with Percolator ( Käll et al . , 2007 ) assigned q-values ≤ 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 00610 . 7554/eLife . 21069 . 007Figure 1—figure supplement 3 . Ndc80 mutations outside of regions A , B , and C do not disrupt the interaction between Dam1 and Ndc80 complexes . ( A ) , ( B ) Survival probability of the data shown in Figure 1D: microtubule residence times of Ndc80-GFP complex with a lethal mutation in region ANdc80p , BNdc80p , or CNdc80p in the absence ( A ) and presence ( B ) of wild-type Dam1 complex . Data here have a range of n = 420 to 2051 . ( C ) Average microtubule residence times of Ndc80-GFP complexes with wild-type Ndc80p , or lethal five amino acid insertional mutation at position 219 or 652 . Bars represent average residence time ± error of the mean ( estimated by bootstrapping analysis; see Materials and methods for additional details ) . Ndc80-GFP complex microtubule residence time raw data are included in Figure 1—figure supplement 3—source data 1 . Refer to Supplementary file 1B for statistical analysis . ( D ) Survival probability curves of the data shown in ( C ) . ( E ) Superdex 200 size exclusion chromatography of the Ndc80 complex with wild-type Ndc80p or insertional mutation in ANdc80p , BNdc80p , or CNdc80p . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 00710 . 7554/eLife . 21069 . 008Figure 1—figure supplement 3—source data 1 . Table of Ndc80-GFP microtubule residence times for Figure 1—figure supplement 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 008 To identify the cognate binding site ( s ) on the Ndc80 complex , we compared the cross-linking data to a map of functional regions on Ndc80p identified in a previous linker-scanning mutagenesis screen ( Tien et al . , 2013 ) . Three of these regions map near or within the regions that cross-link to Dam1p , Ask1p , and Spc34p ( Figure 1A ) . We define these interactions as A , B , and C , and we will refer to the specific regions on each protein that are involved in these interactions as ADam1p and ANdc80p , BAsk1p and BNdc80p , and CSpc34p and CNdc80p , respectively ( Figure 1A and Table 1 ) . 10 . 7554/eLife . 21069 . 009Table 1 . The interacting regions in the Dam1 and Ndc80 complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 009InteractionProtein regionAmino acidsPhosphorylated residuesFive amino acid mutation ( insertion ) positionAADam1p241–330S257 , S265 , S292n/aANdc80p262–322n/a314BBAsk1p133–225S200n/aBNdc80p350–448n/a383CCSpc34p118–274T199n/aCNdc80p532–630n/a563 We tested if the lethal mutations identified in our screen in regions ANdc80p , BNdc80p , and CNdc80p ( Figure 1B ) interfere with interactions between the Ndc80 and Dam1 complexes on microtubules using single-molecule total internal fluorescence ( TIRF ) microscopy ( Figure 1C ) . Importantly , Ndc80 complexes carrying a mutation in regions ANdc80p , BNdc80p , or CNdc80p demonstrated homogeneous behavior in size exclusion chromatography , similar to the wild type Ndc80 complex ( Figure 1—figure supplement 3E ) . These mutations also did not alter the residence time of the Ndc80 complex alone on microtubules , consistent with previous observations ( Tien et al . , 2013 ) . As reported previously , the presence of the Dam1 complex significantly increases the residence time of wild-type Ndc80 complex on microtubules ( Tien et al . , 2010 ) . However , addition of the Dam1 complex only partially increased the residence time of the mutant complexes as compared to the wild-type Ndc80 complex ( Figure 1D and Figure 1—figure supplement 3A , B ) . Ndc80-GFP complex containing Ndc80p insertional mutations outside the regions ANdc80p , BNdc80p , and CNdc80p had similar microtubule and Dam1 complex binding characteristics as the wild-type Ndc80 complex ( Figure 1—figure supplement 3C , D ) . These observations suggest regions ANdc80p , BNdc80p , and CNdc80p each contribute to the interaction of Ndc80 complex with Dam1 complex; and the insertional mutants in these three regions were previously identified as lethal because they disrupted the interaction between the Dam1 and Ndc80 complexes . We then asked if the corresponding regions in the Dam1 complex also contribute to the interaction between Dam1 and Ndc80 complexes . Aurora B kinase phosphorylates the Dam1 complex at six different sites: Dam1p S20 , S257 , S265 , S292; Ask1p S200; Spc34p T199 ( Cheeseman et al . , 2002 ) . Phosphorylation of Dam1p S20 significantly decreases Dam1 complex oligomerization , while phosphorylation of the three C-terminal sites in Dam1p slightly inhibits direct binding of the Dam1 complex to microtubules ( Gestaut et al . , 2008; Zelter et al . , 2015 ) . In our prior work , we showed that phosphorylating all five sites besides S20 fully disrupts interaction between the Dam1 and Ndc80 complexes ( Tien et al . , 2010 ) . As shown in Figure 1A , these five sites fall into the three interaction regions defined above , ADam1p , BAsk1p , and CSpc34p ( Table 1 ) . We therefore systematically phosphorylated combinations of these five sites to dissect their contributions to the interaction between the Dam1 and Ndc80 complexes . We first purified recombinant mutant Dam1 complexes containing different combinations of Ser/Thr to Ala mutations , together with the S20A mutation in all cases . These mutants were treated with Aurora B kinase to produce a series of Dam1 complexes , each with a unique subset of phosphorylated sites . Each phosphorylated Dam1 complex was then tested for its ability to bind the wild-type Ndc80 complex in the single-molecule TIRF assay . Simultaneously phosphorylating all three regions , ADam1p , BAsk1p , and CSpc34p completely disrupted the interaction between the Dam1 and Ndc80 complexes , as reported previously ( Figure 2A and Figure 2—figure supplement 1A; Tien et al . , 2010 ) . Individually phosphorylating regions ADam1p , BAsk1p , or CSpc34p caused only partial disruption . Phosphorylating region BAsk1p together with CSpc34p also only partially disrupted the interaction , but phosphorylating ADam1p in combination with either BAsk1p or CSpc34p was sufficient to fully disrupt the interaction ( Figure 2B and Figure 2—figure supplement 1B , C ) . Neither the alanine mutations themselves , nor phosphorylation of a mutant Dam1 complex with all six sites mutated to alanine had any effect , confirming that the observed disruptions were due to phosphorylation at the known sites ( Figure 2A and Figure 2—figure supplements 1A and 2 ) . Together these data demonstrate that each of these regions contributes to the interaction between the Dam1 and Ndc80 complexes and is regulated by phosphorylation . 10 . 7554/eLife . 21069 . 010Figure 2 . Dam1 and Ndc80 complexes interact through three distinct sites . ( A ) Average microtubule residence time of the wild-type Ndc80-GFP complex alone , in the presence of mock treated , phosphorylated Dam1 complex , or phosphorylated Dam1 6A mutant complex . Dam1 6A: all six Dam1 complex phosphorylation sites mutated to Ala . ( B ) Average microtubule residence time of wild-type Ndc80-GFP complex in the presence of selectively phosphorylated Dam1 complex . Diagram on the left indicates phosphorylated proteins . ( C ) Average microtubule residence time of Ndc80-GFP complex with lethal mutation in region ANdc80p , BNdc80p , or CNdc80p in the presence of selectively phosphorylated Dam1 complex . Diagrams on the left indicate phosphorylated protein ( P ) and Ndc80p region with a lethal mutation ( ∧ ) . On the right are representative kymographs for each experiment . Bars represent average residence time ± error of the mean ( estimated by bootstrapping analysis; see Materials and methods or additional details ) . Ndc80-GFP complex microtubule residence time raw data are included in Figure 2—source data 1 . Refer to Supplementary file 1C for statistical analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01010 . 7554/eLife . 21069 . 011Figure 2—source data 1 . Table of Ndc80-GFP microtubule residence times for Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01110 . 7554/eLife . 21069 . 012Figure 2—figure supplement 1 . Microtubule residence time survival probability plots of experiments in Figure 2 . For all graphs , solid and dashed black lines represent curves for wild-type Ndc80 complex in the absence and presence of wild-type Dam1 complex , respectively . ( A–C ) Wild-type Ndc80-GFP complex microtubule residence time in the presence of variably phosphorylated Dam1 complex . Dam1 6A phos: all Dam1 complex phosphorylation sites mutated to Ala and treated with Aurora B kinase . ( D–F ) Microtubule residence time of Ndc80-GFP complex with a lethal mutation in region ANdc80p , BNdc80p , or CNdc80p in the presence of variably phosphorylated Dam1 complex . Data here have a range of n = 485 to 2051 . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01210 . 7554/eLife . 21069 . 013Figure 2—figure supplement 2 . Mutating Dam1 complex Aurora B kinase phosphorylation sites to Ala does not affect the interaction between Dam1 and Ndc80 complexes . Average microtubule residence time of wild-type Ndc80-GFP complex in the presence of mock treated phosphorylation-blocking Dam1 complex mutant constructs . A red ‘A’ identifies the phosphorylation sites that are mutated to alanine . Mock treatment was carried out by substituting ATP with H20 . Survival probability plots of the average microtubule residence time shown on the right . Ndc80-GFP complex microtubule residence time raw data are included in Figure 2—figure supplement 2—source data 1 . Refer to Supplementary file 1D for statistical analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01310 . 7554/eLife . 21069 . 014Figure 2—figure supplement 2—source data 1 . Table of Ndc80-GFP microtubule residence times for Figure 2—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 014 We next mixed and matched phosphorylation of the Dam1 complex with the mutations in Ndc80p . Full disruption by phosphorylation of the Dam1 complex was recapitulated by substituting a Dam1 complex phosphorylation with its corresponding mutation in Ndc80p . For example , phosphorylation of regions ADam1p and CSpc34p fully disrupted the interaction between the two complexes ( Figure 2B ) . Similarly , phosphorylation at region ADam1p plus mutation in region CNdc80p fully disrupted the interaction between the two complexes ( Figure 2C and Figure 2—figure supplement 1E , F ) . Phosphorylation combinations that partially disrupted the interaction between the Dam1 and Ndc80 complexes were also recapitulated by substituting a phosphorylation event with mutation in the corresponding binding region of Ndc80p . Phosphorylation at regions BAsk1p and CSpc34p partially disrupted the interaction between the two complexes ( Figure 2B ) . Similarly , mutation in region BNdc80pplus phosphorylation at region CSpc34p , or phosphorylation at region BAsk1p plus mutation in CNdc80p partially disrupted the interaction between the two complexes ( Figure 2C and Figure 2—figure supplement 1D ) . Together , our results demonstrate that the Dam1 and Ndc80 complexes interact at three different sites , each of which is phospho-regulated by Aurora B kinase . The Ndc80 complex is predicted to be a long fibril . The coiled-coiled length between regions ANdc80p and CNdc80p is predicted to be 29 nm ( Lupas and Gruber , 2005 ) ( Figure 3D ) . The width of a Dam1 complex ring is only ~7 nm ( Ramey et al . , 2011 ) . Assuming that the Ndc80 complex adopts a conformation roughly parallel to the microtubule axis ( Alushin et al . , 2010; Joglekar et al . , 2009 ) , a single ring could not span this distance . Thus we considered that multiple Dam1 complex rings might bind to the Ndc80 complex . Using negative stain electron microscopy , we imaged the Dam1 complex on microtubules in the absence and presence of the wild-type Ndc80 complex . 10 . 7554/eLife . 21069 . 015Figure 3 . The Ndc80 complex binds two Dam1 complex rings . ( A ) Representative EM images of Dam1 complex rings on microtubules in the absence or presence of the Ndc80 complex . Three experiments include no Ndc80 complex ( Ndc80c ) , wild-type Ndc80c , and Ndc80c10hep . Protein concentrations were 20 nM tubulin and 25 nM Dam1 and Ndc80 complexes . Scale bars: 100 nm . ( B ) Zoomed in images of a Dam1 complex ring doublet for the three experiments . Distances from middle of one ring to that of next ring are indicated . ( C ) Distribution of closest neighboring inter-ring distances measured for the three different conditions . Measurements made between 0–60 nm are shown . The cluster of distance measurements was fitted with a Gaussian distribution and the vertices ± standard deviations for each fit are listed . The raw data of Dam1 complex inter-ring distance measurements are included in Figure 3—source data 1 . ( D ) Diagram demonstrating the predicted coiled-coil distances between the Ndc80p lethal mutations . Diagram below shows the location of additional 10-heptad insertion between the hypothesized binding sites of two Dam1 complex rings . Refer to Supplementary file 1E for statistical analysis of data in part ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01510 . 7554/eLife . 21069 . 016Figure 3—source data 1 . Table of Dam1 complex inter-ring distance measurements for Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01610 . 7554/eLife . 21069 . 017Figure 3—figure supplement 1 . Full distributions of Dam1 complex inter-ring measurements for various experiments shown in Figure 3 and Figure 3—figure supplement 2 . Six conditions include Dam1 complex rings on microtubules in the absence of Ndc80 complex ( Ndc80c ) ( A ) , presence of wild-type Ndc80c ( B ) , Ndc80c10hep ( C ) , Ndc80c lethal mutation ANdc80p ( D ) , BNdc80p ( E ) , and CNdc80p ( F ) . The arrows identify the sharp peaks that contain at least 20% of the events . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01710 . 7554/eLife . 21069 . 018Figure 3—figure supplement 2 . Mutations in regions ANdc80p , BNdc80p , or CNdc80p disrupts the Ndc80 complex’s ability to bind two Dam1 complex rings . ( A ) Representative EM images of Dam1 complex rings on microtubules in the presence of the Ndc80 complex with a lethal mutation in region ANdc80p , BNdc80p , or CNdc80p . Scale bars: 100 nm . Protein concentrations were 20 nM tubulin and 25 nM Dam1 and Ndc80 complexes . ( B ) Zoomed in images of a Dam1 complex ring doublet for the three experiments . Distances from middle of one ring to those of the closest two rings were measured . ( C ) Distribution of closest inter-ring distances measured for the three experiments . Measurements made between 0–60 nm are shown . The cluster of distance measurements was fitted with a Gaussian distribution and the vertex ± standard deviation for each fit is listed . The raw data of Dam1 complex inter-ring distance measurements are included in Figure 3—figure supplement 2–source data 1 . Refer to Supplementary file 1E for statistical analysis of data in part ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 01810 . 7554/eLife . 21069 . 019Figure 3—figure supplement 2—source data 1 . Table of Dam1 complex inter-ring distance measurements for Figure 3—figure supplement 2C . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 019 The Dam1 complex formed rings on microtubules as previously reported ( Figure 3A , B; Miranda et al . , 2005; Westermann et al . , 2005 ) . In the absence of the Ndc80 complex , the Dam1 complex rings tended to bind in doublets of rings , at an inter-ring distance of 13 . 3 ± 2 . 4 nm ( avg ± s . d . ) ( Figure 3A–C and Figure 3—figure supplement 1A ) . This suggests that Dam1 complex rings have an affinity for each other along the longitudinal axis of microtubules and have a tendency to stack together . The stacked double-ring organization may be analogous to the paired ring helices previously reported to form at higher concentrations of Dam1 complex ( Miranda et al . , 2005 ) . The percent of Dam1 ring pairs found in closely spaced doublets was higher in the presence of the Ndc80 complex , 49 . 1% ( estimated by Gaussian fitting to the peak , see Materials and methods ) , as compared to without the Ndc80 complex , 18 . 8% . In addition , the two rings in each doublet were consistently held apart by a distance of 33 . 1 ± 3 . 8 nm , similar to the predicted coiled-coil length between regions ANdc80p and CNdc80p ( Figure 3 and Figure 3—figure supplement 1B ) . This suggests that one Ndc80 complex bridges two Dam1 rings by binding one ring at region ANdc80p and a second ring at region CNdc80p . To further test this hypothesis , 10 coiled-coil heptad repeats were inserted into the Ndc80p and Nuf2p coiled-coil domains to increase the length of the coiled-coil domain between the interaction regions by a predicted 10 . 5 nm ( Figure 3D ) . In the presence of the 10-heptad mutant Ndc80 ( Ndc8010-hep ) complex , Dam1 ring doublets were consistently separated by 42 . 1 ± 2 . 1 nm , 9 nm farther apart than in the presence of wild-type Ndc80 complex ( Figure 3 and Figure 3—figure supplement 1C ) . Finally , mutation in regions A , B or C of Ndc80p abolished its ability to bind two rings as the Dam1 inter-ring distance in these cases was the same in the absence of Ndc80 complex ( Figure 3—figure supplements 1D–F and and 2 ) . These results strongly support the hypothesis that the Ndc80 complex bridges two Dam1 complex rings in vitro . We next explored the possibility that Ndc80 complex might also bridge across two Dam1 complex rings in vivo . In particular , our observations raise the question of whether the specific 33 nm inter-ring distance , delimited by the Ndc80 complex , is important . To test this we constructed yeast expression plasmids for ndc8010-hep and nuf210-hep , encoding Ndc80p and Nuf2p each with 10-heptad repeats inserted . Using a plasmid shuffle assay , we found that even when introduced together , ndc8010hep and nuf210hep could not support growth , even though the encoded proteins were expressed ( Figure 4A , B ) . The Ndc8010hep complex is competent to assemble in vitro , to bind microtubules , to interact with the Dam1 complex on microtubules , and to bind two Dam1 complex rings ( Figure 3; Figure 4C ) . The only known difference is that the inter-ring distance is increased by 9 nm . These results suggest that the Ndc8010hep complex cannot support viability due to an altered Dam1 complex inter-ring distance . 10 . 7554/eLife . 21069 . 020Figure 4 . The Ndc8010hep complex does not support growth . ( A ) Red/white plasmid shuffle assay to test the viability of the Ndc8010hep complex . Solid red colonies indicate the inability of the Ndc8010hep complex and empty vector to support growth . Sectoring white colony indicates the ability of the wild-type Ndc80 complex to support growth . ( B ) α-FLAG immunoblot of crude lysate of cells expressing either wild-type Ndc80p , Ndc80p10hep , or empty vector . α-actin immunoblot of the same lysates shows equal loading of the lysates . ( C ) Wild-type or Ndc8010hep complex average residence time ( top ) on microtubules in the absence and presence of Dam1 complex as derived from the survival probability curves ( bottom ) . Bars represent average residence time ± error of the mean ( estimated by bootstrapping analysis; see Materials and methods or additional details ) . Ndc80-GFP complex microtubule residence times raw data are included in Figure 4—source data 1 . Refer to Supplementary file 1F for statistical analysis of data . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 02010 . 7554/eLife . 21069 . 021Figure 4—source data 1 . Table of Ndc80-GFP microtubule residence times for Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 021 The Ndc80p mutants in regions ANdc80p , BNdc80p , and CNdc80p cannot bind two Dam1 complex rings , and consequently have decreased interaction with the Dam1 complex on microtubules . To determine if these Ndc80p mutants confer defects in chromosome biorientation and attachment , we used an auxin-inducible degron system for degrading the wild-type Ndc80-AID ( Nishimura et al . , 2009 ) . Non-degradable wild-type or mutant NDC80 was integrated at the URA3 locus . Ndc80-AID was efficiently degraded in the presence of auxin ( Figure 5—figure supplement 1 ) . Budding index analyses revealed that cells carrying wild-type NDC80 progressed through mitosis as expected . Cells without NDC80 also budded and divided on schedule , despite major defects in kinetochore attachment ( see below ) , as expected since activation of the SAC requires the Ndc80 complex ( Dou et al . , 2015; Hiruma et al . , 2015; Ji et al . , 2015 ) . In contrast , cells carrying Ndc80p mutations in regions ANdc80p , BNdc80p , and CNdc80p arrested as large-budded cells ( Figure 5—figure supplement 2 ) suggesting defects in kinetochore attachment and indicating that the checkpoint remained functional . We performed a detailed analysis of the phenotype conferred by the NDC80 mutant alleles by imaging wild-type , depleted , and mutant cells with CEN3 and spindle pole bodies ( SPB ) tagged with GFP and mCherry , respectively . Cultures were synchronized at G1 , released from synchrony and then imaged after 60 min . In the majority of wild-type NDC80 cells , CEN3 was bioriented with both CEN3-GFP puncta on the spindle axis; 78% were in metaphase and 22% were in anaphase as judged by the spindle length . In the majority of the Ndc80-depleted cells , CEN3-GFP was off of the spindle axis , consistent with complete detachment . A mutation in region ANdc80p , BNdc80p , or CNdc80p , resulted in 61 , 59 , and 75% of the cells with monooriented CEN3 and 34 , 39 , and 21% of cells with detached CEN3 , respectively . ( Figure 5A , B ) . In addition , in the majority of cells depleted of Ndc80 or with a mutation in regions ANdc80p , BNdc80p , or CNdc80p the SPBs were separated by more than 2 μm ( Figure 5C ) . These long spindles and the high frequency of detached CEN3 suggest that few if any of the kinetochores were bioriented . Thus , we conclude that Ndc80p carrying a mutation in region ANdc80p , BNdc80p , or CNdc80p disrupts kinetochore biorientation and attachment in vivo . 10 . 7554/eLife . 21069 . 022Figure 5 . Lethal mutations in ANdc80p , BNdc80p , and CNdc80p have biorientation and microtubule attachment defects in vivo . ( A ) Representative images from Ndc80-AID degron experiments with cells carrying an extra copy of NDC80 wild-type , no NDC80 ( null ) , or a mutations in ANdc80p , BNdc80p , or CNdc80p . Cells were treated with 6 μM α-factor for two hours prior to treatment with 6 μM α-factor and 0 . 5 mM auxin for one additional hour . At time 0 , cells were released from α-factor and incubated in YPD medium containing 0 . 5 mM auxin . Images were taken at 60 min after release from α-factor arrest . Representative images are shown . ( B ) Stacked bar graphs showing proportion of the cells containing bioriented , monooriented or detached CEN3-GFP . Spindles with two CEN3-GFP puncta positioned between the two SPBs were counted as bioriented . Spindles with one CEN3-GFP puncta positioned between the two SPBs were counted as monooriented and spindles with one ( or rarely two ) CEN3-GFP puncta positioned off the spindle axis were counted as detached . Only cells with two Spc110-mCherry puncta and at least one CEN3-GFP punctum were included in the analysis . ( C ) Stacked bar graph showing the distance between SPBs . Wild-type NDC80 n = 236 cells; no NDC80 n = 210 cells; ANdc80pn = 210 cells; BNdc80pn = 190 cells; CNdc80pn = 260 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 02210 . 7554/eLife . 21069 . 023Figure 5—figure supplement 1 . Wild-type Ndc80-AID is degraded upon the addition of auxin . ( A ) Western blot images from Ndc80-AID degron experiments , blotting for Ndc80-AID and β-actin . Whole cell protein precipitation was carried out for Ndc80-AID cells carrying an extra copy of NDC80 wild-type , no NDC80 ( null ) , lethal mutation in ANdc80p , BNdc80p , or CNdc80p . Samples were collected at various time points after adding auxin ( Images were taken 120 min after auxin addition ) . ( B ) Normalized Ndc80-AID protein levels analyzed by western blot band intensities . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 02310 . 7554/eLife . 21069 . 024Figure 5—figure supplement 2 . Lethal mutation in region ANdc80p , BNdc80p , or CNdc80p causes mitotic arrest . ( A ) Budding index analyses of Ndc80-AID degron experiments . Ndc80-AID cells carrying an addition copy of NDC80 wild-type , no NDC80 ( null ) , lethal mutation in region ANdc80p , BNdc80p , or CNdc80p were fixed at various time points after release from α-factor arrest ( time 0 ) . ( B ) Stacked bar graphs showing the percentage of cells with zero , one , or two CEN3-GFP spots for the various cell types . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 024 The Ndc80 complex is responsible for recruiting the Dam1 complex to the kinetochore ( Janke et al . , 2002 ) . Since Ndc80 complexes carrying a mutation in region ANdc80p , BNdc80p , or CNdc80p are defective in binding the Dam1 complex in vitro , we asked if Dam1 complex kinetochore localization is also compromised in vivo . We again used the auxin-inducible degron system for degrading the wild-type Ndc80-AID , while wild-type or mutant NDC80 was integrated at the URA3 locus; Mtw1-YFP , Dad4-CFP , and Spc110-mCherry marked the kinetochore , Dam1 complex , and SPB , respectively . Cells expressing wild-type NDC80 had two distinct clusters of Mtw1-YFP between two SPBs , as expected for bioriented sister kinetochores . Consistent with previous studies ( Cheeseman et al . , 2001; Hofmann et al . , 1998; Li et al . , 2002 ) , the Dam1 complex localized along the spindle but also showed two distinct clusters that colocalized with the Mtw1-YFP . ( Figure 6A , B ) . None of the mutants showed the organized kinetochores seen in wild-type cells . Instead , in the milder mutants ( mutation in region BNdc80p; CNdc80p; or BNdc80p and CNdc80p ) kinetochores were spread along the spindle and a few were detached . In the more severe mutants ( mutation in region ANdc80p; ANdc80p and BNdc80p; ANdc80p and CNdc80p; ANdc80p , BNdc80p and CNdc80p; or Ndc80-depleted ) few kinetochores were associated with the spindle . Instead , most were detached from the microtubules between the two SPBs ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 21069 . 025Figure 6 . Lethal mutations in ANdc80p , BNdc80p , and CNdc80p confer defects in Dam1 complex recruitment to the kinetochore . ( A ) Representative images from NDC80-AID degron experiments with cells also containing wild-type NDC80 or mutation in region BNdc80p . The inner kinetochore was visualized with Mtw1-YFP , the Dam1 complex with Dad4-CFP , and the SPB with Spc110-mCherry . Cells were treated with 6 μM α-factor for two hours prior to the addition of 0 . 5 mM auxin for one additional hour . At time 0 , cells were released from α-factor and incubated in YPD medium containing 0 . 5 mM auxin . Images were taken at 60 min after release from α-factor arrest . ( B ) Line scan traces of Mtw1-YFP and Dad4-CFP from the images in ( A ) . Only cells with two SPBs in the same plane of focus were selected for analysis . Each line scan extended from one SPB to the other . Each point represents one pixel . ( C ) Summary of correlation analysis carried out between Mtw1-YFP and Dad4-CFP . For each line scan , a correlation of the positive or negative changes of pixel intensity along the line between the Mtw1-YFP and Dad4-CFP channels was calculated to examine co-localization along the spindle ( see Materials and methods ) . Between 18 and 33 cells were analyzed for each mutant . Mtw1-YFP and Dad4-CFP fluorescence intensity raw data are included in Figure 6—source data 1 . Refer to Supplementary file 1G for statistical analysis of data . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 02510 . 7554/eLife . 21069 . 026Figure 6—source data 1 . Tables of Mtw1-YFP and Dad4-CFP fluorescence intensities for cells analyzed in Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 02610 . 7554/eLife . 21069 . 027Figure 6—figure supplement 1 . Mutations in regions ANdc80p , BNdc80p , or CNdc80p results in severe defects in kinetochore biorientation and attachment . Examples of commonly observed phenotypes in experiments with NDC80-AID degron cells carrying an additional copy of ndc80 containing single or multiple mutations in regions ANdc80p , BNdc80p , and CNdc80p , or not carrying a second copy of NDC80 ( null ) . Each example shows Mtw1-YFP , Dad4-CFP , and Spc110-mCherry and the same experiment procedure was carried out as in Figure 6 . The majority of mutant cells with two SPBs exhibited detached kinetochores or had kinetochores spread along spindle microtubules . All scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 027 Unlike wild-type NDC80 , Dad4-CFP in mutant Ndc80 cells decorated the spindle but concentrated poorly at the Mtw1-YFP clusters that did occur ( Figure 6B and Figure 6—figure supplement 1 ) . In wild-type NDC80 cells , an increase or decrease in Mtw1-YFP fluorescence correlated with an increase or decrease in Dad4-CFP fluorescence , respectively; however , Ndc80-depleted cells showed poor correlation . One or multiple mutations in region ANdc80p , BNdc80p , and CNdc80p resulted in significantly decreased correlation between Mtw1-YFP and Dad4-CFP ( Figure 6C ) . Single mutations in region BNdc80p or CNdc80p resulted in less severe defects . These observations suggest that mutations in regions ANdc80p , BNdc80p , and CNdc80p had little effect on binding of Dam1 complex to the microtubules , but disrupted the ability of Ndc80 to recruit Dam1 complex to the kinetochore in vivo , thereby causing severe defects in kinetochore attachment and biorientation .
The discovery that the Dam1 complex forms oligomers and rings around microtubules galvanized the mitosis field by providing a molecular explanation for how kinetochores are able to maintain processive attachment to the flared ends of a depolymerizing microtubule . Rings are not required for the Dam1 complex to track depolymerizing microtubules in the absence of tension ( Gestaut et al . , 2008; Grishchuk et al . , 2008 ) . However , oligomerization of the Dam1 complex is required to maintain attachments under tension as would be experienced during metaphase ( Umbreit et al . , 2014; Volkov et al . , 2013 ) . Prior models of the kinetochore have all assumed that each kinetochore contains only one Dam1 complex ring . Here we show that the Ndc80 complex bridges two rings of the Dam1 complex in vitro . Lethal mutations in Ndc80p that block binding of either one of the two rings result in loss of the Dam1 complex from the kinetochore and failure in biorientation and attachment of kinetochores to the mitotic spindle . These results suggest that faithful microtubule attachments require two Dam1 complex rings per kinetochore in vivo . Counting of kinetochore components by fluorescence microscopy has yielded conflicting results for the number of Dam1 complexes at the yeast kinetochore depending on the standard used . Assuming two Cse4 histone molecules per kinetochore , Joglekar and coworkers report enough Dam1 complexes to form one ring at a kinetochore during metaphase and a partial ring during anaphase ( Aravamudhan et al . , 2013; Joglekar et al . , 2006 ) . Lawrimore and coworkers measured 5 . 5 Cse4 histone molecules per kinetochore , giving enough Dam1 complex for two rings at a kinetochore during anaphase ( Lawrimore et al . , 2011 ) ; whereas Coffman and coworkers measured 8 Cse4 histone molecules per kinetochore ( Coffman et al . , 2011 ) . Our results show that the Ndc80 complex can bind two Dam1 rings in vitro . Given the results of the latter two groups , we propose that there are two Dam1 complex rings at the kinetochore . The region before the hairpin and the loop region of Ndc80p have both been identified as binding sites for the Dam1 complex ( Lampert et al . , 2013; Maure et al . , 2011 ) . Our results indicate that the interaction between the Ndc80 and Dam1 complexes is more extensive forming a tripartite network . The C-terminal domains of Dam1p , Ask1p , and Spc34p interact with three distinct regions of Ndc80p , one of which is the hairpin consistent with the previous report ( Lampert et al . , 2013 ) . Our results are also consistent with a previous study showing an interaction between Ndc80p and the C-terminal region of Dam1p ( Kalantzaki et al . , 2015 ) . However , we did not find any evidence for an interaction between the Ndc80p ‘loop’ region and Dam1p ( Maure et al . , 2011 ) . The inability of the loop deletion mutant to assemble the Dam1 complex into the kinetochore might be due to an indirect effect . Comparing the predicted distances between regions ANdc80p , BNdc80p , and CNdc80p , and the 33 nm Dam1 complex inter-ring distance from EM experiments suggests that regions A and B contribute to the Ndc80 complex interaction with the ring closer to the N-terminus of Ndc80p/Nuf2p , while region C contributes to the Ndc80 complex interaction with the second ring ( Figure 7 ) . Our TIRF microscopy and live cell imaging results suggest that disrupting interaction A is more severe than disrupting interaction B or C . A recent computational model of the structure of the Dam1 complex bound to microtubules positions region ADam1p and BAsk1p near the microtubule surface , while region CSpc34p is positioned away from the microtubule surface ( Zelter et al . , 2015 ) . Accordingly , our model of the interaction between Dam1 and Ndc80 complexes shows the Ndc80 complex positioned under the first Dam1 ring and above the second Dam1 complex ring further from Ndc80p N-terminus ( Figure 7 ) . 10 . 7554/eLife . 21069 . 028Figure 7 . Model of the Ndc80 complex bridging two Dam1 complex rings , separated by 33 nm . Results in this study support ring 1 interaction being comprised of interaction A and B , and ring 2 interaction being comprised of interaction C ( refer to Figure 1 ) . This image depicts only one Ndc80 complex across two Dam1 complex rings , but multiple Ndc80 complexes would be present in vivo . DOI: http://dx . doi . org/10 . 7554/eLife . 21069 . 028 Aberrant kinetochore-microtubule attachments are phosphorylated and destabilized by Aurora B kinase ( Biggins et al . , 1999; Cheeseman et al . , 2002; Hauf et al . , 2003; Pinsky et al . , 2006; Tanaka et al . , 2002 ) . Aurora B kinase phosphorylation of the Dam1p , Ask1p , and Spc34p C-terminal sites together fully disrupts the interaction between the Dam1 and Ndc80 complexes ( Tien et al . , 2010 ) . This regulation mechanism has been thought of as binary , either full or no disruption of the interaction . We show that phosphorylation of certain combinations results in only partial disruption and full disruption does not require phosphorylation of all three proteins . Our results suggest how the Aurora B kinase might fine tune the interaction between Dam1 and Ndc80 complexes , resulting in a range of disruption . The inability of ndc8010hep and nuf210hep together to support growth in vivo might be explained in several ways . The addition of extra heptad repeats in the coiled-coil domains might destabilize formation of the heterotetrameric Ndc80 complex , for example . However , we do not currently favor this hypothesis because the extra-long mutant version of Ndc80p is expressed in vivo , and because the Ndc8010hep complex is stable and binds to both microtubules and to the Dam1 complex in vitro , similar to wild-type Ndc80 complex . Unnatural orientation of the Ndc80p/Nuf2p globular head in the Ndc8010hep complex , due to the addition of extra heptad repeats , might also disrupt the specific geometry that the Ndc80 complex takes in relation to other kinetochore proteins . However given the presence of the flexible loop in Ndc80p , it seems unlikely that the two ends of the complex are held in a rigid orientation relative to each other . Compared to the wild-type Ndc80 complex , the Ndc8010hep complex further separates the Dam1 complex rings in vitro . We propose that the specific 33 nm Dam1 complex inter-ring distance is vital . The transition from 13 nm , in the absence of the Ndc80 complex , to 33 nm Dam1 complex inter-ring distance could signal the establishment of kinetochore-microtubule attachment or biorientation . Deviation from the specific 33 nm distance might disrupt a yet unknown mechanism for detecting kinetochore-microtubule attachments , or bioriented attachments , or both . Future work will focus on characterizing the specific effects of the extra-long Ndc80 complex and the role of the Dam1 complex inter-ring distance in vivo .
The S . cerevisiae Ndc80 and Dam1 complexes were independently expressed in E . coli using polycistronic vectors , as previously described ( Gestaut et al . , 2008; Miranda et al . , 2005; Powers et al . , 2009; Tien et al . , 2010; Wei et al . , 2005 ) . The Dam1 complex component Spc34p C-terminus was tagged with Hisx6 and the Ndc80 complex component Spc24 N-terminus was tagged with FLAG or Hisx6 . Each complex was affinity-purified before further purification through gel filtration . For TIRF microscopy , Nuf2p C-terminus of the Ndc80 complex was tagged with GFP . GST-Ipl1 and GST-Sli15 ( residues 554–698 ) were purified as previously described ( Gestaut et al . , 2008; Tien et al . , 2010; Zelter et al . , 2015 ) . GST-Ipl1 ( pSB196 , Sue Biggins , Fred Hutchinsin Cancer Research Center , Seattle , WA ) and GST-Sli15 ( residues 554–698 ) ( pSB503 , Sue Biggins ) were expressed at 23°C and 37°C , respectively for 2 hr . GST-Ipl1 was purified using GSTrap HP ( GE Healthcare Biosciences , Pittsburgh , PA ) following manufacturer’s instructions , except that the elution buffer was 50 mM Tris buffer ( pH 8 . 0 ) , 250 mM KCL , 10 mM glutathione . HiTrap 26/10 Desalting column ( GE Healthcare ) was used to exchange the buffer to 50 mM HEPES buffer ( pH 7 . 4 ) , 100 mM NaCl . GST-Sli15 was purified with glutathione-Sepharose 4B resin ( GE Healthcare ) following manufacturer’s instructions , except that the elution buffer was 20 mM Tris buffer ( pH 8 . 0 ) , 200 mM NaCl , 1 mM β-mercaptoethanol , 1 mM EDTA , 10 mM glutathione . 4 μM recombinant Dam1 complex was incubated with 0 . 5 μM GST-Ipl1 , 0 . 5 μM GST-Sli15 , 200 mM NaCl , 10 mM ATP , 25 mM MgCl2 , and 50 mM HEPES buffer , pH 7 . 4 . Reaction mixtures were incubated for 90 min at 30°C . Under these conditions , we achieve nearly stoichiometric phosphorylation of the complex ( Gestaut et al . , 2008 ) . Mock treated ( non-phosphorylated ) controls of the Dam1 complex was carried out by substituting ATP with dH2O . Glass slides and functionalized coverslips were used to construct flow chambers , as reported previously ( Gestaut et al . , 2008 , 2010; Tien et al . , 2010; Zelter et al . , 2015 ) . A coverslip was adhered to a glass slide with double-sided tape , forming a flow chamber between two adjacent strips of tape . ‘Rigor’ kinesin was flowed through the chamber to non-specifically bind to the coverslip . Taxol-stabilized , Alexa-647-labelled microtubules were then flowed in and incubated for 5 min for immobilization . For all TIRF microscopy experiments , Ndc80-GFP complex was incubated at 50 pM for single-molecule imaging . Each phosphorylated or mock-treated Dam1 complex mutant construct was incubated at 2 . 5 nM . GFP and Alexa-647 fluorescence channels were simultaneously recorded using a custom TIRF imaging system ( Gestaut et al . , 2010 ) . All TIRF microscopy experiments were carried out in BRB80 ( 80 mM PIPES buffer ( pH . 6 . 8 ) , 1 mM EGTA , 1 mM MgCl2 ) in the presence of 8 mg/ml BSA , 0 . 04 mg/ml κ-casein , and an oxygen scavenger system ( 200 μg/ml glucose oxidase , 35 μg/ml catalase , 25 mM glucose , and 5 mM DTT ) . Each experimental condition was assayed between three and seven times yielding between 401 and 2051 measurements . This sample size adequately identified possible differences between different conditions . Analysis of the single particle tracking was carried out as previously described ( Tien et al . , 2010; Umbreit et al . , 2014; Zelter et al . , 2015 ) . Custom analysis software was developed in Labview ( National Instruments ) ( RRID: SCR_014325 ) ( Source code files 1–3 ) and Igor Pro ( Wavemetrics ) ( RRID: SCR_000325 ) ( Source code files 4–6 ) . Bootstrapping analysis was used to calculate the mean residence times ( Umbreit et al . , 2014 ) . Randomly resampling each dataset ( residence times from each TIRF microscopy experimental condition ) with replacement was repeated 1000 times . Carrying out this method with each dataset yielded normal distributions . Gaussian fits to these distributions were used to estimate the mean residence times and the width of the fit was used as an estimate of error . Statistical analysis of the TIRF data was performed with pairwise z-tests . P-values were computed form the pairwise z-scores: z = ( μ1− μ2 ) • ( δ12 + δ22 ) −0 . 5 . μ1 and μ2 are bootstrap average residence times of Ndc80-GFP complex . δ12 and δ22 are the corresponding estimates of error . Tables of pairwise comparisons can be found in Supplementary file 1A-D and F . XL-MS experiment was performed twice , using EDC and DSS cross-linking agents ( biological replicate ) . The two experiments identified similar interaction regions between the Dam1 and Ndc80 complexes . XL-MS was performed as described in Zelter et al . ( 2015 ) . Briefly , 10 µg of taxol stabilized microtubules were mixed with 10 µg Dam1 complex and 10 ug Ndc80 complex in 100 µL BRB80 at 25°C and allowed to stand for 5 mins . For DSS cross-linking , 3 µL 14 . 5 mM DSS in DMSO was added , and the mixture allowed to cross-link for 2 mins at 25°C before quenching by addition of 10 µL 0 . 5 M NH4HCO3 . For EDC cross-linking 7 . 5 µL 145 mM EDC plus 3 . 75 µL 145 mM Sulfo-NHS were added to the reaction , and the mixture allowed to cross-link for 30 mins at 25°C before quenching by addition of 10 µL 0 . 5 M NH4HCO3 . Quenched reactions were spun at 58 , 000 rpm in a TLA100 rotor for 10 min at 37°C . The pellet was resuspended in 100 µL ice cold buffer reduced with 10 mM dithiothreitol ( DTT ) at 37°C for 30 min followed by 30 mins alkylation at RT with 15 mM iodoacetamide ( IAA ) . Digestion with trypsin at a substrate to enzyme ratio of 60:1 was performed overnight at room temperature with shaking . Digested samples were acidified with 5 M HCL prior to being stored at −80°C until analysis . Mass spectrometry and data analysis was performed on either a Q-Exactive or Q-Exactive HF ( Thermo Fisher Scientific , Waltham , MA ) as previously described ( Zelter et al . , 2015 ) . Each cross-linked sample was run four times and the data were combined before analysis ( technical replicate ) . Mass spectra were converted into mzML using msconvert from ProteoWizard ( Chambers et al . , 2012 ) . Standard linear peptide searches were performed using Comet to identify all proteins in the sample ( Eng et al . , 2013 ) . Cross-linked peptides were identified using the Kojak version 1 . 4 . 2 cross-link identification software ( Hoopmann et al . , 2015 ) following the author’s instructions ( http://www . kojak-ms . org ) . Kojak results were exported to Percolator version 2 . 08 ( Käll et al . , 2007 ) to produce a statistically validated set of cross-linked peptide spectrum matches ( PSMs ) at the desired false discovery rate ( FDR ) threshold . Percolator ( RRID: SCR_005040 ) is a semi-supervised algorithm that assigns a statistically meaningful q value to each PSM through analysis of the target and decoy PSM distributions . Decoy PSMs derive from peptide sequences known to be false . In the current work the target databases consisted of all proteins identified in the sample analyzed , while the decoy databases consisted of the corresponding set of reversed protein sequences . Cross-link PSMs were considered to be false if at least one of the peptides was from a decoy protein sequence . The data presented in this paper was filtered to show only hits to the target proteins that had a Percolator assigned peptide level q value ≤ 0 . 05 . The complete , unfiltered list of all PSMs and their Percolator assigned q values , are available on the ProXL web application ( Riffle et al . , 2016 ) at: http://proxl . yeastrc . org/proxl/viewProject . do ? project_id=24 along with the raw MS spectra and search parameters used . Cleared tubulin was polymerized in a total volume of 40 μl BRB80 ( 80 mM PIPES buffer ( pH 6 . 8 ) , 1 mM EGTA , 1 mM MgCl2 ) containing 1 . 75 mM GTP , 1 mM MgCl2 , and 3 . 5% DMSO at 37°C for 30 min . Microtubules were pelleted and resuspended in BRB80 containing 10 μM taxol . All samples were prepared in BRB80 +10 μM taxol by mixing 20 nM microtubules , and 25 nM Dam1 in the absence or presence of 25 nM Ndc80 complex . Samples were incubated at room temperature for 15 min . Carbon-coated copper grids were negatively discharged in a glow discharge device . A 5 μl volume of sample was applied on a discharged grid for 20 s before being blotted . 6 μl 2% uranyl acetate was then applied on the grid and incubated for 1 min . The grid was blotted and air dried . EM samples were viewed on a transmission electron microscope ( Morgagni; FEI , Hillsboro , OR ) operating at 100 kV . Images were recorded on a bottom-mounted Orius ( Gatan , Pleasanton , CA ) camera at 22 , 000x magnification . ImageJ was used to measure the Dam1 complex inter-ring distances . Distances from middle of a ring to the middle of the next closest ring , on both sides , were measured . The EM microscopy experiments were performed with two independent copper grids . Similarity between the two replicates allowed us to combine the data . Statistical analysis of the EM data was performed with pairwise z-tests . P-values were computed from the pairwise z-scores: z = ( μ1− μ2 ) • ( δ12 + δ22 ) −0 . 5 . μ1 and μ2 are the center of the Gaussian fits around the single large cluster in each sample and δ12 and δ22 are the corresponding estimates of error . A table of pairwise comparisons can be found in Supplementary file 1E . Strains used in this study are derivatives of W303 ( Supplementary file 2A ) . We have previously verified the genotype of the W303 by whole-genome tiling as described ( Gresham et al . , 2006 ) . For strains constructed for this study , NDC80 alleles were integrated at the URA3 locus as previously described ( Widlund and Davis , 2005 ) . Plasmids containing the NDC80 alleles were sequenced before transformation . Successful integration was verified by PCR . Genes were tagged with genes encoding fluorescent proteins as described ( Wach et al . , 1997 ) and verified by PCR . To test if the Ndc8010hep complex is functional in vivo , we used a red/white plasmid shuffle assay ( Davis , 1992; Muller , 1996 ) using strain JTY5-8B ( ade2-1oc ade3Δ−100 leu2-3 , 112 ura3-1 ndc80Δ:: natMX ) , harboring pJT12 ( NDC80 ADE3 in a 2 µm vector ) . JTY5-8B was transformed with pJOK13 ( URA3 ndc8010hep ) and selected for growth on SD-ura low adenine plates . Transformation with pJOK13 yielded non-sectoring red colonies ( strain JOKY3 ) demonstrating that the ndc8010hep is not viable when paired with wild-type NUF2 ( data now shown ) . JOKY3 was then grown in SD-lys-ura media before being transformed with pJOK018 ( LEU2 nuf210hep ) , pJOK019 ( LEU2 NDC80 ) , or pRS315 ( LEU2 ) and plated on SD-ura-leu plates . Four colonies from each transformation were streaked onto SD-ura low-ade plates to screen for sectoring colonies ( technical replicate ) . Two transformations were performed for each condition ( biological replicate ) , both trials gave the same result . Strains for live-cell imaging were constructed using previously described strains: SPC110-mCherry , pCUP1-GFP-LacI12 and CEN3::33LacO ( Wargacki et al . , 2010 ) ; NDC80-3V5-IAA7 and pGPD1-TIR1 ( Miller et al . , 2016 ) . Sue Biggins kindly provided the Ndc80-AID strain . Strains containing NDC80 , ndc80-314 ( ins ANdc80p ) , ndc80-383 ( ins BNdc80p ) , or ndc80-563 ( ins CNdc80p ) were constructed through an integration method previously described ( Widlund and Davis , 2005 ) . YFP and CFP were integrated into Mtw1 and Dad4 , respectively , through homologous recombination with PCR products amplified with primers containing homologous sequence of its target locus . Asynchronously growing cells ( 30 Klett units ) were arrested with 6 μM α-factor for 2 hr . Auxin ( IAA ( 0 . 5 mM ) was then added and the incubation continued for an additional hour . The cells were released from α-factor into YPD +0 . 5 mM IAA medium and imaged at 15 min time-intervals from time 0 to 120 min . At each time point , cells were also fixed in formaldehyde for budding index analysis . Budding index was determined twice ( from two different colonies ) and yielded the same conclusion for each of the five yeast strains shown in Figure 5 . Data is shown for one of the two experiments . Cells were placed on agar pads prior to imaging , as previously described ( Muller et al . , 2005 ) . Cells were imaged using a DeltaVision system ( Applied Precision , Issaquah , WA ) equipped with IX70 inverted microscope ( Olympus , Center Valley , PA ) , a Coolsnap HQ digital camera ( Photometrics , Tucson , AZ ) , and a U Plan Apo 100X objective ( 1 . 35 NA ) . For CEN3 tracking biorientation assay , CEN3-LacI was visualized with GFP-LacO and the spindle pole bodies were visualized withSPC110-mCherry . Both GFP and mCherry were imaged with 0 . 4 s exposures with 16 , 0 . 2 µm z-sections . Images were binned 2 × 2 with a final resolution of 512 × 512 . Biorientation analysis was done manually for those cells with 2 mCherry puncta using Imaris software ( Bitplane , Switzerland ) ( RRID: SCR_007370 ) . For all the strains tested , the 60 min time point was chosen for data analysis . The biorientation assay was performed twice for each strain , each time with a different colony and on a different day . The combination of two experiments yielded sample sizes between 190 and 260 cells . For Dam1 complex kinetochore localization assay , kinetochore , Dam1 complex , and spindle pole bodies were visualized with Mtw1-YFP , Dad4-CFP , and Spc110-mCherry respectively . All three proteins were imaged with single focal plane 0 . 4 s exposures . Images were not binned with a final resolution of 1024 × 1024 . Line scans analyses were obtained of Mtw1-YFP and Dad4-CFP fluorescence using Imaris software ( Bitplane ) ( RRID: SCR_007370 ) and extended from one SPB to the other . For each line scan , a correlation of the sign ( positive or negative ) of pixel intensity changes across the line between the Mtw1 and Dad4 channels was calculated to examine co-localization along the spindle . This was calculated as: R = Nagree - Ndisagree / N − 1 , where Nagree is the number of pixels that changed with the same sign ( both channels increased or decreased ) , Ndisagree is the number of pixels that changed with opposite signs ( one channel increased and one decreased ) and N is the length of the line scan . For each strain , a mean R was calculated from all line scans . The localization assay was performed twice for each strain , each time with a different colony and on a different day . The combination of two experiments yielded sample sizes between 18–33 cells . Statistical analysis of the Mtw1-YFP and Dad4-CFP mean R-scores was performed with Student’s t-test . A table of the pairwise comparisons can be found in Supplementary file 1G . During the yeast live-cell imaging , samples were collected for western blot analysis . After the addition of auxin , cell samples were collected at various time points . 0 . 7 ml cells and 0 . 7 ml 20% tricholoroacetic acid were incubated at 4°C for 1 hr . Samples were pelleted for at 20 , 000xg for 10 min at 4°C . The pellet was washed by resuspending in 1 ml 4°C ethanol . The sample was then pelleted and resuspended once more before a final pelleting step . Supernatant was discarded and sample was air dried before resuspending in 50 μl 0 . 1N NaOH , SDS-PAGE sample buffer . Western blot analysis was performed , probing for Ndc80p ( gift from Arshad Desai , Ludwig Cancer Research Center , University of California San Diego ) and β-Actin ( Abcam , Cambridge , MA ) . Protein levels analysis was carried out using Image Studio Lite ( Li-Cor , Lincoln , Nebraska ) ( RRID: SCR_014211 ) . Each western blot was performed twice , yielding the same conclusion for each experiment . | The genetic material inside yeast , human and other eukaryotic cells is stored within structures called chromosomes . Every time a cell divides to make two “daughter” cells , all the chromosomes in the cell must be copied and then separated into two equal sets , with each set delivered to a different daughter cell . If the copied chromosomes divide into unequal sets the daughter cells may die or not work properly . Chromosomes are attached to tube-like structures called microtubules , which pull the two sets of chromosomes to opposite ends of the cell just before it divides . Microtubules are constantly shrinking and growing , and chromosomes must stay attached the whole time , or they will not be correctly separated . Many proteins are involved in attaching chromosomes to microtubules , including two groups known as the Dam1 complex and the Ndc80 complex . The Dam1 complex forms a ring around microtubules , while the Ndc80 complex forms a long flexible rod that can bend in the middle . However , it was not clear how these shapes allow these complexes to perform their roles in cells . Kim et al . studied these two complexes from yeast cells . The experiments show that the Ndc80 rod bridges two Dam1 rings , not one as previously assumed . Both of these rings and the rod are required for chromosomes to attach to microtubules . Cells with defective Ndc80 rods that can only bind to one ring do not distribute their chromosomes correctly when they divide . Kim et al . also show that the Ndc80 rod holds the Dam1 rings a specific distance apart , which also appears to be important for the chromosomes to be correctly divided between daughter cells . The next step following on from this work is to find out exactly why both rings are needed and why they seem to need to be a set distance apart . Human and other eukaryotic cells divide in a similar way to yeast cells , so these findings may help us to understand what goes wrong in Down’s syndrome and other diseases caused by cells having the wrong number of chromosomes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2017 | The Ndc80 complex bridges two Dam1 complex rings |
The σ subunit of bacterial RNA polymerase ( RNAP ) confers on the enzyme the ability to initiate promoter-specific transcription . Although σ factors are generally classified as initiation factors , σ can also remain associated with , and modulate the behavior of , RNAP during elongation . Here we establish that the primary σ factor in Escherichia coli , σ70 , can function as an elongation factor in vivo by loading directly onto the transcription elongation complex ( TEC ) in trans . We demonstrate that σ70 can bind in trans to TECs that emanate from either a σ70-dependent promoter or a promoter that is controlled by an alternative σ factor . We further demonstrate that binding of σ70 to the TEC in trans can have a particularly large impact on the dynamics of transcription elongation during stationary phase . Our findings establish a mechanism whereby the primary σ factor can exert direct effects on the composition of the entire transcriptome , not just that portion that is produced under the control of σ70-dependent promoters .
The σ subunit of bacterial RNA polymerase ( RNAP ) is an essential initiation factor that specifies the recognition of promoter sequences in the context of the RNAP holoenzyme ( Feklistov et al . , 2014 ) . All bacteria contain a primary σ factor that directs transcription from the major class of bacterial promoters; in addition , most bacteria contain multiple alternative σ factors that direct transcription from specialized promoters in response to stress or alterations in growth state ( Gruber and Gross , 2003; Osterberg et al . , 2011; Guo and Gross , 2014 ) . Among the best-studied primary σ factors is Escherichia coli σ70 , which recognizes promoters that are defined by two conserved hexameric DNA sequence elements termed the −10 and the −35 elements ( consensus sequences: TATAAT and TTGACA , respectively ) . Members of the σ70 family share a conserved 4-domain architecture , with domain 2 contacting the −10 element and domain 4 contacting the −35 element ( Gross et al . , 1998; Paget and Helmann , 2003; Feklistov et al . , 2014; Paget , 2015 ) . E . coli also has six alternative σ factors , five of which are members of the σ70 family and recognize similarly positioned promoter elements using the counterparts of σ70 domains 2 and 4 . Most alternative σ factors exhibit highly restricted promoter specificity ( Koo et al . , 2009b; Rhodius et al . , 2013 ) . Thus , genes that are responsive to disparate physiological inputs often carry two or more promoters that are recognized by distinct σ factors ( Wade et al . , 2006; Gama-Castro et al . , 2008; Cho et al . , 2014 ) . Although σ factors were historically identified as promoter specificity factors , it has become clear that their roles are not limited to the initiation phase of transcription . In particular , multiple studies have shown that the release of σ from the transcription complex is not required for entry into the elongation phase of transcription ( reviewed in Mooney et al . , 2005; Perdue and Roberts , 2011 ) . Furthermore , the functional properties of a transcription elongation complex ( TEC ) containing σ differ from the properties of a TEC that does not contain σ . For example , TEC-associated σ70 can induce transcription pausing by engaging promoter −10-like sequence elements within transcribed regions ( Ring et al . , 1996; Brodolin et al . , 2004; Nickels et al . , 2004; Hatoum and Roberts , 2008; Deighan et al . , 2011; Perdue and Roberts , 2011 ) , a phenomenon that was first uncovered in the context of the bacteriophage λ late gene promoter ( reviewed in Roberts et al . , 1998; Perdue and Roberts , 2011 ) . This pausing occurs due to an interaction between the −10-like element and domain 2 of TEC-associated σ70 ( the same domain of σ70 that binds the promoter −10 element during transcription initiation ) . In addition , the presence or absence of σ can alter the accessibility of the TEC to elongation factors , including the λ Q protein and RfaH ( Roberts et al . , 1998; Nickels et al . , 2002 , 2006; Sevostyanova et al . , 2008 ) , and can influence the ability of RNAP to reinitiate transcription at certain promoters ( Bar-Nahum and Nudler , 2001 ) . Initial-transcribed-region −10-like elements , such as those associated with the λ late promoters and the late promoters of other lambdoid phages , induce early elongation pausing because they are recognized by TECs that have not yet released the σ70 that was used during initiation ( Marr et al . , 2001; Mukhopadhyay et al . , 2001; Nickels et al . , 2004; Kapanidis et al . , 2005 ) . In prior work , we showed that such promoter-proximal σ70-dependent pause elements also function to inhibit σ70 loss during the earliest stages of elongation , increasing the σ70 content of downstream TECs ( Deighan et al . , 2011 ) . This effect can be detected using a template that carries a second pause element positioned downstream of a promoter-proximal pause element; specifically , the presence of the promoter-proximal pause element facilitates the retention of σ70 in the TEC and thus substantially enhances the extent of pausing induced by the downstream pause element both in vitro and in vivo ( Deighan et al . , 2011 ) . Although promoter −10-like elements that induce transcription pausing can be recognized by a σ70 subunit that has been retained in the TEC after promoter escape , in vitro studies have established that transcribed region −10-like elements can also be recognized by a σ70 subunit that was not present during initiation , but rather joined the TEC by loading in trans during elongation . Thus , it has been shown that the efficiency of pausing induced by transcribed region −10-like elements can be increased in vitro by increasing the concentration of free σ70 in the transcription reactions ( Mooney and Landick , 2003; Sevostyanova et al . , 2008; Deighan et al . , 2011; Sevostyanova et al . , 2011 ) . A key question that emerges from these in vitro findings is whether or not cellular conditions permit σ70 to gain access to the TEC through this ‘trans-acting pathway’ in vivo . Here we address this question by employing an assay that enables us to measure the extent of TEC pausing induced by a −10-like element within a transcribed region in vivo . We find that the extent of pausing induced by a transcribed-region −10-like element is sensitive to the intracellular concentration of σ70 , indicating that σ70 can gain access to the TEC in trans . We further establish that σ70 can gain access to the TEC in trans and engage −10-like elements within transcribed regions that are expressed under the control of either a σ70-dependent promoter or a promoter that is recognized by an alternative σ factor . In addition , we show that the extent of TEC pausing mediated by σ70 trans loading varies as a function of growth-phase . Our findings imply that distinct σ factors can control initiation and elongation on the same transcription unit in vivo , and that the functional consequences of σ70 trans loading vary as a function of growth state .
To determine whether or not σ70 can bind in trans to the TEC in vivo , we took advantage of the fact that TEC-associated σ70 can induce transcription pausing by engaging transcribed-region −10-like elements . We therefore sought to determine whether or not the efficiency of pausing induced by a transcribed-region −10-like element was sensitive to the concentration of σ70 present in vivo . To do this , we introduced into E . coli cells a plasmid carrying a σ70-dependent promoter , λPR′ , fused to a transcribed region containing a −10-like element that has the potential to induce σ70-dependent pausing at a nascent RNA length of ∼35 nt ( Deighan et al . , 2011 ) ( Figure 1A , top ) ; the transcription unit also contains an intrinsic terminator element ( positioned to terminate transcription after the synthesis of an ∼116 nt transcript ) . Pausing induced by the −10-like element on this template in vitro is sensitive to the concentration of free σ70 in the transcription reactions ( Deighan et al . , 2011 ) ; furthermore , because the template lacks a promoter-proximal −10-like element , engagement of the pause element by σ70 that is retained during the transition from initiation to elongation contributes minimally to the observed pausing ( Deighan et al . , 2011 ) . 10 . 7554/eLife . 10514 . 003Figure 1 . σ70 trans loading on a σ70-dependent transcription unit in vivo ( MG1655 ) . ( A ) Top: schematic of DNA template carrying λPR' , transcribed-region consensus extended –10 element ( wild-type or mutant ) and terminator ( see ‘Materials and methods’ for the λPR′ promoter sequence ) . Transcribed-region sequences that are complementary to the LNA probe are underlined ( grey bar ) and the positions corresponding to pause sites are indicated . middle Analysis of RNA transcripts in vivo by LNA probe-hybridization . RNA was isolated from MG1655 cells harvested at an OD600 of 0 . 8–1 . 0 ( see ‘Materials and methods’ ) . Pausing is quantified by dividing the signal in the ∼35-nt pause RNA band by the sum of this signal and the signal in the terminated ( full-length ) band; this ratio is expressed as a percentage ( relative abundance ) . Mean and SEM of six independent measurements are shown . Asterisks ( * ) designate values that were too low ( <approximately threefold above background ) for accurate quantification . M , 10-nt RNA ladder . bottom Analysis of σ70 levels by Western blot . Amount of soluble σ70 is normalized to the amount in cells carrying the experimental template ( wt ) and a vector that does not direct σ70 over-production . Mean and SEM of three independent measurements are shown . ( B ) Top: schematic of DNA template carrying λPR′ , initial-transcribed-region σ70-dependent pause element , transcribed-region consensus −10 element and terminator . middle Analysis of RNA transcripts in vivo by locked-nucleic-acid ( LNA ) probe-hybridization , as in panel A . bottom Analysis of σ70 levels by Western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 003 We tested whether or not the efficiency of pausing at a nascent RNA length of ∼35 nt on this template was sensitive to the concentration of σ70 present in vivo by introducing into the cells a second plasmid that did or did not direct the production of excess σ70 . To detect nascent RNAs associated with paused TECs ( pause RNAs ) and full-length terminated transcripts produced from this template , we isolated total RNA and used Northern blotting with a locked-nucleic-acid ( LNA ) probe , as described previously ( Deighan et al . , 2011 ) . We quantified the extent of pausing by dividing the signal associated with a pause RNA by the sum of this signal and the signal associated with the full-length terminated transcript ( hereafter termed relative abundance ) . We found that the relative abundance of a ∼35-nt pause RNA ( see Deighan et al . , 2011 ) increased ∼fivefold when σ70 was overproduced by a factor of ∼7 , compared to that observed in cells containing chromosomally encoded σ70 only ( Figure 1A ) . Furthermore , the ∼35-nt pause RNA was barely detected with or without excess σ70 using a control template carrying base-pair substitutions that disrupt sequence-specific recognition of the transcribed-region −10-like element by σ70 region 2 ( Deighan et al . , 2011 ) ( Figure 1A ) . We conclude that pausing of the TEC under the control of a −10-like element within a transcribed region is sensitive to the intracellular concentration of σ70 , suggesting that σ70 can access the TEC in trans , in vivo . Next , we investigated whether or not σ70 trans loading could augment the effect of a promoter-proximal pause element on the σ70 content of downstream TECs . To do this , we used LNA probe-hybridization to detect transcripts produced from the template shown in Figure 1B . This λPR′ template bears the same −10-like element as the template shown in Figure 1A , but in addition carries a promoter-proximal −10-like element ( positioned between +1 and +6 ) that induces σ70-dependent pausing at a nascent RNA length of ∼16 nt . Consistent with previous findings ( Deighan et al . , 2011 ) , the presence of the promoter-proximal −10-like element resulted in a substantial increase ( ∼ninefold ) in the relative abundance of the ∼35-nt pause species ( compare Figure 1A , B ) . Nonetheless , when σ70 was overproduced , the relative abundance of the ∼35-nt pause species increased further ( ∼1 . 5 fold; Figure 1B , middle and bottom panels ) , indicating that the effect of the promoter-proximal −10-like element on the σ70 content of downstream TECs is not saturating . We next sought to determine whether or not free σ70 can bind to TECs on a transcription unit controlled by an alternative σ factor . To address this possibility we generated a new template that carried a promoter recognized by RNAP holoenzyme carrying σ28 , an alternative σ factor that controls the expression of genes involved in flagellar synthesis ( Chilcott and Hughes , 2000; Koo et al . , 2009a ) . This σ28 dependent promoter ( Ptar ) was fused to the same transcribed region sequences that are present on the λPR′ template shown in Figure 1A starting at position +6 ( including the −10-like element; Figure 2A ) . We first performed in vitro transcription experiments to determine whether or not σ70 could access the TEC and induce pausing on this template . We formed open complexes on Ptar using RNAP holoenzyme containing σ28 and then allowed a single round of transcription to occur in the absence or presence of excess σ70 . We monitored the RNA content of each reaction at three time points after the initiation of transcription . Addition of σ70 to the transcription reactions resulted in the appearance of a cluster of RNAs ∼35-nt in length ( Figure 2 , compare lanes 4–6 with lanes 1–3 ) . These RNAs were not observed when reactions were performed using a control template carrying disruptive base-pair substitutions within the transcribed-region −10-like element ( Figure 2B , lanes 7–12 ) . A set of reactions performed in the presence of σ70 but in the absence of σ28 confirmed that appearance of the cluster of ∼35-nt RNAs is strictly dependent on transcription that initiates from Ptar under the control of σ28 ( Figure 2B , lanes 13–15 ) . In addition , the distribution of RNA species within this cluster closely resembles that within a similar cluster produced when reactions were performed using the λPR′ template ( Figure 1A ) and RNAP holoenzyme containing σ70 ( Figure 2B , lanes 16–18 ) . We conclude that the ∼35-nt RNAs are pause RNAs that arise due to the ability of σ70 to bind TECs generated via transcription initiating at Ptar under the control of σ28 . These findings therefore indicate that free σ70 can bind to TECs on a σ28-controlled transcription unit in vitro . 10 . 7554/eLife . 10514 . 004Figure 2 . σ70 trans loading on a σ28-dependent transcription unit in vitro . ( A ) . Schematic of DNA template carrying Ptar , transcribed-region consensus −10 element ( wild-type or mutant ) and terminator . Template positions corresponding to pause sites are indicated . Note that the pause sites and terminated transcripts emanating from the Ptar promoter are located one base closer to the transcription start site ( +1 ) than on the λPR′ template ( Figure 1A ) . ( See ‘Materials and methods’ for the Ptar promoter sequence . ) ( B ) . Analysis of RNA transcripts in vitro . Single-round in vitro transcription reactions were performed with reconstituted RNA polymerase ( RNAP ) holoenzyme containing σ28 ( lanes 1–12 ) , RNAP core enzyme ( lanes 13–15 ) or reconstituted RNAP holoenzyme containing σ70 ( lanes 16–18 ) and three different templates: Ptar with a wild-type ( wt ) transcribed-region −10 element ( lanes 1–6 & 13–15 ) or a mutated ( mut ) transcribed-region −10 element ( lanes 7–12 ) and λPR′ with a wild-type transcribed-region −10 element ( lanes 16–18 ) . The reactions were performed as a time course with samples taken at 1 , 6 and 18 min after transcription was initiated; these reactions were performed in the absence of transcript cleavage factors GreA and GreB , resulting in a characteristic pattern of long-lived pause species ( Deighan et al . , 2011 ) . Where indicated , excess σ70 ( 1 μM ) was added with the ‘start mix’ after open complex formation . RNAs associated with paused transcription elongation complexes ( TECs ) ( pause ) and terminated transcripts ( full length ) are labeled . The asterisk ( * ) indicates a shorter terminated transcript that is the result of transcription initiating under the control of the transcribed-region −10 element when the σ70-containing holoenzyme is present in the reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 004 We then sought to determine whether or not σ70 can bind to TECs on a σ28-controlled transcription unit in vivo . For this experiment we introduced into cells three compatible plasmids . The first plasmid carried either the wild-type Ptar template or a mutant Ptar template with base-pair substitutions that disrupt sequence-specific recognition of the transcribed-region −10-like element by σ70 . The second plasmid did or did not direct the production of excess σ70 and the third plasmid did or did not direct the production of excess σ28 . We isolated total RNA and soluble protein from cells and used LNA probe-hybridization to detect transcripts emanating from the Ptar promoter ( Figure 3A , top ) and Western blotting to assess the concentrations of σ70 ( Figure 3A , middle ) and σ28 ( Figure 3A , bottom ) . We found that transcripts emanating from Ptar were detected only in cells carrying the plasmid that directed the synthesis of excess σ28 ( Figure 3A , compare lanes 2–5 with lane 6 ) . Furthermore , in the presence of excess σ28 but in the absence of excess σ70 , we detected a small amount of an RNA species that migrated between the 30-nt and 40-nt RNA markers ( Figure 3A , lane 2 ) . This RNA species was similar in size to the ∼35-nt pause RNA detected by LNA probe-hybridization with the λPR′ template in vivo ( Figure 1A ) and to the cluster of ∼35-nt pause RNAs produced from the Ptar template in vitro in the presence of excess σ70 ( Figure 2B , lanes 4–6 ) . We found that the relative abundance of this ∼35-nt RNA was increased ∼sevenfold when σ70 was overproduced by a factor of ∼3 ( Figure 3A , compare lanes 2 and 3 ) . In addition , the ∼35-nt RNA was not detected in cells containing the mutant Ptar template carrying base-pair substitutions in the transcribed-region −10-like element ( Figure 3A , lanes 4 and 5 ) . 10 . 7554/eLife . 10514 . 005Figure 3 . σ70 trans loading on a σ28-dependent transcription unit in vivo . ( A ) . top Detection of RNA transcripts in vivo from the templates shown in Figure 2A by LNA probe-hybridization . Transcribed-region sequences that are complementary to the LNA probe are as in Figure 1A . RNA was isolated from MG1655 cells harvested at an OD600 of 0 . 8–1 . 0 . Pausing is quantified by dividing the signal in the ∼35-nt pause RNA band by the sum of this signal and the signal in the terminated ( full-length ) band . Mean and SEM of three independent measurements are shown . Asterisks ( * ) designate values that were too low for accurate quantification . M , 10-nt RNA ladder . middle Analysis of σ70 levels by Western blot . Amount of soluble σ70 is normalized to the amount in cells carrying the experimental template ( wt ) and a vector that does not direct σ70 over-production . Mean and SEM of three independent measurements are shown . bottom Analysis of σ28 levels by Western blot . ( B ) . Analysis of RNAP-associated transcripts produced from the wild-type Ptar template . RNA was isolated from the lysate fraction ( lys ) or the immunoprecipitated fraction ( IP ) of SG110 cells ( OD600 ∼0 . 5 ) and analyzed by LNA probe-hybridization . The cells contained a vector directing the synthesis of σ28 , as well as a vector that did or did not direct σ70 overproduction . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 005 Next , we sought to determine whether or not the ∼35-nt RNA species produced under the control of the Ptar promoter was RNAP-associated , as would be expected for a pause RNA . To carry out this experiment , we used a strain carrying a chromosomal rpoC-3xFLAG gene , encoding the RNAP β′ subunit with a C-terminal 3xFLAG tag , which enables us to isolate RNAP-associated transcripts by immunoprecipitating RNAP with an antibody against FLAG . We introduced into this strain the plasmid carrying the wild-type Ptar template , the plasmid directing the production of excess σ70 or the corresponding empty vector , and the plasmid directing the production of excess σ28 . We isolated RNA from cell lysates ( Figure 3B , lys ) or from 3×FLAG-tagged TECs immunoprecipitated with an antibody against FLAG ( Figure 3B , IP ) and used LNA probe-hybridization to detect transcripts emanating from the Ptar promoter . The results indicate that a major fraction of the ∼35-nt RNA species , but not the full-length terminated transcript , was immunoprecipitated with an antibody against FLAG whether the cells lacked or contained plasmid encoded overproduced σ70 ( Figure 3B ) . Thus , we conclude that a major fraction of the ∼35-nt RNA species , but not the full-length terminated transcript , is RNAP-associated . Taken together , the results of Figure 3 establish that the appearance of the ∼35-nt RNA depends both on the presence of σ28 and on an intact −10-like element , that the relative abundance of the ∼35-nt RNA is increased upon overproduction of σ70 , and that the ∼35-nt RNA is RNAP-associated . We therefore conclude that the ∼35-nt RNA produced from the Ptar template in vivo represents a pause RNA that arises due to the ability of σ70 to bind TECs generated under the control of σ28 . Furthermore , our ability to detect σ70-dependent pause species produced under the control of a promoter that is recognized by an alternative σ factor enables us unambiguously to identify pausing that is mediated by trans-loaded σ70 . Thus , our findings establish that σ70 can access the TEC in trans , in vivo . Although experiments using the Ptar template revealed that σ70 trans loading is detectable even in the absence of σ70 overproduction , we found that during the exponential phase of growth the extent of pausing due to chromosomally encoded trans-loaded σ70 appeared to be low ( Figures 3A and 4A; the relative abundance of the ∼35-nt RNA was <5% ) . However , when we harvested RNA from stationary phase cells containing the Ptar template , we found that the relative abundance of the ∼35-nt RNA was ∼50% ( Figure 4A , lane 3 ) , which was reduced to ∼10% when the transcribed-region −10-like element was mutated ( Figure 4A , lane 5 ) . Furthermore , like those detected during exponential phase , the ∼35-nt RNAs detected from the Ptar template during stationary phase were RNAP-associated ( Figure 4—figure supplement 1A ) . Thus , the ∼35-nt RNAs detected during both exponential phase and stationary phase exhibit hallmarks of a σ70-dependent pause species ( stable association with RNAP and sensitivity to mutations in the transcribed region −10-like element ) . We conclude that the relative abundance of pause RNAs that arise due to σ70 trans loading varies with growth-phase . 10 . 7554/eLife . 10514 . 006Figure 4 . Growth phase dependent σ70 trans loading on a σ28-dependent transcription unit in vivo . ( A ) . Detection of RNA transcripts in vivo from the templates shown in Figure 2A by LNA probe-hybridization . Transcribed-region sequences that are complementary to the LNA probe are as in Figure 1A . RNA was isolated from SG110 cells harvested at an OD600 of ∼0 . 5 ( log ) or ∼2 . 5 ( sta ) . Pausing is quantified by dividing the signal in the ∼35-nt pause RNA band by the sum of this signal and the signal in the terminated ( full-length ) band . Mean and SEM of six independent measurements are shown . Asterisks ( * ) designate values that were too low for accurate quantification . M , 10-nt RNA ladder . ( B ) . top Detection of RNA transcripts derived from the wt template in vivo after treatment with rifampicin . bottom Percent of transcript remaining relative to T = 0 at indicated time points after addition of rifampicin . Mean and SEM of ten ( log , 1 m ) , eight ( sta , 1 m ) , or six ( log and sta , 3 m ) independent measurements are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 00610 . 7554/eLife . 10514 . 007Figure 4—figure supplement 1 . ( A ) . Analysis of RNAP-associated transcripts produced from the wild-type Ptar template . RNA was isolated from the lysate fraction ( lys ) or the immunoprecipitated fraction ( IP ) of SG110 cells ( OD600 ∼2 . 5 ) and analyzed by LNA probe-hybridization . The cells contained a vector directing the synthesis of σ28 . ( B ) . Analysis of σ70 levels by Western blot . Relative quantification of σ70 ( top ) is normalized to the abundance of rpoA ( α ) in each sample ( bottom ) . Mean and SEM of six independent measurements are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 007 To investigate the basis for the growth phase-dependent change in the abundance of the pause RNAs that arise due to σ70 trans loading , we first performed Western blot analysis to compare the amounts of σ70 in exponential and stationary phase cells . The results indicated that the cell extracts prepared from exponential and stationary phase cultures contained comparable amounts of σ70 ( Figure 4—figure supplement 1B ) . We conclude that the growth phase-dependent increase in the abundance of the ∼35-nt pause RNAs is not a consequence of an increase in the total cellular concentration of σ70 . ( We note that these data do not exclude the possibility that growth phase-dependent changes in the amount of free σ70 available to bind the TEC contribute to changes in the abundance of the pause RNAs that arise due to σ70 trans loading . ) We next used the RNAP inhibitor rifampicin to analyze the half-life of pause RNAs that arise due to σ70 trans loading during exponential phase or stationary phase . To do this , we isolated RNA from cells either just before or 1 and 3 min after rifampicin treatment and used LNA-probe hybridization to measure the decay of the ∼35-nt RNAs and full-length transcripts . We found that the half-life of the ∼35-nt pause RNA was greater in stationary phase than in exponential phase ( Figure 4B ) . In addition , the full-length terminated transcript was at least as stable in stationary phase as in exponential phase ( Figure 4B ) , excluding the possibility that the increase in the relative abundance of the pause RNA might simply reflect a destabilization of the full-length transcript in stationary phase . Thus , our findings indicate that the extent of pausing on the Ptar template due to trans-loaded σ70 varies with growth phase , at least in part , due to an increase in the half-life of the pause .
The results presented here , coupled with prior work ( Shimamoto et al . , 1986; Ring et al . , 1996; Bar-Nahum and Nudler , 2001; Mukhopadhyay et al . , 2001; Brodolin et al . , 2004; Nickels et al . , 2004; Wade and Struhl , 2004; Kapanidis et al . , 2005; Raffaelle et al . , 2005; Reppas et al . , 2006; Mooney et al . , 2009; Deighan et al . , 2011 ) define two pathways whereby σ70 can access the TEC in vivo , a pathway that operates in cis and a pathway that operates in trans ( Figure 5 ) . The cis-acting pathway depends on retention in the TEC of the σ70 that was used during initiation , with the extent of σ70 retention being modulated by the sequence of the initial transcribed region ( Figure 5A ) ( Deighan et al . , 2011 ) . Thus , the cis-acting ( retention ) pathway is necessarily restricted to transcription units controlled by σ70-dependent promoters . In contrast , the trans-acting pathway identified in this study , which can be functionally defined by its sensitivity to the intracellular concentration of σ70 , is potentially operative on all transcription units ( Figure 5B ) . Moreover , the two pathways can function in concert . Thus , we found that σ70 trans loading can increase the σ70 content of TECs generated under the control of a σ70-dependent promoter even in the presence of an initial-transcribed-region σ70-dependent pause element that augments σ70 retention . 10 . 7554/eLife . 10514 . 008Figure 5 . Dual pathways for σ70 to associate with the TEC in vivo . ( A ) . Cis-acting pathway ( Deighan et al . , 2011 ) . The retention in the TEC of the σ70 that was used during initiation enables pausing at transcribed-region −10-like elements on transcription units that are expressed under the control of σ70-dependent promoters . Presence of an initial-transcribed-region σ70-dependent −10-like element increases the σ70 content of downstream TECs and increases the efficiency of pausing at a second σ70-dependent pause element further downstream . Promoter , grey rectangle; σ70-dependent pause elements , black rectangles; RNA , wavy red line . ( B ) . Trans-acting pathway . Binding of σ70 to TECs that have lost the σ factor used during initiation ( here , σ28 ) increases the efficiency of pausing at a transcribed-region σ70-dependent pause element . Promoter , blue rectangle; σ70-dependent pause element , black rectangle; RNA , wavy red line . DOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 008 The use of a transcription unit expressed under the control of an alternative σ factor enabled us to analyze the trans-acting pathway independent of the cis-acting pathway . We found that the effects of trans-loaded σ70 on pausing varied with growth phase . In particular , pausing mediated by chromosomally encoded trans-loaded σ70 was detectable during the exponential phase of growth and this pausing became more prominent during stationary phase . Our experiments revealed that this increase in the relative abundance of the pause species during stationary phase was explained at least in part by an increase in pause half-life in stationary phase cells as compared to exponentially growing cells . We speculate that this increase in pause half-life might be due to a drop in the intracellular NTP concentrations as nutrients are depleted and the cells enter stationary phase ( Buckstein et al . , 2008 ) . It is intriguing to consider our findings in light of a prior report of growth-phase dependent changes in the ability of purified RNAP holoenzyme to retain σ70 during transcription elongation as assayed in vitro ( Bar-Nahum and Nudler , 2001 ) . In particular , the authors of this study found that RNAP holoenzyme purified from stationary phase cells produced a substantially higher fraction of σ70-containing TECs than did RNAP holoenzyme purified from exponentially growing cells , possibly suggesting that the stationary phase RNAP core enzyme binds σ70 more tightly . We note that the effect of growth phase on the relative abundance of pause RNAs may not be limited to σ70–dependent pausing . In fact , our experiments revealed a potential pause species that was detectable above background in stationary phase cells even when the transcribed-region −10-like element was mutated ( Figure 4A , lane 5 ) . We suggest that this RNA arises due to the presence of an overlapping consensus pause element that is recognized by the core RNAP ( G–10Y–1G+1; see Figure 2A ) ( Herbert et al . , 2006; Larson et al . , 2014; Vvedenskaya et al . , 2014 ) and is not disrupted by the mutations in the −10-like element . Two or more different σ factors often control the expression of a given gene by directing initiation from distinct upstream promoters ( Wade et al . , 2006; Gama-Castro et al . , 2008; Cho et al . , 2014 ) . Our findings illustrate another mechanism whereby the combined input of multiple σ factors can modulate gene expression . Specifically , we show that distinct σ factors can direct initiation and modulate elongation on the same transcription unit . Such ‘σ cross regulation’ might enable the cell to integrate signals transmitted via σ70 and an alternative σ factor to modulate gene expression within a single transcription unit under the control of a non σ70–dependent promoter . In principle there are several ways that σ70 trans loading might modulate gene expression . First , as shown here , σ70 trans loading can cause the TEC to pause , which is expected to influence transcription output directly in a manner that depends on pause half-life ( and , as suggested by our results shown in Figure 4 , may become particularly relevant in stationary phase ) . σ70–dependent pausing might also influence gene expression indirectly , by facilitating engagement of regulatory factors , influencing formation of RNA secondary structures , or influencing translation ( Roberts et al . , 1998; Wickiser et al . , 2005; Landick , 2006; Pan and Sosnick , 2006; Lemay et al . , 2011; Perdrizet et al . , 2012; Larson et al . , 2014; Nechooshtan et al . , 2014; Belogurov and Artsimovitch , 2015 ) . Second , σ70 trans loading might impede the accessibility of the TEC to other elongation factors such as NusG or RfaH , which share the same primary binding site on RNAP ( Sevostyanova et al . , 2008; Mooney et al . , 2009 ) . Future work will be required to investigate the extent to which σ70 trans loading contributes to gene expression through these or other mechanisms . In this regard , the application of sequencing-based methodologies such as native elongating transcript sequencing ( NET-seq ) ( Churchman and Weissman , 2011; Larson et al . , 2014; Vvedenskaya et al . , 2014 ) and chromatin immunoprecipitation sequencing ( ChIP-seq ) ( Myers et al . , 2015 ) should enable the identification of transcription units that manifest growth phase-dependent pausing attributable to trans loaded σ70 . Nevertheless , our findings add to a growing body of evidence that the functions of σ are not limited to the initiation phase of transcription . Furthermore , they establish a mechanism whereby the primary σ factor can extend its reach by exerting direct effects on the composition of the entire transcriptome , not just that portion that is produced under the control of σ70-dependent promoters .
All experiments were performed in E . coli strain MG1655 or SG110 ( Vvedenskaya et al . , 2014 ) in which the chromosomal rpoC gene is fused to a 3xFLAG epitope tag-encoding sequence . Plasmids used in this study are listed in Table 1 . Promoter sequences are as follows . λPR′ : TTGACTtattgaataaaattgggTAAATTtgactcA and Ptar: TAAAGTTTcccccctccttGCCGATAAcgagatcA , where the −10 and −35 elements and the +1 nucleotide are capitalized . 10 . 7554/eLife . 10514 . 009Table 1 . PlasmidsDOI: http://dx . doi . org/10 . 7554/eLife . 10514 . 009PlasmidDescriptionSourcepLHN12-HispT7-His6-rpoD ( Panaghie et al . , 2000 ) pET15b-His-fliApT7-His6-fliAThis workpFW11tet-PR′_+19λPR′ promoter and native σ70-dependent pause element with a second σ70-dependent pause element located 19 bp downstream of the +1 transcription start site . The tR′ intrinsic terminator is positioned to terminate transcription ∼116 bp downstream of +1 . ( Deighan et al . , 2011 ) pFW11tet-mutPR′_+19Same as pFW11tet-PR′_+19 but with A+2 G/T+6 G mutations in the native σ70-dependent pause element . ( Deighan et al . , 2011 ) pFW11tet-Ptar_+19 ( pNUN175 ) Same as pFW11tet-PR′_+19 except that the promoter ( up to and including +1 ) has been replaced with the σ28-dependent Ptar promoter . This workpFW11tet-Ptar_mut+19 ( pNUN176 ) Same as pFW11tet-Ptar_+19 but with mutations in the pause element . This workpBR-fliApSG585-fliAThis workpSG585colE1 origin plasmid with lacUV5 upstream of multiple cloning siteThis workpNUN191pCDFlacMUT3-rpoDThis workpCDFlacMUT3pCDFlac with attenuated −35 element ( AATACA ) This workpCDFlacderivative of pCDF-1b into which the lacUV5 promoter has been inserted ( Montero-Diez et al . , 2013 ) Single colonies of E . coli strains bearing the appropriate plasmids were used to inoculate individual 5 ml aliquots of LB broth ( Miller ) ( 10 g tryptone , 5 g yeast extract , 10 g NaCl per liter ) ( EMD-Millipore , Billerica , MA ) containing antibiotics ( spectinomyin [50 μg/ml] and streptomycin [25 μg/ml] were used together to maintain vectors bearing the aadA1 [SmR] allele; carbenicillin [100 μg/ml]; chloramphenicol [25 μg/ml] ) in 18 × 150 mm glass culture tubes which were incubated , rolling , overnight at 37°C . Aliquots of these cultures were diluted 1:100 into 25 ml of LB containing antibiotics and IPTG ( 1 mM ) in 125 ml DeLong flasks with Morton-style closures ( Bellco Glass , Vineland , NJ ) , shaken at 37°C on an orbital platform shaker at 220–250 RPM . For the experiments shown in Figure 4 , cultures were grown as described above , except that cells were initially back-diluted into a volume of 75 ml of media containing antibiotics and IPTG , mixed , and then 25 ml aliquots were transferred into each of two 125 ml flasks . One aliquot was used for each harvest time-point . 50 pmol of LNA probe ( 5′ agCaaAttAacCc 3′ ) , where LNA bases are capitalized; Exiqon , Woburn , MA ) was incubated in a 25 μl volume with 5 μl γ-32P-ATP ( EasyTide; Perkin Elmer , Waltham , MA ) , 2 . 5 μl 10X T4 PNK buffer , 13 . 5 μl nuclease free water ( Life Technologies ) , and 2 μl T4 PNK ( NEB , Ipswich , MA ) at 37°C for 1 hr followed by 95°C for 10 min . Labeled probe was separated from unincorporated radiolabeled nucleotide using a size-exclusion spin column ( SigmaSpin; Sigma–Aldrich , St . Louis , MO ) . RNAs generated in vivo were detected by hybridization as described in ( Pall and Hamilton , 2008; Goldman et al . , 2009; Deighan et al . , 2011 ) using a 5′ radiolabeled LNA probe . RNA was loaded onto 0 . 4 mm thick 20% denaturing polyacrylamide slab gels cast and equilibrated in 50 mM MOPS ( pH 7 with NaOH ) , transferred to neutral nylon membrane ( Whatman Nytran N; GE Healthcare Life Sciences , Piscataway , NJ ) using a semi-dry electroblotting apparatus ( Biorad , Hercules , CA ) operating at 20V for 25 min using chilled 20 mM MOPS ( pH 7 with NaOH ) as conductive medium . RNA was crosslinked to the membrane using 157 mM N- ( 3-dimethylaminopropyl ) -N′-ethylcarbodiimide hydrochloride ( EDC ) ( Sigma–Aldrich ) in 0 . 97% 1-methylimidazole ( pH 8 ) ( Alfa Aesar , Ward Hill , MA ) ( as described in Pall and Hamilton , 2008 ) for 80 min at 55°C . Crosslinking solution was rinsed from the membrane by immersion in 20 mM MOPS ( pH 7 with NaOH ) at 25°C , the membrane was placed onto nylon hybridization mesh , the membrane-mesh stack was placed into a 70 × 150 mm hybridization bottle at 50°C and 50 ml of pre-hybridization solution ( 5× SSC , 5% SDS , 2× Denhardt's solution , 40 μg/ml sheared salmon sperm DNA solution [Life Technologies] , 20 mM Na2HPO4 [pH 7 . 2] in diethylpyrocarbonate ( DEPC ) treated water ) at 50°C was added . The hybridization bottle was rotated in a hybrization oven at 50°C for 30 min , the solution was decanted and replaced by a 50 ml portion of pre-warmed hybridization solution that had been thoroughly mixed with the entire volume of the radiolabeled LNA probe prepared above . The bottle was then returned to the 50°C oven for 16 hr . The membrane was washed twice for 10 min in non-stringent wash buffer ( 3× SSC , 5% SDS , 10× Denhardt's solution , 20 mM Na2HPO4 [pH 7 . 2] in DEPC treated water ) , twice for 30 min in non-stringent wash buffer , and once for 5 min in stringent wash buffer ( 1× SSC , 1% SDS , in DEPC treated water ) before it was blotted dry , wrapped in plastic film , and radiolabeled bands were visualized by storage phosphor screen ( GE Healthcare ) and phosphorimagery ( Storm 830 imager or Typhoon 9400 variable mode imager , GE Healthcare ) . All wash buffers were equilibrated to 55°C prior to use . Hybridization oven was operated at 50°C throughout . With the exception of Figure 4—figure supplement 1B , protein isolation for immunoblotting was performed as follows . 1 ml of cell suspensions was pelleted by centrifugation at 10 , 000 × g for 2 min at ambient temperature , supernatants were carefully removed by vacuum aspiration and pellets were immediately frozen on dry ice before being stored at −80°C . To extract soluble protein , cell pellets were thawed on ice for ∼30 s and resuspended by pipetting in lysis solution normalized to 50 μl per 1 ml of OD600 = 0 . 6 . Lysis solution consisted of 1 ml B-PER reagent ( Thermo Scientific Pierce , Rockland , IL ) , 1/4 protease inhibitor tablet ( Comlete-mini [EDTA-free]; Roche , Indianapolis , IN ) , 2 μl 0 . 5M EDTA ( pH 8 ) , 2 μl lysozyme ( 50 mg/ml ) , 120 μl TurboDNase ( Life Technologies ) , and 200 μl 10× TurboDNase buffer . Lysis mixture was incubated 10 min on ice . Lysates were centrifuged at 21 , 000 × g for 10 min at 4°C to pellet insoluble material . 40 μl of clarified supernatant was then mixed with an equal volume of 2× loading buffer prepared by mixing 500 μl 4× NuPAGE LDS sample buffer ( Life Technologies ) , 50 μl β-mercaptoethanol and 450 μl water . Samples were heated at 70°C for 2 min and centrifuged at 21 , 000 × g for 2 min at ambient temperature prior to electrophoresis . For the experiment of Figure 4—figure supplement 1B , total cellular protein was isolated as follows . Cell pellets , obtained and stored as described above , were resuspended directly into 50 μl per 1 ml of OD600 = 0 . 6 of 1× Laemmli SDS sample buffer ( pH 7 . 4 ) and heated 90°C for 5 min . Debris was pelleted by centrifugation at 21 , 000 × g for 5 min and the supernatants were transferred to fresh tubes and analyzed by gel electrophoresis . With the exception of Figure 4—figure supplement 1B , immunoblots were performed as follows . 10 μl of each soluble protein sample was loaded onto a 4–12% gradient NuPAGE Novex Bis-Tris precast gel ( Life Technologies ) and run in 1X NuPAGE MOPS SDS running buffer until the dye front exited the gel . The gel cassette was then opened and the gel was equilibrated into transfer buffer ( 192 mM glycine , 25 mM Tris , 10% methanol ) for 5–10 min . PVDF membrane ( Immobilon-FL; EMD-Millpore ) was wetted in 100% methanol and equilibrated into transfer buffer prior to transfer-stack assembly . Semi-dry electro transfer was performed using a Trans-Blot SD apparatus ( Bio-Rad ) operating at 10V for 1 hr . After transfer , membranes were placed into blocking solution ( 5% non-fat dry milk in 1× PBS ) and gently agitated at ambient temperature for 30 min . Blocking solution was decanted and replaced with 10 ml of a 1:5000 dilution of affinity purified mouse monoclonal antibody recognizing σ70 ( clone 2G10; Neoclone , Madison , WI ) or σ28 ( clone 1RF18; Neoclone ) in blocking solution and gently agitated for 1 hr as above . The primary antibody solution was decanted and the membrane washed quickly in 10 sequential portions of blocking solution containing 0 . 1% TWEEN-20 . Goat anti-mouse IRDye 680LT secondary antibody ( Li-Cor Biosciences , Lincoln , NE ) was diluted 1:20 , 000 into 20 ml of blocking solution containing 0 . 1% TWEEN-20 and 0 . 02% SDS and 10 ml was added to the membrane which was then incubated and washed as above except that the membrane was kept in the dark during incubation and several quick washes in 1× PBS were performed to remove residual milk prior to imaging . Data was acquired using an Odyssey Classic infra-red imager ( Li-Cor Biosciences ) . For the blot shown in Figure 4—figure supplement 1B , total cellular protein was electrophoresed and transferred as above except that nitrocellulose membrane ( Protran NC , GE Healthcare ) was used . Detection of protein was performed using a 1:20 , 000 dilution of Goat anti-Mouse HRP conjugated secondary antibody , ECL reagents ( SuperSignal West , Pierce ) and a ChemiDoc XRS + instrument ( Bio-Rad ) . Quantification was performed using ImageLab software . His-tagged σ70 and σ28 were purified from BL21 ( DE3 ) cells transformed with pLHN12-His and pET15b-His-fliA , respectively , as described previously ( Panaghie et al . , 2000 ) . E . coli core RNAP was purchased from Epicentre ( Madison , WI ) . Holoenzymes were formed by mixing core RNAP and a twofold molar excess of σ70 or a fivefold molar excess of σ28 in transcription buffer ( 20 mM Tris–HCl pH 8 . 0 , 0 . 1 mM EDTA , 100 mM K-acetate , 100 μg/ml BSA , 10 mM DTT , 5% glycerol , and 0 . 025% Tween-20 ) and incubating at 37°C for 10 min . Linear transcription templates were synthesized by PCR using plasmid DNAs ( pFW11tet-Ptar_+19 , pFW11tet-Ptar_mut+19 or pFW11tet-mutPR′_+19 ) as template and oligonucleotides that anneal to sequences ∼100 bp upstream of the +1 transcription start site ( 5′ CCTATAAAAATAGGCGTATCACGAG 3′ ) and ∼135 bp downstream of the transcription termination site ( 5′ CAGGGTTTTCCCAGTCACGACGTTG 3′ ) . 20 nM PCR template was mixed with 15 nM of the appropriate RNAP holoenzyme or RNAP core enzyme in transcription buffer containing 200 μM ATP , 200 μM GTP , 200 μM CTP , 25 μM UTP ( supplemented with 0 . 5 μCi/μL [α-32P]-UTP; Perkin Elmer ) , and 0 . 5 units/μl Murine RNase Inhibitor ( NEB ) for 5 min at 37°C to form open complexes . A single round of transcription was initiated by adding MgCl2 ( 4 mM final concentration ) and rifampicin ( 10 µg/ml final concentration ) , as described previously ( Grayhack et al . , 1985; Shankar et al . , 2007; Hollands et al . , 2012 ) . When present , excess σ70 was added together with the MgCl2 and rifampicin to a final concentration of 1 μM . Aliquots of the reaction were removed at 1 , 6 , and 18 min and mixed with 1 . 2 × stop buffer ( 600 mM Tris–HCl pH 8 . 0 , 12 mM EDTA , and 100 μg/mL Ambion Yeast RNA [Life Technologies] ) . Samples were extracted with acid phenol:chloroform and RNA transcripts were recovered by ethanol precipitation and resuspended in gel loading buffer ( 95% formamide , 18 mM EDTA , 0 . 025% SDS , 0 . 025% xylene cyanol , 0 . 025% bromophenol blue , 0 . 025% amaranth ) . Samples were heated at 95°C for 5 min , cooled to room temperature , and run on 12% TBE-Urea polyacrylamide gels ( UreaGel system; National Diagnostics , Atlanta , GA ) . Autoradiography of gels was performed using storage phosphor screens and a Typhoon 9400 variable mode imager ( GE Healthcare ) and quantified using ImageQuant software . | Proteins are made following instructions that are encoded by sections of DNA called genes . In the first step of protein production , an enzyme called RNA polymerase uses the gene as a template to make molecules of messenger ribonucleic acid ( mRNA ) . This process—known as transcription—starts when RNA polymerase binds to a site at the start of a gene . The enzyme then moves along the DNA , assembling the mRNA as it goes . This stage of transcription is known as elongation and continues until the RNA polymerase reaches the end of the gene . In bacteria , RNA polymerase needs a family of proteins called sigma factors to help it identify and bind to the start sites associated with the genes that will be transcribed . In the well studied bacterium known as E . coli , the primary sigma factor that is required for transcription initiation on most genes is called sigma 70 . Recent research has shown that sigma 70 also influences the activity of RNA polymerase during elongation . During this stage , the RNA polymerase and several other proteins interact to form a complex called the transcription elongation complex ( or TEC for short ) . However , it is not clear how sigma 70 gains access to this complex: does it simply remain with RNA polymerase after transcription starts , or is it freshly incorporated into the TEC during elongation ? Goldman , Nair et al . found that sigma 70 is able to incorporate into TECs during elongation and causes them to pause at specific sites in the gene . Sigma 70 can even incorporate into TECs on genes where transcription was initiated by a different sigma factor . These findings indicate that sigma 70 can directly influence the transcription of all genes , not just the genes with start sites that are recognized by this sigma factor . Goldman et al . also observed that in cells that were growing and dividing rapidly , the pauses that occurred due to sigma 70 associating with TECs were of shorter duration than those in cells that were growing slowly . This implies that the growth status of the cells modulates the pausing of RNA polymerase during transcription . In the future , it will be important to understand how much influence the primary sigma factor has on RNA polymerase during elongation in E . coli and other bacteria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | The primary σ factor in Escherichia coli can access the transcription elongation complex from solution in vivo |
Neurons in the hippocampus and adjacent brain areas show a large diversity in their tuning to location and head direction , and the underlying circuit mechanisms are not yet resolved . In particular , it is unclear why certain cell types are selective to one spatial variable , but invariant to another . For example , place cells are typically invariant to head direction . We propose that all observed spatial tuning patterns – in both their selectivity and their invariance – arise from the same mechanism: Excitatory and inhibitory synaptic plasticity driven by the spatial tuning statistics of synaptic inputs . Using simulations and a mathematical analysis , we show that combined excitatory and inhibitory plasticity can lead to localized , grid-like or invariant activity . Combinations of different input statistics along different spatial dimensions reproduce all major spatial tuning patterns observed in rodents . Our proposed model is robust to changes in parameters , develops patterns on behavioral timescales and makes distinctive experimental predictions .
Neurons in the hippocampus and the adjacent regions exhibit a broad variety of spatial activation patterns that are tuned to position , head direction or both . Common observations in these spatial dimensions are localized , bell-shaped tuning curves ( O'Keefe , 1976; Taube et al . , 1990 ) , periodically repeating activity ( Fyhn et al . , 2004; Hafting et al . , 2005 ) and invariances ( Muller et al . , 1994; Burgess et al . , 2005 ) , as well as combinations of these along different spatial dimensions ( Sargolini et al . , 2006a; Krupic et al . , 2012 ) . For example , head direction cells are often invariant to location ( Burgess et al . , 2005 ) , and place cells are commonly invariant to head direction ( Muller et al . , 1994 ) . The cellular and network mechanisms that give rise to each of these firing patterns are subject to extensive experimental and theoretical research . Several computational models have been suggested to explain the emergence of grid cells ( Fuhs and Touretzky , 2006; McNaughton et al . , 2006; Franzius et al . , 2007a; Burak and Fiete , 2009; Couey et al . , 2013; Burgess et al . , 2007; Kropff and Treves , 2008; Bush and Burgess , 2014; Castro and Aguiar , 2014; Dordek et al . , 2016; Stepanyuk , 2015; Giocomo et al . , 2011; Zilli , 2012; D'Albis and Kempter , 2017; Monsalve-Mercado and Leibold , 2017 ) , place cells ( Tsodyks and Sejnowski , 1995; Battaglia and Treves , 1998; Arleo and Gerstner , 2000; Solstad et al . , 2006; Franzius et al . , 2007b; Burgess and O'Keefe , 2011; Franzius et al . , 2007a ) and head direction cells ( McNaughton et al . , 1991; Redish et al . , 1996; Zhang , 1996; Franzius et al . , 2007a ) . Most of these models are designed to explain the spatial selectivity of one particular cell type and do not consider invariances along other dimensions , although the formation of invariant representations is a non-trivial problem ( DiCarlo and Cox , 2007 ) . In view of the variety of spatial tuning patterns , the question arises of whether differences in tuning of different cells in different areas reflect differences in microcircuit connectivity , single cell properties or plasticity rules , or whether there is a unifying principle . In this paper we suggest that both the observed spatial selectivities and invariances can be explained by a common mechanism – interacting excitatory and inhibitory synaptic plasticity – and that the observed differences in the response profiles of grid , place and head direction cells result from differences in the spatial tuning of excitatory and inhibitory synaptic afferents . Here , we explore this hypothesis in a computational model of a feedforward network of rate-based neurons . Simulations as well as a mathematical analysis indicate that the model reproduces the large variety of response patterns of neurons in the hippocampal formation and adjacent areas and can be used to make predictions for the input statistics of each cell type .
We first simulate a rat that explores a linear track ( Figure 1 ) . The spatial tuning of each input neuron is stable in time and depends smoothly on the location of the animal , but is otherwise random ( e . g . Figure 1a ) . As a measure of smoothness , we use the spatial autocorrelation length . In the following , this is the central parameter of the input statistics , which is chosen separately for excitation and inhibition . In short , we assume that temporally stable spatial information is presynaptically present but we have minimal requirements on its format , aside from the spatial autocorrelation length . At the beginning of each simulation , all synaptic weights are random . As the animal explores the track , the excitatory and inhibitory weights change in response to pre- and postsynaptic activity , and the output cell gradually develops a spatial activity pattern . We find that this pattern is primarily determined by whether the excitatory or inhibitory inputs are smoother in space . If the inhibitory tuning is smoother than the excitatory tuning ( Figure 1b ) , the output neuron develops equidistant firing fields , reminiscent of grid cells on a linear track ( Hafting et al . , 2008 ) . If instead the excitatory tuning is smoother , the output neuron fires close to the target rate of 1 Hz everywhere ( Figure 1c ) ; it develops a spatial invariance . For spatially untuned inhibitory afferents ( Grienberger et al . , 2017 ) , the output neuron develops a single firing field , reminiscent of a one-dimensional place cell ( Figure 1d ) ; ( cf . Clopath et al . , 2016 ) . The emergence of these firing patterns can be best explained in the simplified scenario of place field-like input tuning ( Figure 1e , f ) . The spatial smoothness is then given by the size of the place fields . Let us assume that the output neuron fires at the target rate everywhere ( see Materials and methods ) . From this homogeneous state , a small potentiation of one excitatory weight leads to an increased firing rate of the output neuron at the location of the associated place field ( highlighted red curve in Figure 1e ) . To bring the output neuron back to the target rate , the inhibitory learning rule increases the synaptic weight of inhibitory inputs that are tuned to the same location ( highlighted blue curve in Figure 1e ) . If these inhibitory inputs have smaller place fields than the excitatory inputs ( Figure 1c ) , this restores the target rate everywhere ( Vogels et al . , 2011 ) . Hence , inhibitory plasticity can stabilize spatial invariance if the inhibitory inputs are sufficiently precise ( i . e . not too smooth ) in space . In contrast , if the spatial tuning of the inhibitory inputs is smoother than that of the excitatory inputs , the target firing rate cannot be restored everywhere . Instead , the compensatory potentiation of inhibitory weights increases the inhibition in a spatial region at least the size of the inhibitory place fields . This leads to a corona of inhibition , in which the output neuron cannot fire ( Figure 1e , blue region ) . Outside of this inhibitory surround the output neuron can fire again and the next firing field develops . Iterated , this results in a periodic arrangement of firing fields ( Figure 1f and Figure 7b for a depiction of the input currents ) . Spatially untuned inhibition corresponds to a large inhibitory corona that exceeds the length of the linear track , so that only a single place field remains . From a different perspective , spatially untuned input can also be understood as a limit case of vanishing spatial variation in the firing rate rather than a limit of infinite smoothness . Consistent with this view , a development of grid patterns or invariance requires a sufficiently strong spatial modulation of the inhibitory inputs ( Materials and methods ) . The argument of the preceding paragraph can be extended to the scenario where input is irregularly modulated by space . For non-localized input tuning ( Figure 1b , c , d ) , any weight change that increases synaptic input in one location will also increase it in a surround that is given by the smoothness of the input tuning ( see Materials and methods for a mathematical analysis ) . In the simulations , the randomness manifests itself in occasional defects in the emerging firing pattern ( Figure 1h , bottom , and Figure 1—figure supplement 1 ) . The above reasoning suggests that the width of individual firing fields is determined by the smoothness of the excitatory input tuning , while the distance between grid fields , that is , the grid spacing , is set by the smoothness of the inhibitory input tuning . Indeed , both simulations and a mathematical analysis ( Materials and methods ) confirm that the grid spacing scales linearly with the inhibitory smoothness in a large range , both for localized ( Figure 1g ) and non-localized input tuning ( Figure 1h ) . The analysis also reveals a weak logarithmic dependence of the grid spacing on the ratio of the learning rates , the mean firing rates and the number of afferents of the excitatory and inhibitory population ( Equation 78 and Figure 8b ) . In summary , the interaction of excitatory and inhibitory plasticity can lead to spatial invariance , spatially periodic activity patterns or single place fields depending on the spatial statistics of the excitatory and inhibitory input . When a rat navigates in a two-dimensional arena , the spatial firing maps of grid cells in the medial entorhinal cortex ( mEC ) show pronounced hexagonal symmetry ( Hafting et al . , 2005; Fyhn et al . , 2004 ) with different grid spacings and spatial phases . To study whether a hexagonal firing pattern can emerge from interacting excitatory and inhibitory plasticity , we simulate a rat in a quadratic arena . The rat explores the arena for 10 hr , following trajectories extracted from behavioral data ( Sargolini et al . , 2006b ) ; Materials and methods . To investigate the role of the input statistics , we consider three different classes of input tuning: ( i ) place cell-like input ( Figure 2a ) , ( ii ) sparse non-localized input , in which the tuning of each input neuron is given by the sum of 100 randomly located place fields ( Figure 2b and ( iii ) dense non-localized input , in which the tuning of each input is a random function with fixed spatial smoothness ( Figure 2c ) . For all input classes , the spatial tuning of the inhibitory inputs is smoother than that of the excitatory inputs . Initially , all synaptic weights are random and the activity of the output neuron shows no spatial symmetry . While the rat forages through the environment , the output cell develops a periodic firing pattern for all three input classes , reminiscent of grid cells in the mEC ( Fyhn et al . , 2004; Hafting et al . , 2005 ) and typically with the same hexagonal symmetry . This hexagonal arrangement is again a result of smoother inhibitory input tuning , which generates a spherical inhibitory corona around each firing field ( compare Figure 1e ) . These center-surround fields are arranged in a hexagonal pattern – the closest packing of spheres in two dimensions; ( cf . Turing , 1952 ) . We find that the spacing of this pattern is determined by the inhibitory smoothness . The similarity between cells in terms of orientation and phase of the grid depends – in decreasing order – on whether they receive the same inputs , on the trajectories on which the tuning was learned and on the initial synaptic weights ( Figure 2—figure supplement 1 ) . Two grid cells can thus have different phase and orientation , even if they share a large fraction or all of their inputs . For the linear track , the randomness of the non-localized inputs leads to defects in the periodicity of the grid pattern . In two dimensions , we find that the randomness leads to distortions of the hexagonal grid . To quantify this effect , we simulated 500 random trials for each of the three input scenarios and plotted the grid score histogram ( Appendix 1 ) before and after 10 hr of spatial exploration ( Figure 2d , e , f ) . Different trials have different trajectories , different initial synaptic weights and different random locations of the input place fields ( for sparse input ) or different random input functions ( for dense input ) . For place cell-like input , most of the output cells develop a positive grid score during 10 hr of spatial exploration ( 33% before to 86% after learning , Figure 2d ) . Even for low grid scores , the firing rate maps look grid-like after learning but exhibit a distorted symmetry ( Figure 2d ) . For sparse non-localized input , the fraction of output cells with a positive grid score increases from 35% to 87% and for dense non-localized input from 16% to 68% within 10 hr of spatial exploration ( Figure 2e , f ) . The excitatory and inhibitory inputs are not required to have the same tuning statistics . Grid patterns also emerge when excitation is localized and inhibition is non-localized ( Figure 2—figure supplement 2 ) . In summary , the interaction of excitatory and inhibitory plasticity leads to grid-like firing patterns in the output neuron for all three input scenarios . The grids are typically less distorted for sparser input ( Figure 2g ) . In unfamiliar environments , neurons in the mEC exhibit grid-like firing patterns within minutes ( Hafting et al . , 2005 ) . Moreover , grid cells react quickly to changes in the environment ( Fyhn et al . , 2007; Savelli et al . , 2008; Barry et al . , 2012 ) . These observations challenge models for grid cells that require gradual synaptic changes during spatial exploration . In principle , the time scale of plasticity-based models can be augmented arbitrarily by increasing the synaptic learning rates . For stable patterns to emerge , however , significant weight changes must occur only after the animal has visited most of the environment . To explore the edge of this trade-off between speed and stability , we increased the learning rates to a point where the grids are still stable but where further increase would reduce the stability ( Figure 3—figure supplement 1 ) . For place cell-like input , periodic patterns can be discerned within 10 min of spatial exploration , starting with random initial weights ( Figure 3a , b ) . The pattern further emphasizes over time and remains stable for many hours ( Figure 3c and Figure 3—figure supplement 2 ) . To investigate the robustness of this phenomenon , we ran 500 realizations with different trajectories , initial synaptic weights and locations of input place fields . In all simulations , a periodic pattern emerged within the first 30 min , and a majority of patterns exhibited hexagonal symmetry after 3 hr ( increasing from 33% to 81% , Figure 3c , d ) . For non-localized input , the emergence of the final grids typically takes longer , but the first grid fields are also observed within minutes and are still present in the final grid , as observed in experiments ( Hafting et al . , 2005 ) ; ( Figure 3—figure supplement 3 ) . Above , we modeled the exploration of a previously unknown room by assuming the initial synaptic weights to be randomly distributed . If the rat had previous exposure to the room or to a similar room , a structure might already have formed in some of the synaptic weights . This structure could aid the development of the grid in similar rooms or hinder it in a novel room . To study this , we simulate a network that first learns the synaptic weights in one room . We then introduce a graded modification of the room by remapping the firing fields of a fraction of input neurons to random locations . We find that the output firing pattern is robust to such perturbations , even if more than half of the inputs are remapped ( Figure 3—figure supplement 2 ) . If all inputs are changed , corresponding to a novel room , a grid pattern is learned anew . The strong initial pattern in the weights does not hinder this development ( Figure 3—figure supplement 2 ) . Recently , Wernle et al . , 2018 discovered that in an arena separated by a wall , single grid cells form two independent grid patterns — one on each side of the wall — that coalesce once the wall is removed . They find that grid fields close to the partition wall move to establish a more coherent pattern . In contrast , fields far away from the partition wall do not change their locations . Rosay et al . reproduced this experimental finding by simulating grid fields as interacting particles ( Rosay et al . , in preparation ) . They also demonstrated how it could be reproduced by a feedforward model for grid cells based on firing rate adaptation ( Rosay et al . , in preparation; Kropff and Treves , 2008 ) . Inspired by these experiments and simulations , we simulate a rat that first explores one half of a quadratic arena and then the other half , for 2 . 5 hr each ( Figure 4a ) . A grid pattern emerges in each compartment ( Figure 4b , c ) . We then remove the partition wall and the rat explores the entire arena for another 5 hr ( Figure 4a ) . As observed experimentally , grid fields close to the former partition line rearrange to make the two grids more coherent and grid fields far away from the partition line basically stay where they were ( Figure 4d ) . In summary , periodic patterns emerge rapidly in our model and the associated time scale is limited primarily by how quickly the animal visits its surroundings , that is , by the same time scale that limits the experimental recognition of the grids . In addition to grids , the mEC and adjacent brain areas exhibit a plethora of other spatial activity patterns including spatially invariant ( Burgess et al . , 2005 ) , band-like ( periodic along one direction and invariant along the other ) ( Krupic et al . , 2012 ) , and spatially periodic but non-hexagonal patterns ( Krupic et al . , 2012; Hardcastle et al . , 2017; Diehl et al . , 2017 ) . Note that it is currently debated whether or not some of the observed spatially periodic but non-hexagonal firing patterns are artifacts of poorly isolated single cell data in multi-electrode recordings ( Navratilova et al . , 2016; Krupic et al . , 2015b ) . In contrast to spatially periodic tuning , place cells in the hippocampus proper are typically only tuned to a single or few locations in a given environment ( O'Keefe and Dostrovsky , 1971; Moser et al . , 2008; Leutgeb et al . , 2005 ) . If the animal traversed the environment along a straight line , all of these cells would be classified as periodic , localized or invariant ( Figure 1 ) , although the classification could vary depending on the direction of the line . Based on this observation , we hypothesized that all of these patterns could be the result of an input autocorrelation structure that differs along different spatial directions . We first verified that also in a two-dimensional arena , place cells emerge from a very smooth inhibitory input tuning ( Figure 5a , b ) . The emergence of place cells is independent of the exact shape of the excitatory input . Non-localized inputs ( Figure 5a ) lead to similar results as those from grid cell-like inputs of different orientation and grid spacing ( Figure 5b , Methods and materials ) ; for other models for the emergence of place cells from grid cells see ( Solstad et al . , 2006; Franzius et al . , 2007b; Rolls et al . , 2006; Molter and Yamaguchi , 2008; Ujfalussy et al . , 2009; Savelli and Knierim , 2010 ) . Next we verified that also in two dimensions , spatial invariance results when excitation is broader than inhibition ( Figure 5c ) . We then varied the smoothness of the inhibitory inputs independently along two spatial directions . If the spatial tuning of inhibitory inputs is smoother than the tuning of the excitatory inputs along one dimension but less smooth along the other , the output neuron develops band cell-like firing patterns ( Figure 5d ) . If inhibitory input is smoother than excitatory input , but not isotropic , the output cell develops stretched grids with different spacing along two axes ( Figure 5e ) . For these anisotropic cases , stretched hexagonal grids and rectangular arrangements of firing fields appear similarly favorable ( compare Figure 5e , second row and column ) . A hexagonal arrangement is favored by a dense packing of inhibitory coronas , whereas a rectangular arrangement would maximize the proximity of the excitatory centers , given the inhibitory corona ( Figure 5—figure supplement 1 ) . In summary , the relative spatial smoothness of inhibitory and excitatory input determines the symmetry of the spatial firing pattern of the output neuron . The requirements for the input tuning that support invariance , periodicity and localization apply individually to each spatial dimension , opening up a combinatorial variety of spatial tuning patterns . Many cells in and around the hippocampus are tuned to the head direction of the animal ( Taube et al . , 1990; Taube , 1995; Chen et al . , 1994 ) . These head direction cells are typically tuned to a single head direction , just like place cells are typically tuned to a single location . Moreover , head direction cells are often invariant to location ( Burgess et al . , 2005 ) , just like place cells are commonly invariant to head direction ( Muller et al . , 1994 ) . There are also cell types with conjoined spatial and head direction tuning . Conjunctive cells in the mEC fire like grid cells in space , but only in a particular head direction ( Sargolini et al . , 2006a ) , and many place cells in the hippocampus of crawling bats also exhibit head direction tuning ( Rubin et al . , 2014 ) . To investigate whether these tuning properties could also result in our model , we simulated a rat that moves in a square box , whose head direction is constrained by the direction of motion ( Appendix 1 ) . Each input neuron is tuned to both space and head direction ( see Figure 6 for localized and Figure 6—figure supplement 1 for non-localized input ) . In line with the previous observations , we find that the spatial tuning of the output neuron is determined by the relative spatial smoothness of the excitatory and inhibitory inputs , and the head direction tuning of the output neuron is determined by the relative smoothness of the head direction tuning of the inputs from the two populations . If the head direction tuning of excitatory input neurons is smoother than that of inhibitory input neurons , the output neuron becomes invariant to head direction ( Figure 6a ) . If instead only the excitatory input is tuned to head direction , the output neuron develops a single activity bump at a particular head direction ( Figure 6b , c ) . The concurrent spatial tuning of the inhibitory input neurons determines the spatial tuning of the output neuron . For spatially smooth inhibitory input , the output neuron develops a hexagonal firing pattern ( Figure 6a , b ) , and for less smooth inhibitory input the firing of the output neuron is invariant to the location of the animal ( Figure 6c ) . In summary , the relative smoothness of inhibitory and excitatory input neurons in space and in head direction determines whether the output cell fires like a pure grid cell , a conjunctive cell or a pure head direction cell ( Figure 6d ) . We find that the overall head direction tuning of conjunctive cells is broader than that of individual grid fields ( Figure 6e ) . This results from variations in the preferred head direction of different grid fields . Typically , however , these variations remain small enough to preserve an overall head direction tuning of the cell , because individual grid fields tend to align their head direction tuning ( compare with Figure 5—figure supplement 1 , but in three dimensions ) . Whether or not a narrower head direction of individual grid fields or a different preferred direction for different grid fields is present also in rodents is not resolved ( Figure 6—figure supplement 2 ) .
The origin of synaptic input to spatially tuned cells is not fully resolved ( van Strien et al . , 2009 ) . Given that our model is robust to the precise properties of the input , it is consistent with input from higher sensory areas ( Tanaka , 1996; Quiroga et al . , 2005 ) that could inherit spatial tuning from their sensory tuning in a stable environment ( Arleo and Gerstner , 2000; Franzius et al . , 2007a ) . This is in line with the observation that grid cells lose their firing profiles in darkness ( Chen et al . , 2016; Pérez-Escobar et al . , 2016 ) and that the hexagonal pattern rotates when a visual cue card is rotated ( Pérez-Escobar et al . , 2016 ) . The input could also stem from within the hippocampal formation , where spatial tuning has been observed in both excitatory ( O'Keefe , 1976 ) and inhibitory ( Marshall et al . , 2002; Wilent and Nitz , 2007; Hangya et al . , 2010 ) neurons . For example , the notion that mEC neurons receive input from hippocampal place cells is supported by several studies: Place cells in the hippocampus emerge earlier during development than grid cells in the mEC ( Langston et al . , 2010; Wills et al . , 2010 ) , grid cells lose their tuning pattern when the hippocampus is deactivated ( Bonnevie et al . , 2013 ) and both the firing fields of place cells and the spacing and field size of grid cells increase along the dorso-ventral axis ( Jung et al . , 1994; Brun et al . , 2008b; Stensola et al . , 2012 ) . Moreover , entorhinal stellate cells , which often exhibit grid-like firing patterns , receive a large fraction of their input from the hippocampal CA2 region ( Rowland et al . , 2013 ) , where many cells are tuned to the location of the animal ( Martig and Mizumori , 2011 ) . Inhibition is usually thought to arise from local interneurons – but see ( Melzer et al . , 2012 ) – suggesting that spatially tuned inhibitory input to mEC neurons originates from the entorhinal cortex itself . Interneurons in mEC display spatial tuning ( Buetfering et al . , 2014; Savelli et al . , 2008; Frank et al . , 2001 ) that could be inherited from hippocampal place cells , other grid cells ( Couey et al . , 2013; Pastoll et al . , 2013; Winterer et al . , 2017 ) or from entorhinal cells with nongrid spatial tuning ( Diehl et al . , 2017; Hardcastle et al . , 2017 ) . The broader spatial tuning required for the emergence of spatial selectivity could be established , for example by pooling over cells with similar tuning or through a non-linear input-output transformation in the inhibitory circuitry . If inhibitory input is indeed local , the increase in grid spacing along the dorso-ventral axis ( Brun et al . , 2008b ) suggests that the tuning of inhibitory interneurons gets smoother along this axis . For smoother tuning functions , fewer neurons are needed to cover the whole environment , in accordance with the decrease in interneuron density along the dorso-ventral axis ( Beed et al . , 2013 ) . The excitatory input to hippocampal place cells could originate from grid cells in entorhinal cortex ( Figure 5b ) , which is supported by anatomical ( van Strien et al . , 2009 ) and lesion studies ( Brun et al . , 2008a ) . The required untuned inhibition could arrive from interneurons in the hippocampus proper that often show very weak spatial tuning ( Marshall et al . , 2002 ) . In addition to grid cell input , place cells are also thought to receive inputs from other cell types , such as border cells ( Muessig et al . , 2015 ) and other brain regions such as the medial septum ( Wang et al . , 2015 ) . The observed spatial tuning patterns have also been explained by other models . In continuous attractor networks ( CAN ) , each cell type could emerge from a specific recurrent connectivity pattern , combined with a mechanism that translates the motion of the animal into shifts of neural activity on an attractor . How the required connectivity patterns – which lie at the core of any CAN model – could emerge is subject to debate ( Widloski and Fiete , 2014 ) . Our model is qualitatively different in that it does not rely on attractor dynamics in a recurrent neural network , but on experience-dependent plasticity of spatially modulated afferents to an individual output neuron ( Mehta et al . , 2000 ) . A measurable distinction of our model from CAN models is its response to a rapid global reduction of inhibition . While a modification of inhibition typically changes the grid spacing in CAN models of grid cells ( Couey et al . , 2013; Widloski and Fiete , 2015 ) , the grid field locations generally remain untouched in our model . The grid fields merely change in size , until inhibition is recovered by inhibitory plasticity ( Figure 7a ) . This can be understood by the colocalization of the grid fields and the peaks in the excitatory membrane current ( Figure 7b , c ) . A reduction of inhibition leads to an increased protrusion of these excitatory peaks and thus to wider firing fields . Grid patterns in mEC are temporally stable in spite of dopaminergic modulations of GABAergic transmission ( Cilz et al . , 2014 ) and the spacing of mEC grid cells remains constant during the silencing of inhibitory interneurons ( Miao et al . , 2017 ) . Both observations are in line with our model . Moreover , we found that for localized input tuning , the inhibitory membrane current typically also peaks at the locations of the grid fields . This co-tuning breaks down for non-localized input ( Figure 7b ) . In contrast , CAN models predict that the inhibitory membrane current has the same periodicity as the grid ( Schmidt-Hieber and Häusser , 2013 ) , but possibly phase shifted . The grid patterns of topologically nearby grid cells in the mEC typically have the same orientation and spacing but different phases ( Hafting et al . , 2005 ) . Moreover , the coupling between anatomically nearby grid cells – for example their difference in spatial phase – is more stable to changes of the environment than the firing pattern of individual grid cells ( Yoon et al . , 2013 ) . These properties are immanent to CAN models . In contrast , single cell models ( Burgess et al . , 2007; Kropff and Treves , 2008; Castro and Aguiar , 2014; Stepanyuk , 2015; Dordek et al . , 2016; D'Albis and Kempter , 2017; Monsalve-Mercado and Leibold , 2017 ) require additional mechanisms to develop a coordination of neighboring grid cells . The challenge for any mechanism is to correlate the grid orientations , but leave the grid phases uncorrelated . The most obvious candidate , recurrent connections among different grid cells ( Si et al . , 2012 ) , requires an intricate combination of mechanisms to perform this balancing act . We assume that an appropriate recurrent connectivity would not be simpler in our model . CAN models predict that all grid fields in a conjunctive ( grid x head direction ) cell have the same head direction tuning , whereas our model predicts that there could be differences between different grid fields ( Figure 6e ) . Our preliminary analysis suggests that an in-depth evaluation would require data for central grid fields without trajectory biases ( Figure 6—figure supplement 2 ) , which are at present not publicly available . In addition , CAN models require that conjunctive ( grid x head direction ) cells are positively modulated by running speed . Such modulation has been observed in experiments ( Kropff et al . , 2015 ) . In our model , we could introduce a running speed dependence , for example as a global modulation of the input signals . We expect that in this case , the output neuron would inherit speed tuning from the input but would otherwise develop similar spatial tuning patterns . A recent analysis has shown that periodic firing of entorhinal cells in rats that move on a linear track can be assessed as slices through a hexagonal grid ( Yoon et al . , 2016 ) , which arises naturally in a two-dimensional CAN model . In our model , we would obtain slices through a hexagonal grid if the rat learns the output pattern in two dimensions and afterwards is constrained to move on a linear track that is part of the same arena . If the rat learns the firing pattern on the linear track from scratch , the firing fields would be periodic . Models that learn grid cells from spatially tuned input do not have to assume a preexisting connectivity pattern or specific mechanisms for path integration ( Burgess et al . , 2007 ) , but are challenged by the fast emergence of hexagonal firing patterns in unfamiliar environments ( Hafting et al . , 2005 ) . Most plasticity-based models require slow learning , such that the animal explores the whole arena before significant synaptic changes occur . Therefore , grid patterns typically emerge slower than experimentally observed ( Dordek et al . , 2016 ) . This delay is particularly pronounced in models that require an extensive exploration of both space and movement direction ( Kropff and Treves , 2008; Franzius et al . , 2007a; D'Albis and Kempter , 2017 ) . In contrast to these models , which give center stage to the temporal statistics of the animal’s movement , our approach relies purely on the spatial statistics of the input and is hence insensitive to running speed . For the mechanism we suggested , the self-organization was very robust and allowed rapid pattern formation on short time scales , similar to those observed in rodents ( Figure 3 ) . This speed could be further increased by accelerated reactivation of previous experiences during periods of rest ( Lee and Wilson , 2002 ) . By this means , the exploration time and the time it takes to activate all input patterns could be decoupled , leading to a much faster emergence of grid cells in all trajectory-independent models with associative learning . Other models that explain the emergence of grid patterns from place cell input through synaptic depression and potentiation also develop grid cells in realistic times ( Castro and Aguiar , 2014; Stepanyuk , 2015; Monsalve-Mercado and Leibold , 2017 ) . These models differ from ours in that they do not require inhibition , but instead specific forms of rate-dependent synaptic depression and potentiation that change the synaptic weights such that place cell-like input leads to grid cell-like output . How these models generalize to potentially non-localized input is yet to be shown . Learning the required connectivity in CAN models can take a long time ( Widloski and Fiete , 2014 ) . However , as soon as the required connectivity and translation mechanism is established , a grid pattern would be observed immediately , even in a novel room . For different rooms this pattern could have different phases and orientations , but similar grid spacing ( Fyhn et al . , 2007 ) . Similarly , we found that room switches in our model lead to grid patterns of the same grid spacing but different phases and orientations . The pattern emerges rapidly , but is not instantaneously present ( Figure 3—figure supplement 2 ) . It would be interesting to study whether rotation of a fraction of the input would lead to a bimodal distribution of grid rotations: No rotation and co-rotation with the rotated input , as recently observed in experiments where distal cues were rotated but proximal cues stayed fixed ( Savelli et al . , 2017 ) . Recently , it was discovered that in an arena separated by a wall , single grid cells form two independent grid patterns – one on each side – that coalesce once the wall is removed ( Wernle et al . , 2018; Rosay et al . , in preparation ) . This coalescence is local , that is , grid fields close to the partition wall readjust , whereas grid fields far away do not change their locations . Feedforward models like ours can explain such a local rearrangement ( Figure 4; Rosay et al . , in preparation ) . Experiments show that the pattern and the orientation of grid cells is influenced by the geometry of the environment . In a quadratic arena , the orientation of grid cells tends to align – with a small offset – to one of the box axes ( Stensola et al . , 2015 ) . In trapezoidal arenas , the hexagonality of grids is distorted ( Krupic et al . , 2015a ) . We considered quadratic and circular arenas with rat trajectories from behavioral experiments and found that the boundaries also distort the grid pattern in our simulations , particularly for localized inputs ( Figure 2—figure supplement 3 ) . In trapezoidal geometries , we expect this to lead to non-hexagonal grids . However , we did not observe a pronounced alignment to quadratic boundaries if the input place fields were randomly located ( Figure 2—figure supplement 3 ) . We found that interacting excitatory and inhibitory plasticity serves as a simple and robust mechanism for rapid self-organization of stable and symmetric patterns from spatially modulated feedforward input . The suggested mechanism ports the robust pattern formation of attractor models from the neural to the spatial domain and increases the speed of self-organization of plasticity-based mechanisms to time scales on which the spatial tuning of neurons is typically measured . It will be interesting to explore how recurrent connections between output cells can help to understand the role of local inhibitory ( Couey et al . , 2013; Pastoll et al . , 2013 ) and excitatory connections ( Winterer et al . , 2017 ) and the presence or absence of topographic arrangements of spatially tuned cells ( O'Keefe et al . , 1998; Stensola et al . , 2012; Giocomo et al . , 2014 ) . We illustrated the properties and requirements of the model in the realm of spatial representations . As invariance and selectivity are ubiquitous properties of receptive fields in the brain , the interaction of excitatory and inhibitory synaptic plasticity could also be essential to form stable representations from sensory input in other brain areas ( Constantinescu et al . , 2016; Clopath et al . , 2016 ) .
The code for reproducing the essential findings of this article is available at https://github . com/sim-web/spatial_patterns ( Weber , 2018 ) under the GNU General Public License v3 . 0 . A copy is archived at https://github . com/elifesciences-publications/spatial_patterns . We study a feedforward network where a single output neuron receives synaptic input from NE excitatory and NI inhibitory neurons ( Figure 1a ) with synaptic weight vectors wE ∈ RNE , wI ∈ RNI and spatially tuned input rates rE ( x ) ∈RNE , rI ( x ) ∈RNI , respectively . Here x∈Rdimensions denotes the location and later also the head direction of the animal . For simplicity and to allow a mathematical analysis we use a rate-based description for all neurons . The firing rate of the output neuron is given by the rectified sum of weighted excitatory and inhibitory inputs: ( 1 ) rout ( x ( t ) ) =[∑i=1NEwiE ( t ) riE ( x ( t ) ) −∑j=1NIwjI ( t ) rjI ( x ( t ) ) ]+ , where [⋅]+ denotes a rectification that sets negative firing rates to zero . To comply with the notion of excitation and inhibition , all weights are constrained to be positive . In most simulations we use NE=4NI . Simulation parameters are shown in Tables 1–3 for the main figures and in Tables 4–6 for the supplementary figures . In each unit time step ( Δt=1 ) , the excitatory weights are updated according to a Hebbian rule: ( 2 ) ΔwE=ηErE ( x ) rout ( x ) ( and normalization ) . The excitatory learning rate ηE is a constant that we chose individually for each simulation . To avoid unbounded weight growth , we use a quadratic multiplicative normalization , that is , we keep the sum of the squared weights of the excitatory population ∑i=1NE ( wiE ) 2 constant at its initial value , by rescaling the weights after each unit time step . However , synaptic weight normalization is not a necessary ingredient for the emergence of firing patterns ( Figure 2—figure supplement 4 ) . We model inhibitory synaptic plasticity using a previously suggested learning rule ( Vogels et al . , 2011 ) : ( 3 ) ΔwI=ηIrI ( x ) ( rout ( x ) −ρ0 ) , with inhibitory learning rate ηI and target rate ρ0 = 1 Hz . Negative inhibitory weights are set to zero . In the linear track model ( one dimension , Figures 1 and 7 ) , we create artificial run-and-tumble trajectories x ( t ) constrained on a line of length L with constant velocity v = 1 cm per unit time step and persistence length L/2 ( Appendix 1 ) . In the open arena model ( two dimensions , Figures 2 , 3 , 5 and 7 ) , we use trajectories x ( t ) from behavioral data ( Sargolini et al . , 2006b ) of a rat that moved in a 1 m × 1 m quadratic enclosure ( Appendix 1 ) . In the simulations with a separation wall ( Figure 4 ) , we create trajectories as a two-dimensional persistent random walk ( Appendix 1 ) . In the model for neurons with head direction tuning ( three dimensions , Figure 6 ) , we use the same behavioral trajectories as in two dimensions and model the head direction as noisily aligned to the direction of motion ( Appendix 1 ) . The firing rates of excitatory and inhibitory synaptic inputs riE , rjI are tuned to the location 𝐱 of the animal . In the following , we use x and y for the first and second spatial dimension and z for the head direction . For place field-like input , we use Gaussian tuning functions with standard deviation σE , σI for the excitatory and inhibitory population , respectively . In Figure 5 the standard deviation is chosen independently along the x and y direction . The centers of the Gaussians are drawn randomly from a distorted lattice ( Figure 2—figure supplement 5 ) . This way we ensure random but spatially dense tuning . The lattice contains locations outside the box to reduce boundary effects . For sparse non-localized input with NPf fields per neuron of population P , we first create NPf distorted lattices , each with NP locations . We then assign NPf of the resulting NPfNP locations at random and without replacement to each input neuron ( see also Appendix 1 ) . For dense non-localized input , we convolve Gaussians with white noise and increase the resulting signal to noise ratio by setting the minimum to zero and the mean to 0 . 5 ( Appendix 1 ) . The Gaussian convolution kernels have different standard deviations for different populations . For each input neuron we use a different realization of white noise . This results in arbitrary tuning functions of the same autocorrelation length as the – potentially asymmetric – Gaussian convolution kernel . For grid cell-like input , we place Gaussians of standard deviation σE on the nodes of perfect hexagonal grids whose spacing and orientation is variable . In Figure 5b we draw the grid spacing of each input from a normal distribution of mean 6σE and standard deviation σE/6 . The grid orientation was drawn from a uniform distribution between -30 and 30 degrees . For input with combined spatial and head direction tuning , we use the Gaussian tuning curves described above for the spatial tuning and von Mises distributions along the head direction dimension ( Appendix 1 ) . For all input tunings , the standard deviation of the firing rate is of the same order of magnitude as the mean firing rate ( Appendix 1 ) . We specify a mean for the initial excitatory and inhibitory weights , respectively , and randomly draw each synaptic weight from the corresponding mean ±5% . The excitatory mean is chosen such that the output neuron would fire above the target rate everywhere in the absence of inhibition; we typically take this mean to be 1 ( Table 1 and Appendix 1 ) . The mean inhibitory weight is then determined such that the output neuron would fire close to the target rate , if all the weights were at their mean value ( Table 2 and Appendix 1 ) . Choosing the weights this way ensures that initial firing rates are random , but neither zero everywhere , nor inappropriately high . We model a global reduction of inhibition by scaling all inhibitory weights by a constant factor , after the grid has been learned . In the following , we derive the spacing of periodic firing patterns as a function of the simulation parameters for the linear track . We first show that homogeneous weights , chosen such that the output neuron fires at the target rate , are a fixed point for the time evolution of excitatory and inhibitory weights under the assumption of slow learning . We then perturb this fixed point and study the time evolution of the perturbations in Fourier space . The translational invariance of the input overlap leads to decoupling of spatial frequencies and leaves a two-dimensional dynamical system for each spatial frequency . For smoother spatial tuning of inhibitory input than excitatory input , the eigenvalue spectrum of the dynamical system has a unique maximum , which indicates the most unstable spatial frequency . This frequency accurately predicts the grid spacing . We first consider place cell-like input ( Gaussians ) and then non-localized input ( Gaussians convolved with white noise ) . At the end of the analysis , you will find a glossary of the notation . Whenever we use P as a sub- or superscript instead of E or I , this implies that the equation holds for neurons of the excitatory and the inhibitory population . The analysis is written as a detailed and comprehensible walk-through . The reader who is interested only in the result can jump to Equations 78 and 104 . A summary of notation:The rat′s position at time t:x ( t ) Spatial dimensions x , y and head direction z:x= ( x , y , z ) Population label; can be E ( excitatory ) or I ( inhibitory ) :PStandard deviation of Gaussian tuning of population P:σPSpatial autocorrelation length of input of population P:σP , corrNumber of input neurons of population P:NPNumber of place fields per input neuron of population P:NPfFiring rate of output neuron:rout ( x ) Firing rate of input neuron i of population P:riP ( x ) Synaptic weight of input neuron i of population P to output neuron:wiP ( t ) Learning rates of excitation and inhibition:ηE , ηITarget rate of the output neuron:ρ0Length of linear track:LHeight of the Gaussian input fields:αE , αIValue of Gaussian with standard deviation σP at location x:𝒢P ( x ) Von Mises distribution with width σP that is periodic in [−L/2 , L/2]:ℳP ( x ) | Knowing where you are never hurts , be it during a holiday in New York or on a hiking trip in the Alps . Our sense of location seems to depend on a structure deep within the brain called the hippocampus , and its neighbor , the entorhinal cortex . Studies in rodents have shown that these areas act a little like an in-built GPS for the brain . They contain different types of neurons that help the animal to work out where it is and where it is going . Among those are place cells , present within the hippocampus , and grid cells and head direction cells , found within the entorhinal cortex and other areas . Place cells fire whenever an animal occupies a specific location in its environment , with each place cell firing at a different spot . Grid cells generate virtual maps of the surroundings that resemble grids of repeating triangles . Whenever an animal steps onto a corner of one of these virtual triangles , the grid cell that generated that map starts to fire . Head direction cells increase their firing whenever an animal’s head is pointing in a specific direction . These cell types thus provide animals with complementary information about their location . But how do the cells first become selective for specific places or head directions ? Weber and Sprekeler propose that a single mechanism gives rise to the spatial characteristics of all these different types of cells . Like all neurons , these cells communicate with their neighbors at junctions called synapses . These may be either excitatory or inhibitory . Cells at excitatory synapses activate their neighbors , whereas cells at inhibitory synapses deactivate them . Weber and Sprekeler used a computer to simulate changes in excitatory and inhibitory synapses in a virtual rat exploring an environment . Interactions between the two types of synapses gave rise to virtual cells that behaved like place , grid or head direction cells . Which cell type emerged depended on whether the excitatory or the inhibitory synapses were more sensitive to the virtual rat’s location . This idea adds to a range of others proposed to explain how the brain codes for locations . Whether any of these ideas or a combination of them is correct remains to be determined . Further pieces are needed if we are to solve the puzzle of how the brain supports navigation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Learning place cells, grid cells and invariances with excitatory and inhibitory plasticity |
Despite numerous studies , there is little agreement about what brain changes accompany motor sequence learning , partly because of a general publication bias that favors novel results . We therefore decided to systematically reinvestigate proposed functional magnetic resonance imaging correlates of motor learning in a preregistered longitudinal study with four scanning sessions over 5 weeks of training . Activation decreased more for trained than untrained sequences in premotor and parietal areas , without any evidence of learning-related activation increases . Premotor and parietal regions also exhibited changes in the fine-grained , sequence-specific activation patterns early in learning , which stabilized later . No changes were observed in the primary motor cortex ( M1 ) . Overall , our study provides evidence that human motor sequence learning occurs outside of M1 . Furthermore , it shows that we cannot expect to find activity increases as an indicator for learning , making subtle changes in activity patterns across weeks the most promising fMRI correlate of training-induced plasticity .
Humans have the remarkable ability to learn complex sequences of movements . While behavioural improvements in sequence learning tasks are easily observable , the underlying neural processes remain elusive . Understanding the neural underpinnings of motor sequence learning could provide clues about more general mechanisms of plasticity in the brain . This motivation has led numerous functional magnetic resonance imaging ( fMRI ) studies to investigate the brain changes related to motor sequence learning . However , there is little agreement about how and where in the brain learning-related changes are observable . Previous studies include reports of signal increases across various brain regions ( Floyer-Lea and Matthews , 2005; Grafton et al . , 1995; Hazeltine et al . , 1997; Karni et al . , 1995; Lehéricy et al . , 2005; Penhune and Doyon , 2002 ) , as well as signal decreases ( Jenkins et al . , 1994; Peters et al . , 2017; Toni et al . , 1998; Ungerleider et al . , 2002; Wiestler and Diedrichsen , 2013 ) , nonlinear changes in activation ( Ma et al . , 2010; Xiong et al . , 2009 ) , spatial shifts in activity ( Lehéricy et al . , 2006; Steele and Penhune , 2010 ) , changes in multivariate patterns ( Wiestler and Diedrichsen , 2013; Wymbs and Grafton , 2015 ) , and changes in inter-regional functional connectivity ( Bassett et al . , 2015; Bassett et al . , 2011; Doyon et al . , 2002; Mattar et al . , 2016 ) . Additionally , some experiments have matched the speed of performance ( Karni et al . , 1995; Penhune and Doyon , 2002; Steele and Penhune , 2010; Lehéricy et al . , 2005; Seidler et al . , 2002; Seidler et al . , 2005 ) , while others have not ( Bassett et al . , 2015; Lutz et al . , 2004; Wiestler and Diedrichsen , 2013; Wymbs and Grafton , 2015 ) . Given that fMRI analysis has many degrees of freedom , these inconsistencies may not be too surprising . However , the implicit pressure in the publication system to report findings may also have contributed to a lack of coherency . To address this issue , we designed a comprehensive longitudinal study of motor sequence learning that allowed us to systematically reinvestigate previous findings . In order to increase transparency , we pre-registered the design , as well as all tested hypotheses on the Open Science Framework ( Berlot et al . , 2017; https://osf . io/etnqc ) , and make the full dataset available to the research community ( Berlot , 2020; OpenNeuro accession number ds002776 ) . The main aim of our study was to systematically evaluate different ideas of how learning-related changes are reflected in the fMRI signal . In the context of motor sequence learning , the most commonly examined brain region is the primary motor cortex ( M1 ) . Previous reports of increased M1 activation after long-term learning have been interpreted as additional recruitment of neuronal resources for trained behavior , taken to suggest the skill is represented in M1 ( Floyer-Lea and Matthews , 2005; Karni et al . , 1995; Karni et al . , 1998; Lehéricy et al . , 2005; Penhune and Doyon , 2002; for a review see Dayan and Cohen , 2011; Figure 1a ) . Since then , several pieces of evidence have suggested that sequence-specific memory may not reside in M1 ( Beukema et al . , 2019; Wiestler and Diedrichsen , 2013; Yokoi and Diedrichsen , 2019 ) . However , some of these reports studied skill acquisition over a course of a few days , while human skill typically evolves over weeks ( and months ) of practice . Therefore , including several weeks of practice might be more suitable to test whether , and at what time point , M1 develops skill-specific representations . Outside of M1 , learning-related activation changes have been reported in premotor and parietal areas ( Grafton et al . , 2002; Hardwick et al . , 2013; Honda et al . , 1998; Penhune and Doyon , 2002; Tamás Kincses et al . , 2008; Vahdat et al . , 2015 ) , with activation increases commonly interpreted as increased involvement of these areas in the skilled behavior . Yet , recent studies have mostly found that , as the motor skill develops , activation in these areas predominantly decreases ( Penhune and Steele , 2012; Wiestler and Diedrichsen , 2013; Wu et al . , 2004 ) . Such reductions are harder to interpret as they could reflect a reduced areal involvement in skilled performance or , alternatively , more energy efficient implementation of the same function ( Figure 1b; Picard et al . , 2013; Poldrack et al . , 2005 ) . To complicate things further , regional activity increases and decreases could occur simultaneously in the same area ( Figure 1c; Steele and Penhune , 2010 ) . In such a scenario , the net activation in the region would not change , yet , the trained sequences would engage slightly different subpopulations of the region than untrained sequences . A variant of this idea is that each specific sequence becomes associated with dedicated neuronal subpopulation ( and hence fMRI activity pattern ) . Such a representation would form the neural correlate of sequence-specific learning – the part of the skill that does not generalize to novel , untrained motor sequences ( Karni et al . , 1995 ) . Sequence-specific activation patterns should change early in learning ( Figure 1d ) , when behavior improves most rapidly , and stabilize later , once the skill has consolidated and an optimal pattern is established ( Peters et al . , 2017 ) . One possible way in which sequence-specific patterns could reorganize is by becoming more distinct from one another ( Figure 1e; Wiestler and Diedrichsen , 2013 ) . Having a distinctive code for each sequence might be of particular importance to the system in a trained state , allowing it to produce different dynamical sequences , while avoiding confusion or ‘tangling’ of the different neural trajectories ( Russo et al . , 2018 ) . To systematically examine the cortical changes associated with motor sequence learning , we carried out a longitudinal study over 5 weeks of training with 4 sessions of high-field ( 7 Tesla ) fMRI scans . Behavioural performance in the first three scanning sessions was imposed to the same speed of performance . This allowed us to inspect whether examined fMRI metrics reflect brain reorganization , independent of behavioral change . However , controlling for speed incurs the danger of not tapping into neural resources that are necessary for skilled performance ( Orban et al . , 2010; Poldrack , 2000 ) . We therefore compared the fMRI session with paced performance at the end of behavioural training with one acquired with full speed performance ( Figure 2 ) . This manipulation allowed us to systematically assess the role of speed on the fMRI metrics of learning , thereby addressing an important methodological problem faced by virtually every study on motor learning .
We trained 26 participants to perform six 9-digit sequences with their right hand on a keyboard device ( Figure 2a ) . During training , they received visual feedback ( green for correct and red for incorrect presses ) and were rewarded for both accuracy and speed ( see Materials and methods ) . Over the course of 5 weeks , participants practiced ~4000 trials ( Figure 2b ) . This led to substantial performance improvement , with the average movement time ( MT ) to complete a sequence decreasing from an initial 3 . 2 s to 1 . 2 s at the end of the training ( Figure 2c ) . The training regime was complemented with behavioral assessments on four occasions designed to specifically assess participants’ performance on trained sequences relative to untrained sequences ( Figure 2d , yellow underlay ) . Prior to training ( test day 1 ) , the speed of sequence execution did not differ between trained and untrained sequences . For all subsequent sessions , MTs were significantly faster for trained than untrained sequences ( p<0 . 001 ) , implying sequence-specific learning . Additionally , performance of trained sequences improved between all subsequent sessions , even after week 3 ( week 3–5: t ( 25 ) =5 . 49 , p=1 . 1e−5 ) . Thus , participants’ performance of trained sequences improved across the five weeks . To assess fMRI changes with learning , participants underwent four fMRI scans ( 1st scan: before the main training; 2nd scan: week 2; 3rd and 4th scan: week 5 ) , performing both trained and untrained sequences ( Figure 2d – grey underlay ) . The first session served as a baseline , where all of the presented sequences ( trained and untrained ) were novel to participants . Both trained and untrained sequences were always cued by presenting the corresponding digits on the screen ( Figure 2a ) . During the first three sessions , participants were paced with a metronome so that all sequences , trained and untrained , were performed at the same speed as in the first scan . Performance in the fourth session was at maximum speed , resulting in significantly lower MTs for trained compared to untrained sequences ( Figure 2d ) . To assess different neural signatures of observed behavioral learning , we first examined how the overall evoked activation changed over weeks of training for the same speed of movement . First , we re-investigated the classical finding that activity , measured as the percent BOLD signal change relative to rest , increased in M1 for matched performance after long-term training ( Karni et al . , 1995; Figure 1a ) . Our task elicited activation in a range of cortical areas ( Figure 3a for session 1 – that is , prior to learning ) . A region of interest ( ROI ) analysis of the hand area of M1 , contralateral to the performing hand , however , showed no significant change across weeks ( F ( 2 , 50 ) =0 . 44 , p=0 . 85 ) . There was a significant main effect of sequence type ( F ( 1 , 25 ) =6 . 32 , p=0 . 019 ) , but none of the post-hoc t-tests revealed a significant difference . Additionally , the interaction between the two factors was also not significant ( F ( 2 , 50 ) =0 . 17 , p=0 . 84 ) . The absence of overall activity changes , however , should not be taken as evidence for an absence of plasticity in the region . It is possible that some subregions of M1 increased in activation for learned sequences , while other decreased , as suggested by Steele and Penhune , 2010 . Such mixed changes would result in a shift of the overall pattern , which would lead to an increase in the angle between the mean activity pattern for trained and untrained sequences ( Figure 1c ) . Because we calculated the angle between activity patterns for each participant separately , this criterion does not assume that the observed shift is spatially consistent across individuals – any idiosyncratic shift could be detected . Therefore it serves as a sensitive statistical criterion to detect shifts in spatial location of activation , which were previously reported only descriptively ( Steele and Penhune , 2010 ) . However , in M1 , the averaged cosine angle ( Figure 3c ) remained unchanged across the weeks ( F ( 2 , 50 ) =1 . 71 , p=0 . 19 ) , indicating that the average activity pattern remained comparable across trained and untrained sequences . In sum , we found no evidence for activation increases ( Karni et al . , 1995 ) , decreases , or relative shifts in activation patterns ( Steele and Penhune , 2010 ) in M1 . To investigate activation changes in areas outside of M1 , we calculated changes in activity between the weeks in a map-wise approach ( Figure 4a ) . Over the three measurement time points , we found no reliable activation increases in any cortical area that was activated by the task in week 1 . Instead , we observed widespread learning-related reductions in activity in premotor and parietal areas ( Figure 4a ) , in line with our pre-registered prediction . These activation reductions were observed across both subsequent sessions ( i . e . weeks 1–2 , weeks 2–5 ) for trained and untrained sequences , with bigger reductions for trained sequences . In weeks 2 and 5 , trained sequences elicited overall lower activity than untrained sequences ( Figure 4b; see Figure 4—figure supplement 1 for statistical maps ) . These learning-related reductions in activity were also statistically significant in our predefined ROIs in premotor ( dorsal premotor cortex – PMd ) and parietal cortices ( anterior superior parietal lobule – SPLa ) ( Figure 3b ) : In a 3 ( week ) x 2 ( sequence type ) ANOVA on observed activation both main effects and interaction were highly significant in PMd ( week: F ( 2 , 50 ) =17 . 47 , p=1 . 77e−6; sequence type: F ( 1 , 25 ) =11 . 86 , p=2 . 03e−3; interaction: F ( 2 , 50 ) =13 . 22 , p=2 . 46e−5 ) as well as in SPLa ( week: F ( 2 , 50 ) =19 . 14 , p=6 . 73e-7; sequence type: F ( 1 , 25 ) =19 . 36 , p=1 . 77e−4; interaction: F ( 2 , 50 ) =21 . 59 , p=1 . 74e−7 ) . In contrast , no main effect of week was observed in S1 ( F ( 2 , 50 ) =1 . 82 , p=0 . 17 ) . Neither was there a significant effect of sequence type ( F ( 1 , 25 ) =0 . 19 , p=0 . 66 ) , or interaction between the two factors ( F ( 2 , 50 ) =2 . 01 , p=0 . 14 ) . This pattern of results on changes in overall activation remained unchanged after excluding error trials from the analyses ( see Figure 3—figure supplement 1a ) . Thus , we observed widespread activation decreases with learning across secondary and association cortical areas . In a few smaller areas , activation increased with learning ( red patches in Figure 4a–b ) . This was observed uniformly in areas with activity at or below baseline – thus these changes reflect decreased suppression of activity rather than increases . It is likely that these activity increases are not task relevant , but instead reflect the increasing automaticity and lower need for central attentional resources with learning ( see Discussion ) . We also examined whether there were , in addition to the overall activity decreases , shifts in the average activity patterns in the predefined regions of interest ( Figure 1c ) . As for M1 , we calculated the cosine angle dissimilarity ( see Materials and methods ) between the average activity patterns for trained and untrained sequences , separately for each scanning session . Figure 5a shows cosine angle dissimilarities between trained and untrained sequences in PMd , displayed using multidimensional scaling ( MDS ) . Patterns for trained sequences moved away from the starting point over weeks , and became more different from untrained patterns . Both in parietal and premotor areas there was clear evidence for a shift – cosine angular dissimilarity between the average trained and untrained sequence activation increased significantly across weeks ( PMd: F ( 2 , 50 ) =23 . 63 , p=5 . 98e−8; SPLa: F ( 2 , 50 ) =23 . 19 , p=7 . 49e−8 ) ( Figure 3c ) . S1 also showed a significant increase in cosine dissimilarity between trained and untrained patterns with learning ( F ( 2 , 50 ) =8 . 68 , p=5 . 79e−4 ) . These changes , however , were much less pronounced than those observed in premotor and parietal areas . This observed increase in dissimilarity between average trained and untrained pattern in PMd and SPLa , and to a lesser extent in S1 , was also observed when analyzing only trials with correct performance ( see Figure 3—figure supplement 1b ) . To investigate whether the observed changes in the overall activity patterns in premotor and parietal areas were spatially consistent across individuals , we normalized ( z-scored ) activation maps in each region and assessed the relative contribution of subregions to overall activation in weeks 1 and 5 ( Figure 5b ) . Comparing the pattern of activation revealed that before training ( week 1 , blue ) sequences elicit relatively more activation in rostral parts of the premotor and supplementary motor areas , and that activity was more caudal after training ( week 5 , red; Figure 5c displays the cross-section of relative activation changes ) . Some differences were also observed in the posterior parietal cortex , with activation shifting from more posterior to anterior subregions after learning ( Figure 5c ) . Altogether , these results show that with learning , the execution of sequences relies on slightly different subareas within premotor and parietal regions . Our analyses so far have been concerned with changes in the overall pattern of trained vs . untrained sequences , and showed widespread reductions in activation and some more subtle changes in relative location . The sequence-specific performance advantage , however , indicates that the brain must represent specific sequences – i . e . , there should be activity patterns that are unique to each individual sequence . Sequence-specific learning should then be reflected in changes of these sequence-specific activity patterns with learning ( Figure 1d ) . Consistent with previous results ( Wiestler and Diedrichsen , 2013; Yokoi and Diedrichsen , 2019 ) , we detected sequence-specific activity patterns , i . e . activity patterns that differentiate between the tested motor sequences , in various cortical regions , even in session 1 ( Figure 6a ) . This allowed us to assess their reorganization across sessions . Our pre-registered hypothesis ( https://osf . io/etnqc ) was that earlier in learning sequence-specific activity patterns would change more for trained than untrained sequences , and would stabilize later in learning . In contrast to the other ideas tested in this paper , this was a novel hypothesis and not based on previous reports . Specifically , we predicted that the correlation of each sequence-specific pattern between weeks 1 and 2 should be lower for trained as compared to untrained sequences . The problem with performing a simple correlation analysis on the patterns , however , is that the estimated correlation will be biased by noise – that is , more within-session variability for one set of sequences will result in a lower correlation ( Diedrichsen et al . , 2018 ) . To address this problem , we used the pattern component modelling ( PCM ) framework which explicitly models and estimates the signal and noise for each session explicitly . Using this approach , we estimated the likelihood of participants’ data under a series of models , each assuming a true correlation in the range between 0 ( uncorrelated patterns ) and 1 ( perfect positive correlation; see Materials and methods for details ) . Figure 6b shows the log-likelihood for each specific correlation model relative to the mean across all models . In SPLa , the most likely correlation of the activity patterns for the trained sequences between weeks 1 and 2 was r = 0 . 37 . For week 2–5 , the likelihood peaked at r = 0 . 6 . In contrast , the likelihood functions for untrained sequences indicated that the most likely model was between r = 0 . 6–0 . 7 for both week 1–2 and 2–5 . The advantage of this analysis is that we can be sure that the observed low correlation in week 1–2 for trained sequence was not due to increased noise . In fact , if the noise in one or both sessions was too high , then the model would be unable to distinguish between any of the correlation models – i . e . , the likelihood curve would be a flat line . To statistically assess the difference in correlations across trained and untrained sequences , we compared the likelihood of the data of trained sequences between two models: the best-fitting model for the trained sequences ( r = 0 . 37 in SPLa ) and the correlation model best fitting the data of untrained sequences ( r = 0 . 6 ) ( black dots and projections onto y-axis in Figure 6b ) . To avoid double-dipping , the ‘best-fitting’ model was chosen on 25 participants ( n-1 ) and the likelihood assessed on the left-out subject ( see Materials and methods ) . The difference in model evidence was significant for correlation between weeks 1–2 in SPLa ( t ( 25 ) =2 . 88 , p=4 . 0e−3 ) . In contrast , no difference in correlation was observed later in learning , between weeks 2 and 5 ( t ( 25 ) =1 . 21 , p=0 . 24 ) . A similar pattern of results was observed in PMd , with correlation of trained sequences significantly lower than that of untrained sequences between weeks 1 and 2 ( t ( 25 ) =2 . 93 , p=3 . 6e−3 ) , but not between weeks 2 and 5 ( t ( 25 ) =0 . 88 , p=0 . 39 ) . No such change in correlation across weeks 1–2 was observed in M1 ( t ( 25 ) =0 . 43 , p=0 . 34 ) . In S1 , the effect was just significant ( t ( 25 ) =1 . 72 , p=0 . 049 ) . To ensure that the observed lower correlation for trained patterns was not due to larger difference in error rate between weeks 1 and 2 for trained than for untrained sequences , we repeated the analysis excluding error trials . The pattern of results remained unchanged in PMd and SPLa ( see Figure 6—figure supplement 1a ) , with lower correlation for trained than untrained patterns across weeks 1–2 . In S1 , after accounting for error trials , the correlation across weeks 1–2 did no longer differ between trained and untrained patterns . Overall , we found significant evidence that sequence-specific trained patterns in SPLa and PMd reorganize more in weeks 1–2 as compared to the untrained sequences , and stabilize later on with learning , in line with our new pre-registered prediction . To determine more generally where in the neocortex sequence-specific plasticity could be detected , we fit PCM correlation models to regularly tessellated regions spanning the cortical surface . Figure 6c displays the correlation with the highest evidence for activity patterns across weeks 1–2 and 2–5; separately for trained and untrained sequences . In general , the highest correlations were found in core sensorimotor areas . Across weeks 1–2 for trained sequences , correlations were significantly lower in a number of dorsal premotor , inferior frontal , and parietal regions ( Figure 6c ) . Across the cortex , correlation for trained patterns increased for weeks 2–5 , resulting in similar values which did not differ significantly between trained and untrained sequences for most tessels ( see Figure 6—figure supplement 1b ) . Together , these results confirmed that sequence-specific activation patterns in secondary association areas show less stability early in learning , but stabilize later on . Can we obtain further insight into how the sequence-specific patterns change in these areas ? One specific preregistered prediction was that there would be an increase in distinctiveness ( dissimilarity ) between fMRI patterns underlying each trained sequence ( Wiestler and Diedrichsen , 2013; Figure 1e ) . To test this hypothesis , we calculated crossnobis dissimilarities ( Walther et al . , 2016 ) between sequence-specific activations , separately for trained and untrained sequences . In contrast to our prediction , no significant change in dissimilarity across weeks was observed in any of the predefined regions ( Figure 6d ) . This suggests that the reorganization observed for trained sequences early in learning did not increase the average distinctiveness of the sequence-specific patterns . In the last part of the experiment , we asked whether some of the negative findings ( e . g . no changes in M1 , no increase in dissimilarities for trained sequences ) might have been due to the fact that participants were paced at a relatively slow speed . Matching the speed across sessions allows for the comparisons of changes in neural activity for exactly the same behavioral output ( Karni et al . , 1995; Lehéricy et al . , 2005 ) . However , it could be that controlling for speed impairs our ability to study brain representations of motor skill; simply because after learning , the system is not challenged enough to activate the neuronal representations supporting skilled performance . Consequently , several studies have not ( Bassett et al . , 2011; Wymbs and Grafton , 2015 ) , or not strictly ( Wiestler and Diedrichsen , 2013 ) , matched performance across sessions or levels of training . To examine the effect of performance speed , we added a fourth scanning session ( full speed - fs ) , just a day after from the third session in week 5 , in which participants were instructed to perform the sequence as fast as possible . Performance during the 4th scan was 1010 ms faster than in the first session ( t ( 25 ) =15 . 7 , p=1 . 82e−14 ) and also 338 ms ( t ( 25 ) =9 . 92 , p=4 . 58e−10 ) faster for trained than for untrained sequences . Averaged over trained and untrained sequences , we found that the faster performance in this session led to an increase in activity across premotor and parietal areas ( Figure 7a , b ) . Although trained sequences were executed faster than untrained sequences , activation was still lower for trained compared to untrained sequences , similar to what we observed for paced performance ( Figure 7c; see Figure 7—figure supplement 1a for statistical maps ) . In M1 and S1 , we found no difference in activation between trained and untrained sequences ( Figure 7a; M1: t ( 25 ) =1 . 78 , p=0 . 09; S1: t ( 25 ) =1 . 69 , p=0 . 10 ) . Overall , the pattern of results for evoked activation did not change qualitatively when participants performed at full speed . Next , we examined whether the brain representations of individual sequences are similarly engaged at slow and fast speeds . The correlation between sequence-specific patterns was relatively high ( r = 0 . 62 ) across our regions of interest . We found no differences between the different regions ( F ( 3 , 75 ) =1 . 47 , p=0 . 23 ) , or sequence types ( trained vs . untrained: F ( 1 , 25 ) =0 . 25 , p=0 . 62 ) . Thus , the sequence-specific representations activated during performance at high skill level ( full speed ) are at least partly activated even when performance slowed down . Having established that the mean activation results are replicated across paced and full-speed performance , and that similar sequence-specific representations are activated in both cases , we tested whether activation patterns for different trained sequences are more distinct during full speed performance , as reported in Wiestler and Diedrichsen , 2013 . Overall , crossnobis dissimilarities increased at full speed for trained sequences in PMd and SPLa ( Figure 7e ) . No such changes were found in M1 or S1 . Moreover , trained sequences showed larger dissimilarities than untrained at full-speed performance across premotor and parietal cortices ( Figure 7f ) , which was not the case for the last paced session . In our predefined ROIs , this difference was significant for PMd ( Figure 7d ) , but also parietal areas showed significantly higher dissimilarities between trained sequences at full speed ( Figure 7—figure supplement 1b ) . This suggests that while activity patterns at full speed are correlated to those during paced performance , they are more distinguishable for trained sequences . Could this effect be driven by behavioral performance , with trained sequences performed more differently at full speed ( i . e . different speeds across trained sequences ) , while untrained sequences were performed at a more equal speed ? To test for this , we calculated crossnobis dissimilarities between movement times associated with different trained and untrained sequences . The dissimilarities based on speed of performance did not differ significantly across trained and untrained sequences ( t ( 25 ) =0 . 57 , p=0 . 57 ) . Therefore , increased dissimilarity of trained compared to untrained patterns in premotor and parietal areas could not be explained by a difference in execution speed . Instead , this effect likely reflects changes in activity patterns underlying full speed skilled performance . We observed learning-related changes in cortical association areas , but not in the primary motor cortex . Of course , learning could also be driven by neuronal changes in subcortical brain regions ( Ashby et al . , 2010; Graybiel , 2000; Graybiel and Grafton , 2015; Hikosaka et al . , 1999; Yin et al . , 2009 ) . The striatum in particular has been proposed as a structure where motor skills are stored ( Kawai et al . , 2015; Lehéricy et al . , 2006 ) . Inspecting changes in overall activity across sessions , we observed no difference in activity between trained and untrained sequences in either putamen or caudate nucleus ( Figure 8a ) . Previous experiments have reported that with learning , activation moves from more ‘cognitive’ areas of the striatum ( i . e . caudate nucleus ) to more ‘motor’ areas ( i . e . putamen ) ( Coynel et al . , 2010; Lehéricy et al . , 2005; Reithler et al . , 2010 ) . Our data fail to replicate this result: Both the visual inspection ( Figure 8b ) , and statistical quantification of the mean pattern difference for trained and untrained sequences across sessions revealed no such learning-specific shift of mean striatal activation pattern with learning . Lastly , we examined if the striatum represents individual sequences . During the paced sessions , activity patterns for different sequences were not distinguishable in either caudate nucleus or putamen ( Figure 8c ) . However , during full speed performance trained sequences elicited distinct activity patterns in both regions ( i . e . crossnobis dissimilarity >0: caudate nucleus: t ( 25 ) =2 . 27 , p=0 . 032; putamen: t ( 25 ) =2 . 44 , p=0 . 022; Figure 8c ) . This effect was specific to the trained sequences , with untrained sequences still exhibiting undistinguishable patterns of activity at full speed . Thus , we found some evidence that trained motor sequences are represented in the form of distinct activity patterns in the striatum during full speed skilled performance . To examine whether the speed purely pulls the signal out of the noise better , or qualitatively changes the representation , we , similarly to the analyses in the cortical regions , performed the PCM correlation model across the paced and full speed sessions in week 5 . The correlation between sequence-specific patterns in both regions was higher than 0 ( putamen: t ( 25 ) =9 . 56 , p=8 . 0e−10; caudate: t ( 25 ) =6 . 37 , p=1 . 1e−6 ) , but lower than 1 ( putamen: t ( 25 ) =8 . 85 , p=3 . 6e−9; caudate: t ( 25 ) =5 . 86 , p=4 . 1e−6 ) . Similarly as for the cortical regions , we found no differences between the caudate nucleus and putamen ( F ( 3 , 75 ) =0 . 19 , p=0 . 66 ) , or sequence types ( trained vs . untrained: F ( 1 , 25 ) =0 . 05 , p=0 . 83 ) . Thus , the sequence-specific representations activated during performance at high skill level ( full speed ) are at least partly activated even when performance slowed down . This suggests that moving faster engages similar representations as moving slower , but helps to increase the signal-to-noise ratio .
The search for neural substrates of learning is a daunting task: the acquisition of longitudinal data sets is work intensive , and the large dimensionality of possible brain metrics makes the search difficult ( Poldrack , 2000 ) . Historically , the question was simplified by studying activation increases in single areas as proxies for motor ‘engram’ localization ( Lashley , 1950; Berlot et al . , 2018 ) . Here we found no evidence for such activation increases; instead we observed widespread and distributed decreases in activation across cortical areas . In contrast , subtler changes in the distributed patterns of fMRI activity have the potential to provide more direct metrics of plasticity . Increased pattern reorganization ( across weeks ) , and larger pattern separation for trained sequences was found across prefrontal , parietal , and striatal regions . These metrics may be useful as general fMRI correlates of neural reorganization beyond the domain of motor learning .
Twenty-seven volunteers participated in the experiment . One of them was excluded because field map acquisition was distorted in one of the four scans . The average age of the remaining 26 participants was 22 . 2 years ( SD = 3 . 3 years ) , and the sample included 17 women and 9 men . All participants were right-handed and had no prior history of psychiatric or neurological disorders . They provided written informed consent to all procedures and data usage before the study started . The experimental procedures were approved by the Ethics Committee at Western University . Participants performed finger sequences with their right hand on an MRI-compatible keyboard ( Figure 2a ) , with keys numbered 1–5 for thumb-little finger . The keys had a groove for each fingertip and were not depressible . The force of isometric finger presses was measured by the force transducers ( FSG-15N1A , Sensing and Control , Honeywell; dynamic range 0–25 N ) mounted underneath each key with an update rate of 2 ms . A key press was recognized when the sensor force exceeded 1 N . The measured signal was amplified and sampled at 200 Hz . Participants were trained to execute six 9-digit finger sequences over a period of five weeks ( Figure 2a ) . They were randomly split into two groups , with trained sequences of one group constituting the untrained sequences for the other group and vice versa . Finger sequences of both groups were matched as closely as possible in terms of the starting finger , number of finger repetitions in a sequence and first-order finger transitions . This counterbalancing between the groups ensured that any of the observed results were not specific to a set of chosen trained sequences . In the pre-training session prior to the first scan ( Figure 2b ) , participants were acquainted with the apparatus and task performed during scanning . Sequences executed during this pre-training session were not encountered later on in the experiment . During the training sessions , participants were trained to perform the six sequences as fast as possible . They received visual feedback for the correctness of their presses with digits turning green for a correct finger press and red for an incorrect one . After each trial , participants received points based on the accuracy and their movement time ( MT – time from the first press until the last finger release in the sequence; Figure 2c ) . Trials executed correctly and faster than participant’s median MT from the previous blocks were rewarded with 1 point . If participants performed correctly and 20% faster than the median MT from previous blocks , they received 3 points . If they made a mistake or performed below their median MT , they received 0 points . Participants performed each sequence twice in a row: digits were written on the screen for the first execution , but removed for the second execution so that participants had to perform the finger sequence from memory . Training sessions were broken into several blocks , each consisting of 24 trials ( 4 trials per trained sequence ) , with time between blocks to rest . At the end of each block , participants received feedback on their error rate , median MT and points obtained during the block . If participants performed with an error of <15% and faster than the previous median MT , the MT threshold was updated . This design feature was chosen to maintain participants’ motivation to execute the sequences as fast as possible , within the allowed error range . During the behavioral test sessions ( Figure 2d ) , participants executed both the trained sequences they were trained on , as well as matched untrained sequences , with all sequences randomly interspersed . As in training , each sequence was performed twice in a row – however , the 9-digit sequence numbers were presented on the screen present on both executions . Therefore , the requirement to remember the sequences from the first to second execution , which was present in training sessions , was omitted for test sessions . For this reason , performance in training and test sessions ( Figure 2c–d ) cannot be directly compared . As an additional feature of the four behavioral test sessions , we examined participants’ performance with their left hand . Specifically , we tested them on execution of intrinsically-matched trained sequences ( i . e . producing the same finger combinations ) , extrinsically-matched trained sequences ( i . e . producing the same external consequences using mirrored fingers ) and random sequences . This was added to probe to what extent learning generalized to the other effector in intrinsic or extrinsic coordinate frames , at different stages of learning . Per session , participants performed 4 repetitions of each trained sequence in intrinsically-matched space and 4 repetitions in extrinsically-matched space . Participants underwent four scanning sessions ( Figure 2d ) – with the first one before learning regime started , the second after a week and two more scans after completion of the 5 training weeks . Each scanning session consisted of eight functional runs . We employed an event-related design , randomly intermixing execution of trained and untrained sequences . Each sequence was repeated twice in a row with digits present on the screen during both executions . Thus , there was no need to memorize either trained or untrained sequences from first to second execution in the scanner . Each sequence was repeated for a total of six times in every run . Each trial started with 1 s preparation time , during which the sequence was presented on the screen . After that time , a ‘go’ signal was displayed as short pink line underneath the sequence numbers . In scanning sessions 1–3 , this line started expanding below the written numbers , indicating the speed at which participants were required to press along . In scanning session 4 , only a short line was presented in front and underneath the sequences . When the line disappeared , this signaled a ‘go’ cue for participants to execute the presented sequence as fast as possible . The execution phase including the feedback on overall performance lasted for 3 . 5 s , and the inter-trial interval was 0 . 5 s ( see Figure 2—figure supplement 1 ) . Each trial lasted for 5 s . Participants always received 3 points upon correct execution of the sequence , and 0 points otherwise . Five periods of rest , each 10 s long , were added randomly between trials in each run to provide a better estimate of baseline activation . Participants performed the task inside the scanner for approximately 75 min . After each scanning session , they filled out a recall and recognition questionnaires on trained and untrained sequences performed inside the scanner . Data was acquired on a 7-Tesla Siemens Magnetom scanner with a 32-receive channel head coil ( 8-channel parallel transmit ) . Anatomical T1-weighted scan was acquired at the beginning of the first scanning session , using a magnetization-prepared rapid gradient echo sequence ( MPRAGE ) with voxel size of 0 . 75 × 0 . 75×0 . 75 mm isotropic ( field of view = 208×157 x 110 mm [A-P; R-L; F-H] , encoding direction coronal ) . Functional data were acquired using a sequence ( GRAPPA 3 , multi-band acceleration factor 2 , repetition time [TR]=1 . 0 s , echo time [TE]=20 ms , flip angle [FA]=30 deg ) . We acquired 44 slices with isotropic voxel size of 2 × 2 × 2 mm . For estimating magnetic field inhomogeneities , we additionally acquired a gradient echo field map . Acquisition was in the transversal orientation with field of view 210 × 210 × 160 mm and 64 slices with 2 . 5 mm thickness ( TR = 475 ms , TE = 4 . 08 ms , FA = 35 deg ) . To monitor the use of 7T for human research , participants filled out a questionnaire rating their levels of dizziness , wellbeing , alertness and focus after each imaging session . Functional data were analyzed using SPM12 and custom written MATLAB code . Functional runs were corrected for geometric distortions using fieldmap data ( Hutton et al . , 2002 ) , and head movements during the scan ( 3 translations: x , y , z; 3 rotations: pitch , roll , yaw ) , and aligned across sessions to the first run of the first session . The functional data were then co-registered to the anatomical scan . No smoothing or normalization to an atlas template was performed . Preprocessed data were analyzed using a general linear model ( GLM; Friston et al . , 1994 ) . Each of the performed sequences was defined as a separate regressor per imaging run , resulting in 12 regressors per run ( 6 trained , 6 untrained sequences ) , together with intercept for each of the functional runs . All instances of sequence execution were included into estimating regressors , regardless of whether the execution was correct or erroneous ( see section Treatment of error trials below ) . The regressor was a boxcar function starting at the beginning of the trial and lasting for trial duration . The boxcar function was convolved with a hemodynamic response function , with a time to peak of 5 . 5 s , and time to undershoot of 12 . 5 s . We adjusted the hrf onset individually per participant . For that , we defined a combined region of interest between PMd and M1 , and averaged the response across all voxels in the combined ROI for all performed sequences ( i . e . trained and untrained sequences together ) in session 1 . We then performed a grid-search with delay values of 0 , 0 . 5 and 1 s , and chose the one that maximally fit the evoked response for each subject . The same delay was used across all sessions . Ultimately , this analysis resulted in one activation estimate ( beta image ) for each of the 12 conditions per run , in each scanning session . We reconstructed individual subjects’ cortical surfaces using FreeSurfer ( Dale et al . , 1999 ) . All individual surfaces were aligned to the FreeSurfer’s Left-Right symmetric template ( Workbench , 164 k nodes ) via spherical registration . To detect sequence representation across the cortical surface , we used a surface-based searchlight approach ( Oosterhof et al . , 2011 ) , where for each node we selected a circular region of 120 voxels in the grey matter . The resulting analyses ( dissimilarities between sequence-specific activity patterns , see below ) was assigned to the center node . As a slightly coarser alternative to searchlights , we performed regular tessellation of cortical surface into 162 tessels per hemisphere . This allowed us to fit correlation models ( see below ) across the cortical surface , while not being as computationally intensive as searchlight analyses . We defined four regions of interest to cover primary somato-motor regions as well as secondary associative regions . M1 was defined using probabilistic cytoarchitectonic map ( Fischl et al . , 2008 ) by including nodes with the highest probability of belonging to Brodmann area ( BA ) 4 , while excluding nodes more than 2 . 5 cm from the hand knob ( Yousry et al . , 1997 ) . Similarly , S1 was defined as nodes related to hand representation in BA 1 , 2 and 3 . Additionally , we included dorsal premotor cortex ( PMd ) as the lateral part of the middle frontal gyrus . The anterior part of the superior parietal lobule ( SPLa ) was defined to include anterior , medial and ventral intraparietal sulcus . We also defined caudate nucleus and putamen as striatal regions of interest . The definition was carried out in each subject using FSL’s ( Jenkinson et al . , 2012 ) subcortical segmentation . We calculated the average percent signal change for trained and untrained sequences ( averaged across the 6 trained and 6 untrained sequences ) relative to the baseline for each voxel . The resulting volume map was projected to the surface for each subject , and a group statistical t-map was generated across subjects . Statistical maps were thresholded at p<0 . 01 , uncorrected , and the family-wise error corrected p-value for the size of the peak activation and activation cluster size was determined using a permutation test . Specifically , we ran 1000 simulations where we randomly flipped the sign of the contrast for subjects ( chosen at random out of 226 possible permutations ) . The rationale behind this is that under the null hypothesis , there should be no difference between the two conditions , and the sign of each contrast should be interchangeable . As for the data , we thresholded the statistical map from each permutation , and recorded the peak t-value ( across the map ) and the size of the largest cluster . The real data was then compared against this distribution to assess the probability of the observed t-value and cluster-size under the null hypothesis . Additionally , we assessed changes in percent signal in predefined regions of interest ( M1 , S1 , PMd , SPLa ) . This was performed in the native volume space of each subject . To do so , we averaged the percent signal change of voxels belonged to a defined region per subject and quantified activation changes across subjects using ANOVAs and t-tests across subjects . Besides overall activation , we also examined relative changes in elicited activation for trained sequences across sessions . This was done by normalizing ( z-scoring ) the percent signal change surface maps across voxels , separately for each subject . Normalization was applied both map-wise ( for Figure 5b ) , as well as for each of the pre-defined ROIs separately ( for cross-sections in Figure 5c ) . Statistical assessment of the difference between relative evoked activation pattern for trained vs . untrained sequence was carried out by calculating cosine angle dissimilarities between the mean evoked patterns . Cosine angle dissimilarity was chosen because it is not sensitive to overall magnitude in activation , and therefore assesses the difference in the relative activation distribution . To evaluate which cortical areas display sequence-specific encoding , we performed a searchlight analysis calculating the dissimilarities between evoked beta patterns of individual sequences . Beta patterns were first multivariately prewhitened ( standardized by voxels’ residuals and weighted by the voxel covariance matrix ) , which has been found to increase the reliability of dissimilarity estimates ( Walther et al . , 2016 ) . We then calculated the cross-validated squared Mahalanobis dissimilarities ( i . e . crossnobis dissimilarities ) between evoked sequence patterns ( 66 dissimilarity pairs for 6 trained and 6 untrained sequences ) . These dissimilarities were then averaged overall , as well as separately for pairs within trained sequences , and within untrained sequences . This metric was used both for searchlight analysis and calculation of metric within predefined regions ( cortical and striatal ) . The cortex surface maps contrasting dissimilarities between trained and untrained sequences were corrected for multiple comparisons using permutations , as described above for percent signal change surface maps . Correspondence of sequence-specific patterns across sessions was quantified using pattern component modelling ( PCM; Diedrichsen et al . , 2018 ) . This framework is superior at estimating correlations than simply performing Pearson’s correlation on raw activity patterns , or even in a crossvalidated fashion . The main problem with estimating correlations on data is that activation patterns are biased by noise , which varies across scanning sessions , and would therefore underestimate the true correlation . PCM separately models the noise and signal component , and can in this way combat the issue more than by simply performing crossvalidation . We designed 30 correlation models with correlations between 0 and 1 in equal step sizes and assessed the group likelihood of the observed data under each model . Subsequent group inferences were performed using crossvalidated approach on assessing individual log-Bayes factors ( model evidence ) . A crossvalidated approach was used to ensure that our choice of ‘best-fitting models’ and the evidence associated was independent and did not involve double-dipping . Specifically , we used n-1 subjects to determine the best-fitting models for trained and untrained patterns and recorded the log-Bayes factors for those two correlation models on the left-out subject . This was repeated across all subjects and a t-test was performed on the recorded log-Bayes factors ( i . e . out-of-sample model evidences ) . The same evaluation was performed for pre-defined regions of interest ( Figure 6b ) , as well as a regular tessellation across the cortical surface ( Figure 6c ) . As in behavioral sessions , participants were instructed to keep their error rate below 15% also inside the scanner . This was on average achieved , with the following error rate for trained vs . untrained sequences across the 4 scanning sessions: Week 1: 0 . 14 ± 0 . 02 vs . 0 . 15 ± 0 . 02 , week 2: 0 . 08 ± 0 . 01 vs . 0 . 14 ± 0 . 02 , week 5: 0 . 06 ± 0 . 01 vs . 0 . 09 ± 0 . 01 , speeded scan week 5: 0 . 14 ± 0 . 01 vs . 0 . 13 ± 0 . 01 . The number of errors varied significantly across sessions , and between sequence types . A session x sequence type ANOVA was significant for week ( F ( 3 , 75 ) =9 . 19 , p=2 . 97e−5 ) , sequence type ( F ( 1 , 25 ) =11 . 16 , p=2 . 63e−3 ) , as well as for their interaction ( F ( 3 , 75 ) =8 . 39 , p=7 . 00e−5 ) . Post-hoc t-tests revealed that the error rate differed between trained and untrained sequences in week 2 ( t ( 25 ) =4 . 20 , p=2 . 95e−4 ) and 5 ( t ( 25 ) =4 . 81 , p=6 . 1e−5 ) , but not in week 1 and for the speeded session of week 5 . To control for difference in error rate , we performed an additional first-level analysis with error trials excluded to ensure that our results were not due to inclusion of errors . Indeed , our results did not differ qualitatively when excluding errors , therefore we here report only the analyses with all trials included . The code is available at https://github . com/eberlot/motor_sequence_learning . git ( Berlot , 2020; copy archived at https://github . com/elifesciences-publications/motor_sequence_learning/ ) . | It has famously been claimed that it takes 10 , 000 hours to become an expert at something . But while most of us will never become concert pianists , we can all learn new motor skills and improve existing ones – from touch-typing to tennis – by practicing . What happens in the brain to produce these improvements in performance ? Researchers have tried to answer this question by scanning the brains of people as they practice motor skills , but the results have proved inconsistent . Some studies find that specific brain areas become more active as people practice . This could indicate that these areas are ‘storing’ new skills . But others report that brain activity decreases with practice . This might indicate that practice instead makes certain brain areas work more efficiently . It is also unclear where in the brain these learning-related changes occur . Some studies suggest that most occur in the primary motor cortex , or M1 – the area that sends commands to muscles . Others suggest that most changes take place outside of M1 , in areas that plan movements . Berlot et al . set out to resolve these inconsistencies by scanning the brains of healthy volunteers as they learned to play six 9-digit sequences on a keyboard . Each volunteer completed about 4 , 000 training trials over 5 weeks , and had their brain scanned four times . As the weeks passed , the volunteers became faster and more accurate at playing the sequences . However , the activity of their primary motor cortex did not change . By contrast , the activity of areas involved in planning movements decreased throughout training . The patterns of activity for each individual sequence reorganized throughout learning in the areas outside of the M1 . This happened most quickly during the early stages of training when the volunteers showed the fastest improvements in performance . Overall , these findings suggest that when we learn a new skill , activity in the brain areas supporting that skill decrease as the brain becomes more efficient . Increases in brain activity are thus unlikely to reflect the acquired skill . Instead , more subtle changes , in which the brain uses more specific patterns of activity to encode the skill , may underlie improved performance . This may also be true for other types of learning , such as acquiring a new language . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | A critical re-evaluation of fMRI signatures of motor sequence learning |
A core prediction of the vesicular transport model is that COPI vesicles are responsible for trafficking anterograde cargoes forward . In this study , we test this prediction by examining the properties and requirements of inter-Golgi transport within fused cells , which requires mobile carriers in order for exchange of constituents to occur . We report that both small soluble and membrane-bound secretory cargo and exogenous Golgi resident glycosyl-transferases are exchanged between separated Golgi . Large soluble aggregates , which traverse individual stacks , do not transfer between Golgi , implying that small cargoes ( which can fit in a typical transport vesicle ) are transported by a different mechanism . Super-resolution microscopy reveals that the carriers of both anterograde and retrograde cargoes are the size of COPI vesicles , contain coatomer , and functionally require ARF1 and coatomer for transport . The data suggest that COPI vesicles traffic both small secretory cargo and steady-state Golgi resident enzymes among stacked cisternae that are stationary .
The Golgi apparatus is a central feature of the secretory pathway in all eukaryotic cells . In higher eukaryotes , the Golgi stack consists of four to six flattened cisternae , which contain a series of glycosyl-transferases and other resident membrane proteins . These are localized in the order of their function in distinct steady-state distributions along the axis between the cis ( entry ) and the trans ( exit ) face , as a result of a dynamic equilibrium resulting from a balance of anterograde ( ER → cis → trans ) and retrograde ( trans → cis and Golgi → ER ) flows . Proteins secreted from the cell , as well as constituents of the plasma membrane and a broad variety of membrane-enclosed compartments pass through the ER–Golgi system in the anterograde direction , diverging only as they depart the Golgi stack at its trans face ( also termed the TGN ) . During this anterograde passage they are typically glycosylated in a step-wise fashion as they encounter the responsible enzymes ( Emr et al . , 2009; Klumperman , 2011 ) . There are two broadly opposing alternative mechanisms ( with many variations ) to explain anterograde transport ( cis → trans ) across the Golgi ( Emr et al . , 2009; Rothman , 2010 ) : ( 1 ) mobile cisternae ( also termed cisternal progression ) , in which the cisterna themselves move from cis → trans , being continuously remodeled by retrograde flow of resident enzymes in the process ( Mironov et al . , 2001; Losev et al . , 2006; Matsuura-Tokita et al . , 2006 ) . ( 2 ) The vesicular transport model , in which cisterna are viewed as static and the anterograde cargo must then be mobile , moving forward from cisterna-to-cisterna by a carrier mechanism . COPI-coated vesicles are the principal candidates for anterograde carriers , as they form and fuse copiously throughout the Golgi stack , and many contain anterograde cargo ( Balch et al . , 1984b; Rothman and Wieland , 1996; Orci et al . , 1997; Rothman , 2010 ) . COPI vesicles are known to carry retrograde cargo from Golgi to ER ( Letourneur et al . , 1994; Emr et al . , 2009 ) . However , dynamic tubular connections between cisternae have also been proposed ( Trucco et al . , 2004 ) . In this study , we seek to gain insight into how small anterograde ( and retrograde ) directed cargoes traverse stacked Golgi . We also test core predictions of the COPI vesicular model for anterograde transport among static cisternae . For these purposes we have employed a simple assay for inter-Golgi transport within fused cells , which requires mobile carriers in order for exchange of Golgi constituents to occur . Over 25 years ago , long before dynamic imaging in live cells with GFP tags was possible , we reported the surprising finding that VSV-encoded G protein is capable of rapid exchange between two different Golgi populations within fused cells ( Rothman et al . , 1984a , b ) . These results recapitulated a previous discovery made in cell-free extracts ( Fries and Rothman , 1980; Balch et al . , 1984a ) , which later allowed the identification of key vesicle transport machinery ( such as coatomer , NSF , and SNAREs ) needed for inter-Golgi transport ( Malhotra et al . , 1988; Waters et al . , 1991; Serafini et al . , 1991a , b ) . Although the data generated with this biochemical approach strongly indicated that VSV-G containing COPI vesicles were employed in anterograde transport in the Golgi ( Orci et al . , 1989 ) , this interpretation was debated ( Mellman and Simons , 1992; Pelham , 1994 ) because the role of COPI vesicles in retrograde transport was subsequently uncovered ( Letourneur et al . , 1994 ) , and the then available data on inter-Golgi traffic could not rule out that Golgi glycosyl-transferases are also transported between Golgi and could contribute an unknown portion of the assay signals ( Love et al . , 1998; Walter et al . , 1998 ) . Other groups used a cell fusion assay to observe the behavior of fluorescently labeled Golgi markers , but were only able to do so hours after fusion had taken place , when the Golgi were already reassembled at the center of the newly-formed poly-karyons ( Ho et al . , 1990; Deng et al . , 1992 ) . Here , we have reinvestigated inter-Golgi transport with much more precise and penetrating tools of modern cell and molecular biology and optics , and find that both small anterograde cargo and Golgi resident glycosyl-transferases are briskly exchanged . Large soluble aggregates , which traverse the stack , do not transfer between Golgi , implying that small cargo ( which can fit in a typical transport vesicle ) are transported by a different mechanism . Super-resolution microscopy and live cell imaging reveal that the carriers of both anterograde and retrograde cargo are the size of COPI vesicles , contain coatomer , and are Arf1-dependent .
To assess inter-Golgi exchange in real time we fused two populations of HeLa cells , one containing GFP tagged galactosyltransferase ( GT-GFP ) with a second containing RFP tagged sialyltransferase ( ST-RFP ) ( Schaub et al . , 2006 ) ( Figure 1 ) . The cells were separately transfected , combined and allowed to spread as a mixed population on a cover slip . At the steady state in HeLa cells , GT and ST have partially overlapping distributions in Golgi cisternae ( Rabouille et al . , 1995; Schaub et al . , 2006 ) . GT is localized in the trans Golgi cisternae , while ST is localized in the trans and trans-most/TGN cisternae . The cell population containing GT-GFP was also transfected with VSV-G protein to enable us to trigger cell–cell fusion with a brief acidic exposure ( Florkiewicz and Rose , 1984 ) . Following pH 5 exposure , the fused cells were immediately monitored by confocal video-microscopy , enabling us to record the time dependence of exchange of these markers between the Golgi populations in the fused cells from the onset of fusion . Importantly , all the experiments were performed at 20°C and in the presence of cycloheximide ( unless mentioned otherwise ) , to block new synthesis of GT-GFP and ST-RFP in the ER during the experiment and to block exit of these markers from the Golgi while enabling transport in the Golgi to proceed ( Matlin and Simons , 1983 ) . Each Golgi area remains tightly associated with its original nucleus for at least 2 hr , longer than the maximum time course of the exchange experiments . 10 . 7554/eLife . 01296 . 003Figure 1 . General procedure . HeLa cells co-expressing a GFP-labeled Golgi localized protein and VSV-G are mixed with HeLa cells expressing a RFP-labeled Golgi localized protein . Cell fusion is triggered by acidic exposure of cell surface targeted VSV-G . Cycloheximide was added 1 hr prior to fusing the cells and during the imaging procedure , to prevent de novo protein synthesis . Golgi-content mixing is assessed by live imaging confocal imaging , visualization and characterization of the putative transport intermediates are assessed by STED microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 003 We observed that the Golgi originally containing the GT-GFP marker progressively acquired the ST-RFP marker ( Figure 2 , Golgi 3 and 4 ) , the average fluorescence intensity of such Golgi areas increasing by a factor 10 . 2 ± 3 . 7 ( n = 5 ) at 30 min post-fusion at 20°C . Reciprocally , the initially red-labeled Golgi areas acquired the GT-GFP marker ( Figure 2 , Golgi 1 and 2 ) with a 12 . 1 ± 6 . 1 ( n = 5 ) fold increase at 30 min post-fusion . An interesting case arises in a polykaryon when a GT-GFP transfected cell fuses with an ST-RFP transfected cell and a third cell that failed to be transfected with either marker ( Figure 2 , Golgi 5 ) . Here , the initially non-fluorescent Golgi simultaneously acquires markers from both red and green Golgi . Altogether , these initial studies establish that some exogenous resident Golgi enzymes traffic between separate Golgi populations in fused cells . 10 . 7554/eLife . 01296 . 004Figure 2 . Inter-Golgi transport of Golgi resident glycosyl-transferase proteins . HeLa cells expressing either GT-GFP and VSV-G , or ST-RFP were mixed and fused by acidic exposure ( 1 min , pH 5 ) and then monitored by confocal video-microscopy at 20°C . Cells were treated with CHX ( 100 μg/ml ) 2 hr prior to fusion and during the imaging . Graphs show fluorescence intensity of markers within Golgi 2 , 4 , and 5 over time . Results are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 004 Most proteins require about 10–20 min to fully traverse the Golgi stack at 37°C , and this process is expected to be several fold slower at 20°C . Therefore , the time course of exchange between Golgi , remarkably enough , seems to be similar to that for transit of a Golgi stack in a single cell . Moderate rate differences and variations in this exchange can be expected due to the variability of Golgi numbers per polykaryon as well as to their respective distances to each other . After establishing our assay with Golgi resident enzymes , we evaluated transport of model anterograde soluble and membrane cargo proteins . We specifically chose model proteins whose aggregation state could be readily controlled , allowing us to test the cargo size dependence of the exchange mechanism . To control the size of a chimeric anterograde cargo , we took advantage of a well-established drug-dependent aggregation system ( Rivera et al . , 2000 ) . We tested both a soluble cargo ( hGH ) and a transmembrane cargo ( CD8lumenal ) , each of which also contained four repeats of a self-aggregation domain ( FM ) and a fluorescent protein tag . The drug AP21998 maintains these proteins in the monomeric state; removal of AP21998 triggers reversible aggregation ( Volchuk et al . , 2000 ) . Both the soluble and the membrane-attached version behave as bona fide anterograde cargo , moving rapidly from ER through Golgi to cell surface ( Volchuk et al . , 2000; Lavieu et al . , 2013 ) . However , the soluble and membrane-attached aggregates behave differently . Whereas soluble aggregates of FM4-hGH traverse the Golgi stack , aggregated CD8lumenal morphologically ‘staples’ Golgi cisternal membranes and remains static in the centers of Golgi cisternae ( Lavieu et al . , 2013 ) . Confocal video-microscopy of fused cells revealed that the green-labeled CD8lumenal was transported to the originally red-labeled acceptor Golgi in the presence of the AP21998 disaggregating drug ( Figure 3A , Golgi 1 ) , with an average of 12 . 9 ± 6 . 5 ( n = 4 ) fold increase at 30 min post-fusion . Monomeric soluble cargo ( hGH ) also exchanged at similar rate ( Figure 3B , Golgi 1 ) . This showed that exchange is not limited to resident Golgi proteins and suggests that exchange of anterograde cargo can occur during the process of anterograde transport , at least in fused cells . 10 . 7554/eLife . 01296 . 005Figure 3 . Inter-Golgi transport of small anterograde cargo . ( A ) HeLa cells expressing either ssGFP-FM4-CD8 or ST-RFP ( +VSV-G ) were mixed and fused . Before fusion , cells were incubated at 20°C for 2 hr in the presence of CHX ( 100 μg/ml ) and AP21998 ( 500 nM ) to trigger the release of the cargo from the ER and its accumulation in the Golgi . Both drugs were maintained during the imaging . For the study of the aggregated cargo ( staples ) , AP21988 was removed 30 min before fusion , and cells were imaged in an AP21988-depleted medium at 20°C . Graphs show quantification of both markers over time for Golgi 1 and 2 . Results are representative of three independent experiments . ( B ) HeLa cells expressing either ssDsred-FM4-hGH or GT-GFP ( +VSV-G ) were mixed and fused . As for ( A ) AP21988 was removed to trigger the formation of soluble aggregates within the Golgi . Graphs show quantification of both markers over time for Golgi 1 to 4 . Results are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 00510 . 7554/eLife . 01296 . 006Figure 3—figure supplement 1 . Inter-Golgi transport is microtubule independent . HeLa cell expressing GT-GFP ( +VSV-G ) and ST-RFP were mixed and fused . When required , nocodazole ( 2 . 10−3 μg/ml ) was added 2 hr before the fusion and was maintained during the time course of the experiment . Upper panel , fixed cell fixed 1 hr post-fusion . Lower panel , live imaging of inter-Golgi exchange in nocodazole-treated poly-karyon . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 006 The removal of AP21998 to trigger aggregation of cargo in the Golgi completely abolished the inter-Golgi transport of the membrane-bound cargo CD8lumenal ( Figure 3A , Golgi 2 ) , with a 0 . 9 ± 0 . 3 ( n = 3 ) fold increase at 30 min post-fusion . Similarly , aggregated soluble FM4-hGH was not exchanged efficiently with GT-GFP labeled acceptor Golgi ( Figure 3B , Golgi 4 ) . These results are important because they speak to the mechanism of transport between Golgi . Because re-aggregated CD8lumenal permanently resides in the Golgi membranes ( Lavieu et al . , 2013 ) and is not transported between Golgi areas , we can rule out that the traffic process somehow results from fragments of Golgi stacks ( or entire mini-stacks ) that detach by fission from one Golgi area and travel to another nucleus to join its Golgi area . The movement of Golgi mini-stacks over the required distances ( approximately 5–50 μm ) would be expected to be very slow without motor-based motility . We have extensively tested for the possible requirements of microtubules , which do not appear to be required ( Figure 3—figure supplement 1 ) . Surprisingly , since they are rapidly transported within a single Golgi stack , large-soluble cargo aggregates that concentrate in the dilated rims of cisternae ( Volchuk et al . , 2000 ) ‘are not’ transported between separated Golgi ( Figure 3B , lower panel ) . This suggests that the ( unknown ) carriers of small cargo are mobile and capable of traveling a long distance , whereas the ‘carriers’ of large aggregates of soluble cargo are less mobile . We performed all the experiments in the presence of the protein synthesis inhibitor cycloheximide to block production of new GFP-tagged cargo and to allow the ER to be drained of the vast majority of GFP-tagged cargo present before cells were fused . Nonetheless , in theory some of the cargo or resident protein transported to an exogenous Golgi could have originated from a small amount remaining in the ER , reflecting the established process of ER → Golgi rather than the novel process of Golgi → Golgi transport . To place rigorous limits on this possibility we performed photo-bleaching experiments . In one approach , we measured the fluorescence recovery of photo-bleached Golgi within single or fused cells treated with H89 , a kinase inhibitor that prevents ER export ( Aridor and Balch , 2000 ) . In single cells , we anticipate the absence of recovery of photo-bleached Golgi in presence of H89 . Within fused cells , we reasoned that H89 should abolish the fluorescence recovery of photo-bleached Golgi only if the ER is the main source of cargo . As described above , one population of cells expressing the GFP-tagged anterograde cargo and a second population containing a RFP-tagged resident protein were mixed and fused in the presence of AP21998 ( to maintain the FM-linked cargo in a monomeric state ) to allow inter-Golgi exchange of both markers . 30 min post-fusion , the fused cells were treated with H89 for 15 min before and during the FRAP of Golgi of fused cells ( Figure 4 , Golgi 2 ) . The average recovery at 30 min after-photo-bleaching was 34 ± 9 . 9% ( n = 5 ) . As a control , in the same population of cells that had not in fact fused ( Figure 4 , Golgi 1 ) no significant recovery was measured ( 7 . 2 ± 6 . 5% , [n = 4] ) , validating that ER → Golgi is abolished under H89 treatment . It is known that H89 also inhibits PKD ( Jamora et al . , 1999 ) and prevents exit from the TGN ( Muniz et al . , 1996 ) . This broad spectrum of action of H89 , which blocks ER → Golgi as well as TGN → PM , also suggests that inter-Golgi transport occurs exclusively between cis/medial/trans cisternae of separated Golgi . 10 . 7554/eLife . 01296 . 007Figure 4 . ER as a minor source of anterograde cargo during inter-Golgi transport . ( A ) FRAP performed on Golgi within fused or single cells treated with H89 ( 50 μM ) . Top panel , Golgi 1 ( green ) from a single cell and Golgi 2 ( green and red ) from fused cells 30 min post-fusion in presence of CHX and AP21988 at 20°C . Lower panel , 15 min after the H89 treatment ( t = 0 min ) , Golgi 1 and 2 were photobleached and the fluorescence recovery was monitored by confocal video-microscopy . The graph shows the fluorescence intensity over time of the GFP marker within fused cells ( di- or polykaryons ) or single cells treated with H89 . The values are the mean of three to five independent experiments . ( B ) Sar1H79G does not inhibit inter-Golgi transport of anterograde cargo . HeLa cells expressing either ssDsRed-FM4-CD8 ( +VSVG ) or YFP-SarH79 G were mixed and fused . Before fusion , cells were incubated at 20°C for 2 hr in the presence of CHX ( 100 μg/ml ) and AP21998 ( 500 nM ) to trigger the release of the cargo from the ER and its accumulation in the Golgi . Graphs show quantification of ssDsRed-FM4-CD8 over time at both donor ( Golgi 1 ) and acceptor ( Golgi 2 ) compartments . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 00710 . 7554/eLife . 01296 . 008Figure 4—figure supplement 1 . Inter-Golgi transport does not involved transit through endosomes . HeLa cells expressing ( A ) Rab7-GFP ( +VSV-G ) or ssDsRed-FM4-CD8 ( +VSV-G ) , ( B ) ssGFP-FM4-CD8 ( +VSV-G ) or Rab5A-RFP ( +VSV-G ) , ( C ) ssGFP-FM4-CD8 ( +VSV-G ) or ssDsRed-FM4-CD8 ( +VSV-G ) , were incubated at 20°C in the presence of AP21998 and CHX for 2 hr , fused and fixed after 15 min . Cells were imaged and the Pearson's correlation was measured for each fused cell using Volocity . ( D ) Box plots showing the distributions of the calculated Pearson's coefficient for experimental conditions described in A–D . Rab7-GFP and ssDsRed-Fm4-CD8 = 12 , ssGFP-FM4-CD8 and Rab5A-RFP n+10 , ssGFP-FM4-GFP and ssDsRed-FM4-CD8 n = 36 , ssGFP-FM4-CD8 and COPI n = 42 . Fused outlined in white , nuclei in blue . Red box shows zoom area in lower right of overlay . Scale bar = 10 μm , zoom scale bar = 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 008 Finally , to rule out any involvement of the ER , we tested if a Sar1 dominant mutant ( Sar1H79G , a GTP-locked mutant ) had any effect on inter-Golgi transport of disaggregated CD8 . Although this mutant inhibits ER → Golgi transport and triggers the retention of Golgi resident enzymes within the ER ( Ward et al . , 2001 ) , Golgi matrix proteins remain associated with membranes within the peri-nuclear area ( Ward et al . , 2001 ) . This area could serve as an acceptor compartment during inter-Golgi exchange in the cell–cell fusion assay . We fused YFP-Sar1H79G with cells expressing disaggregated-DsRed-CD8 ( +VSVG ) accumulated in the Golgi ( using the 20°C temperature block ) . DsRed-CD8 , initially labeling the donor Golgi ( Golgi 1 , Figure 4B ) progressively labeled the peri-nuclear area ( presumably the Golgi remnant ) of the acceptor cell that initially expressed the Sar1 mutant . The rate of the fluorescence increase in the acceptor cell was not significantly different from the one reported in the previous experiments performed in the absence of the Sar1 mutant ( Figure 3 ) , and was directly proportional to the decrease of the fluorescence at the donor Golgi , as expected if the donor Golgi is the only source of fluorescence within a freshly formed dikaryons . Note that the Sar1 mutant spreading into the dikaryons showed a peri-nuclear labeling reinforcement ( Figure 4B , arrow ) consistent with the localization ( near the cis-Golgi/ERGIC ) previously reported for Sar1 and GFP-Sar1H79 G ( Kuge et al . , 1994; Venditti et al . , 2012 ) . We concluded that the contribution of ER ( if any ) during the inter-Golgi transport assay could only be minor . Another possibility is that inter-Golgi transport of anterograde cargo is mediated via indirect transit through the endosomal pathway . To assess this possibility , we evaluated within fused cells the degree of co-localization between the anterograde cargo and early/late endosomal markers . When cell expressing small GFP-CD8lumenal were fused for 15 min with cells expressing the early ensosomal marker Rab5A-RFP , no co-localization was observed by confocal microscopy , the two signals being clearly separated even when observed at high magnification ( Figure 4—figure supplement 1A , red inset ) . Similarly , the late endosomal marker Rab7-GFP was clearly distinguishable from DsRed–CD8lumenal . ( Figure 4—figure supplement 1B ) . As a positive control , fused cells expressing DsRed–CD8 and GFP–CD8lumenal showed a strong co-localization of both signals ( Figure 4—figure supplement 1C ) . When quantified , the average Pearson’s coefficient for the positive control was >0 . 8 , whereas the coefficient for both endosomal markers samples were <0 . 4 , synonymous of poor to no co-localization . These results suggest that endosomes are not involved in inter-Golgi trafficking . In order to establish the existence of putative transport intermediates carrying markers between Golgi populations during the exchange process , we performed two types of photo-bleaching experiments in fused cells expressing disaggregated CD8lumenal accumulated in the Golgi . In the first type of experiment , photo-bleaching served to lower the cytoplasmic background during exchange allowing the presumed intermediates to be imaged in transit between two Golgi areas . To simplify the analysis , we chose only heterokaryons containing only two cells including the ( initially non-fluorescent ) acceptor Golgi . At 30 min post-fusion , the entire volume of the heterokaryon was photo-bleached , sparing only its donor Golgi . Thus , the unbleached donor Golgi ( Figure 5A , Golgi 1 ) is the only potential source of fluorescent cargo for trafficking to the acceptor Golgi . As exchange continues , the fluorescence of the donor Golgi should decrease and the fluorescence of the acceptor Golgi should correspondingly increase . During this process transport intermediates carrying the fluorescent cargo between the two Golgi , if such exist , should in principle be present in between the two Golgi areas . 10 . 7554/eLife . 01296 . 009Figure 5 . Diffusible inter-Golgi transport intermediates . ( A ) 30 min post fusion , the entire volume a dikaryon ( ER and acceptor Golgi 2 ) was photobleached except for the Golgi 1 that remains the only source of fluorescence . Small diffusing fluorescent dots ( gray arrow heads ) were observed throughout the cytoplasm of fused cells . The graph shows the intensity of both Golgi ( acceptor in blue and donor in red ) . Values are the mean of three independent experiments . ( B ) Cells expressing ssGFP-FM4-CD8 and ST-RFP ( +VSV-G ) were incubated at 20°C for 2 hr in the presence of CHX ( 100 µg/ml ) and AP21998 ( 500 nM ) to trigger the release of the cargo from the ER and its accumulation in the Golgi . Both drugs were maintained during the imaging . ( 1 ) Cells were fused and imaged 30 min post fusion . ( 2 ) Entire cell volume ( Inter-Golgi space ) was bleached , except the two Golgi and was allowed to ( 3 ) recover for 30 min . ( 4 ) Donor and acceptor Golgis were bleached ( dashed white boxes ) and allowed to ( 5 ) recover for 30 min . Red box indicates zoom area ( ssGFP-FM4-CD8 in green and ST-RFP in red ) . Images are representative of three independent experiments . Images displayed are single slices of z-stacks ( 21 one µm slices ) acquired . LUTs displayed below each grayscale image . Graph , quantification of total fluorescence intensity in region outside of the Golgis ( blue line ) and in Golgi areas ( red line ) . These values represent the mean of three independent experiments . Fluorescence intensity for each timepoint was calculated over entire z-stack . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 00910 . 7554/eLife . 01296 . 010Figure 5—figure supplement 1 . ARF1Q71L abolished fluorescence recovery in the inter-Golgi zone and at the Golgi . Cells expressing either ssGFP-FM4-CD8 or ARF1Q71L–DsRed ( +VSV-G ) were incubated at 20°C for 2 hr in the presence of CHX ( 100 µg/ml ) and AP21998 ( 500 nM ) to trigger the release of the cargo from the ER and its accumulation in the Golgi . Both drugs were maintained during the imaging . ( 1 ) Cells were fused and imaged 30 min post fusion . CD8-GFP remained at the Golgi 2 and 3 ( donor Golgi ) whereas ARF1 bound each Golgi within the polykaryon ( including Golgi 1 , the theoretical acceptor Golgi for the cargo ) . ( 2 ) Entire cell volume ( Inter-Golgi space ) was bleached , except the two Golgi and was allowed to ( 3 ) recover for 30 min . ( 4 ) Donor and acceptor Golgis were bleached ( dashed white boxes ) and allowed to ( 5 ) recover for 30 min . Graph shows quantification of total fluorescence intensity in region outside of the Golgis ( blue line ) and in Golgi areas ( red line ) compare to control ( when ARF1 mutant is absent ) . Fluorescence intensity for each timepoint was calculated over entire z-stack . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 010 30 min after the photo-bleaching , a 23 . 9 ± 6 . 4% ( n = 3 ) fluorescence recovery of the pre-bleaching fluorescence was observed at the acceptor Golgi ( consistent with the 34% previously measured in the H89-treated fused cells ) accompanied by a 24 . 8 ± 9 . 03% ( n = 3 ) decrease in fluorescence of the donor Golgi . This independently confirms that the donor Golgi is a sufficient source for inter-Golgi exchange and established the pre-conditions to look for transport intermediates . In fact , during the exchange period we readily observed numerous fluorescent “dots” present in the intervening cytoplasm ( Figure 5A , gray arrows ) . We estimate that there may be as many as ∼20 , 000 of such dots per heterokaryon over 30 min considering that we monitored ∼6 dots/frame ( with 1 frame/5 s over 30 min , in a 500 nm focal plane within HeLa cells that are ∼5 μm thick ) . These particles were not easily distinguishable above background fluorescence when the inter-Golgi assay was performed without prior photo-bleaching , probably because the background signal was is too high . To further test if fluorescently-labeled cargo present in the dots is a bona fide intermediate that contain cargo in transit and if they carry sufficient fluorescent cargo to account for the bulk of exchanged cargo , we performed a sequential ‘double FRAP’ experiment . This was performed by spinning-disk microscopy to track all the fluorescent objects within the entire volume of a dikaryon . First , 30 min post-fusion , we photo-bleached the entire volume of the fused cell ( including the ER and the diffusing fluorescent dots [Figure 5B panel 2] ) except the two Golgi , which remained the only source of fluorescence . As described above , fluorescent dots that emanated from the Golgi rapidly appeared , increased in number and spread throughout the cytoplasm , resulting in an increase of the fluorescence intensity measured in the inter-Golgi space ( Figure 5B panel 3 , Graph ) . Then , after a 30 min recovery period , a second photo-bleaching was performed on the two Golgi ( panel 4 ) , the only source of fluorescence being now exclusively provided by the freshly formed cytoplasmic dots . We reasoned that if these fluorescent dots are truly the transport intermediates , then they should be able to repopulate the photo-bleached Golgi . After 30 min recovery , the gain of fluorescence observed at the Golgi ( +56 ± 14% of the total fluorescence initially emanating from the dots before the second photo-bleaching ) was very similar to the loss ( −63 ± 14% ) of the fluorescence intensity monitored in the inter-Golgi space ( and emanating from the dots ) ( Graph , Figure 5B ) . The subtle difference ( around 7% ) between the loss ( in the Golgi inter-space ) of and the gain ( at the Golgi ) of fluorescence that suggests a loss of fluorescent material during the experiment , could be fairly attributed to moderate photo-bleaching that usually occurs during the time course of our imaging . Note that in these experiments , ST-RFP was coexpressed with GFP–CD8lumenal , and allowed for rigorous identification of the Golgi area and to measure the GFP fluorescence recovery ( Figure 5B , red squares ) . These results strongly suggest that the Golgi-derived small fluorescent dots are the diffusible transport intermediates responsible for the inter-Golgi transport . One possibility is that the small diffusible transport intermediates , which we visualize by confocal microscopy as fluorescent dots , are COPI vesicles , which are known to copiously bud from Golgi membranes and contain both anterograde and retrograde cargoes . The small GTPase ARF1 promotes vesicle budding from various donor membranes , including Golgi membranes ( Serafini et al . , 1991a; Bremser et al . , 1999 ) . To test the possible requirement of ARF1 in inter-Golgi transport , we introduced into the fused cell a dominant-interfering mutant of ARF1 ( ARF1Q71L ) that can bind but not hydrolyze GTP , thereby accumulating coated COPI vesicles ( Tanigawa et al . , 1993 ) . To do this we included a third cell population expressing the DsRed–ARF1Q71L mutant , with cells co-expressing VSV-G and GT-CFP and cells expressing GT-YFP . After this three-way fusion , confocal video-microcopy revealed that ARF1Q71L–DsRed spread effectively throughout the polykaryons and localized to exogenous Golgi membranes ( Figure 6A , Golgi 1 and 2 ) . However , inter-Golgi exchange of yellow and cyan-labeled GT no longer took place . Each marker remained within its parental Golgi even 1 hr after fusion . This suggests that the GTPase activity of ARF1 is required for inter-Golgi exchange of resident Golgi proteins . 10 . 7554/eLife . 01296 . 011Figure 6 . Role of ARF1 and ε−COP on inter-Golgi transport . ( A ) Mixed and fused HeLa cells expressing GT-CFP ( +VSV-G ) , GT-YFP ( +VSV-G ) or ARF1Q71L–DsRed , in presence of CHX . Graphs show the fluorescence intensity over time for each fluorescent marker within Golgi 1 and 2 . Results are representative of two independent experiments . ( B ) Mixed and fused HeLa cells expressing either ssGFP-FM4-CD8 ( +VSV-G ) or ARF1Q71L–DsRed in the presence of AP21988 and CHX at 20°C . The graphs show the fluorescence intensity overtime of each marker for each Golgi . Results are representative of three different experiments . ( C ) Mixed and fused WT-CHO cells or LdlF-CHO cells expressing GT-GFP or ST-RFP ( +VSV-G ) , in presence of CHX . After fusion , cells were incubated at 39°C for 15 or 60 min , fixed , and monitored by confocal microscopy . Graph shows the relative Pearson’s correlation coefficients for each cells type at each time point . Values are the mean of three independent experiments , with 10–15 polykaryons being analyzed for each condition . Pictures are representative of three experiments and illustrate cells at 1-hr post fusion . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 01110 . 7554/eLife . 01296 . 012Figure 6—figure supplement 1 . ARF1 WT does not prevent inter Golgi exchange of small anterograde cargo . Mixed and fused HeLa cells expressing either ssDsRed-FM4-CD8 ( +VSV-G ) or ARF1WT-GFP in the presence of AP21988 and CHX at 20°C . The graphs show the fluorescence intensity overtime of each marker for each Golgi . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 012 To test the possible ARF1 dependence of inter-Golgi transport of anterograde cargo , we mixed and fused two cell populations , one expressing the disaggregated GFP–CD8lumenal cargo within their Golgi , the other expressing DsRed–ARF1Q71L . Once again , the ARF1 mutant diffused throughout the polykaryon and bound to the exogenous Golgi membranes ( Figure 6B , Golgi 1 ) and prevented inter-Golgi transport of this small anterograde cargo , which remained in its original Golgi . Remarkably , when the ARF1 mutant is replaced with ARF1WT , inter-Golgi transport of the cargo is unaltered ( Figure 6—figure supplement 1 ) , demonstrating that the transport inhibition was specific to ARF1 mutant . To test if the dots ( transport intermediates ) are ARF GTPase-dependent , we performed sequential double FRAP experiments in a heterokaryon resulting from the fusion between cells containing disaggregated GFP–CD8lumenal cargo accumulated in the Golgi and cells that contained DsRed–ARF1Q71L ( Figure 5—figure supplement 1 ) . As shown above ( Figure 6 ) , GFP–CD8 remained at the donor Golgi 30 min–post fusion , whereas the ARF1 mutant decorated each Golgi of the polykaryon . After the first photo-bleaching , virtually none of the GFP-fluorescent dots that were normally observed ( Figure 5B ) could be visualized when ARF1-GTP locked mutant was present ( Figure 5—figure supplement 1 ) . Consistent with this lack of repopulation of transport intermediates in the inter-Golgi zone , the GFP-fluorescence recovery in that zone was less than 5% of the recovery measured in polykaryons lacking the ARF1 mutant ( Figure 5B and Figure 5—figure supplement 1 ) . As expected , and consistent with the absence of production of transport intermediates , no significant GFP-fluorescence recovery was observed after the second round of photo-bleaching at the Golgi area ( Figure 5—figure supplement 1 ) . Because of the lack of resolution , we do not know if in the presence of the ARF1-locked mutant , transport intermediates were actually produced but remained in the vicinity of the donor Golgi in a coated form , or if the ARF1 mutant prevented complete formation of COPI vesicles . In any case , this result suggests that the GTPase activity of ARF1 involved in the formation of COPI vesicles is required for generation of diffusible transport intermediates that are responsible for inter-Golgi exchange . To test the possible requirement for coatomer , we examined inter-Golgi exchange of Golgi resident proteins ( ST-RFP and GT-GFP ) at the restrictive temperature of 39°C within fused wild-type CHO cells ( WT ) or Ldl-F mutant cells , a well characterized mutant that is thermo-sensitive for the ε−COP subunit of the coatomer ( Guo et al . , 1994 ) . This process could not be monitored in real time at this temperature , because our microscope was not equipped with an appropriate temperature-controlled chamber . Therefore , cells were fixed 15 min and 60 min after fusion to determine if inter-Golgi exchange had occurred . To estimate the degree of colocalization of the two Golgi markers , we calculated the Pearson’s coefficient for each picture ( Figure 6C ) . At 60 min post-fusion , the average Pearson’s coefficient in fused WT cells was 0 . 69 ± 0 . 11 ( n = 30 ) , indicative of strong co-localization . In contrast , in fused mutant Ldl-F cells , the negative value of the Pearson’s coefficient ( −0 . 29 ± 0 . 09 [n = 30] ) illustrated the absence of significant co-localization , indicating that ε-COP is required for the inter-Golgi transport of resident proteins . All the previous experiments with the anterograde cargo were performed using a 20°C temperature block to retain the cargo within the Golgi . Unfortunately , this is not possible in the mutant CHO cells if we wish to retain the restriction on coatomer function to test for a requirement . When analogous experiments were performed with the CHO cells to assess the requirement of the coatomer , the 39°C temperature shift of course triggers rapid exit of the anterograde cargo from the Golgi to the plasma membrane . Therefore , we could not perform a suitable control to rigorously validate the absence of inter-Golgi transport of anterograde cargo . Our results show that inter-Golgi transport requires both ARF1 and COPI , suggesting that COPI vesicles are the transport intermediates . To further test the possibility that the putative carriers are COPI vesicles , we used STED microscopy to pinpoint their size and their content at ∼80 nm resolution ( Pellett et al . , 2011 ) . In these experiments we used a primary monoclonal antibody against β-COP and a secondary antibody coupled to the STED compatible dye , ATTO647 , to identify the putative COPI coated carrier . We fused donor cells that expressed both a tagged resident protein ( ST-RFP ) and small anterograde membrane cargo ( GFP–CD8lumenal in presence of AP21998 ) with untransfected acceptor cells . 15 min after fusion , when the markers were in mid-transit to the acceptor Golgi , we fixed the samples and performed immunofluorescence against β-COP to label COPI containing membranes . This results in labeling of the Golgi itself and the dots/putative carriers ( Figure 7 and Figure 7—figure supplement 2 ) . 10 . 7554/eLife . 01296 . 013Figure 7 . Inter-Golgi transport intermediates are compatible with COPI vesicles . ( A ) In vitro prepared COPI vesicles labeled with Alexa488 were attached to glass coverslips . Immunofluorescence against COPI was performed , and samples were imaged by both confocal and STED microscopy . The size of COPI vesicles was determined by STED microscopy using a custom Matlab routine by fitting to a 2D Lorenztian function and confocal images were fit to a 2D Gaussian function . Graph , size distribution of 64 in vitro COPI vesicles fit with a 2D Lorentzian function . The mean is 114 nm . ( B ) HeLa cells expressing ssGFP-FM4-CD8 , ST-RFP and VSV-G were incubated at 20°C in the presence of AP21998 and CHX for 2 hr , fused , and fixed after 30 min . Immunofluorescence was performed against β-COP to label all the Golgi related structures . COPI like transport intermediates imaged by STED microscopy were identified based on size and the contents were evaluated . Numbered boxes illustrating types of cargo containing COPI like intermediates . Red squares , different objects at low and high magnification . The sizes of the intermediates were determined using a 2D Lorentzian function . Graph ( C ) shows the size distribution of 179 carriers fit with a 2D Lorentzian function , The mean is 128 nm . Graph ( D ) shows the distribution of COPI spots positive for either ssGFP-FM4-CD8 cargo or ST-RFP , or both ( gray columns ) . As a positive control we used cells co-expressing ssRFP-FM4-CD8 instead of ST-RFP ( white columns ) . Control: 48 cells , 138 spots analyzed; Experimental: 31 cells , 179 spots analyzed . Results are representative of two independent experiments , each one being duplicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 01310 . 7554/eLife . 01296 . 014Figure 7—figure supplement 1 . Inter-Golgi carriers are COPI positive and COPII negative . ( A ) Cells expressing CD8-GFP-FM4 , ST-RFP and VSV-G were incubated at 20°C in the presence of AP21998 and CHX for 2 hr , fused , and fixed after 30 min . Immunofluorescence was performed against β-COP to label all the Golgi related structures . The CD8-GFP-FM4/β-COP and ST-RFP/β-COP co-localizations are highlighted with a white mask generated with ImageJ . Red squares , different objects at low and high magnification . Results are representative of three different experiments . ( B ) Cells expressing CD8-GFP-FM4 , ST RFP and VSV-G were incubated at 20°C in presence of AP21998 for 2 hr , fused and fixed 30 min after . Immunofluorescence was performed against sec31 to label COPII related structures ( COPII vesicle and ER exit sites ) . The CD8-GFP-FM4/sec31 and ST-RFP/sec31 co-localizations are highlighted with a white mask generated with ImageJ . Due to the resolution limit , the ER exit sites that are at the vicinity of the Golgi appear as co-localizing with the Golgi . However , virtually none of the small particles carrying the cargoes , which are distant from the donor Golgi , were positive for sec31 . The red squares illustrated the different objects at low and high magnification . Results are representative of two different experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 01410 . 7554/eLife . 01296 . 015Figure 7—figure supplement 2 . Discriminating tubules from vesicles with STED . Vesicles and tubules with different dimensions simulated in a set of different 3D orientations to represent an isotropic distribution ( supplementary method section ) . Pictures illustrate the object according to their different 3D orientation . For both in vitro COPI vesicles and intracellular COPI objects ( imaged in Figure 7 ) , the size was determined by STED microscopy using a custom Matlab routine by fitting to a 2D Lorenztian function and the radial symmetry ( major/minus axis ratio ) was calculated to assess the degree of symmetry of the objects . The simulated objects were analyzed with the same custom-written MATLAB routine used for the experimental data . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 015 As a standard for comparison with the fluorescent dots found between Golgi , we measured the size of in vitro COPI vesicles by STED microscopy . We determined their mean diameter ( Figure 7A ) to be 114 ± 11 nm n = 64 , consistent with the outer diameter reported from electron microscopy ( Orci et al . , 1986 ) . We then measured the size of the β-COP positive transport intermediates ( dots ) within fused cells by STED microscopy ( Figure 7B ) . The mean size of the dots containing COPI was 126 ± 15 nm , n = 181 , and approximately the same size as the in vitro vesicles . Note that the vast majority ( around 90% ) of the dots containing either the anterograde cargo or the Golgi resident enzymes were negative for the COPI staining , consistent with the predictions of the vesicular model . Interestingly these uncoated structures showed a slightly reduced diameter 105 ± 13 nm , n=60 , which may reflect the absence of coatomer . Analysis of the shape ( defined as the length of the major axis/length of the minor axis ) of the transport intermediates and the in vitro carriers revealed similar radial symmetry ( Figure 7—figure supplement 2 ) . However , according to our 3D modeling , both the in vivo putative transport intermediates and the in vitro COPI vesicles were not as symmetrical as modeled vesicles ( Figure 7—figure supplement 2 ) . This could be attributed to variation generated by the length and the spatial orientation of the antibodies ( both primary and secondary ) that bound heterogeneously to the coatomer . This subtle asymmetry fit with a recent high-resolution 3D study of in vitro COPI vesicles , which showed variable shapes of the vesicles , as well as a non-homogenous surface distribution of the coatomer ( Faini et al . , 2012 ) . Finally , our shape analysis and 3D modeling clearly establish that the COPI positive transport intermediates are not long tubules ( 30 nm × 300 nm ) , which show a significantly higher degree of asymmetry . However , it is impossible to rule out from such modeling that the carrier is a hypothetical , partially COPI coated small tubule . With a resolution of 80 nm , a 30 nm diameter tubule that is 150 nm long is indistinguishable from a 110 nm vesicle ( as confirmed by simulations of STED images of vesicles and tubules; Figure 7—figure supplement 2 ) . Are the anterograde cargo and the resident protein cargo present in the same or different COPI vesicles ? Quantification using size filtering based on the in vitro COPI vesicles results ( size < 150 nm ) showed the following heterogeneous distribution when GFP–CD8lumenal was co-expressed with ST-RFP ( Figure 7B ) : 66 ± 20% of the COPI-positive objects carried the anterograde cargo alone , 20 ± 11% carried the retrograde cargo alone and 15 ± 8 % carried both ( Figure 7B ) , consistent with the three populations observed by confocal microscopy ( Figure 7—figure supplement 1 ) . When the same anterograde cargoes tagged with either green or red fluorophores were co-expressed , 87 ± 3% of the objects carried both cargoes , as expected for a co-localization positive control . None of the dots ( again , containing one or both markers ) could be stained with an antibody to the COPII coat subunit SEC31 ( Figure 7—figure supplement 1 ) , yet again ruling out again any involvement of ER export . Our inter-Golgi transport assay has two main limitations . First , all the experiments presented in this study were performed at 20°C , and one could argue that this non-physiological temperature could create missorting because of accumulation of over-expressed cargo at the TGN . This so-called 20°C block has never been reported to prevent intra-Golgi trafficking ( neither anterograde or retrograde ) . Instead it has been shown that 20°C slows-down the trafficking , and results in retaining a large portion of the anterograde cargo within the trans-Golgi , with a significant amount remaining in the upstream cisternae that could serve as a back-up for inter-transport ( Van Deurs et al . , 1988 ) . However , to rule out that our results were reflecting non-specific sorting of the anterograde cargo at the TGN , we co-expressed the anterograde cargo with a dominant interfering mutant of the protein-Kinase D ( PKD ( DI ) ) , known to block the TGN exit of anterograde cargo ( Baron and Malhotra , 2002; Figure 8C ) . Cells expressing CD8-GFP and PKD ( DI ) were pre-incubated at 37°C for 15 min in the presence of AP91988 and CHX , to allow for Golgi targeting of the cargo . As expected , we did not observe any leak to the PM . These cells were fused with other cells expressing DsRed-tagged resident enzyme and PKD ( DI ) , and the inter-Golgi transport was recorded at 32°C for 1 hr . The anterograde cargo was efficiently transported to the acceptor Golgi ( Figure 8A , Golgi 1 and Graph ) that originally contained only the resident enzyme . Note that 1 hr post-fusion , cells were fixed and processed for immunofluorescence against GST-PKD ( DI ) to rigorously assess for the presence of the PKD mutant within the very same polykarion ( Figure 8B ) . This result suggests that inter-Golgi transport is not due to a missorting of over-expressed cargo at the TGN . 10 . 7554/eLife . 01296 . 016Figure 8 . Inter-Golgi exchange of anterograde cargo at 37°C . ( A ) HeLa cells expressing ss-GFP-FM4-CD8 and GST-PKD ( DI ) were mixed and fused with cells expressing ST-RFP , GST-PKD ( DI ) and VSV-G . Prior to fusion , cells were incubated at 37°C for 15 min in the presence of the disaggregating drug and CHX . Cells were monitored by video confocal microscopy at 32°C in the presence of the disaggregating drug and CHX . Graphs show quantification of both markers over time for Golgi 1 . Results are representative of two independent experiments . ( B ) 1 hr post-fusion , cells were fixed and prepared for immunofluorescence against GST to assess the presence of the PKD ( DI ) . Note that the confocal micrograph showed the very same field of cells ( shown in A ) after fixation . ( C ) PKD ( DI ) inhibits plasma membrane targeting of ss-GFP-FM4-CD8 . HeLa cells expressing ss-GFP-FM4-CD8 alone ( upper panel ) or with GST-PKD ( DI ) ( lower panel ) were incubated for 1 hr at 37°C in the presence of the disaggregating drug . Non-permeabilized cells were incubated on ice and incubated with an anti-GFP antibody ( detected with a secondary antibody labeled with Alexa-546 ) to assess for cell surface exposure of ss-GFP-FM4-CD8 . Then , cells were fixed , permeabilized and processed for immunofluorescecne against GST ( using a secondary antibody labeled with Atto-647 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01296 . 016 The second limitation is the over-expression of the cargoes . It is possible that we artificially increased the probability of loading the highly concentrated cargo within the carriers . For instance , the excess of anterograde cargo localized in the trans-Golgi may now be transported to the cis-Golgi of the acceptor Golgi through a retrograde pathway . On the other hand , over-expression of the resident enzymes may alter their proper localization , resulting now in an anterograde transport of these resident proteins when they are in large excess . We cannot rule out that these phenomena happened to some extent in our assay . However , if these assumptions are correct , the two types of cargo should always be carried within the same transport intermediates when the cargoes are released from the same donor Golgi . Our results clearly establish that this in not the case ( Figure 7B ) . Another interpretation would be to consider that the inter-Golgi transport intermediates would be COPI vesicles , which contain exclusively anterograde cargo in ∼60% of the cases . ∼20% would be COPI vesicles containing only Golgi resident enzymes that are retrograde-directed , needed in any model to maintain the steady-state distribution of resident Golgi enzymes in a dynamic equilibrium . It is tempting to suggest that the ∼20% of COPI vesicles containing both anterograde cargo and resident enzymes would be anterograde-directed , contributing to the steady-state distribution of resident enzymes by forward flow . As is often the case , the truth may lie in between these two most extreme interpretations .
This report describes a study specifically intended to test the predictions of the vesicular transport model . Our reasoning was that only mobile carriers , such as COPI vesicles , should be able to travel the long distance between well-separated Golgi of fused cells . Our results clearly establish that only small cargoes ( anterograde or resident proteins ) are exchanged , showing that inter-Golgi carriers are capable of size-filtration , and that small cargo transport must utilize a distributive mechanism different from the mechanism used by large soluble aggregates . More than half of the carriers exclusively contained anterograde cargo , suggesting sub-populations of COPI vesicles as previously documented by immuno-EM ( Orci et al . , 1997 ) . Because the imaged carriers contained coatomer , are indistinguishable in size from authentic COPI vesicles , and require both coatomer and ARF1 to function , the simplest interpretation is that the bulk of inter-Golgi transport of small cargo is mediated by COPI vesicles . Certainly , the double photo-bleaching experiments show that the bulk of the cargo flux between Golgi is mediated by fluorescent ‘dots’ that function as carriers . However , we cannot strictly rule out that the genetic effects of ARF1 and ε−COP mutations on inter-Golgi transport are somehow an indirect coincidence or that the carrier is some kind of very short tubule that is partially coated and capable of fission ( at one Golgi ) and fusion ( at another ) . Correlative electron microscope imaging of the fluorescent dots , including immunolabeling of coatomer , would theoretically be needed to address this distinction , which is probably semantic in any case since many COPI vesicles have uncoated portions ( Faini et al . , 2012 ) . Because the dots/COPI vesicles are highly dilute in cytoplasm , the rarity of finding such vesicles in single EM sections together with the inherent inefficiency of immunolabeling at the EM level , makes such an effort quixotic at best . Although we cannot completely rule out the possibility that exogenous glycosyltransferases exceed the capacity of the Golgi retention machinery and thus begin to behave like soluble cargoes , the relative abundance of COPI containing carriers harboring Golgi resident glycosyltransferase ( ∼40% of the total ) suggests that these proteins are engaged in a dynamic equilibrium of anterograde and retrograde flow to maintain their differential steady-state cis-trans distributions . Their presence in transport vesicles is a required feature of the mobile cisterna model , as incorporated in the concept of cisternal ‘maturation’ ( Glick and Malhotra , 1998 ) , but may also be important for establishing and maintaining their cis-trans distributions by bi-directional transport ( Pelham and Rothman , 2000 ) among static cisternae . We note that previous reports on the prevalence of glycosyl-transferases in COPI vesicles have yielded variable results ( Orci et al . , 2000; Martinez-Menarguez et al . , 2001; Cosson et al . , 2002 ) . The present results naturally suggest that transport of small cargo is bi-directional and mediated by COPI-coated vesicles . The data suggest that COPI vesicles traffic both secretory cargo and steady-state Golgi resident enzymes among stacked cisternae that are stationary over many hours in non-dividing cells . Further , they suggest that large soluble cargo are transported across the Golgi by a specialized mechanism that involves carriers which remain closely linked to the stack as the progress across it . While these inferences are as strong as current imaging technology permits , super-resolution methods are rapidly advancing and it will be important to further test these models as it becomes possible to visualize the content and movements of individual carriers in real time in living cells .
GT-GFP and ST-RFP were a gift from J Roher ( Schaub et al . , 2006 ) . The VSV-G-encoding vector was provided by JK Rose ( Florkiewicz and Rose , 1984 ) . ss-GFP-FM4-CD8 ( CD8lumenal ) was generated by sequential insertion of GFP and CD8 encoding sequences into a pC4-FM4 backbone vector ( ARIAD ) , using XbaI/SpeI compatibility , and BamH1 restriction site . CD8-GFP was amplified from pC4-CD8-GFP as previously described ( Lavieu et al . , 2010 ) . ss-DsRed-FM4-hGH was created by adding DsRed into pC4S1-FM4-FCS-hGH ( ARIAD ) using XbaI/SpeI compatibility . ARF1Q71L was cleaved from ARF1Q71L–GFP , a gift from G Romero ( Vasudevan et al . , 1998 ) and inserted into a mDsRed vector ( Clontech ) . GST tagged PKD- ( DI ) was a gift from Vivek Malhotra . GT-CFP and GT-YFP and YFP-Sar1H79 G ( Quintero et al . , 2010 ) were a gift from C Giraudo . RAb5A-RFP and RAb7-GFP were a gift from P De Camilli . HeLa cells were maintained at 37°C in 5% CO2 in DMEM ( Gibco , Grand Island , NY ) supplemented with 10% FBS ( Gibco ) . CHO cells were maintained at 34°C in 5% CO2 in F-12 Glutamax ( Gibco ) supplemented with 10% FBS . Cells were transfected using lipofectamine 2000 ( Invitrogen , Grand Island , NY ) as recommended by the manufacturer . When cells were co-transfected ( VSV-G/other fluorescent-tagged protein ) , we used a 3/1 ratio . Cells were plated at 6 × 104 ( HeLa ) or 1 × 105 ( CHO ) cells/well in 24-well plates . Cells were transfected with the plasmids of interest for 6 hr . Transfection media was removed , and cells were washed with complete culture media prior to being detached ( trypsin 0 . 05% 2 min ) and mixed overnight in glass bottom microwell dishes ( HeLa cells ) , or on coverslips coated with collagen type I ( CHO cells ) . 2 hr before fusion , HeLa cells were incubated at 20°C in HBSS ( Gibco ) supplemented with 10% FBS and 100 μg/ml CHX ( Sigma , St . Louis , MO ) ( imaging media ) . CHO cells were kept in F12 media containing 10% FBS and 100 μg/ml CHX . For live imaging , HeLa cells were incubated on the stage of the microscope for 1 min in 37°C pre-warmed fusion buffer ( 10 mM Na2HPO4 , 10 mM NaH2PO4 , 150 mM NaCl , 10 mM 2- ( Nmorpholino ) ethanesulfonic acid [MES] , 10 mM N-2-hydroxyethylpiperazine-N9-2-ethanesulfonic acid [HEPES] , pH 5 ) . Subsequently , the fusion buffer was replaced by the imaging media . CHO cells were incubated 1 min at 37°C in the fusion buffer before being incubated at 39°C with pre-warmed complete F12 media . Unless mentioned otherwise , all the experiments were performed in the presence of 100 μg/ml CHX at 20°C . HeLa cells expressing CD8-GFP-FM4 , ST-RFP , and VSV-G were fused with untransfected cells . Cells were fixed with PFA 4% , permeabilized with 0 . 1% Triton X100 , and incubated with an anti-β−COP antibody ( 1/1000 , EAGE clone , rabbit ) or an anti-hsec31 antibody ( 1/1000 , rabbit; Shugrue et al . , 1999 ) , and an anti-rabbit Alexa-633 conjugated secondary antibody ( 1/2000; Invitrogen ) . GST antibody was purchase from Abcam , Cambridge , MA . Cells were then processed for confocal microscopy imaging , which was performed in multi-tracking mode on either a Zeiss LSM510 or a Zeiss LSM510 META . Images were analyzed using Zeiss LSM510 software or using ImageJ ( co-localization finder plugin ) . Cells were fixed with 4% paraformaldehyde for 15 min at room temperature , permeabilized for 3 min with 0 . 3% NP40 and 0 . 05% Triton-X , blocked with 10% normal goat serum for 1 hr at room temperature and incubated with primary antibodies overnight at 4°C . Cells were incubated with secondary antibodies for 1 hr at room temperature and mounted in mowiol . All antibodies were used at a 1:1000 dilution: α-COPI , CM1A10 , α-RFP rabbit polyclonal ( Invitrogen ) ; α-mouse ATTO647N ( Active Motif , Carlsbad , CA ) and α-rabbit Alexa543 ( Invitrogen ) . STED images were obtained using a commercial Leica TCS STED microscope ( Pellett et al . , 2011 ) . The size of in vitro COPI vesicles were measured using a custom MATLAB routine by fitting STED images to a 2D Lorenztian function and confocal data to a 2D Gaussian function . For each channel , 100 background regions of interest were randomly selected A normal distribution of the brightest pixel in each ROI was created and the mean and standard deviation was calculated . A thresholding value of the mean plus one standard deviation was used as a cutoff . Using Volocity , COPI positive transport intermediates were identified based on the size data from in vitro generated COPI vesicles . Intermediates identified were evaluated for the presence of GFP or RFP signal above the calculated thresholding value . Only intermediates positive for another marker were included in the STED image simulations . We simulated vesicles of 110-nm diameter and tubules of 30 nm diameter and 80 , 110 , 150 and 300 nm lengths on a three-dimensional grid of 2003 cubic voxels of 5 nm size with the software Imspector ( written by Dr Andreas Schoenle , Max Planck Institute for Biophysical Chemistry , Goettingen ) . All structures were assumed to be surface-labeled , that is to carry signal only on a shell defined by the delimiters of the structures ±5 nm . The tubules were simulated in a set of different orientations to represent an isotropic distribution . Given that the size of the organelles is significantly smaller than the axial resolution of our STED microscope ( maximum 300 nm vs ∼700 nm ) , the rotated structures were projected into the focal plane before convolving them with a radially symmetric Lorentzian-shaped STED PSF with 80 nm full-width at half maximum which matches to the experimental performance of the used STED microscope . Poisson-distributed background and signal shot noise was applied and adjusted to experimentally observed signal-to-noise levels creating 150 structures of each shape . This data was analyzed with the same custom-written MATLAB routine used for the experimental data analysis . | All eukaryotic cells contain an organelle called the Golgi apparatus , which consists of a series of four to six flattened structures called cisternae . Proteins that are intended for secretion from the cell , or proteins that go on to become part of the cell membrane , must pass through the Golgi , where they undergo modifications that ensure they are targeted to the correct place . There are two main models for how proteins are transported from the entry side of the Golgi , known as the cis face , to the exit side ( trans face ) , through a process known as anterograde transport . One possibility is that the cargo protein matures within a single cisterna , which gradually moves from the cis to the trans face without the protein ever leaving it . Alternatively , the cisternae may remain fixed in position , while individual proteins are carried between them by specialized transport vesicles called COPI vesicles . Now , Pellett et al . have used modern molecular biology techniques to revisit this question , more than 25 years after members of the same group first obtained evidence suggesting the involvement of COPI vesicles . To do this , they labelled the proteins that reside within the Golgi of one cell green , and those within the Golgi of another cell , red . They then fused the two cells together , and traced the movement of labelled proteins between the two organelles . Proteins that are known to undergo anterograde transport were also transported between the two Golgi , whereas large protein aggregates were not . Super-resolution microscopy revealed that the transported proteins were carried in vesicles the size of COPI vesicles and surrounded by a coat protein that resembles COPI . Moreover , transport involved the adaptor protein ARF , which helps to load cargo into COPI vesicles . By providing evidence that Golgi resident proteins and proteins that normally undergo anterograde transport can be carried by COPI vesicles between two physically separate Golgi , Pellett et al . increase the weight of evidence that COPI vesicles may also be responsible for both retrograde and anterograde transport within the Golgi itself . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2013 | Inter-Golgi transport mediated by COPI-containing vesicles carrying small cargoes |
Identification of the host genetic factors that contribute to variation in vaccine responsiveness may uncover important mechanisms affecting vaccine efficacy . We carried out an integrative , longitudinal study combining genetic , transcriptional , and immunologic data in humans given seasonal influenza vaccine . We identified 20 genes exhibiting a transcriptional response to vaccination , significant genotype effects on gene expression , and correlation between the transcriptional and antibody responses . The results show that variation at the level of genes involved in membrane trafficking and antigen processing significantly influences the human response to influenza vaccination . More broadly , we demonstrate that an integrative study design is an efficient alternative to existing methods for the identification of genes involved in complex traits .
Influenza remains one of the major threats to human health worldwide and is responsible for an estimated 250 , 000–500 , 000 deaths each year ( World Health Organization , 2009 ) . Attempts at immunization pre-dated the isolation of the virus from humans in 1933 ( Smith et al . , 1933 ) and vaccination remains the cornerstone of prevention strategies . Since 1977 , strains of influenza A ( H3N2 ) , influenza A ( H1N1 ) , and influenza B have been responsible for the majority of documented human infections and trivalent vaccines are updated annually to contain the circulating strains . Animal models have demonstrated that immune responses and susceptibility to influenza infection can be strongly influenced by host genetic factors ( Trammell and Toth , 2008; Srivastava et al . , 2009 ) . As with viral infection , variability in the immune response to vaccination is likely to be influenced by genotype . Accordingly , twin and sibling studies have shown heritability estimates as high as 45% for a varicella vaccine ( Klein et al . , 2007 ) and 90% for a measles vaccine ( Tan et al . , 2001 ) . Studies investigating influenza vaccine immunogenicity in humans have consistently shown large inter-individual variability , but the genetic contribution to this variability remains poorly understood . Gene expression is strongly controlled by common genetic variants ( Morley et al . , 2004; Stranger et al . , 2007 ) with both broad ( Bullaughey et al . , 2009 ) and tissue-specific effects ( Innocenti et al . , 2011; Rotival et al . , 2011 ) , referred to as expression quantitative trait loci ( eQTL ) . Moreover , genome-wide association studies have identified hundreds of variants associated with human disease risk that are also eQTL , implying that the mechanism by which they influence risk involves variation in transcriptional responses ( Emilsson et al . , 2008; Cookson et al . , 2009; Naukkarinen et al . , 2010; Nicolae et al . , 2010; Rotival et al . , 2011; Barreiro et al . , 2012 ) Finally , integrative genomic studies in model organisms ( Schadt et al . , 2005; Amit et al . , 2009 ) have demonstrated that the combination of genetic and transcriptional information can allow direct tests of causal mechanisms in controlled experiments . We hypothesized that integrating genome-wide genotype data with serial measurements of the transcriptional and humoral responses to an influenza vaccine in a clinical study could be used to identify loci that influence vaccine responsiveness and subsequent immunity to influenza in humans . We immunized an ethnically homogeneous group of 119 healthy adult male volunteers with licensed trivalent influenza vaccine . DNA was obtained from peripheral blood and genome-wide SNP genotyping was performed . We also measured global transcript abundance in peripheral blood RNA specimens before and at three time points ( days 1 , 3 , and 14 ) after vaccination . Type-specific antibody measurements ( H1N1 , H3N2 , and FluB ) were made in serum samples before and at two time points ( days 14 and 28 ) after vaccination . An identical study was then carried out with an independent validation cohort of 128 ethnically homogeneous healthy adult female volunteers . This experimental design allowed us to search for loci that show evidence of a transcriptional response to vaccination , genetic regulation of gene expression ( cis-acting eQTL ) , and correlation between gene expression and the magnitude of the antibody response .
We performed mixed model regression analysis with SNPs located in 1-Mb intervals around each expression reporter sequence . We began by identifying SNP-transcript pairs with both significant evidence of a cis-acting eQTL and significant changes in gene expression in response to vaccination . Thresholds for local significance were initially explored , since only SNPs flanking each reporter sequence were tested for cis association . In the discovery cohort , 3229 SNP-transcript pairs , corresponding to 408 unique genes , exhibited significant genotype-expression association ( genotype effect p<1 × 10−4 ) and concomitant evidence of a transcriptional response to the vaccine ( day effect p<0 . 01 ) . Of these , 2606 SNP-transcript pairs , corresponding to 256 genes , were validated in the independent cohort of female volunteers ( genotype effect p<0 . 05 and day effect p<0 . 01 ) . When more stringent thresholds were applied , 756 SNP-transcript pairs , corresponding to 114 unique genes , exhibited significant genotype-expression association ( genotype effect p<5 × 10−8 ) and concomitant evidence of a transcriptional response to the vaccine ( day effect p<0 . 01 ) in the discovery cohort . Of these , 654 SNP-transcript pairs , corresponding to 93 genes , were validated in the second cohort ( genotype effect p<0 . 05 and day effect p<0 . 01 ) . A majority of these ( 467 SNP-transcript pairs , corresponding to 78 unique genes ) would pass equally stringent thresholds in both cohorts ( genotype effect p<5 × 10−8 , day effect p<0 . 01 ) . A Manhattan plot of these results is presented in Figure 1 . Data for the individual SNP-transcript pairs that passed equally stringent thresholds in both cohorts , including results of significance testing and gene identifiers , are provided in Table 1 via the Interactive Results Tool ( which is also available to download from Zenodo and shown within Supplementary file 1 ) . 10 . 7554/eLife . 00299 . 003Figure 1 . Multiple genes show both a transcriptional response to the vaccine and evidence of genetic regulation of gene expression ( cis-acting eQTL ) in both cohorts . Manhattan plots of the genotype-expression—log10 p-values across the genome for the discovery ( inner circle ) and validation ( outer circle ) cohorts . Each dot represents a SNP-transcript pair . Red dots indicate SNP-transcript pairs for which there is evidence of significant genotype-expression association ( genotype p<5 × 10−8 ) and evidence of a transcriptional response to the vaccine ( day effect p<0 . 05 ) . The 78 genes that showed both properties in the two cohorts are shown in the outer margin . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 003 We hypothesized that , at some loci , the magnitude of the genetic effect could be different before and at different time points after vaccination . This type of effect , which would not be observed in a cross-sectional study design , could be directly examined with our serial expression data . We analyzed the additive effect of genotype on expression at each day in the study . Using a cis-effect significance threshold of p<1 × 10−4 in the discovery cohort and p<0 . 05 in the validation cohort , this analysis identified 5155 validated eQTL SNP-transcript pairs ( 3011 at baseline and 3417 , 2496 , and 3043 at days 1 , 3 , and 14 , respectively ) . These SNP-transcript pairs correspond to 543 unique genes . We then identified the SNP-transcript pairs in which the expression variance explained was most strongly increased after vaccination ( highest change in genetic variance explained , which we termed delta-Rg2 ) . This analysis revealed multiple loci at which the genetic effect was either enhanced or only apparent after the experimental perturbation . An example is presented in Figure 2A , which displays local Manhattan plots for the NECAB2 locus before and 3 days after vaccination in both cohorts . 10 . 7554/eLife . 00299 . 004Figure 2 . At some loci , the magnitude of the genetic effect changes after the experimental perturbation . ( A ) A specific example of this phenomenon: local Manhattan plots for the gene NECAB2 before and on day 3 after vaccination in each of the two cohorts , showing an increase in the magnitude of the genotype effect ( R2g ) after the experimental perturbation . ( B ) An increase in R2g after the experimental perturbation is a general feature of the SNP-transcript pairs that show a strong cis-eQTLs and a transcriptional response to vaccination ( left ) . The within-genotype variance is unchanged ( MSE , center ) , while the strength of the genotype effect on expression ( slope of the additive association; β , right ) increases , suggesting that the latter is the main driver for the observed increase in the genetic effect after vaccination . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 004 Theoretically , the observed temporal changes in the estimated genotype effect after vaccination could be driven by an increase in the effect size , a relative decrease in the variability within genotype strata , or both . We analyzed all SNP-transcript pairs for loci at which we observed both a strong cis-acting eQTL and a transcriptional response to vaccination , calculating the relative magnitude of slope and within-genotype variance between the pre-vaccination and maximal Rg2 time points . Figure 2B shows that an increase in the strength of the genotype effect ( slope of the additive association ) was the main driver for the observed change in Rg2 , and that this amplitude change was a general feature of the loci in which we observed both a strong cis-acting eQTL and a transcriptional response to the vaccine stimulus . The delta-Rg2 values were consistent between the cohorts when evaluated by Spearman’s rank correlation analysis using all SNP-transcript pairs ( Cor = 0 . 25 , p<2 × 10−16 ) . To select a conservative set of candidate loci based on this property for further analysis , we identified the SNP-transcript pairs that were in the top 1% of the delta-Rg2 distribution and also showed evidence of a strong cis-acting eQTL ( genotype effect p<5 × 10−8 ) , in both cohorts . Data for the resulting set of 146 SNP-transcript pairs , including Rg2 values , are provided in Table 2 via the Interactive Results Tool ( which is also available to download from Zenodo and shown within Supplementary file 1 ) . Of the 78 genes that had the strongest validated evidence of a genotype effect and a transcriptional response to the vaccine , 14 were also in the list of 34 genes with the strongest evidence of an increase in the magnitude of the genetic effect after vaccination . Content analysis on the union of the two sets ( 98 genes ) showed significant enrichment for genes involved in antigen processing and presentation , cytotoxic T-lymphocyte-mediated apoptosis of target cells , dendritic cell maturation and function , and membrane trafficking ( Figure 3 ) . 10 . 7554/eLife . 00299 . 005Figure 3 . Content analysis shows enrichment for genes involved in membrane trafficking , antigen processing , and antigen presentation . Barplots show categories with significant overrepresentation in the list of 98 genes with a strong cis-eQTL and a response to vaccination expressed as either a transcriptional response or a change in the genetic effect in both cohorts . The negative log ( p-value ) is plotted on the x-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 005 We and others have shown that for some transcripts there is significant correlation between the magnitude of the transcriptional and antibody responses to the vaccine stimulus ( Zhu et al . , 2010; Bucasas et al . , 2011; Nakaya et al . , 2011 ) In a combined analysis of the two cohorts in the present study , 301 transcripts were found to correlate with the magnitude of the antibody response ( Figure 4 ) . Additional details of these 301 transcripts , including correlation coefficients and days of maximum correlation , are provided in Table 3 via the Interactive Results Tool ( which is also available to download from Zenodo and shown within Supplementary file 1 ) . We imposed an additional selection threshold based on this correlation , and identified 20 genes that show evidence of significant genotype-expression association ( genotype effect p<5 × 10−8 ) , a significant correlation between the transcriptional and antibody responses ( expression-antibody effect p<0 . 05 ) , and either a transcriptional response to the vaccine ( day effect p<0 . 01 ) or evidence of a change in the magnitude of the genetic effect after vaccination ( top 1% of the delta-Rg2 distribution ) in the two independent cohorts . These loci have the strongest evidence of genetic variation influencing the immune response to the vaccine , and include TAP2 , SNX29 , FGD2 , NAPSA , NAPSB , GM2A , C1orf85 , JUP , FBLN5 , CHST13 , DIP2A , PAM , D4S234E , C3AR1 , HERC2 , LST1 , LRRC37A4 , OAS1 , RPL14 , and DYNLT1 . Remarkably , seven of these encode proteins involved in intracellular antigen transport and processing ( Figure 5 ) . 10 . 7554/eLife . 00299 . 006Figure 4 . Gene expression at specific loci correlates with the antibody response to vaccination . ( A ) Examples of positive ( DYNLT1 ) and negative ( ANKRD33 ) correlation between gene expression on day 1 and the magnitude of the antibody response to the vaccine . Data points and regression lines in the scatterplots display the results for the discovery ( blue ) and validation ( magenta ) cohorts . ( B ) A total of 301 genes showed evidence of significant correlation between gene expression and the antibody response to the vaccine in both cohorts . Of these , 281 showed evidence of positive correlation and 83 of negative correlation . Each individual is represented by a column in the heatmaps . The top heatmaps display the magnitude of the antibody response ( titer response index ) . The bottom heatmaps display the deviations around the expression mean for each gene . Individual gene identifiers and correlation coefficients are presented in the Interactive Results Tool . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 00610 . 7554/eLife . 00299 . 007Figure 5 . Genetic variation in intracellular antigen transport and processing influences the human immune response to influenza vaccination . 20 genes show evidence of a transcriptional response to vaccination , significant genotype effects on gene expression , and correlation between the transcriptional and antibody responses . Remarkably , seven of these are involved in intracellular antigen transport , antigen processing , and antigen presentation . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 007 We determined genetic associations to the antibody response using 137 eQTL SNPs from these 20 loci . The quantile-quantile plot from the association tests performed on these SNPs shows marked deviation from the empirical null distribution for QTL associations ( Figure 6 ) , supporting the idea that these loci are enriched for true genetic associations . 10 . 7554/eLife . 00299 . 008Figure 6 . SNPs at the 20 loci identified show evidence of association with the antibody response to the vaccine . 137 SNP-transcript pairs with evidence of a strong cis-eQTL , a dynamic response to the vaccine ( a change in transcript abundance or in the magnitude of the genetic effect ) , and correlation between the transcriptional and antibody responses were selected ( result SNPs , in red ) . The empirical quantile-quantile plot of the result SNPs shows significant deviation from the empirical distribution of the entire data set ( background SNPs , in blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 008 We explored three types of associations in our work: genotype to gene expression ( eQTL ) , gene expression to antibody titer , and genotype to antibody titer ( QTL ) . We now considered alternative models for the relationships between these distinct types of association ( Figure 7A ) , and we evaluated our data to determine which of these alternatives appears most consistent with our observations . The alternative models considered were: ( i ) genotype association with gene expression is independent of genotype association or trends of association with antibody response ( independent model ) ; ( ii ) genotype association or trends of association with antibody response are mediated by gene expression patterns that are strongly correlated with genotype ( causal model ) ; and ( iii ) genotype associations to antibody response are not mediated by expression , but instead gene expression patterns are a response to the antibody trait or its early correlates ( reactive model ) . To perform a comparative analysis of these alternatives we extended the framework for causal modeling ( Pearl , 2010 ) in eQTL data recently developed by others ( Millstein et al . , 2009 ) and applied the method to our time-course gene expression study . We used the 137 eQTL SNP-transcript pairs from the 20 loci with the strongest evidence of genetic variation influencing the immune response to the vaccine , as described above . We found that the patterns in the data trend toward the causal model compared to the reactive model ( Figure 7B ) , but a power analysis based on the distribution of the empirical effect sizes of our observed associations also indicates that our sample size is too modest to support definitive conclusions ( Figure 7C ) . 10 . 7554/eLife . 00299 . 009Figure 7 . The study design permits causal and reactive model analyses . ( A ) Three models were evaluated , each showing a candidate hypothesis for the three-way association between genotype ( G ) , expression ( E ) and trait ( T ) . In the independent model , expression and trait each associate with genotype but are not themselves directly related . In the causal model , expression mediates the association between genotype and trait . In the reactive model , genotype and expression relate through the trait , so that gene expression changes are a downstream response to the trait . ( B ) p-values for independent-versus-reactive and independent-versus-causal hypothesis tests . Each point shows the result for one SNP-transcript pair . Points to the right of the solid vertical line are significant ( p<0 . 05 ) for the reactive hypothesis and points above the solid horizontal line are significant for the causal hypothesis . The dashed line shows a p=0 . 1 threshold . ( C ) Power for rejection of the independent hypothesis . Non-independent data were simulated with effect sizes and variances similar to those in the enrichment set ( the set of SNP-transcript pairs that were found to be significant in our study ) . The curve shows the proportion of cases in which the simulated data rejected the independent ( null ) hypothesis . The dotted line indicates the combined sample size in our study . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 009
The results provide an unbiased integrative survey of the genetic and transcriptional components of the humoral immune response to influenza vaccination in humans . They suggest that variation at the level of genes involved in antigen processing and intracellular trafficking is an important determinant of vaccine immunogenicity . Even in healthy , young individuals , there are a significant number of people who do not develop a protective antibody response after influenza vaccination . If these individuals could be identified prior to vaccination , modifications to the type or dose of vaccine could be attempted , with the goal of reducing the number of unprotected vaccinated individuals . The genes identified in this study as playing a role in variation in the humoral response to vaccination would be a logical starting point for the development of DNA- or RNA-based predictive biomarkers . Prospective evaluation of such biomarkers would be the next step towards clinical implementation . Understanding the mechanisms that underlie variation in response to the vaccine may also direct modification of factors that enhance the response . Most of the efforts to date have focused on vaccine adjuvants that activate known immunologic mechanisms . Surprisingly , many of the genes identified in this study encode proteins that are not specifically immune but play a more general role in membrane trafficking and intracellular transport . Interventions aimed at increasing vaccine antigen affinity to these proteins or altering their intracellular concentrations could represent new avenues in vaccine development . More broadly , the results demonstrate that a longitudinal , integrative genomic analysis study design , applied to a clinical intervention , is an efficient alternative to cross-sectional methods for the identification of genes involved in medically relevant complex traits . By making repeated measurements on the same individual over time after a controlled experimental perturbation , we were able to account for individual variation in a way that would not have been possible otherwise . The dynamic nature of the measurements also allowed us to uncover genetic effects that are either enhanced by or only evident after the experimental perturbation . The specificity of gene identification in this study emerges from the genome’s acute response to the perturbation , which cannot be assessed by a cross-sectional eQTL analysis or a genome-wide association study . This approach could be used for a broad variety of medically important problems whenever there is the opportunity to test a well-controlled intervention such as drug , dietary , or vaccine responses . Several limitations of the study are worth noting . First , we studied two samples of healthy young adults , thereby excluding the segments of the population that are most likely to have a poor response to influenza vaccination: children , the elderly , and individuals with severe illnesses . Second , in order to minimize the risk of false associations related to population stratification , we studied an ethnically homogeneous group of individuals . Third , while an interesting aspect of our study design is that it could open the door for direct comparisons of causal and reactive models , the sample size in this study was not sufficient to establish whether or not there is a causal relationship between the loci for which an association was identified and the antibody response to the vaccine . Finally , while antibody titers have historically been used to evaluate vaccine responsiveness , it is clear that they do not capture the complexity of the human immune response to vaccination . Additional studies would be necessary to determine whether the genes identified are also related to variation in influenza vaccine responses in groups other than the one chosen for this study , whether there is a causal relationship between these genes and the antibody response , or whether they also influence the cell-mediated immune response to the vaccine .
A visual representation of the study design , the resulting data sets , and the integrative analysis scheme , is presented in Figure 8 . 10 . 7554/eLife . 00299 . 010Figure 8 . Study design and integrative analysis scheme . ( A ) Individuals were immunized on day 0 and peripheral blood RNA samples were obtained on days 0 , 1 , 3 , and 14 . Antibody titers were measured on pre-immune sera and on days 14 and 28 . Genotyping was carried out on a peripheral blood genomic DNA sample obtained on day 1 . Identical sample collection schemes were used , 1 year apart , for the discovery ( males ) and validation ( females ) cohorts . ( B ) Sample sizes and data generation platforms . ( C ) Integrative analysis involved identification of loci that exhibit a transcriptional response to vaccination , evidence of genetic regulation of expression ( constitutive eQTL ) , evidence of correlation between gene expression and the antibody response , and evidence of correlation between genotype and the antibody response ( QTL ) . Because transcript abundance was measured serially , we were able to evaluate changes in the magnitude of the genetic effect on expression at different time points following vaccination . In addition , the study design permitted QTL analysis conditional on gene expression , which led to the identification of loci whose genetic effects on the antibody response are causally linked through the eQTL . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 010 Healthy volunteers ages 18 to 40 years were enrolled . Individuals who were known to have received an influenza vaccine in the previous 3 years or who had signs or symptoms of an active infection at the time of enrollment were excluded . To minimize false-positive results related to population stratification , enrollment was limited to individuals of self-reported Caucasian ancestry . Enrollment , vaccination and sample collections were conducted at a university campus . The initial ( discovery ) cohort was restricted to males . The validation cohort was enrolled approximately 18 months after the initial cohort and was restricted to females . The protocol was approved by the institutional review boards of all participating institutions . Informed consent was obtained from each subject prior to enrollment . Study participants were immunized on day 0 . Those enrolled in the initial cohort received the 2008–2009 inactivated trivalent influenza vaccine ( A/Brisbane/59/2007[H1N1] , A/Brisbane/10/2007[H3N2] , B/Florida/4/2006; Sanofi-Pasteur , Lyon , France ) . The validation cohort received the 2009–2010 vaccine , which came from the same manufacturer and included ( A/Brisbane/59/2007 ) , ( A/Brisbane/10/2007[H3N2] ) , and ( B/Brisbane/60/2008 ) strains . Whole blood samples ( 7 ml ) for DNA purification were collected in Vacutainer acid citrate dextrose ( ACD ) tubes ( Beckton-Dickinson , Franklin Lakes , NJ ) , on day 1 after vaccination . DNA was purified using Qiagen Gentra Puregene Blood Kits ( Qiagen Sciences , Germantown , MD ) . Quantitation and quality control were performed with a NanoDrop-1000 spectrophotometer ( Thermo Fisher Scientific , Waltham , MA ) and using the Quant-iT PicoGreen dsDNA Reagent ( Life Technologies , Carlsbad , CA ) in a Tecan GENios microplate reader ( Tecan Group , Mannendorf , Switzerland ) . Peripheral blood samples for RNA purification were obtained immediately before ( day 0 ) and on days 1 , 3 , and 14 after vaccination . To minimize changes in gene expression induced by sample handling and processing , whole-blood samples ( 2 . 5 ml ) were collected in PAXgene RNA stabilization tubes ( Qiagen Inc . , Valencia , CA ) and frozen at −80°C . RNA purification was performed using the PAXgene Blood RNA system ( Qiagen Inc . ) . All RNA samples from an individual were consistently purified in the same batch . Spectrophotometry ( NanoDrop-1000 Spectrophotometer; Thermo Fisher Scientific ) and microfluidic electrophoresis ( Experion Automated Electrophoresis System; Bio-Rad Laboratories , Hercules , CA ) were used for quality control . Genotyping was performed on Illumina HumanOmniExpress microarrays ( Illumina , Inc . , San Diego , CA ) following the manufacturer’s instructions . The arrays include 730 , 525 SNP markers . Basic quality control of the genotyping data was performed on GenomeStudio software , version 2010 ( Illumina , Inc . ) . All microarrays had call rates >0 . 99 . In vitro transcription was performed using Ambion Illumina TotalPrep RNA Amplification Kits ( Applied Biosystems/Ambion , Austin , TX ) . cRNA was hybridized onto Illumina HumanHT-12v3 or HumanHT-12v4 Expression BeadChips ( Illumina , Inc . ) , following the manufacturer’s protocol . All samples for a given individual were processed on the same slide . The arrays have 48 , 742 ( v3 ) and 47 , 301 ( v4 ) reporters , representing approximately 25 , 000 genes and non-annotated gene candidates . Whole blood ( 10 ml ) was collected in Vacutainer Serum Separator Tubes ( Beckton–Dickinson ) . Serum was separated by centrifugation prior to storage at −20°C . Hemagglutination inhibition ( HAI ) tests were performed as previously described ( Dowdle et al . , 1979 ) , except for a starting serum dilution of 1:4 and the use of turkey red blood cells . HAI test antigens were allantoic fluid harvests from infected embryonated hen’s eggs ( whole-virus antigens ) . Neutralizing antibody tests were performed as previously described ( Frank et al . , 1980 ) except that hamster serum was not included . Test strains were the same as those used in the vaccine . HAI and neutralizing antibody titers were measured for each viral antigen included in the vaccine . For all antigens , the change in antibody titer post-vaccination is known to be negatively correlated with the pre-vaccination titer . The responses to the individual viral antigens were correlated within individuals . For each individual , we computed a Titer Response Index ( TRI ) , as previously described ( Bucasas et al . , 2011 ) . The TRI characterizes the magnitude of an individual’s antibody rise accounting for their pre-vaccination titer and integrating the responses across the three vaccine components to provide an overall measure of vaccine responsiveness . Raw signal intensity data from all individuals and all time points in the discovery cohort were first processed in a single batch . Background adjustment , variance stabilization transformation ( Lin et al . , 2008 ) and robust spline normalization were performed using the lumi package ( Du et al . , 2008 ) in R ( R Development Core Team , 2009 ) . Eight individuals had missing expression data from two or more time points and were excluded . We required a detection p-value of <0 . 01 in at least 80% of the samples for a transcript to be considered detected . We also aligned the entire set of expression reporter sequences to the human genome reference sequence ( Build 36 [March 2006]/hg18 ) by applying the BLAT algorithm in BlatSuite34 software ( Kent , 2002 ) , and excluded any reporters that did not map or mapped to more than one region . Using these two thresholds for the data in the discovery cohort , the final data set included 9809 detected transcripts . This data set was then used for eQTL analysis in the discovery cohort . Once data generation for the validation cohort was completed , the expression microarray signal intensity data from all individuals and all time points in that cohort were processed in a single batch . Background adjustment , variance stabilization transformation , robust spline normalization , and the application of detection thresholds were performed identically to the discovery cohort . Five individuals had missing expression data from two or more time points and were excluded . Because two different array versions were used ( HT12-v3 and HT12-v4 ) , unique reporter identifiers ( ProbeID and nuID ) for the 9809 reporters selected in the discovery cohort were used to subset the data from the validation cohort . This data set was then used for eQTL analysis in the validation cohort . The analysis of correlation between gene expression and antibody titer in the discovery cohort was previously published ( Bucasas et al . , 2011 ) . As part of the integrative genomic analysis described in the present study , we performed a similar analysis of this expression/titer correlation , but included expression data from both cohorts . For this purpose , the two data sets described above were combined , and an additional quantile normalization step was performed to account for batch effects between cohorts . Array quality was initially assessed using GenomeStudio software ( Illumina , Inc . ) . Default algorithms were used to normalize , generate SNP clusters , and make genotype calls . SNPs with minor allele frequency ( MAF ) <0 . 05 and Hardy-Weinberg Equilibrium ( HWE ) χ2 < 1 × 10−7 were removed prior to analysis . For local eQTL mapping , we restricted the analyses to SNPs within 1 Mb ( 500 kb upstream or downstream ) of each unambiguously mapped expression reporter sequence . We used the genotype information to reliably establish that the individuals in each cohort were indeed unrelated and ethnically matched . For this , we estimated pairwise identity-by-descent ( IBD ) metrics on the basis of identity-by-state ( IBS ) information from their genome-wide SNP genotyping results , followed by multidimensional scaling ( MDS ) analysis on the resulting matrix of pairwise distances , using PLINK ( Purcell et al . , 2007 ) . The first three components of the MDS analysis on our study samples and the founders from six HapMap populations were plotted with R package Scatterplot3D ( Ligges and Machler , 2003 ) and are shown in Figure 9 . As expected , our study samples from both cohorts cluster with the HapMap CEU population . In the male cohort , we identified four pairs of individuals with cryptic or undisclosed familial relationships ( pi-hat ≥ 0 . 125 ) . 12 outliers were identified by component 2; two of the four pairs of cryptic or undisclosed relatives were among the outliers . One individual from each related pair and all remaining outliers were removed prior to analysis . In the female cohort , we identified one pair of individuals with cryptic or undisclosed familial relationship ( pi-hat ≥ 0 . 125 ) . 11 outliers were identified by component 2 and one by component 1 . One individual from the related pair and all outliers were removed prior to analysis . 10 . 7554/eLife . 00299 . 011Figure 9 . Study samples cluster with the HapMap CEU population . Pairwise identity-by-descent metrics were estimated based on genotype data from our two study samples and six HapMap populations . Multidimensional scaling analysis was performed on the resulting pairwise distances . Components 1–3 of this analysis are plotted for the male ( top ) and female ( bottom ) cohorts , and the comparison populations . As expected , the study samples cluster with the HapMap CEU population . 12 outliers were identified in each cohort and were removed prior to analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 011 Analyses were carried out separately for each cohort and in the combined data set . Because we have repeated gene expression measurements on the same individuals over a series of time points , we performed a random effects linear model analysis . For each SNP-transcript pair , we fit a model with terms for day , additive effects for genotype and day–genotype interactions , and a random effect for each person:Where Y = observed expression value , i = day ( 0 , 1 , 3 , 14 ) , j = individual , D = day effect , g = genotype effect , P = person effect ( random ) , and e = random error . We fit this model using the R package nlme and the estimation method of the same name . The day effect ( D ) is a shift term which considers time-ordered change in expression . The genotype effect ( g ) is a main effect for genotype that considers the expression differences between genotypes to be additively related to the number of copies of a reference allele . The day–genotype interaction allows the additive effect to vary by day ( see Figure 10 for a visual representation of the analysis scheme ) . The person effect ( P ) is modeled as a random effect; this term helps to efficiently account for the fact that the same individual appears in each time point , without overfitting the expression model to the particular sample ascertained in our study . We explored different parametrizations , such as allowing the person effect to vary by day , and found that the simplest model with a single term for each person was effective . This person effect can account for inter-individual or batch differences which are stable within individuals for the duration of the study . Yi , j=Di+βgj+βigj+Pj+ei , j10 . 7554/eLife . 00299 . 012Figure 10 . Genetic and transcriptional analysis on a prospective cohort . The figure displays hypothetical results for a single SNP-transcript pair . For any such pair , one may observe changes in transcript abundance at different time points after the experimental perturbation ( lower box plots ) . In addition , gene expression at each time point may be different for different genotypes when there is evidence of an eQTL ( upper box plots ) . Finally , the slope of the expression-genotype association ( β ) , as well as the proportion of the variance in expression explained by genotype ( R2g ) , may vary across time points . DOI: http://dx . doi . org/10 . 7554/eLife . 00299 . 012 In addition to the random effects linear model , we analyzed each expression trait with respect to genotype at each time point usingwhere Yj denotes the matrix of expression traits of individuals j at a given time point ( days 0 , 1 , 3 , 14 ) and Gi , k is a matrix of genotypes for individuals j at SNP locus k such that each element is assigned 0 , 1 or 2 according to the number of minor alleles at the kth locus of the jth individual . This allowed estimates of the proportion of expression variance explained by individual genotypes from the model’s coefficient of determination ( R2g ) . This model permitted a more detailed examination of the changes in the strength of the genotypic association with expression at each time point . To detect SNP-transcript pairs where the magnitude of the genotype effect varied after immunization , we took the difference in R2g measures ( ΔR2g ) from pairwise linear models fitted after ( days 1 , 3 and 14 ) and before vaccination ( day 0 ) . We then took the top 1% ΔR2g values between day 0 and a later time point . This cis-acting eQTL subset represents the loci that show significant changes in the genetic effect as a result of the vaccine . Yj=βGi , k+e The titer response index was treated as a dependent variable and modeled as a linear combination of expression values for each transcript across the time course . A single F-statistic p-value was determined to evaluate the explanatory value of the expression data using all days . Separate evaluations were made in each cohort and in the combined data . The 301 genes displayed in Figure 4B had F-statistic p-values in the combined data <0 . 01 and absolute values of the maximum average correlation between expression and titer response in the two cohorts >0 . 15 . Enriched biological and functional pathways were analyzed using DAVID Bioinformatics Resources ( Dennis et al . , 2003 ) , and Ingenuity Pathway Analysis ( IPA ) software . To evaluate the relationship between the associations identified in our study , we extended a recently published analysis framework for causal modeling in eQTL data ( Millstein et al . , 2009 ) . The relationship is modeled aswhere T is trait , E is gene expression , and G is genotype at a locus . In our experiment , the term E includes a separate value for each day . The term G has a separate effect on the trait under the independent null model , while under the causal alternative , where it acts through gene expression , its influence is entirely transmitted through E . The statistic used to assess the influence of G is the partial F reported by the ‘aov’ function in R ( version 2 . 15 . 1 ) . The analyses utilize an equivalence test approach , where the null hypothesis is that the second independent variable has an effect on the response conditional on the first variable , while the alternative is complete co-linearity in the associations . The null distribution of the F statistic–which under parametric assumptions would have a non-central F distribution—is derived using a permutation test procedure , as follows: The relationship between E and T is decoupled by regressing E on G , permuting the resulting residuals , then adding the permuted residuals to the predicted values to arrive at E* , which is independent of T but maintains the marginal variance and G-correlation of E . This procedure enforces the assumption of the independent model , while leaving other properties unchanged . The partial F statistic for G inrepresents a sample from the distribution under the independent null hypothesis . This process , repeated 1000 times , provided a null distribution . The p-value for the observed F is the proportion of the null distribution that is below the observed value . T∼E+GT∼E*+G The reactive model is evaluated using the same approach , but with the expression and trait terms swapped , that isE∼T+G This scheme tests the extent to which the trait value mediates the relationship between genotype and gene expression . In this analysis , the permutation step decouples the expression-to-trait relationship from the eQTL association and asks if the eQTL association disappears after accounting for the expression-to-trait association . Power analyses for these models apply the same scheme to simulated data . The term G is simulated as binomial , while E and T are normal , with parameters chosen to match the minor allele frequency , variance , and correlations observed in our data set . The power to reject the false ( independent ) model is shown using a threshold of p<0 . 05 . The power properties for the tests used to evaluate both the reactive and causal models are equivalent because the model structures and r2 relationships are symmetric between the two alternatives . | Vaccines increase resistance to disease by priming the immune system to respond to specific viruses or microorganisms . By presenting a weakened ( or dead ) form of a pathogen , or its toxins or surface proteins , to the immune system , vaccines trigger the production of antibodies against the virus or microorganism . If a vaccinated individual then encounters the pathogen , their immune system should be able to recognize and destroy it . Many vaccines also include a secondary agent , known as an adjuvant , to further stimulate the immune response . Influenza , an RNA virus commonly referred to as the ‘flu’ , is an infectious disease that affects both birds and mammals . Seasonal epidemics occur each year affecting 2–7% of the population . According to the World Health Organization , influenza leads to nearly 5 million hospitalizations each year and causes up to half a million deaths . Vaccination is a primary strategy for the prevention of seasonal influenza , but responses to the vaccine vary markedly , partly because of variation in the genetic makeup or genotype of individuals . However , the details of how genes influence response to vaccination , and indeed susceptibility to influenza , remain unclear . To investigate the genetic basis of variation in the immune response of healthy adults to the seasonal influenza vaccine , Franco et al . combined information about the genotypes of individuals with measurements of their gene transcription and antibody response to vaccination . They identified 20 genes that contributed to differential immune responses to the vaccine . Almost half of these encode proteins that are not specifically associated with the immune system , but have more general roles in processes such as membrane trafficking and intracellular transport . Focusing on these genes may enable researchers to spot those individuals who are less likely to respond to a vaccine . It could also open up new avenues of research for vaccine development: rather than designing adjuvants that target known immune mechanisms , researchers should develop adjuvants that target the proteins encoded by these 20 genes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2013 | Integrative genomic analysis of the human immune response to influenza vaccination |
Susceptibility to cancer is heritable , but much of this heritability remains unexplained . Some ‘missing’ heritability may be mediated by epigenetic changes in the parental germ line that do not involve transmission of genetic variants from parent to offspring . We report that deletion of the chromatin regulator Kdm6a ( Utx ) in the paternal germ line results in elevated tumor incidence in genetically wild type mice . This effect increases following passage through two successive generations of Kdm6a male germline deletion , but is lost following passage through a wild type germ line . The H3K27me3 mark is redistributed in sperm of Kdm6a mutants , and we define approximately 200 H3K27me3-marked regions that exhibit increased DNA methylation , both in sperm of Kdm6a mutants and in somatic tissue of progeny . Hypermethylated regions in enhancers may alter regulation of genes involved in cancer initiation or progression . Epigenetic changes in male gametes may therefore impact cancer susceptibility in adult offspring .
Intergenerational inheritance of epigenetic state may significantly impact disease susceptibility in animals , including humans . In addition to genetic information , male and female gametes transmit epigenetic regulatory information , in the form of covalent DNA modification , histone modification , and small RNAs , to the zygote at fertilization . Accumulating evidence indicates that the epigenetic information inherited from both maternal and paternal gametes can modulate gene expression and phenotype in progeny throughout the metazoan lineage ( Arico et al . , 2011; Carone et al . , 2010; Ciabrelli et al . , 2017; Greer et al . , 2011; Morgan et al . , 1999; Siklenka et al . , 2015 ) . In mammals , these transcriptional effects can manifest phenotypically as defects in early development ( Chong et al . , 2007; Siklenka et al . , 2015 ) or as altered metabolic or behavioral states during adulthood ( Carone et al . , 2010; Dias and Ressler , 2014; Ng et al . , 2010 ) . Consistent with these findings , there is mounting evidence that mature mammalian sperm carry an information-rich epigenome . Although the final stages of testicular sperm development involve extensive nuclear rearrangement , including widespread replacement of histones with protamines and nucleus-wide chromatin compaction , mammalian sperm are more than motile packages of DNA . Mature mammalian sperm exhibit relatively high levels of DNA methylation ( Monk et al . , 1987 ) , contain populations of small RNAs ( Sharma et al . , 2016; Siklenka et al . , 2015 ) , and retain 5–10% of their histones ( Hammoud et al . , 2009; Jung et al . , 2017 ) , which bear extensive post-translational modifications ( Erkek et al . , 2013; Hammoud et al . , 2009; Luense et al . , 2016 ) . Specific histone modifications at some loci in the male germ line have been conserved during mammalian evolution , implying a biologically important function ( Lesch et al . , 2016 ) . Recent evidence suggests that mature mouse sperm also retain elements of large-scale three-dimensional genomic domains found in somatic cells ( Jung et al . , 2017; Ke et al . , 2017 ) . Altered epigenetic states play a significant role in cancer pathogenesis , making cancer a strong candidate for sensitivity to intergenerational epigenetic effects . Many cancers are highly heritable , meaning that the presence of a given tumor or set of tumors in one individual increases the risk of developing the same tumors among close relatives ( Goldgar et al . , 1994; Lichtenstein et al . , 2000 ) . However , despite extensive genetic studies of many human cancers , a large fraction of this heritability remains unexplained by specific genetic mutations or variants ( Lichtenstein et al . , 2000; Mucci et al . , 2016 ) . Meanwhile , investigations into tumor biology have revealed that cancer is in part a disease of epigenetic dysregulation . Many tumors are characterized by significantly perturbed gene regulatory states and exhibit abnormal genome-wide histone methylation and DNA methylation profiles ( Dawson and Kouzarides , 2012 ) . Cancer genetics studies over the last decade have revealed that , when susceptibility genes can be identified , many encode chromatin regulators , implying that epigenetic changes contribute either to tumor initiation or tumor progression ( Baylin and Jones , 2011; Dawson and Kouzarides , 2012 ) . Kdm6a ( Utx ) has been identified as a candidate tumor suppressor in cancer genetics studies . The KDM6A protein has histone demethylase activity toward lysine 27 on histone H3 ( H3K27 ) , as well as demethylase-independent functions in establishment of enhancer regions ( Hong et al . , 2007; Lan et al . , 2007; Wang et al . , 2017 ) . KDM6A mutations are found in a variety of human cancers , including multiple myeloma , renal cell carcinoma , bladder carcinoma , acute myeloid leukemia ( AML ) , acute lymphoid leukemia ( ALL ) , prostate cancer , and medulloblastoma ( Jones et al . , 2012; Ntziachristos et al . , 2014; Van der Meulen et al . , 2014; van Haaften et al . , 2009 ) . KDM6A has been mechanistically implicated as a tumor suppressor in ALL ( Ntziachristos et al . , 2014; Van der Meulen et al . , 2015 ) , AML ( Gozdecka et al . , 2018 ) , and lung cancer ( Wu et al . , 2018 ) , and also plays important developmental roles , especially in the heart ( Lee et al . , 2012; Welstead et al . , 2012 ) and blood ( Beyaz et al . , 2017; Thieme et al . , 2013 ) . Because KDM6A functions primarily as a chromatin regulator , these studies imply that the epigenetic sequelae of KDM6A loss contribute to tumor initiation or progression . Here , we delete Kdm6a specifically in the male germ line of the mouse , and evaluate gene regulatory and phenotypic effects in genetically wild type offspring . We find that offspring of Kdm6a male germline knockouts exhibit an increased incidence of tumors , and that this effect is enhanced when Kdm6a is deleted in the germ line in two successive generations . Because these effects are provoked by a single genetic lesion in the parent , we were able to define specific epigenetic changes resulting from this manipulation . We find widespread perturbation of H3K27 methylation state in the Kdm6a mutant male germ line , as well as increased levels of DNA methylation at specific loci . Some of the changes in DNA methylation observed in the mutant germ line are retained in somatic tissue of wild type progeny , and may affect the transcriptional regulation of genes involved in cancer susceptibility .
We designed a breeding strategy to produce genetically wild type male offspring from a Kdm6a mutant male germ line . We generated a germline-specific Kdm6a conditional knockout ( Kdm6a cKO ) in male mice by crossing a Cre recombinase driven by the Ddx4 ( Mvh ) promoter ( Gallardo et al . , 2007 ) to a conditional allele of Kdm6a ( Welstead et al . , 2012 ) . The Cre transgene is expressed in the prenatal germ line , and excision of the conditional allele is complete by the time postnatal spermatogenesis begins ( Hu et al . , 2013 ) . Kdm6a is encoded on the X chromosome , so recombination of a single allele is sufficient to generate a complete knockout . Because developing spermatogenic cells are linked by cytoplasmic bridges until just before sperm are released , and therefore share cytoplasmic factors , loss of Kdm6a expression from the X chromosome affects both X- and Y-bearing spermatogenic cells even after meiosis ( Braun et al . , 1989 ) . Mating Kdm6a cKO males to wild-type females produced genetically wild type male offspring ( ‘Kdm6a F1’ ) and heterozygous female offspring ( Figure 1A ) . Cre-negative littermates of Kdm6a cKO males were mated to age-matched wild type females , and the male offspring of these crosses were used as controls ( ‘control F1’ ) . Critically , Kdm6a F1 males are genetically wild type , but generated from a paternal germ line lacking KDM6A activity . Kdm6a cKO males were fertile , produced male and female offspring at Mendelian ratios , and exhibited normal spermatogenesis ( Figure 1—figure supplement 1 ) . We confirmed the high efficiency of Cre recombinase activity ( >97% ) by genotyping the heterozygous female offspring of these crosses ( Figure 1—figure supplement 2 ) . Male Kdm6a and control F1s were housed with littermates and followed until natural death or morbidity requiring euthanasia , at which time all animals underwent complete necropsy . Mice surviving past 24 months of age were considered healthy survivors . We found that lifespans of Kdm6a F1 males were shorter than those of control F1s ( Figure 1B , Figure 1—figure supplement 3 ) . This effect was independent of whether the animal carried the Ddx4-Cre transgene ( Figure 1—figure supplement 4 ) and of mode of death ( Figure 1—figure supplement 5 ) . There was no significant difference in weight and a mild reduction in body length for Kdm6a F1s compared to control F1s at the time of death ( Figure 1—figure supplement 6 ) . We evaluated cumulative necropsy data to define pathological correlates of the difference in survival between Kdm6a F1s and control F1s ( Figure 1—figure supplement 7 ) . We found that Kdm6a F1s that died between 12 and 18 months of age did not exhibit evidence of a unifying disease process . In contrast , Kdm6a F1s that died between 18 and 24 months of age exhibited an increased tumor burden compared to age-matched control F1s ( Figure 1C , Figure 1—figure supplement 8 ) . The spectrum of tumors identified was similar to that observed in normally aging mice ( Haines et al . , 2001 ) , but appeared earlier and at higher frequencies . The most common cancer type was histiocytic sarcoma , a blood tumor of the monocyte/macrophage lineage ( Figure 1D; 6/22 vs . 1/25 mice , p=0 . 040 , Fisher’s Exact test ) ; this tumor was found in Kdm6a F1s at a mean age of 624 ± 61 days , and in a single control F1 at 722 days . Flow cytometry of bone marrow from these mice revealed expanded populations of monocyte-lineage cells , consistent with histiocytic sarcoma . In addition , Kdm6a F1 mice not identified as having histiocytic sarcoma by histopathology also had a moderate increase in monocyte-lineage cell populations , indicating subtle skewing of hematopoietic lineages even in the absence of full-blown disease ( Figure 1—figure supplement 9 ) . Kdm6a F1 mice also developed a variety of other solid and blood tumors ( Figure 1E , Figure 1—figure supplement 10 , Figure 2—source data 2 ) . We then asked whether this effect could be transmitted to a second generation . We designed a breeding strategy in which wild type males were generated from male germ cells that had passed through two successive generations of Kdm6a conditional deletion ( ‘Kdm6a F2’ ) , or through one generation of Kdm6a deletion followed by one generation with an intact Kdm6a gene ( ‘control F2’ ) ( Figure 2A ) . F2 males were followed under the same protocol as F1 males . We found that , like Kdm6a F1s , Kdm6a F2s exhibited reduced survival relative to the original control F1 cohort , whereas survival of control F2 males was more variable ( Figure 2B , Figure 2—figure supplement 1 ) . Also like Kdm6a F1s , Kdm6a F2s had an increased tumor burden relative to the F1 control cohort ( Figure 2C and D , Figure 2—figure supplement 2 ) . We did not find evidence for increased tumor burden in control F2 males . We conclude that repeated loss of Kdm6a in the male germ line is required to maintain the intergenerational tumor susceptibility phenotype . Notably , the tumor phenotype was more pronounced in Kdm6a F2s compared to Kdm6a F1s: Kdm6a F2s developed more tumors per mouse ( overall tumor rate: 0 . 24 control , 0 . 95 Kdm6a F1 , 1 . 30 Kdm6a F2; Figure 2—source data 2 ) , and when present , tumors were more aggressive . Thirty-eight percent ( 15/40 ) of Kdm6a F2 mice had more than one independent tumor at death , compared to 23% ( 5/22 ) of Kdm6a F1 mice and 4% ( 1/25 ) of control F1 mice ( Figure 2E ) . In addition , a higher fraction of Kdm6a F2 tumors were malignant ( Figure 2F ) . We conclude that exposure of male germ cells to loss of Kdm6a across multiple generations confers a cumulative risk of tumor development on offspring . These findings imply that the molecular changes mediating this effect accumulate across generations , but can be reset when germline Kdm6a expression is restored . We then turned our attention to the molecular mechanism by which loss of Kdm6a in the germ line might affect tumor susceptibility in the next generation . An advantage of our experimental strategy is that any epigenetic perturbation in germ cells is a consequence of a single defined genetic lesion , knockout of Kdm6a . We could therefore predict the nature of epigenetic changes in the Kdm6a cKO germ line based on the known molecular functions of the KDM6A protein . KDM6A is an H3K27me3 histone demethylase , and also plays a demethylase-independent role in promoting assembly of active enhancer regions ( Hong et al . , 2007; Lan et al . , 2007; Shpargel et al . , 2012; Wang et al . , 2017 ) . We first examined the effect of Kdm6a deletion on H3K27 methylation in male germ cells . We collected H3K27me3 ChIP-seq data from two biological replicates of epididymal sperm from Kdm6a cKO and littermate control males ( Figure 3—source data 2 , Figure 3—figure supplement 1 ) . ChIP-seq data were strongly correlated between sperm replicates ( Figure 3—figure supplement 2 ) . We examined H3K27me3 signal in 2-kilobase ( kb ) tiles throughout the genome . Genome-wide , we observed an increase in H3K27me3 signal in Kdm6a cKO sperm relative to control sperm after normalizing for library size , as expected for loss of an H3K27me3 demethylase ( Figure 3A ) . We confirmed a global gain in H3K27me3 by Western blot ( Figure 3B ) . However , this effect was not uniform throughout the genome . While H3K27me3 signal increased in the majority of tiles in Kdm6a cKO sperm , those tiles with the highest overall H3K27me3 signal exhibited a paradoxical loss of H3K27me3 in Kdm6a cKOs ( Figure 3C , Figure 3—figure supplement 3 ) . The result is an apparent flattening of the H3K27me3 profile: a decrease in H3K27me3 at regions with high signal , accompanied by an increase in H3K27me3 in adjacent regions ( Figure 3D–E , Figure 3—figure supplements 4–5 ) . This effect is compatible with several explanations . First , it may reflect genuine loss of signal in some regions accompanied by gain in adjacent regions . Second , widespread gain of H3K27me3 due to loss of KDM6A demethylase activity could result in the false appearance of signal loss at regions where H3K27me3 levels are actually unchanged . Finally , this effect may represent a more homogeneous signal at the population level due to increased variability between individual sperm . Allowing for each of these explanations , we conclude that loss of KDM6A increases H3K27me3 overall and alters the normal pattern of distribution of H3K27me3 during spermatogenesis . Because we deleted Kdm6a early in spermatogenesis , we then considered the possibility that some epigenetic changes carried by Kdm6a cKO sperm might be indirect effects of early KDM6A loss . Deposition of H3K27 methylation has been associated with both gain and loss of cytosine DNA methylation , depending on the genomic and cellular context ( Brinkman et al . , 2012; Neri et al . , 2013; Viré et al . , 2006 ) . DNA methylation is stable across long developmental time periods and is retained at high levels in sperm ( Monk et al . , 1987; Smallwood et al . , 2011; Smith et al . , 2012 ) . We therefore asked if DNA methylation levels changed in regions of the genome where H3K27me3 was most perturbed in Kdm6a cKO relative to control sperm . We collected reduced representation bisulfite sequencing ( RRBS ) data from epididymal sperm of three control and three Kdm6a cKO males ( Figure 4—source data 2 ) . Overall levels of DNA methylation did not differ between control and cKO sperm ( 65% and 66% methylation , respectively ) . However , regions where H3K27me3 was altered , defined as those tiles with log2 fold change >0 . 5 or<−0 . 5 and false discovery rate <0 . 1 in both ChIP-seq replicates and which were not called as different in comparisons between the two control or two cKO datasets , were associated with increased DNA methylation ( Figure 3F and G , Figure 3—figure supplement 6 , Figure 3—figure supplement 7 ) . Both increased and decreased H3K27me3 were associated with a gain in DNA methylation , possibly due to secondary alterations in histone methylation after establishment of an initial change in DNA methylation . These regions were enriched near gene bodies ( p=9 . 898×10−6 for H3K27me3 gain and p=5 . 892×10−4 for H3K27me3 loss , Fisher’s exact test ) , and regions of H3K27me3 loss were also weakly enriched at transcription start sites ( p=0 . 01368 , Fisher’s exact test ) . Genes exhibiting loss of H3K27me3 and gain of DNA methylation were enriched for functions such as ‘negative regulation of myeloid dendritic cell activation’ and ‘positive regulation of immune effector process’ ( Figure 3—figure supplement 8 ) . Together , our results indicate that deletion of Kdm6a early in spermatogenesis induces redistribution of H3K27me3 , and that regions strongly affected by H3K27me3 redistribution gain DNA methylation in mature sperm . We then asked if the changes in DNA methylation evident in Kdm6a cKO sperm could also be detected in somatic tissues of aging Kdm6a F1 adults . We collected RRBS data from bone marrow of Kdm6a F1 and control F1 males ( Figure 4—source data 2 ) , and compared it to the RRBS data from Kdm6a cKO and control sperm . We identified differentially methylated regions ( DMRs: 100 bp tiles with false discovery rate <0 . 05 ) in Kdm6a cKO vs . control sperm and in Kdm6a F1 vs . control F1 bone marrow ( Figure 4A ) . To avoid the confounding effect of disease on DNA methylation , we excluded F1 mice with any histopathological abnormality in the blood lineage . DMRs in both Kdm6a cKO sperm and Kdm6a F1 bone marrow were more likely to be hypermethylated than hypomethylated relative to their respective controls ( 4725 hypermethylated vs . 323 hypomethylated DMRs in sperm and 3156 hypermethylated vs . 1122 hypomethylated DMRs in bone marrow ) . Two hundred and ninety-nine regions were differentially methylated in both Kdm6a cKO sperm and Kdm6a F1 bone marrow , significantly more than expected by chance ( 57 regions expected , p=4 . 22e-121 , hypergeometric test ) ( Figure 4B ) . Considering all 299 shared DMRs , there was a positive correlation between the magnitude of DNA methylation change in sperm and in F1 bone marrow ( R = 0 . 17 , p=0 . 0026 ) ( Figure 4C ) . Two hundred and twenty-six individual DMRs ( 76% ) were positively correlated between sperm and F1 bone marrow , including 207 ( 69% ) hypermethylated and 19 ( 6% ) hypomethylated regions ( Figure 4—source data 3 , Figure 4—source data 2 ) . Given the overall hypermethylation of DMRs in both Kdm6a cKO sperm and Kdm6a F1 bone marrow , we focused our attention on the 207 hypermethylated regions . We considered these positively-correlated hypermethylated DMRs as candidates for direct inheritance of DNA methylation state from the paternal germ line , and refer to them as ‘persistent’ DMRs . We validated our RRBS findings using pyrosequencing in Kdm6a cKO sperm and Kdm6a F1 bone marrow , and confirmed hypermethylation at 12 of 13 tested DMRs in at least one tissue and at seven of 13 DMRs in both tissues ( Figure 4—figure supplement 1 ) . We then asked what genomic and regulatory features were associated with persistent DMRs . We found that persistent DMRs were distributed throughout the genome ( Figure 4D ) and frequently overlapped the regions of greatest H3K27me3 change in Kdm6a cKO sperm ( Figures 3F and 4E ) . In contrast , there was no association between persistent DMRs and various other features , including CpG islands , imprinted regions , and transcription start sites ( TSS ) ( Figure 4F , Figure 4—figure supplement 2 ) . Although repetitive elements such as retrotransposons can be resistant to DNA methylation reprogramming in the germ line ( Guibert et al . , 2012 ) , persistent DMRs were not more likely to overlap repetitive elements compared to the complete set of genomic regions covered by our RRBS data ( Figure 4G ) . We conclude that the location of persistent DMRs is strongly associated with regions of altered H3K27me3 in sperm , implying that loss of Kdm6a in the male germ line sensitizes these regions to DNA hypermethylation . Some of these sensitive regions may retain their methylation state during somatic development in the next generation . We next asked whether persistent DMRs might be functionally important to the tumor susceptibility phenotype observed in Kdm6a F1s . We examined the proximity of persistent DMRs to enhancer regions in whole bone marrow and in sorted bone marrow macrophages ( mouse ENCODE project ) ( Yue et al . , 2014 ) and in round spermatids , the last stage of spermatogenesis at which there is active transcription ( our data , Figure 3—source data 2; Figure 4—figure supplement 3 ) . We found that persistent DMRs were close to or overlapping both poised ( marked by H3K4me1 ) and active ( marked by both H3K4me1 and H3K27ac ) enhancer regions in all three of these tissues or cell types ( Figure 4H , Figure 4—figure supplement 2 ) . We then used GREAT ( Genomic Regions Enrichment of Annotations Tool ) to identify enriched phenotypes , defined by the Mouse Genome Informatics ( MGI ) phenotype ontology , associated with the set of 207 persistent DMRs ( Blake et al . , 2009; McLean et al . , 2010 ) . The top ten most strongly enriched mouse phenotypes were all related to tumorigenesis , including ‘increased classified tumor incidence’ , ‘altered tumor susceptibility’ , and ‘malignant tumors’ ( Figure 4I–J , Figure 4—figure supplements 4–5 ) . We conclude that Kdm6a-dependent hypermethylated persistent DMRs affect enhancer regions relevant to tumorigenesis in mice . We note that the edges of a ChIP-seq peak do not represent precise boundaries for functional enhancer regions , meaning that DMRs that are close to but not directly overlapping enhancers in our analysis may still affect their function , for example by altering local transcription factor binding affinities or long-range chromatin interactions ( Tiwari et al . , 2008; Yin et al . , 2017; Onuchic et al . , 2018 ) . To test the hypothesis that methylation changes persisted from sperm through the early embryo to adult tissue , we also evaluated DNA methylation changes in spleens of five control F1 and three Kdm6a F1 mice , and in liver tumors from two control F1s and two Kdm6a F1s . Of the 207 persistent DMRs detected in bone marrow , 140 ( 67% , OR 87 . 32 , p<2 . 2e-16 ) were also found in liver tumors , and 68 ( 35% , OR 408 . 65 , p<2 . 2e-16 ) were also found in spleen , and the magnitudes of DNA methylation changes were positively correlated: R = 0 . 232 ( liver ) and R = 0 . 786 ( spleen ) . The similarity of methylation changes across different tissues supports the model that these changes were present in the early embryo and persisted during lineage commitment and organ differentiation . One effect of DNA methylation at enhancers is to modulate the binding affinities of recruited transcription factors ( TFs ) , thereby altering downstream regulatory circuitry ( Yin et al . , 2017 ) . We therefore investigated the possibility that the set of persistent DMRs contains methylation-sensitive TF binding sites that can impact expression of nearby genes . We used AME ( Analysis of Motif Enrichment ) ( McLeay and Bailey , 2010 ) to find enriched TF binding motifs in the set of persistent DMRs . We detected enrichment of binding sites corresponding to the ETS transcription factors ELK1 , ELK4 , and GABPA ( Figure 5A ) . DNA methylation reduces the affinity of all three of these factors for their binding sites ( Yin et al . , 2017 ) , implying that persistent hypermethylation at these sites can impact expression of downstream genes in F1 somatic tissue . To evaluate this possibility , we collected RNA-seq data from bone marrow of healthy Kdm6a F1s ( n = 3 ) , Kdm6a F1s with abnormal histiocytic proliferation or sarcoma ( n = 2 ) , and healthy control F1s ( n = 5 ) , and looked for transcriptional signatures consistent with altered regulation by ELK1 , ELK4 , or GABPA . We called differentially expressed genes ( adjusted p-value<0 . 05 ) for healthy Kdm6a F1s vs . control F1s and for diseased Kdm6a F1s vs . control F1s separately ( Figure 5B ) . In keeping with our prediction , four of ten differentially expressed genes in healthy Kdm6a F1 bone marrow and 134 of 1404 differentially expressed genes in diseased Kdm6a F1 bone marrow were targets of the hematopoiesis-associated transcription factor RUNX2 , a direct target of ELK1 ( p=0 . 00492 and p=0 . 00102 for healthy and diseased Kdm6a F1 bone marrow , respectively , Fisher’s exact test ) ( Matys et al . , 2003; Zhang et al . , 2009 ) . An ELK1 binding site in the first intron of Runx2 falls within a persistent hypermethylated DMR and exhibits increased DNA methylation in both Kdm6a cKO sperm and Kdm6a F1 bone marrow ( Figure 5C ) . Expression of Runx2 itself was decreased in diseased Kdm6a F1 compared to control F1 bone marrow ( Figure 5D ) . Principal component analysis of expression data for the 134 differentially expressed RUNX2 target genes placed healthy Kdm6a F1 between diseased Kdm6a F1 and control F1 bone marrow , revealing potential underlying similarities in regulation of the Runx2 transcriptional network among Kdm6a F1 samples ( Figure 5E ) . Although the observed effect was small and should be confirmed in additional tissues , this result implies that altered regulation of transcriptional networks downstream of DNA methylation-sensitive transcription factors could result from persistent DNA hypermethylation transmitted from the Kdm6a cKO germ line to F1 somatic tissue .
We propose a model ( Figure 6 ) wherein loss of Kdm6a results in extensive redistribution of the H3K27me3 mark during male germ cell development . Retention of H3K27me3 at regions where it would ordinarily be turned over leaves some of these loci vulnerable to DNA methylation , leading to hypermethylation in sperm . Early in embryogenesis , when most DNA methylation is removed from the paternal genome , some of these hypermethylated regions may resist reprogramming , such that methylation persists in somatic tissue during development; alternatively , hypermethylation may be lost at these loci during reprogramming and reestablished later in development following transmission through an epigenetic intermediate . When these regions coincide with functional enhancers , the altered epigenetic state inherited from the paternal gamete can have transcriptional consequences . Importantly , because each sperm carries only one copy of a given locus , the relatively modest shift we observe in DNA methylation levels must reflect variability among individual sperm . Such variability is consistent with the heterogeneous tumor profiles and other pathological phenotypes seen in our F1 population . Similarly , in F1 somatic tissues , we propose that the effects of DNA methylation on downstream gene regulation manifest as a shift in the probability of transcription factor binding , resulting in subtle changes to transcriptional networks that impact tissue function only in the context of stressors , or cumulatively over time . Several key questions remain to be answered . First , it will be important to define the role of Uty , the Y-linked homolog of Kdm6a that lacks histone demethylase activity , in this phenomenon ( Hong et al . , 2007 ) . Second , while we have focused on adult cancer phenotypes , it is possible that additional developmental phenotypes are also affected . We did observe some developmental anomalies in adult Kdm6a progeny , including ectopic tissue rests , tail kinks , scoliosis , and a thyroglossal duct cyst . Third , the extent to which premature appearance of age-associated tumors might reflect a more generalized premature aging phenotype should be examined in more depth . It will be critical to dissect the underlying molecular mechanism in more detail . While we suggest that DNA methylation changes induced during spermatogenesis persist during reprogramming in the early embryo , we have not yet directly demonstrated that this is the case . It is also possible that DNA hypermethylation is lost during reprogramming , but that epigenetic information is transmitted through an alternative chromatin mark or RNA intermediate to reestablish DNA hypermethylation later in development . Assessment of DNA methylation in early Kdm6a F1 embryos will help to resolve this question . The relationship between Kdm6a loss , redistribution of H3K27me3 , and gain of DNA methylation also remains to be defined . The simplest explanation for our data is that dysregulation of H3K27me3 leads to DNA hypermethylation at vulnerable loci , but it is also possible that KDM6A acts through an independent mechanism to regulate DNA methylation . Determination of the stage of spermatogenesis ( proliferating spermatogonia , meiotic spermatocytes , or haploid spermatids ) at which the observed changes in H3K27me3 and DNA methylation first appear will help to delineate the relationship between the different epigenetic consequences of Kdm6a loss . Intriguingly , a recent study of H3K27me3 in mouse embryonic stem cells ( mESCs ) grown in 2i compared to serum-containing media described flattening of H3K27me3 signal very similar to the effect we observed in Kdm6a cKO sperm ( van Mierlo et al . , 2019 ) . In 2i mESCs , H3K27me3 flattening was also associated with altered DNA methylation . A closer examination of the relationship between the phenomenon we observe in sperm and that reported in 2i mESCs may shed light on the mechanisms underlying both phenomena . Finally , a critical experiment will be to examine the sperm of Kdm6a F1s to test the prediction that changes in DNA methylation at persistent DMRs are amplified in the second generation of gametes . At least two persistent DMRs are located in or near genes encoding components of the DNA methylation machinery ( Dnmt3a and Tdg ) , raising the possibility that DNA hypermethylation at these sites in sperm amplifies the changes in DNA methylation in offspring . Dnmt3a is frequently mutated in hematological tumors and has been defined as an important tumor suppressor ( Yang et al . , 2015 ) . We restricted our study to progeny of a single male founder in order to limit the amount of genetic variation in the experiment and thereby reduce the potential contribution of genetic heterogeneity , given a moderate number of experimental animals ( ~100 F1s and F2s total ) . However , our findings should be tested in a larger study including several founder males . Likewise , a larger study would allow recovery of more diseased samples for transcriptional analysis . It will also be critical to exclude the possibility that loss of KDM6A in the male germ line leads to increased DNA damage and accumulation of genomic mutations that could contribute to a tumor phenotype in the next generation . Since increased DNA damage during spermatogenesis frequently leads to meiotic arrest and impaired fertility , the normal spermatogenesis and fertility of Kdm6a cKO mice argue against a strong mutator phenotype ( Hunt and Hassold , 2002 ) . However , a more subtle effect should be ruled out by sequencing of genomic DNA in multiple F1 progeny and careful assessment of mutation rates . Virtually nothing is known about the contribution of epigenetic perturbations in the male germ line to human disease susceptibility . Specifically , while increased attention is being paid to the possible impacts of diet and environmental exposure on male fertility and epigenetic inheritance ( Anway et al . , 2005; Carone et al . , 2010; Kaati et al . , 2002; Ly et al . , 2017 ) , the role of mutations that arise in the male germ line but are not transmitted to the next generation is entirely unknown and unexplored . Spermatogenic stem cells continue to divide and to accumulate de novo mutations throughout a man’s lifetime . De novo germline mutations linked to advanced paternal age have been implicated in the pathogenesis of autism and schizophrenia; in these cases , the causative mutations arise in the germ line and are inherited by the affected progeny ( Awadalla et al . , 2010; de Kluiver et al . , 2017; Girard et al . , 2011; Iossifov et al . , 2014; Nybo Andersen and Urhoj , 2017 ) . Our results imply that de novo mutations in the male germ line in genes such as Kdm6a may have phenotypic consequences for progeny , even when they are not inherited . Intergenerational paternal effects on development have also been reported for heterozygous autosomal mutations in genes encoding chromatin regulators in the mouse ( Chong et al . , 2007 ) , suggesting that the effects of non-inherited paternal germline mutations do not depend on complete loss of gene function in the germ cells . Interestingly , a paternal age effect has been reported for ALL , a tumor shown to be sensitive to epigenetic regulation by KDM6A , but increased rates of inherited de novo mutations have not yet been demonstrated for ALL patients ( Sergentanis et al . , 2015 ) . Many patients with cancer are now being treated with drugs that target epigenetic regulators . If these drugs alter the epigenetic state of germ cells , these treatment protocols could have long-term consequences for offspring of fertile patients . Based on the findings reported here and previously ( Carone et al . , 2010; Chong et al . , 2007; Kaati et al . , 2002; Pembrey et al . , 2006; Siklenka et al . , 2015 ) , we suggest that paternal epigenetic state should be evaluated as an important risk factor in human disease susceptibility .
This experiment was designed to test the hypothesis that epigenetic changes in the germ line resulting from loss of KDM6A could induce gross phenotypic or survival changes in genetically wild type offspring . The F1 experiment was 80% powered to detect a survival hazard ratio of 2 . 5 and 90% powered to detect a 2 . 5-fold change in phenotype incidence . The F2 experiment was 90% powered to detect a survival hazard ratio of 2 . 5 and 95% powered to detect a 2 . 5-fold change in phenotype incidence . Type I error rate ( alpha ) was 5% for all power calculations . Survival hazard ratios were calculated using a Cox proportional hazards model . Fisher’s exact test was used to compare proportions . Welch’s t-test was used to compare continuous , normally-distributed variables . A Mann-Whitney U test was used for continuous variables when a normal distribution could not be assumed . All mice were maintained at the Whitehead Institute animal facility . Mice were kept under standard conditions and all experiments were conducted in compliance with the Animal Welfare act and approved by the Animal Care and Use Committee at the Massachusetts Institute of Technology . Kdm6a cKO , control , and all F1 and F2 mice were generated with breeding schemes described in the main text using Ddx4-Cre ( B6-Ddx4tm1 . 1 ( cre/mOrange ) Dcp ) ( Hu et al . , 2013 ) and Kdm6a ( fl ) ( B6;129S4-Kdm6atm1c ( EUCOMM ) Jae/J ) ( Welstead et al . , 2012 ) alleles . Experiments were carried out on a mixed C57BL/6 , 129S4 background . We controlled for background effects by generating all experimental mice from a single founder male , generating experimental F1s and F2s from littermate Kdm6a cKO and control males , and by removing loci containing known B6/129 variants from downstream analysis of DMRs . To generate F1 and F2 mice , single males were continuously co-housed with single C57BL/6 females , and litters were weaned at three weeks of age . All control and experimental mice were housed with littermates in adjacent cages on the same rack and subjected to identical handling protocols . F1 and F2 mice were checked daily for morbidity and mortality beginning at 6 months of age . Mice that died spontaneously were recovered within 24 hr to avoid autolysis and underwent a full necropsy . Mice that were independently identified by the MIT veterinary staff as requiring euthanasia due to morbidity were euthanized using CO2 and then underwent complete necropsy . For each mouse , adrenal gland , bone , bone marrow , brain , heart , small and large intestine , kidney , liver , lungs , pancreas , spleen , testes , thymus , and any additional tumors or gross abnormalities identified were embedded and sectioned , and a single representative slide was stained with hematoxylin and eosin and examined by a trained veterinary pathologist ( R . T . B . ) . The pathologist was blinded to the experimental condition of the animals ( e . g . Kdm6a F1 , control F1 , Kdm6a F2 , or control F2 ) . When possible , the entire organ was included on the slide . The complete set of conditions identified in F1 and F2 mice was tabulated once all mice had undergone necropsy . All IHC was performed on the Leica Bond III automated staining platform . Anti-CD3 ( A0452 , clone F7 . 2 . 38 , Dako , Santa Clara , CA ) was run at 1:250 dilution using the Leica Biosystems Refine Detection Kit with EDTA antigen retrieval . Anti-CD20 ( M0755 , clone L26 , Dako ) was run at 1:500 dilution using the Leica Biosystems Refine Detection Kit with citrate antigen retrieval . Anti-VEGF ( ab52917 , clone EP1176Y , Abcam ) was run at 1:100 dilution using the Leica Biosystems Refine Detection Kit with EDTA antigen retrieval . Anti-F4/80 ( MCA497GA , clone CI:A3-1 , Serotec , Hercules , CA ) was run at 1:5000 dilution using the Leica Biosystems Refine Detection Kit with enzymatic antigen retrieval . To collect bone marrow for flow cytometry analysis , RRBS , and RNA-seq , mice were euthanized by an overdose of carprofen ( 25 mg/kg ) by intraperitoneal injection . The sternum was removed and fixed in 10% formalin for histological analysis . The spinal column , pelvic bone , and both femurs , fibulas , and tibias were stripped of muscle tissue and macerated in wash buffer ( PBS + 2% FBS ) using a mortar and pestle . All liquid was pipetted off of the remaining solid tissue and passed through a 100 micrometer ( um ) filter into a 50 mL Falcon tube , then spun down at 1200 rpm for 5 min at 4C . Supernatant was removed , and cells were resuspended in 10 mL red blood cell lysis buffer ( #555899 , Becton Dickinson , Mountain View , CA ) and incubated for 5 min on ice . 20 mL wash buffer was added and the cells were passed through a 70 um filter into a fresh tube , then spun down again . The supernatant was removed , cells were resuspended in 20 mL wash buffer and passed through a 40 um filter into a fresh tube . Approximately 1 mL of this cell suspension was removed for DNA isolation for RRBS ( see below ) . The remaining suspension was spun down one more time , then resuspended in freeze solution ( 90% FBS +10% DMSO ) , aliquoted to cryotubes and stored in liquid nitrogen . Peripheral blood , bone marrow , spleen and tumor cells were analyzed using the LSRII-Fortessa instrument ( Becton Dickinson ) using anti-mouse CD11b ( clone M1/70 , BioLegend , San Diego , CA ) , anti-mouse Gr1 ( clone RB6-8C5 , BioLegend ) and LIVE/DEAD Fixable Aqua Dead Cell Stain Kit ( Life Technologies , Carlsbad , CA ) . Figures were prepared using FCSalyzer version 0 . 9 . 13 . To collect germ cell-enriched mouse testis tissue , Kdm6a cKO and control male littermates were euthanized and testes transferred to 3 cm culture dish on ice , keeping individuals separate . The tunicae were removed and 450 ul cold collagenase solution ( 0 . 1% ( w/v ) hyaluronidase ( #H3506 , Sigma Aldrich , St . Louis , MO ) , 0 . 2% ( w/v ) collagenase ( #C5138 , Sigma Aldrich ) , 1:500 DNAse I ( #07900 , Stem Cell Technologies , Vancouver , Canada ) in PBS was added . Testes tubules were teased apart using forceps for 7 min at room temperature . Liquid was removed and the sample was washed twice for 3 min in 450 ul wash solution ( 1:1000 DNAse I in PBS ) , with continued teasing . After the last wash , liquid was removed and tubules were resuspended in 700 ul trypsin solution ( 0 . 2% collagenase , 0 . 25% trypsin , 0 . 1 mM EDTA , 1:1000 DNAse I in water ) and pipetted vigorously to break up clumps . Samples were shaken for 10–15 min at room temperature and then quenched with 700 ul Cosmic Calf Serum , and any remaining tissue chunks were allowed to settle . The cell suspension was transferred to a new tube and spun down at 3000xg , 4 min , 4C . The supernatant was removed and cells were resuspended in RIPA buffer containing protease inhibitor cocktail ( sc-24948 , Santa Cruz Biotechnology ) . The protein concentration was measured with Pierce BCA protein assay kit ( #23225 , Thermo Scientific ) . 30 ug of total protein was used for each blot and was incubated overnight with primary antibody against H3K27me3 ( #07–449 , Millipore Sigma ) and GAPDH ( #sc-32233 , Santa Cruz Biotechnology ) . Blots were imaged on a FluorChem E System ( ProteinSimple , San Jose , CA ) . Relative protein expression levels were quantitated using ImageJ and normalized to GAPDH . Blots were performed in triplicate for two biological replicates . Dissociated testis cells were collected from Kdm6a cKO and control littermates as described above . The supernatant was removed and cells were resuspended in 1 mL cold resuspension solution ( 1% BSA in PBS ) . 2 ul DyeCycle Green ( #V35004 , Thermo Fisher , Waltham , MA ) was added , and the cell suspension was mixed by inversion and then incubated for 30 min at 37C in the dark . Cells were then passed through a 40 ul filter . Round spermatids were recovered by flow cytometry using a FACSJazz ( Beckton Dickinson ) after sorting for cells with 1C DNA content and large size ( to differentiate elongating from round spermatids ) . The purity of the cell population was verified by fluorescence microscopy ( ≥95% round spermatids ) and by qPCR ( Figure 4—figure supplement 3 ) . qPCR primer sequences are listed in the Key Resources table . Epididymal sperm for ChIP-seq and RRBS was collected by swim-up as follows: cauda epididymi were recovered from euthanized mice and cut 4–6 times on parafilm , then transferred to 6 cm culture dishes containing 5 mL of Donner’s medium ( 135 mM NaCl , 5 mM KCl , 1 mM MgSO4 , 2 mM CaCl2 , 30 mM HEPES , 25 mM NaHCO3 , 20 mg/mL BSA , 1 mM sodium pyruvate , 0 . 53% ( v/v ) sodium DL-lactate ) , keeping tissue from each mouse separate . Epididymes were incubated at 37C for 1 hr with periodic gentle agitation , then passed through a 40 um filter , washed 1x in cold 0 . 45% NaCl to lyse any red blood cells and 1x in cold PBS . Sperm were resuspended in PBS , and 10 ul were removed for counting following standard procedures . Pyrosequencing for three control sperm samples , three Kdm6a cKO sperm samples , three control F1 bone marrow samples , and three Kdm6a F1 bone marrow samples was performed at 13 loci by EpigenDx ( Hopkinton , MA ) according to the company’s standard protocols . EpigenDx was blinded to tissue and experimental condition . Kaplan-Meier curves were generated in R ( R Development Core Team , 2015 ) using the package rms ( Harrell , 2016 ) . Hazard ratios and p-values for survival were calculated using a Cox proportional hazards model , using the R package survival ( Therneau , 2015 ) . Transcription factor binding motifs enriched in persistent hypermethylated DMRs were identified using AME ( McLeay and Bailey , 2010 ) with the motif databases UniPROBE mouse ( 386 motifs ) ( Badis et al . , 2009; Berger et al . , 2008 ) , JASPAR CORE vertebrates ( 519 motifs ) ( Mathelier et al . , 2016 ) and human/mouse HT-SELEX ( 843 motifs ) ( Jolma et al . , 2013 ) . The full set of 100 bp tiles covered by at least 10 RRBS reads ( 263820 total ) was used as the control set . Fisher’s exact test was used to determine significance . Principal component analysis was carried out using the PCA function in the FactoMineR package in R ( Le et al . , 2008 ) . All sequencing datasets are available at GEO under accession number GSE102313 . | Many diseases , such as certain cancers , run in families . Often , this is because several related individuals inherit a version of a gene that is faulty and causes the condition . But in a number of families with high rates of cancer , scientists are unable to pinpoint such disease-causing gene versions . Instead , it is possible that individuals inherit healthy genes that are not read and interpreted correctly by the cells . This could be because of epigenetic changes , modifications that do not alter the genetic code but can instead turn genes on or off temporarily by adding or removing certain marks on the genetic information . For a long time , researchers thought that epigenetic changes could not be passed from one generation to the next , but recent studies have revealed this is actually possible . However , it had never been shown that this could be associated with having a higher risk of developing cancer . Now , Lesch et al . show that epigenetic changes passed from male mice to their offspring make these animals more likely to develop tumors than typical mice . In the experiments , mouse sperm were genetically engineered to have a mutation in a gene called Kdm6a ( also called Utx by cancer researchers ) , which controls the placement of epigenetic marks . Male mice carrying a defective Kdm6a gene were then mated to normal females . The resulting offspring developed more tumors than mice produced from normal sperm , even though they inherited a normal copy of the Kdm6a gene from their mother . Lesch et al . also show that the offspring have epigenetic marks similar to the ones found in the mutant sperm . This may change whether genes that stop or promote tumor formation are switched on or off . Certain cancer treatments work by targeting epigenetic changes . The results by Lesch et al . therefore call for more research into whether cancer patients exposed to these drugs could transmit these modifications if they have children soon after the end of their treatment . Ultimately , knowing more about how epigenetic changes are involved in inherited diseases may start to provide answers to families affected by cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2019 | Intergenerational epigenetic inheritance of cancer susceptibility in mammals |
Protein–protein interactions are fundamental to many biological processes . Experimental screens have identified tens of thousands of interactions , and structural biology has provided detailed functional insight for select 3D protein complexes . An alternative rich source of information about protein interactions is the evolutionary sequence record . Building on earlier work , we show that analysis of correlated evolutionary sequence changes across proteins identifies residues that are close in space with sufficient accuracy to determine the three-dimensional structure of the protein complexes . We evaluate prediction performance in blinded tests on 76 complexes of known 3D structure , predict protein–protein contacts in 32 complexes of unknown structure , and demonstrate how evolutionary couplings can be used to distinguish between interacting and non-interacting protein pairs in a large complex . With the current growth of sequences , we expect that the method can be generalized to genome-wide elucidation of protein–protein interaction networks and used for interaction predictions at residue resolution .
A large part of biological research is concerned with the identity , dynamics , and specificity of protein interactions . There have been impressive advances in the three-dimensional ( 3D ) structure determination of protein complexes which has been significantly extended by homology-inferred 3D models ( Mosca et al . , 2012; Webb et al . , 2014 ) ( Hart et al . , 2006; Zhang et al . , 2012 ) . However , there is still little , or no , 3D information for ∼80% of the currently known protein interactions in bacteria , yeast , or human , amounting to at least ∼30 , 000/∼6000 incompletely characterized interactions in human and Escherichia coli , respectively ( Mosca et al . , 2012; Rajagopala et al . , 2014 ) . With the rapid rise in our knowledge of genetic variation at the sequence level , there is an increased interest in linking sequence changes to changes in molecular interactions , but current experimental methods cannot match the increase in the demand for residue-level information of these interactions . One way to address the knowledge gap of protein interactions has been the use of hybrid , computational–experimental approaches that typically combine 3D structural information at varying resolutions , homology models , and other methods ( de Juan et al . , 2013 ) , with force field-based approaches such as RosettaDock , residue cross-linking , and data-driven approaches that incorporate various sources of biological information ( Kortemme and Baker , 2002; Dominguez et al . , 2003; Kortemme and Baker , 2004; Kortemme et al . , 2004; Svensson et al . , 2004; Chaudhury et al . , 2011; Schneidman-Duhovny et al . , 2012; Velazquez-Muriel et al . , 2012; Karaca and Bonvin , 2013; Rodrigues et al . , 2013; Webb et al . , 2014 ) . However , most of these approaches depend on the availability of prior knowledge and many biologically relevant systems remain out of reach , as additional experimental information is sparse ( e . g . , membrane proteins , transient interactions , and large complexes ) . One promising computational approach is to use evolutionary analysis of amino acid co-variation to identify close residue contacts across protein interactions , which was first used 20 years ago ( Gobel et al . , 1994; Pazos and Valencia , 2001 ) , and subsequently used also to identify protein interactions ( Pazos et al . , 1997; Pazos and Valencia , 2002 ) . Others have used some evolutionary information to improve a machine learning approach to developing docking potentials ( Faure et al . , 2012; Andreani et al . , 2013; Andreani and Guerois , 2014 ) . These previous approaches relied on a local model of co-evolution that is less likely to disentangle indirect and therefore incorrect correlations from the direct co-evolution , as has been described in work on residue–residue interactions in single proteins ( Marks et al . , 2012 ) . More recently , reports using a global model have been successful in identifying residue interactions from evolutionary co-variation , for instance between histidine kinases and response regulators ( Burger & van Nimwegen , 2008; Skerker et al . , 2008; Weigt et al . , 2009 ) , and this approach has only recently been generalized and used to predict contacts between proteins in complexes of unknown structure , in an independent effort parallel to this work ( Ovchinnikov et al . , 2014 ) . In principle , just a small number of key residue–residue contacts across a protein interface would allow computation of 3D models and provide a powerful , orthogonal approach to experiments . Since the recent demonstration of the use of evolutionary couplings ( ECs ) between residues to determine the 3D structure of individual proteins ( Marks et al . , 2011; Morcos et al . , 2011; Aurell and Ekeberg , 2012; Jones et al . , 2012; Kamisetty et al . , 2013 ) , including integral membrane proteins ( Hopf et al . , 2012; Nugent and Jones , 2012 ) , we reason that an evolutionary statistical approach such as EVcouplings ( Marks et al . , 2011 ) could be used to determine co-evolved residues between proteins . To assess this hypothesis , we built an evaluation set based on all known binary protein interactions in E . coli that have 3D structures of the complex as recently summarized ( Rajagopala et al . , 2014 ) . We develop a score for every predicted inter-protein residue pair based on the overall inter-protein EC score distributions resulting in accurate predictions for the majority of top ranked inter-protein EC pairs ( inter-ECs ) and sufficient to calculate accurate 3D models of the complexes in the docked subset ( Figure 1A ) . This approach was then used to predict evolutionary couplings for 32 complexes of unknown 3D structures that have a sufficient number of sequences . Predictions include the structurally unsolved interactions between the a- , b- , and c-subunits of ATP synthase , which are supported by previously published experimental results . 10 . 7554/eLife . 03430 . 003Figure 1 . Co-evolution of residues across protein complexes from the evolutionary sequence record . ( A ) Evolutionary pressure to maintain protein–protein interactions leads to the co-evolution of residues between interacting proteins in a complex . By analyzing patterns of amino acid co-variation in an alignment of putatively interacting homologous proteins ( left ) , evolutionary couplings between co-evolving inter-protein residue pairs can be identified ( middle ) . By defining distance restraints on these pairs , the 3D structure of the protein complex can be inferred using docking software ( right ) . ( B ) Distribution of E . coli protein complexes of known and unknown 3D structure where both subunits are close on the bacterial genome ( left ) , allowing sequence pair matching by genomic distance . For a subset of these complexes , sufficient sequence information is available for evolutionary couplings analysis ( dark blue bars ) . As more genomic information is created through on-going sequencing efforts , larger fractions of the E . coli interactome become accessible for EVcomplex ( right ) . A detailed version of the workflow used to calculate all E . coli complexes currently for which there is currently enough sequence information is shown in Figure1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00310 . 7554/eLife . 03430 . 004Figure 1—figure supplement 1 . Details of the EVcomplex Pipeline . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 004
To compute co-evolution across proteins , individual protein sequences must be paired up with each other that are presumed to interact , or being tested to see if they interact . Without this condition , proteins could be paired together that do not in fact interact with each other and therefore detection of co-evolution would be compromised . Given that the evolutionary couplings method depends on large numbers of diverse sequences ( Hopf et al . , 2012 ) , some assumption must be made about which proteins interact with each other in homologous sequences in other species . Since it is challenging to know a priori whether particular interactions are conserved across many millions of years in thousands of different organisms , we use proximity of the two interacting partners on the genome as a proxy for this , with the goal of reducing incorrect pairings . To assemble the broadest possible data sets to test the approach and make predictions , we take all known interacting proteins assembled in a published data set that contains ∼3500 high-confidence protein interactions in E . coli ( Rajagopala et al . , 2014 ) . After removing redundancy and requiring close genome distance between the pairs of proteins this results in 326 interactions , see ‘Materials and methods’ ( Figure 1B , Figure 1—figure supplement 1 , Supplementary file 1 and 2 ) , The paired sequences are concatenated and statistical co-evolution analysis is performed using EVcouplings ( Marks et al . , 2011; Morcos et al . , 2011; Aurell and Ekeberg , 2012 ) , that applies a pseudolikelihood maximization ( PLM ) approximation to determine the interaction parameters in the underlying maximum entropy probability model ( Balakrishnan et al . , 2011; Ekeberg et al . , 2013; Kamisetty et al . , 2013 ) , simultaneously generating both intra- and inter-EC scores for all pairs of residues within and across the protein pairs ( Figure 1A ) . Evolutionary coupling calculations in previous work have indicated that this global probability model approach requires a minimum number of sequences in the alignment with at least 1 non-redundant sequence per residue ( Marks et al . , 2011; Morcos et al . , 2011; Hopf et al . , 2012; Jones et al . , 2012; Kamisetty et al . , 2013 ) . Our current approach allows complexes with fewer available sequences to be assessed ( minimum at 0 . 3 non-redundant sequences per residue ) by using a new quality assessment score to assess the likelihood of the predicted contacts to be correct . The EVcomplex score is based on the knowledge that most pairs of residues are not coupled and true pair couplings are outliers in the high-scoring tail of the distribution ( See ‘Materials and methods’ , Figure 2A , B , Figure 2—figure supplement 1 and 2 ) . The score can intuitively be understood as the distance from the noisy background of non-significant pair scores , normalized by the number of non-redundant sequences and the length of the protein ( ‘Materials and methods’ , equations 1 and 2 ) . If the number of sequences per residue is not controlled for , there is a large bias in the results , overestimating performance with low numbers of sequences ( Figure 2B , C ) . The precise functional form of the correction for low numbers of sequences was chosen non-blindly after observing the dependencies in the test set . 10 . 7554/eLife . 03430 . 005Figure 2 . Evolutionary couplings capture interacting residues in protein complexes . ( A ) Inter- and Intra-EC pairs with high coupling scores largely correspond to proximal pairs in 3D , but only if they lie above the background level of the coupling score distribution . To estimate this background noise a symmetric range around 0 is considered with the width being defined by the minimum inter-EC score . For the protein complexes in the evaluation set , this distribution is compared to the distance in the known 3D structure of the complex that is shown here for the methionine transporter complex , MetNI . ( Plots for all complexes in the evaluation set are shown in Figure 2—figure supplement 1 and 2 . ) ( B ) A larger distance from the background noise ( ratio of EC score over background noise line ) gives more accurate contacts . Additionally , the higher the number of sequences in the alignment the more reliable the inferred coupling pairs are which then reduces the required distance from noise ( different shades of blue ) . Residue pairs with an 8 Å minimum atom distance between the residues are defined as true positive contacts , and precision = TP/ ( TP + FP ) . The plot is limited to range ( 0 , 3 ) which excludes the histidine kinase—response regulator complex ( HK–RR ) —a single outlier with extremely high number of sequences . ( C ) To allow the comparison across protein complexes and to estimate the average inter-EC precision for a given score threshold independent of sequence numbers , the raw couplings score is normalized for the number of sequences in the alignment , resulting in the EVcomplex score . In this work , inter-ECs with an EVcomplex score ≥0 . 8 are used . Note: the shown plot is cut off at a score of 2 in order to zoom in on the phase change region and the high sequence coverage outlier HK-RR is excluded . ( D ) For complexes in the benchmark set , inter-EC pairs with EVcomplex score ≥0 . 8 give predictions of interacting residue pairs between the complex subunits to varying accuracy ( 8 Å TP distance cutoff ) . All predicted interacting residues for complexes in the benchmark set that had at least one inter-EC above 0 . 8 are shown as contact maps in Figure 2—figure supplement 3–8 . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00510 . 7554/eLife . 03430 . 006Figure 2—figure supplement 1 . Distribution and accuracy of raw EC scores for all complexes in evaluation set . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00610 . 7554/eLife . 03430 . 007Figure 2—figure supplement 2 . Distribution and accuracy of raw EC scores for all complexes in evaluation set ( 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00710 . 7554/eLife . 03430 . 008Figure 2—figure supplement 3 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 . Predicted coevolving residue pairs with an EVcomplex score ≥0 . 8 and all inter-ECs up to the rank of the last include inter-EC are visualized in complex contact maps ( red dots: inter-ECs , green and blue dots: intra-ECs for monomer 1 and 2 , respectively ) . Top left and bottom right quadrants: intra-ECs; top right and bottom left quadrants: inter-ECs . Inter- and intra-protein crystal structure contacts at minimum atom distance cutoffs of 5/8/12 Å are shown as dark/middle/light gray dots , respectively; missing data in the crystal structure as shaded blue rectangles . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00810 . 7554/eLife . 03430 . 009Figure 2—figure supplement 4 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 ( 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 00910 . 7554/eLife . 03430 . 010Figure 2—figure supplement 5 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 ( 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 01010 . 7554/eLife . 03430 . 011Figure 2—figure supplement 6 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 ( 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 01110 . 7554/eLife . 03430 . 012Figure 2—figure supplement 7 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 ( 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 01210 . 7554/eLife . 03430 . 013Figure 2—figure supplement 8 . Contact maps of all complexes with solved 3D structure with inter-ECs above EVcomplex score of 0 . 8 ( 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03430 . 013
A primary limitation of our current approach is its dependence on the availability of a large number of evolutionarily related sequences . If a protein interaction is conserved across enough sequenced genomes , using a single pair per genome can give accurate predictions of the interacting residues . However , if the protein pair is present in limited taxonomic branches , there may be insufficient sequences at any given time to make confident predictions . A solution to this could be to include multiple paralogs of the interacting proteins from each genome , but this requires correct pairing of the interaction partners , which is in general hard to ascertain . In addition , details of interactions may have diverged for paralogous pairs . Hence , in this current version of the method , we have imposed a genome distance requirement across all genomes for all homolog pairs in order to be less sensitive to these complications . As the need to use genome proximity to pair sequences becomes less important with the increasing availability of genome sequences , there will be a dramatic increase in the number of interactions that can be inferred from evolutionary couplings , including those unique to eukaryotes . With currently available sequences ( May 2014 release of the UniProt database ) , EVcomplex is able to provide information for about 1/10th of the known 3000 protein interactions in the E . coli genome . Once there are ∼10 , 000 bacterial genome sequences of sufficient diversity , one would have enough information to test each potentially interacting pair of homologs for evidence of interaction and , given sufficiently strong evolutionary couplings , infer the 3D structure of each protein–protein pair , as well as of complexes with more than two proteins . For any set of species , e . g . , vertebrates or mammals , one can imagine guiding sequencing efforts to optimize species diversity to facilitate the extraction of evolutionary couplings . This can open the doors for more comprehensive and more rapid determination of approximate 3D structures of proteins and protein complexes , as well as for the elucidation in molecular detail of the most strongly evolutionarily constrained interactions , pointing to functional interactions . Determining the three-dimensional models of complexes from the predicted contacts was successful in many of the tested instances . Using minimal computing resources and a small number of inter-EC-derived contacts , low interface positional RMSDs relative to experimental structures can be achieved . However , a significant number of proteins exist as homomultimers within larger complexes . To determine models of these complexes one must deconvolute homomultimeric inter-ECs from the intra-protein signal , which is an important technical challenge for future work . The analysis of subunit interactions in ATP synthase in this work is a ‘proof of principle’ study showing that methods such as EVcomplex can determine which proteins interact with each other at the same time as specific residue pair couplings across the proteins ( as also shown in the work by the Baker lab on ribosomal protein interactions ( Ovchinnikov et al . , 2014 ) ) . Understanding the networks of protein interactions is of critical interest in eukaryotic systems , such as networks of protein kinases , GPCRs , or PDZ domain proteins . An understanding of the distributions of interaction specificities is of high interest to many fields . Although we do not know how well our evolutionary coupling approach will handle less obligate interactions , results on the two-component signaling system ( histidine kinase/response regulator ) both here and in other work ( Skerker et al . , 2008; Weigt et al . , 2009 ) suggest optimism . The approximately scale-free EVcomplex score is a heuristic based on the distribution of raw EC scores from the statistical model , their dependence on sequence alignment depth and the length of the concatenated sequences . The score provides a simple way of accounting for these dependencies such that a uniform threshold , say 0 . 8 , can be used for any protein pair with the expectation of reasonably accurate predictions . Since cutoff thresholds can be useful but overly sharp , we recommend investigating predicted contacts below the threshold used in this work , especially where there is independent biological knowledge to validate the predictions . The work presented here is in anticipation of a genome-wide exploration and , as a proof of principle , shows the accurate prediction of inter-protein contacts in many cases and their utility for the computation of 3D structures across diverse complex interfaces . As with single protein ( intra-EC ) predictions , evolutionarily conserved conformational flexibility and oligomerization can result in more than one set of contacts that must be de-convoluted . Can evolutionary information help to predict the details and extent for each complex ? A key challenge will be the development of algorithms that can disentangle evolutionary signals caused by alternative conformations of single complexes , alternative conformations of homologous complexes , and effectively deal with false positive signals . Taken together , these issues highlight fruitful areas for future development of evolutionary coupling methods . Despite conditions for the successful de novo calculation of co-evolved residues , the method described here may accelerate the exploration of the protein–protein interaction world and the determination of protein complexes on a genome-wide scale at residue level resolution . The use of co-evolutionary analysis in computational models to determine protein specificity and promiscuity , co-evolutionary dynamics and functional drift will open up exciting future research questions .
The candidate set of complexes for testing and de novo prediction was derived starting from a data set of binary protein–protein interactions in E . coli including yeast two-hybrid experiments , literature-curated interactions and 3D complex structures in the PDB ( Rajagopala et al . , 2014 ) . Three complexes not contained in the list were added based on our analysis of other subunits in the same complex , namely BtuC/BtuF , MetI/MetQ , and the interaction between ATP synthase subunits a and b . Since our algorithm for concatenating multiple sequence pairs per species assumes the proximity of the interacting proteins on the respective genomes of each species ( See below ) , we excluded any complex with a gene distance >20 from further analysis . The gene distance is calculated as the number of genes between the interacting partners based on an ordered list of genes in the E . coli genome obtained from the UniProt database . The resulting list of pairs ( ∼350 ) was then filtered for pseudo-homomultimeric complexes based on the identification of Pfam domains in the interacting proteins ( 330 ) . All remaining complexes with a known 3D structure ( as summarized in Rajagopala et al . , 2014 ) or a homologous interacting 3D structure ( 93 ) ( identified by intersecting the results of HMMER searches against the PDB for both monomers ) were used for evaluating the method , while complexes without known structure ( 236 ) were assigned to the de novo prediction set ( Figure 1—figure supplement 1 ) . The set with protein complexes of known 3D structure was further filtered for structures that only cover fragments ( <30 amino acids ) of one or both of the monomers and structures with very low resolution ( >5 Å ) , which led to the re-assignment of Ribonucleoside-diphosphate reductase 1 ( complex_002 ) , Type I restriction-modification enzyme EcoKI ( complex_012 ) , RpoC/RpoB ( complex_041 ) , RL11/Rl7 ( complex_165 ) , the ribosome with SecY ( complex_226 , complex_250 , and complex_255 ) , and RS3/RS ( complex_254 ) to the set of unknown complexes . Large proteins were run with the specific interacting domains informed by the known 3D structure , when the full sequence was too large for the number of retrieved sequences ( for domain annotation see ‘Supplementary data’ ) . This set could serve as a benchmark set for future development efforts in the community . Each protein from all pairs in our data set was used to generate a multiple sequence alignment ( MSA ) using jackhmmer ( Johnson et al . , 2010 ) to search the UniProt database ( UniProt Consortium , 2014 ) with 5 iterations . To obtain alignments of consistent evolutionary depths across all the proteins , a bit score threshold of 0 . 5 * monomer sequence length was chosen as homolog inclusion criterion ( -incdomT parameter ) , rather than a fixed E-value threshold which selects for different degrees of evolutionary divergence based on the length of the input sequence . In order to calculate co-evolved residues across different proteins , the interacting pairs of sequences in each species need to be matched . Here , we assume that proteins in close proximity on the genome , e . g . , on the same operon , are more likely to interact , as in the methods used previously matching histidine kinase and response regulator interacting pairs ( Skerker et al . , 2008; Weigt et al . , 2009 ) ( ‘Supplementary data’ ) . We retrieved the genomic locations of proteins in the alignments and concatenated pairs following 2 rules: ( i ) the CDS of each concatenated protein pair must be located on the same genomic contig ( using ENA ( Pakseresht et al . , 2014 ) for mapping ) and ( ii ) each pair must be the closest to one another on the genome , when compared to all other possible pairings in the same species . The concatenated sequence pairs were filtered based on the distribution of genomic distances to exclude outlier pairs with high genomic distances of more than 10k nucleotides ( ‘Supplementary data’ ) . Alignment members were clustered together and reweighted if 80% or more of their residues were identical ( thus implicitly removing duplicate sequences from the alignment ) . Supplementary file 1 and 2 report the total number of concatenated sequences , the lengths , and the effective number of sequences remaining after down-weighting in the evaluation and de novo prediction set , respectively . Inter- and intra-ECs were calculated on the alignment of concatenated sequences using a global probability model of sequence co-evolution , adapted from the method for single proteins ( Marks et al . , 2011; Morcos et al . , 2011; Hopf et al . , 2012 ) using a pseudo-likelihood maximization ( PLM ) ( Balakrishnan et al . , 2011; Ekeberg et al . , 2013 ) rather than mean field approximation to calculate the coupling parameters . Columns in the alignment that contain more than 80% gaps were excluded and the weight of each sequence was adjusted to represent its cluster size in the alignment thus reducing the influence of identical or near-identical sequences in the calculation . For the evaluation set , we can then compare the predicted ECs for both within and between the proteins/domains to the crystal structures of the complexes ( for contact maps and all EC scores , see ‘Supplementary data’ ) . In order to estimate the accuracy of the EC prediction , we evaluate the calculated inter-ECs based on the following observations: ( 1 ) most pairs of positions in an alignment are not coupled , i . e . , have an EC score close to zero , and tend to be distant in the 3D structure; ( 2 ) the background distribution of EC scores between non-coupled positions is approximately symmetric around a zero mean; and ( 3 ) higher-scoring positive score outliers capture 3D proximity more accurately than lower-scoring outliers ( See also Figure 2 ) . The width of the ( symmetric ) background EC score distribution can be approximated using the absolute value of the minimal inter-EC score . The more a positive EC score exceeds the noise level of background coupling , the more likely it is to reflect true co-evolution between the coupled sites . For each inter-protein pair of sites i and j with pair coupling strength ECinter ( i , j ) , we therefore calculate a raw reliability score ( 'pair coupling score ratio' , Figure 2B ) defined by ( 1 ) Qinterraw ( i , j ) =ECinter ( i , j ) |mini , j ( ECinter ( i , j ) ) | Since the accuracy of evolutionary couplings critically depends both on the number and diversity of sequences in the input alignment and the size of the statistical inference problem ( Marks et al . , 2011; Morcos et al . , 2011; Jones et al . , 2012 ) , we incorporate a normalization factor to make the raw reliability score comparable across different protein pairs . The normalized EVcomplex score is defined as ( 2 ) EVcomplex−Score ( i , j ) =Qinterraw ( i , j ) 1+ ( NeffL ) −12where Neff is the effective number of sequences in the alignment after redundancy reduction , and L ( total number of residues ) is the length of the concatenated alignment . Previous work on single proteins has shown that the method requires a sufficient number of sequences in the alignment to be statistically meaningful . We thus filter for sequence sufficiency requiring Neff/L > 0 . 3 ( Table 1 , Supplementary files 1 and 2 ) . Predictions of coupled residues in the evaluation set were evaluated against their residue distances in known structures of protein pairs ( Rajagopala et al . , 2014 ) ( See Supplementary file 7 ) in order to determine the precision of the method . To interpret the EVcomplex prediction of interaction between subunits a and b of the ATP synthase as well as UmuC and UmuD , individual monomer models were built de novo for the structurally unsolved a-subunit of ATP synthase and UmuC using the EVfold pipeline as previously published ( Marks et al . , 2011; Hopf et al . , 2012 ) . In both cases coupling parameters were calculated using PLM ( Balakrishnan et al . , 2011; Ekeberg et al . , 2013 ) and sequences were clustered and weighted at 90% sequence identity ( the resulting models are provided in ‘Supplementary data’ ) . Following this same protocol EVcomplex scores were calculated for all possible 28 combinations of the 8 E . coli ATP synthase F0 and F1 subunits . Since we want to compare the computational predictions to some ‘ground truth’ , as with the complexes for the rest of the manuscript , we used known 3D structures of the ATP synthase complex to assign whether or not the subunits interact ( PDB: 3oaa , 1fs0 , 2a7u; Supplementary file 7 ) . Since we are also determining whether the subunits interact , not necessarily knowing full atomic detail residue interactions , we included subunit interactions that have been inferred from cryo-EM , crosslinking or other experiments , but do not necessarily have a crystal structure . These are represented as solid blue boxes , if the interaction is well established ( DeLeon-Rangel et al . , 2013; Schulenberg et al . , 1999; Brandt et al . , 2013; McLachlin and Dunn , 2000 ) , or crosshatched blue if there is a lack of consensus in the community ( Figure 6B , left panel ) . For each possible interaction the EVcomplex score of the highest ranked inter-EC was considered as a proxy for the likelihood of interaction . Pairs with scores above 0 . 8 are considered likely to interact , between 0 . 75 and 0 . 8 weakly predicted , while interactions with scores below 0 . 75 are rejected as possible complexes ( blue boxes , blue crosshatched , and white respectively in right panel of Figure 6B , and ‘Supplementary data’ ) . A diverse set of 15 complexes was chosen from the 22 in the evaluation set that had at least 5 couplings above a complex score of 0 . 8 and was subsequently docked ( Supplementary file 3 ) . Proteins that have been crystallized together in a complex could bias the results of the docking , as they have complementary positions of the surface side chains . Therefore , where possible we used complexes that had a solved 3D structure of the unbound monomer , namely GcsH/GcsT , CyoA , FimC , DhaL , AtpE , PtqA/PtqB , RS10 , and HK/RR , and in all other cases the side chains of the monomers were randomized either by using SCWRL4 ( Krivov et al . , 2009 ) or restrained minimization with Schrodinger Protein Preparation Wizard ( Sastry et al . , 2013 ) before docking . For ubiquinol oxidase ( complex_054 ) the unbound structure of subunit 2 ( CyoA ) only covers the COX2 domain . In this case docking was performed using this unbound structure plus an additional run using the bound complex structure with perturbed side chains . We used HADDOCK ( Dominguez et al . , 2003 ) , a widely used docking program based on ARIA ( Linge et al . , 2003 ) and the CNS software ( Brunger , 2007 ) ( Crystallography and NMR System ) , to dock the monomers for each protein pair with all inter-ECs with an EVcomplex score of 0 . 8 or above implemented as distance restraints on the α-carbon atoms of the backbone . Each docking calculation starts with a rigid-body energy minimization , followed by semi-flexible refinement in torsion angle space , and ends with further refinement of the models in explicit solvent ( water ) . 500/100/100 models generated for each of the 3 steps , respectively . All other parameters were left as the default values in the HADDOCK protocol . Each protein complex was run using predicted ECs as unambiguous distance restraints on the Cα atoms ( deff 5 Å , upper bound 2 Å , lower bound 2 Å; input files available in ‘Supplementary data’ ) . As a negative control , each protein complex was also docked using center of mass restraints alone ( ab initio docking mode of HADDOCK ) ( de Vries et al . , 2007 ) and in this case generating 10 , 000/500/500 models . Each of the generated models is scored using a weighted sum of electrostatic ( Eelec ) and van der Waals ( Evdw ) energies complemented by an empirical desolvation energy term ( Edesolv ) ( Fernandez-Recio et al . , 2004 ) . The distance restraint energy term was explicitly removed from the equation in the last iteration ( Edist3 = 0 . 0 ) to enable comparison of the scores between the runs that used a different number of ECs as distance restraints . All computed models in the docked set were compared to the cognate crystal structures by the RMSD of all backbone atoms at the interface of the complex using ProFit v . 3 . 1 ( http://www . bioinf . org . uk/software/profit/ ) . The interface is defined as the set of all residues that contain any atom <6 Å away from any atom of the complex partner . For the AtpE–AtpG complex , we excluded the 2 C-terminal helices of AtpE as these helices are mobile and take many different positions relative to other ATP synthase subunits ( Cingolani and Duncan , 2011 ) . Similarly , since the DHp domain of histidine kinases can take different positions relative to the CA domain , the HK-RR complex was compared over the interface between the DHp domain alone and the response regulator partner . In the case of the unbound ubiquinol oxidase docking results , only the interface between COX2 in subunit 2 and subunit 1 was considered . Accuracy of the computed models with EC restraints was compared with computed models with center of mass restraints alone ( negative controls ) ( Figure 3—figure supplement 1 , Supplementary file 3 ) . Data analysis was conducted primarily using IPython notebooks ( Perez and Granger , 2007 ) . A webserver and all data are available at EVcomplex . org and the Dryad Digital Repository ( Hopf et al . , 2014 ) . ( Available at http://doi . org/10 . 5061/dryad . 6t7b8 [Hopf et al . , 2014] ) Supplementary data 1: Concatenated alignments for complexes predicted in this work Supplementary data 2: Genome distance distribution of concatenated sequences per alignment Supplementary data 3: EVcomplex predictions for evaluation and de novo set Supplementary data 4: Docking input files and top 10 predicted models for evaluation set Supplementary data 5: ATP synthase predictions , ATP synthase subunit a model Supplementary data 6: UmuC model | DNA is often referred to as the ‘blueprint of life’ , as this molecule contains the instructions that are required to build a living organism from a single cell . But these instructions largely play out through the proteins that DNA encodes; and most proteins do not work alone . Instead they come together in different combinations , or complexes , and a single protein may participate in many complexes with different activities . Proteins are so small that it is difficult to get clear information about what they look like . Visualizing protein complexes is even harder . Most protein–protein interactions remain poorly understood , even in the best-studied organisms such as humans , yeast , and bacteria . Proteins are made from smaller molecules , called amino acids , strung together one after the other . The order in which different amino acids are arranged in a protein determines the protein’s shape and ultimately its function . Like DNA , protein sequences can change over time . Sometimes , the sequence of one protein changes in a way that prevents it binding to another protein . If these two proteins must work together for an organism to survive , the second protein will often develop a compensating change that allows the protein–protein complex to reform . Identifying pairs of changes in the sequences of pairs of proteins suggests that the two proteins interact and gives some information about how the proteins fit together . Different species can have copies of the same proteins that have slightly different sequences . Since the DNA sequences from many different organisms are already known , there are now many opportunities to find sites in pairs of proteins that have evolved together , or co-evolved , over time . To find sites that seem to have co-evolved , Hopf et al . used a computer program based on an approach from statistical physics to look at pairs of proteins that were already known to form complexes . Co-evolving sites were found in over 300 pairs of proteins; including 76 where the structure of the complex was already known . When sites that were predicted to be co-evolving were then mapped to these known complex structures , the co-evolving sites were remarkably close to the true protein–protein contacts . This indicates that the information from the co-evolved sequences is sufficient to show how two proteins fit together . Hopf et al . then turned their attention to 82 pairs of proteins that were thought to interact , but where a structure was unavailable . For 32 of these pairs , structures of the entire complex could be predicted , showing how the two proteins might interact . Furthermore , when other researchers subsequently worked out the structure of one of these complexes , the prediction was a good match to the solved complex structure . The machinery of life is largely made up of proteins , which must interact in ever-changing but precise ways . The new methods developed by Hopf et al . provide a new way to discover and investigate the details of these interactions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Sequence co-evolution gives 3D contacts and structures of protein complexes |
Animals can learn causal relationships between pairs of stimuli separated in time and this ability depends on the hippocampus . Such learning is believed to emerge from alterations in network connectivity , but large-scale connectivity is difficult to measure directly , especially during learning . Here , we show that area CA1 cells converge to time-locked firing sequences that bridge the two stimuli paired during training , and this phenomenon is coupled to a reorganization of network correlations . Using two-photon calcium imaging of mouse hippocampal neurons we find that co-time-tuned neurons exhibit enhanced spontaneous activity correlations that increase just prior to learning . While time-tuned cells are not spatially organized , spontaneously correlated cells do fall into distinct spatial clusters that change as a result of learning . We propose that the spatial re-organization of correlation clusters reflects global network connectivity changes that are responsible for the emergence of the sequentially-timed activity of cell-groups underlying the learned behavior .
The mechanisms of memory formation have been the subject of considerable study ( Morris et al . , 1988; Kandel , 2001 ) . Much evidence points to Hebbian plasticity as the neural mechanism for the association of two co-occurring stimuli ( Bliss and Collingridge , 1993; Morris , 2003 ) . However , this mechanism alone is not sufficient to account for learning under conditions where the two stimuli are separated in time by more than 100 ms ( Levy and Steward , 1983 ) , as has been commonly observed ( Solomon et al . , 1986; Baeg et al . , 2003 ) . One example of such a time-bridging task is trace eyeblink conditioning , where the goal is to associate a neutral tone or conditioned stimulus ( CS ) with a temporally separated , aversive puff of air to the eye , or unconditioned stimulus ( US ) ( McEchron and Disterhoft , 1999 ) . After many pairings of these stimuli , the subject learns to blink in response to the tone , even though the tone and puff never overlap in time ( Tseng et al . , 2004; Kalmbach et al . , 2009 ) . Lesion studies have shown that the hippocampus is required for learning the trace conditioning task , but not a related delay conditioning task , where the CS and US overlap in time ( Büchel et al . , 1999; Tseng et al . , 2004 ) . These observations indicate a role for the hippocampus during the association of temporally discontiguous events ( Wallenstein et al . , 1998 ) , specifically during the ‘trace’ period separating stimulus pairs . How might a network of neurons maintain a representation of the stimulus through time ? Two possible models have been proposed . The first model hypothesizes that the representation of the first stimulus is maintained by the sustained firing of stimulus-selective cells through a trace interval ( Solomon et al . , 1986 ) . Such a model is supported by observations of sustained firing by neurons in the medial prefrontal cortex ( Fuster , 1973; Baeg et al . , 2003 ) and the medial entorhinal cortex ( Egorov et al . , 2002 ) during working memory tasks . An alternative model proposes that sensory representations are maintained by the sequential activation of groups of neurons ( Levy et al . , 2005; Howe and Levy , 2007; MacDonald et al . , 2013 ) . This view arose from modeling studies that trained simple hippocampal area CA3 network models on the trace conditioning task . Rather than observing sustained firing , the authors found that groups of neurons began to show activity in well-timed , sequential bouts . Neurons representing the CS kicked off this ‘relay’ of activation , which eventually activated US representing neurons at the appropriate time ( Levy et al . , 2005; Howe and Levy , 2007 ) . However , this model awaits experimental verification . Sequential activity in hippocampal CA1 cells has been seen previously , albeit in some very different behavioral contexts , including temporal memory tasks ( Louie and Wilson , 2001; Pastalkova et al . , 2008; Gill et al . , 2011 ) . There has also been a series of recent , more closely related studies of hippocampal activity , in which rats or monkeys performed stimulus-retention tasks where they had to remember an odor or visual stimulus in order to receive a reward . Here too , hippocampal CA1 neurons were observed to be active in stimulus-triggered , time-locked sequences ( MacDonald et al . , 2011; Naya and Suzuki , 2011; Kraus et al . , 2013; MacDonald et al . , 2013 ) . It is increasingly clear that hippocampal CA1 cells adopt sequential activity patterns when subjects are placed in a behavioral context requiring the bridging of temporally separated stimuli . However , in all the studies where this has been observed , only well-trained subjects were used , leaving the time course and mechanism of the emergence of sequentially timed activity entirely unknown . Despite sequential activity having been implicated in several temporal memory tasks , there is little experimental data on the network changes that underlie its emergence . Functional connectivity , as measured by correlations between neuronal activity in the absence of stimulus presentation , is one way to monitor such network changes ( Ts’o et al . , 1986; Bair et al . , 2001; Fujisawa et al . , 2008 ) . In a study using two photon calcium imaging of motor cortex pyramidal neuron activity , changes in spontaneous activity correlations have been inferred to indicate learning-related circuit plasticity ( Komiyama et al . , 2010 ) . In another study where rats were exploring a novel track , correlations between pairs of place cells increased with increasing exposure to the novel environment ( Cheng and Frank , 2008; Dragoi and Tonegawa , 2013 ) . With the ability to measure changes in input from upstream circuits using spontaneous activity correlations across many cells in the network , two-photon recordings allow the testing of predictions from the model proposed for the emergence of sequential activity ( Levy et al . , 2005; Howe and Levy , 2007 ) . Furthermore , such recordings provide relative cell locations within the hippocampus , allowing one to examine the spatial organization of activity patterns ( Hampson et al . , 1999; Brivanlou et al . , 2004; Kjelstrup et al . , 2008 ) . As learning progresses , one should see changes in spontaneous correlations reflecting altered inputs and changes in network connectivity . Two key questions remain unanswered in the absence of data recorded from large numbers of neurons during training on a temporal memory task – how do sequential activity representations in the hippocampus emerge during learning , and what are the underlying changes in network connectivity ? In this study , we train mice on a trace eyeblink conditioning task while recording activity from populations of area CA1 neurons using two photon calcium imaging . In order to eliminate the influence of running or changing spatial position on hippocampal activity , we implemented a trace eye-blink conditioning task for head-fixed mice . Further , our activity measurements revealed CA1 network dynamics during learning , as we began with naïve mice and trained them to criterion within the recording session . We found that sequentially timed activity of groups of area CA1 cells emerged progressively during the course of learning . Additionally , mean spontaneous activity correlations at the network level rose transiently , while only the correlations between co-tuned cells remained elevated towards the end of the session . Finally , we observed spatially organized clusters of neurons that had elevated spontaneous activity correlations . These correlation clusters re-organized during learning .
Head-restrained mice were trained on a trace eyeblink conditioning task ( Tseng et al . , 2004 ) , where they learned to associate a neutral tone stimulus ( Conditioned Stimulus–CS ) with an aversive puff of air to the eye ( Unconditioned Stimulus–US ) within a single session ( ‘Materials and methods-Behavioral training’ ) . Tone and puff were non-overlapping and separated by a 250 ms interval , thus requiring the subject to maintain a representation or ‘trace’ of the tone ( CS ) in order to associate it with the puff ( US ) . Behavioral responses were measured by recording deflections in eyelid position ( Figure 1B ) . We found that naïve mice responded to tone presentation with small , but distinct and measurable eyelid movements early in training , even on trials prior to the introduction of the puff stimulus ( Figure 1C ) . Following repeated pairings of tone and puff , however , blink responses to tone increased significantly ( >2 standard deviations [SD] ) in amplitude and duration , ( n = 9 of 18 mice , Figure 1B , D ) . Conditioned Response ( CR ) trials were defined as those trials in the training session whose area under the eyelid–position curve , during the interval spanning tone onset to puff onset , was significantly larger than the pre-training baseline ( ‘Materials and methods-Behavioral training’ ) . We next established a performance score by measuring the ratio of CR blink rates to spontaneous blink rates ( ‘Materials and methods-Data analysis’ , Figure 1E , F ) . As a learning control , we randomized the relative timing of tone and puff from trial to trial , for a different set of mice . These pseudo-conditioned mice did not demonstrate an increase in blink amplitude or duration ( performance scores , trace = 8 . 74 ± 2 . 73 , pseudo = 1 . 96 ± 0 . 44 , mean ± standard error of the mean ( SEM ) ; Figure 1E , F , two-sample t test , p=0 . 012 ) . 9 of 18 trace conditioned mice had performance scores higher than those of the pseudo-conditioned mice . We next examined the performance of each mouse , to assess which individual mice learned the task , and if so , at what trial number in the training session . We used a previously described expectation maximization algorithm to assess whether each individual mouse had learned the association , and to obtain learning curves ( Smith et al . , 2004 ) . Briefly , the algorithm uses the list of CR trials for a given mouse , along with the chance probability of the occurrence of a well-timed , significant blink to estimate the probability of CR production at each trial in the session ( Figure 1G ) . Furthermore , the trial at which a mouse has learned the task is statistically defined . It is the first trial when the lower 95% confidence interval of the probability of CR production rises and remains above chance . As per this criterion , 9 out of 18 trace conditioned mice learned the association . The mean of the individual learning trials was 26 ± 5 ( mean ± SD , n = 9 learners; for six of these learner mice , imaging data was also acquired , Video 1 shows high-speed video of mouse blinks before and after learning ) . These were the same nine mice that also had higher performance scores than pseudo-conditioned controls . 10 . 7554/eLife . 01982 . 004Video 1 . Mouse Behavior . Video of mouse eyeblink behavior acquired at 100 frames per second ( fps ) and played back at 10 fps ( i . e . 0 . 1x speed ) . A yellow spot in the bottom-right corner indicates when tone is being delivered , and a red spot indicates when the air-puff is being delivered . Frames from three trials , at different points in the session are shown , depicting behavior early in the session , prior to learning , and late in the session after learning . The last portion of the video shows eyeblink behavior in a probe-trial , where the tone was presented but no air-puff was delivered . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 004 Half the mice trained on the trace eyeblink conditioning task failed to learn to criterion . In part , this was because we were restricted to training mice for only a single session due to the limited residence time of calcium indicator dye in cells ( Stosiek et al . , 2003 ) . These non-learners most likely represent a heterogeneous population of mice at different stages of learning involving multiple brain regions ( Kalmbach et al . , 2009 ) . In other words , given more training sessions , many of these mice would likely have learned this task to criterion . Consequently , the interpretation of area CA1 calcium imaging data from ‘non-learners’ is complicated by the uncertain state of learning of each mouse . Hence , in most analyses , we have not used data from these mice . However , for completeness , we have included data from non-learners in key figures . We surgically exposed the left , dorsal hippocampus of naïve mice and bolus loaded a synthetic calcium indicator dye . We then implanted cranial windows through which we imaged calcium responses from cells in area CA1 of the hippocampus ( Figure 2A , Figure 2—figure supplement 1A , Video 2 shows calcium responses from a sample field of view ) . Image acquisition was carried out within a field of view covering 96 ± 29 ( mean ± SD ) cells per mouse , imaged at frame rates ranging from 11 Hz to 16 Hz . In parallel with calcium imaging , we simultaneously measured eyeblink responses of mice over the entire pre-training and training sections of the conditioning protocol ( n = 14 trace conditioned and six pseudo-conditioned mice; ‘Materials and methods-Awake , two photon calcium imaging of area CA1 cells’; Figure 2—figure supplement 1B ) . Cell bodies were imaged from the visually identified stratum pyramidale layer of the hippocampus ( Figure 2—figure supplement 1C ) , at depths ranging between 135–150 µm below the hippocampal surface ( supporting Video 2 shows imaged calcium responses and Video 3 shows a stack of optical , z-section images with the densely labeled cell-body layer visible ) . To ensure the reliability of our calcium fluorescence data , we carried out two checks . First , the frequency of calcium transients was observed to be 1 . 3 ± 0 . 2 Hz ( mean ± SD ) . While this need not accurately reflect spike rates in our experimental system , it falls well within the range of spontaneous spike rates previously observed in CA1 cells , and was un-changed between early and late trials in the session ( Figure 2—figure supplement 1D; Czurkó et al . , 1999 ) . Second , the summed area under the calcium curve was also calculated for all datasets , and those datasets that showed a significant drift between early and late trials were discarded ( n = 1 of 21; Figure 2—figure supplement 1E ) . 10 . 7554/eLife . 01982 . 005Figure 2 . Two-photon imaging of calcium-responses in area CA1 neurons from awake mice . ( A ) Schematic of the imaging preparation . o–objective lens , s–skull , cs–cover slip , hb–head bar , ag–agarose . ( B ) Histogram of neuron response widths in ms , calculated as the time for which a given neuron’s trial-averaged , ΔF/F trace remains above 50% of the peak value . The red , dotted line indicates a response width of 600 ms , which is the time of interest between tone-onset and puff-onset . ( C ) Area CA1 cell responses show sequentially timed activity peaks after learning . Calcium response ( ΔF/F ) traces for six exemplar neurons from a single mouse , for sets of three trials before ( panel on the left ) , and after ( panel on the right ) task learning . Neurons have been sorted as per the timing of the peak in the averaged trace . The yellow and red bars at the bottom represent the times of delivery of tone and puff respectively . The gray shading to the left covers the period of spontaneous activity prior to the onset of the tone . The red asterisks indicate the peak in each individual trace . Scale bars indicate 0 . 4 ΔF/F and 500 ms along the time axis . ( D ) Area CA1 cell activity peaks tile the entire CS-on to US-off interval . Area CA1 calcium response traces from an example dataset , sorted by the peak times of the responses . Each response trace has been averaged over all trials following the learning trial ( Figure 1G ) , and has been normalized to the peak ΔF/F response value for each neuron . The yellow and red bars below indicate times of delivery of tone and air-puff respectively . 50% of the neurons from the field of view , with the most reliably timed responses have been shown . This is to make this plot comparable to the ones from subsequent analyses , where neurons have been similarly chosen . ( E and F ) Cell activity peak timings change during learning . In E , pseudo-colored ΔF/F traces for the period of interest during and after tone delivery , are plotted using data acquired during the pre-training session , where tones without air-puff were delivered . Cells have been sorted as per the timings of peaks in pre-training session data . The yellow bar at the bottom indicates time of delivery of tone ( 350 ms ) . In F , the same averaged activity traces as in E have been re-ordered according to each cell’s activity peak timing after learning has occurred , as shown in ( D ) Plotted in Figure 2—figure supplement 1 , are panels depicting the surgical preparation , the numbers of mice from each treatment group , images of dye-loaded tissue taken at multiple depths and basic data quality control analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 00510 . 7554/eLife . 01982 . 006Figure 2—figure supplement 1 . Imaging preparation and calcium response data . ( A ) Left: Top-view of a head-fixed , awake mouse . a: white outline of custom-designed head bar affixed to the dorsal surface of the skull using skull-screws and dental acrylic . s: exposed skull . c: craniotomy of diameter ∼ 1 mm . Green fluorescence is due to OGB-1 dye loaded into the dorsal hippocampus . Right: OGB-1 dye-loading in the dorsal hippocampus in a coronal brain section . ( B ) A Venn-diagram indicating the numbers of mice ( black shapes ) trained on the trace conditioning task ( light blue , dotted border , labeled Tr ) , that learned the task ( blue , solid border , labeled L ) , that were pseudo-conditioned ( green , dotted border , labeled Ps ) and from whom calcium imaging data was obtained ( red , solid border , labeled Im ) . ( C ) A selection of imaged optical slices from a bolus loaded volume in hippocampal area CA1 . Scale to the left indicates the depth of each slice in microns , relative to an image of the hippocampal surface . The scale bar is 50 µm in size . Interneurons superficial to the cell-body layer are visible in the second slice from the top . The third slice depicts densely packed cell bodies ( CA1 pyramidal cells ) in the stratum pyramidale , imaged to acquire calcium response data . ( D and E ) Mean number of detected calcium peak events ( D ) and summed area under the ΔF/F curve ( E ) have been plotted for the first and last quarter of trials , as measures of recording stability . Datasets showing significant changes as per both metrics were discarded . For this , frames from a 4 s time window , starting 5 s after stimulus delivery ( assumed to be background activity ) was used . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 00610 . 7554/eLife . 01982 . 007Video 2 . Calcium responses . Video showing a time-series of images of a single field of view of area CA1 cells . Brighter colors in the gray-scale indicate higher fluorescence intensities . Flashes visible are calcium responses . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 00710 . 7554/eLife . 01982 . 008Video 3 . z-stack of dye-loaded hippocampal tissue . Video showing optical sections through a typical , dye-loaded imaging preparation of the dorsal hippocampus . Optical sections were acquired beginning at the dorsal hippocampal surface and moving ventrally in steps of 2 µm/frame . The scale bar represents 50 µm . The densely-packed cell-bodies of the curved , stratum pyramidale cell-body layer appear near the 7 s time-point in the video . The frames of this video show some motion—they were not motion corrected as each frame was taken at a different depth and thus , differed from the others . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 008 We first looked for evidence of sustained or sequentially timed neuronal activity in calcium fluorescence traces for individual cells ( ΔF/F , ‘Materials and methods: Data analysis’ ) , on trials following the CR learning peak . First , we determined whether or not cells showed sustained activity through the entire period of interest by measuring how wide each cell’s response was at half the amplitude of the peak of the response . The mean response width measured in this manner was found to be 120 ms ± 81 ms ( mean ± SD , n = 542 cells , Figure 2B ) . Only three cells ( ∼0 . 5% , of all cells ) had response widths greater than 600 ms , the length of the period from tone-onset to puff-onset . Based on this analysis , we concluded that sustained cell activation is not the mechanism used by the hippocampus to maintain a stimulus representation . We next looked for indications of sequential activation . Prior to learning , calcium response peaks were not reliably timed across trials , relative to tone onset ( Figure 2C , left panel ) . Post learning , however , we observed that area CA1 cells had activity peaks at fixed time-points relative to tone onset , seen consistently across multiple trials ( Figure 2C , right panel ) . We further characterized this by averaging these ΔF/F activity traces over trials and identifying activity peaks within the time window from tone onset to 200 milliseconds after puff onset . We rank ordered cells based on the timings of their activity peaks and found that small groups of cells firing at each time-point , tiled the entire interval of interest ( Figure 2D ) . When the same procedure ( averaging followed by sorting by timing ) was carried out on traces from the pre-training dataset , tiling was markedly skewed with most cells ( 66% in this case ) showing activity peaks during the tone period ( Figure 2E ) . Furthermore , if neurons in the pre-training dataset were ordered as per timing of peaks in the training session , no clear tiling was visible ( Figure 2F ) . This indicated that timings of peak activity of area CA1 cells , relative to the onset of tone stimulus , changed during training . Furthermore , at the population level , these peak times appeared to tile the interval of interest between tone and puff . Having observed that area CA1 cells were active at fixed time-points relative to tone onset , we next wanted to quantify the reliability with which cells fired at these times . For each cell , we defined the timing of the peak in its averaged ΔF/F trace as its peak response time ( PT ) . We then quantified the reliability with which cells fired at their respective PTs , from trial to trial ( Figure 3A shows a cell reliably firing at a particular PT ) . First , we reasoned that if cells show time-aligned activity , the peak of the trial-averaged ΔF/F trace would be higher than if activity were not reliably timed . Hence , we computed the peaks of the averaged ΔF/F traces during the tone-onset to puff-onset period ( Figure 3—figure supplement 1A ) . The mean peak amplitude for cells from trace conditioned mice was significantly higher than for cells from pseudo-conditioned mice or for spontaneous activity data ( trace = 0 . 022 ± 0 . 007 , pseudo = 0 . 017 ± 0 . 009 , spontaneous = 0 . 015 ± 0 . 008; mean ± SEM; one way Analysis of Variance ( ANOVA ) , followed by Tukey Kramer honest significant difference ( h . s . d . ) p<0 . 01 ) . The difference seen was a small one , but this measure does not control for differences in cell response sizes or average activity . Hence , to more rigorously characterize activity timing reliability , we computed a reliability score for each cell . 10 . 7554/eLife . 01982 . 009Figure 3 . After training , area CA1 cells show reliably timed , sequential calcium responses . ( A ) Calcium response traces for an example neuron , aligned to the time of stimulus delivery ( tone and puff indicated by yellow and red bars at bottom respectively ) . Warmer colors in the traces indicate higher ΔF/F values . The blue rectangle on the ‘trial number’ axis indicates the trial averaging window comprising all trials after the learning trial . ( B ) The same data as in A , except with each trial’s ΔF/F trace given a random time offset . ( C ) Averaged calcium response curves obtained from aligned , as well as random time-offset traces . The averaged curve from time-aligned traces ( blue curve ) has a prominent peak , which is absent in the random time offsets case ( red curve ) . This indicates that the neuron fires reliably at a fixed time relative to stimulus delivery . The area under the shaded region ( peak ± 1 frame ) was used to calculate the reliability score . ( D ) Area CA1 cells from trace learners show significantly higher activity-timing reliability scores . Average reliability scores for all neurons in the entire dataset for learners of the trace-conditioning task ( blue ) , pseudo-conditioned mice ( green ) spontaneous activity data ( red ) and data from non-learners ( cyan; * indicates p<0 . 01 ) . ( E ) Change in reliability score with learning for neurons from trace conditioned ( blue ) , pseudo-conditioned ( green ) and spontaneous activity data ( red ) respectively . Reliability of firing at the final peak response time ( PT ) gradually increased over the training session . Reliability scores were computed in five-trial bins . The shaded regions represent SEM . ( F ) Change in reliability scores between early and late blocks of training trials . The increase over the training session for trace-learners was significantly higher than for spontaneous activity data or for pseudo-conditioned mice ( * indicates p<0 . 01 ) . This increase was calculated by subtracting reliability scores of the first half of the session from those of the second half . ( G and H ) Distributions of single-cell peak response times ( PT ) at different stages of learning: pre-training ( G ) and early in training ( H ) . The distribution shifts from one showing distinct peaks at the time of tone delivery in the pre-training stage ( G ) , to one showing a peak at the time of the air-puff during early training ( H ) . Only the PTs of cells with high reliability scores were included in these plots . The red and yellow bars at the bottom indicate times of delivery of tone and air puff respectively . ( I ) Distribution of single cell peak response times ( PT ) after the learning trial in trace conditioned mice . For trials after the learning trial , PT is uniformly distributed across all times in the trace interval between tone and puff . Only the PTs of cells with significant reliability scores were included in this plot . ( J ) Average , time-decoder performance score , pooled across all trace conditioned mice ( blue ) , pseudo-conditioned mice ( green ) , spontaneous background activity ( red ) and non-learners ( cyan ) . Dotted line indicates chance level scores , error bars indicate SEM ( * indicates p<0 . 01 ) . Figure 3—figure supplement 1 , presents further characterization of the reliability score increase . A schematic explaining the time-decoder algorithm is depicted in Figure 3—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 00910 . 7554/eLife . 01982 . 010Figure 3—figure supplement 1 . Activity peak times change during learning . ( A ) Peak heights in trial-averaged ΔF/F traces are greater for cells from trace conditioned mice than controls . ΔF/F traces from tone onset to air-puff onset were averaged across all trials after learning and peak heights in these traces were determined and plotted for trace ( blue bar ) , pseudo-conditioned ( green bar ) and spontaneous activity data ( red bar ) respectively ( * indicates p<0 . 01 ) . ( B ) Distribution of reliability scores for cells pooled from all mice that learned the trace conditioning task . ( C ) Reliability score depends on the number of trials averaged or the bin-width . In Figure 3D , F of the main text , reliability scores were calculated by averaging different numbers of trials ( trial bin widths of 24 [on average] and 5 respectively ) . In this figure , we plotted the cell-averaged reliability score calculated using trial a range of trial bin-widths . The point highlighted with a blue circle indicates the bin-width used for Figure 3D and the red-highlighting indicates the bin-width used for Figure 3E . In Figure 3D , E of the main text , reliability scores were calculated by averaging different numbers of trials ( trial bin widths of 24 [on average] and 5 respectively ) . The size of the highest peak in the averaged trace from randomly time-shuffled trials depends on the number of time-shuffled trials averaged . Hence , the reliability score depends on the trial-bin width used to calculate it . As seen in this panel , the highest cell-averaged reliability scores for trial bin-widths of 5 and 25 closely match the reliability scores seen in Figure 3D , F respectively . ( D ) Reliability scores for neuronal responsiveness at locally determined peak times do not change with learning . This plot shows the mean ( solid line ) and standard error ( shaded areas ) of the reliability score for neuron activity at a peak-timing determined locally , within each five-trial window . The blue , green and red curves are for neurons from trace learners , pseudo-conditioned mice and spontaneous activity respectively . ( E ) Activity peak timings progressively change towards the final peak timings . We computed the local activity peak timing within 10-trial windows over the entire session . Plotted here is the mean , absolute difference between these local peak timings and the final activity peak timings for cells from trace learners ( blue ) , pseudo-conditioned mice ( green ) , spontaneous activity ( red ) and non-learners ( cyan ) . The solid curves depict mean values and the shaded areas , SEM . Final activity peak timings were computed using trials after the learning trial . ( F ) Activity peak timings early in the session are different from final , learning-related peak timings . Activity peak timings were calculated for the early half of the session by averaging activity traces for trials 1 to 25 . The final peak timing for each cell was determined by averaging every alternate trial from trials 26 to 50 . These two sets of peak-timings were then compared with the final , post-learning peak timings determined by averaging a non-overlapping set of trials from the 26–50 trial window . The absolute differences in peak timing have been plotted for trace learners ( blue ) , pseudo-conditioned mice ( green ) , spontaneous activity ( red ) and non-learners ( cyan ) . For trace learners , the mean timing-difference is significantly higher for peak timings determined early in the session ( solid colored bars ) than for those determined in the latter half ( bars with black hatch pattern ) . Peak timing difference does not decrease for all controls . The red line indicates the average frame time across datasets ( * indicates p<0 . 01 ) . ( G ) Cell-groups sharing peak activity timing are randomly distributed in the field of view . Neuron masks from an example field of view in a trace learner mouse , color-coded as per the timing ( frame number ) of peak calcium fluorescence within the tone to puff period . Scale bar represents 50 microns . ( H ) Average , pair-wise distances between neurons sharing the same peak timing ( left ) and random neuron pairs ( right ) are plotted , with error-bars marking SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 01010 . 7554/eLife . 01982 . 011Figure 3—figure supplement 2 . Time-decoder algorithm . Schematic explaining the use of a template matching time decoder . Trials after peak behavioral trial were alternately split into two groups—one training set averaged to obtain a template of expected population response vectors for each time-point and another test group of trials used to calculate performance . To compute the performance score , we first measured the similarity ( dot-product ) of every test trial frame , with all expected population response vectors . Performance scores ( PS ) were then determined by computing the ratio of the dot product for the correct frame to a weighted mean of the dot products for the incorrect ones ( ‘Experimental Procedures’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 011 The reliability score was computed by comparing each cell’s time-tuning to that of control data obtained by artificially disrupting any trial to-trial timing relationships in calcium activity . We achieved this by pseudo-randomly shifting single-trial ΔF/F traces for each cell in time ( Figure 3B ) . For every cell , a reliability score was then computed as the ratio of the peak area of averaged ΔF/F traces to the peak area of averaged , random time-offset traces ( Figure 3C ) . Hence , the reliability score refers to the size of a time-aligned peak , expressed in multiples of that seen after random shifting . To aid in the intuitive understanding of this score , we report here the reliability score of the example neuron whose traces have been shown in Figure 3A , which is clearly firing reliably at a particular time ( RS = 2 . 5 ) . That is , the area under the peak of the averaged curve for this cell is more than double that obtained after giving random offsets to individual trials . Reliability scores for area CA1 cells in mice that learned the task ( n = 542 cells from 6 mice ) , were significantly higher than reliability scores for neurons from pseudo-conditioned mice ( n = 498 cells from 6 mice ) , non-learners ( n = 612 cells from 8 mice ) and for spontaneous activity data from the trace learners ( trace = 3 . 12 ± 0 . 20 , pseudo = 1 . 35 ± 0 . 08 , spontaneous = 0 . 84 ± 0 . 07 , non- learners = 1 . 72 ± 0 . 13; mean ± SEM; ANOVA followed by h . s . d . , p<0 . 01; Figure 3D ) . Since the peak timing reliability scores for cells from trace learners were significantly higher than those for controls , we concluded that only neurons from trace conditioned , learner mice had reliably time-tuned peaks in the ΔF/F trace . On plotting the distribution of all reliability scores obtained for cells from trace learners , we found that 44% of neurons had reliability scores close to 1 , that is they displayed no time-locked activity ( Figure 3—figure supplement 1B ) . Hence , for all subsequent analysis involving time-tuned activity , 50% of the neurons with the lowest reliability scores were discarded so that we could focus on the cells that did have significant time-tuning . The same treatment ( calculation of reliability scores and then discarding the poorest 50% ) was also given to control data from the spontaneous activity period , from pseudo-conditioned mice , or from non-learner mice . Having seen that reliable time-tuning emerges only in trace conditioned mice we next examined the emergence of time-tuning during training . We quantified how the reliability in peak-timing changed as the training session progressed . The same reliability score ( Figure 3A–C ) was calculated for the 50% most reliable neurons in five-trial blocks ( ‘Materials and methods: Data analysis’ ) . We found that the reliability scores for neurons from trace learners increased over the training session ( Figure 3E ) and that this increase was significantly greater than that seen in pseudo-conditioned mice or in spontaneous activity data from trace conditioned mice ( Figure 3F , ANOVA followed by Tukey–Kramer h . s . d . , p<0 . 01 ) . Interestingly , the first trial bin where the increase in reliability scores is seen is for trials 21–25 . This coincides well with the mean behavioral learning trial ( 24 ± 1 for imaged mice , 26 ± 2 for all mice mean ± SEM ) . The peak reliability score seen in Figure 3E is lower than that seen in Figure 3D . This is because the number of trials used to calculate the score in each case is different ( bin-widths of 5 and 24 respectively ) . The size of the peaks in the averaged trace from randomly time-shuffled trials depends on the number of trials averaged together since averaging time-shuffled traces would cause isolated peaks to collapse in proportion to the number of trials averaged . On the other hand , peaks aligned in time across trials would reinforce each other upon averaging . We computed reliability scores for different bin sizes and found that mean reliability score increases with the number of trials used to compute it ( Figure 3—figure supplement 1C ) . This explains the difference in peak reliability scores seen in Figure 3D , E . We concluded that time-selective area CA1 cell activity emerged progressively , as the mouse learned an association between two stimuli separated in time . Furthermore , the time-course of their emergence matched the average time taken by mice to learn the association . We next considered two possible phenomena that might account for the improvement of reliability scores with learning . First , a cell may have a fixed activity peak timing from the beginning of the session , but not fire on every trial . In such a case , if the cell gradually started firing at the same , fixed peak-timing on a greater fraction of trials , its reliability score would increase . The cell in Figure 3A is such a cell . Alternatively , the timing of peak activity of the cell might change ( e . g . , the cells in Figure 2C ) . Such a cell too , would show an increase in reliability score . We therefore computed scores that distinguished between these phenomena , to establish that the predominant effect of learning involves changes in cell activity peak timings , and not their response probabilities . First , we assessed the extent to which increases in trial to trial response probability affected reliability scores , ignoring changes in peak timing . We calculated local reliability scores as training progressed ( Figure 3—figure supplement 1D ) . Here , the timings at which reliability scores were calculated were re-estimated for each five-trial bin . Hence , if peak-timing were to shift during training , it would not affect the reliability score . On the other hand , if cells began to show activity on a larger proportion of trials , the score would increase . On average , no training-related increases in local reliability score were detectable . This indicates that most cells do not increase their response probability with training . We next quantified how local activity peak timings for each neuron differed from the peak timings found after learning . Half the trials after the learning trials were used to estimate post-learning activity peak timings for all cells . Then , we computed local peak timings for 10-trial bins through the entire session and computed the absolute difference between these local peak timings and the post-learning peak timings . These differences served to estimate how far , on average , cell peak timings were from their values after learning . We observed a gradual decrease in absolute peak timing difference as training progressed ( Figure 3—figure supplement 1E ) for cells from learners . Such a decrease was not seen for peak timing differences computed for cells from pseudo-conditioned mice , spontaneous activity data or non-learners . We also calculated the same peak timing differences , binning all trials before and after learning . The difference in peak timings from the post-learning peak timings was high prior to learning , and reduced to the order of a single frame-time post learning ( Figure 3—figure supplement 1F ) for trace learners . For mice that learned the association , the absolute difference in peak timings prior to , and post learning , were significantly different ( ANOVA followed by Tukey Kramer h . s . d , p<0 . 01 ) . No such decrease in peak timing difference was seen for any of the control cases . Hence , we concluded that reliability scores increase as learning progresses and that this increase is mainly due to changes in the peak-timings of cell activity during learning . To characterize the changes in peak timings of cell populations due to learning , we next examined the distributions of the PTs of all neurons pooled across trace conditioned mice , at various stages of learning . The distribution was clearly non-uniform both before ( χ2 goodness of fit test with uniform expected distribution; χ2 = 69 . 14 , p<10−4 , Figure 3G ) as well as at an early stage of trace conditioning ( χ2 = 61 . 92 , p<10−4 , Figure 3H ) . However , for the group of trials after the learning trial , the distribution was much closer to a uniform distribution ( χ2 = 20 . 53 , p=0 . 015 , Figure 3I ) . Thus , the highly non-uniform PT distributions prior to and early in training became more uniform , effectively tiling the entire period of interest as a result of trace conditioning . If neuronal activity is indeed sequential and time-locked , it should be possible to use activity data to decode time since tone onset . Thus , as an independent test of time-locking of activity , we formulated a simple time-decoding algorithm based on template matching ( Fenton and Muller , 1998; Zhang et al . , 1998; ‘Materials and methods: Data analysis’; Figure 3—figure supplement 2 ) . The decoder uses single-frame , ΔF/F fluorescence values from cell-populations on single trials to decode time elapsed since tone presentation . The time-decoder prediction accuracy scores ( normalized to decoder accuracy for control , time-shuffled data ) for data from trace learners were significantly higher than those for data from pseudo-conditioned mice , spontaneous activity data from trace learners and data from non-learners ( trace = 2 . 28 ± 0 . 27 , pseudo = 1 . 26 ± 0 . 10 , spontaneous = 1 . 37 ± 0 . 06 , non-learners = 1 . 59 ± 0 . 16 , mean ± SEM; ANOVA , followed by Tukey Kramer h . s . d . , p<0 . 01; Figure 3J ) . As with the activity timing reliability score , the time decoder prediction score is also expressed in multiples of the score achieved using temporally randomized , control data . Hence , a decoder score of 2 would indicate prediction capability twice as good as that obtained with randomized data . To aid an intuitive understanding of the score , we also calculated the percentage of decoder frame-number predictions that were within one frame of the correct frame . For activity data from trace learners , the percentage was 70 . 3% , for pseudo-conditioned mice it was 32 . 4% , for spontaneous activity data it was 47 . 6% , and for non-learners it was 53 . 1% . The mean correct prediction percentage by random chance was 28 . 8% . The mean performance scores reported in Figure 3J closely match the values obtained by taking the ratio of these prediction percentages to the chance prediction percentage . We also checked if neurons sharing time-tuning were spatially clustered within the hippocampus . The mean pair-wise distance between neurons sharing time-tuning was not significantly different from the mean distance between random pairs of neurons , that is neurons sharing time-tuning were not spatially clustered together ( mean ± SEM distance for same time-tuning 71 . 27 ± 1 . 53 µm , random 70 . 89 ± 0 . 62 µm; Figure 3—figure supplement 1G , H; two-sample t test p=0 . 73 ) . Together , these data suggested that prior to training , area CA1 cells responded in a manner that was stimulus locked . These cells began to show reliably timed , sequential activity when mice were learning a task that required the maintenance of a sensory representation through a trace time-interval . Training caused a shift in the distribution of these activity peak times towards more uniform coverage of the interval of interest , spanning tone and puff stimulus durations as well as the intervening trace period . Trial by trial correlations in spontaneous activity ( noise correlations or NC ) have been used previously as a measure of functional connectivity and have been thought to indicate the presence of direct ( monosynaptic ) or indirect , polysynaptic connections between correlated neurons ( Ts’o et al . , 1986; Bair et al . , 2001; Fujisawa et al . , 2008 ) . Noise correlations , as measured by slower calcium recordings , are also thought to be indicative of shared presynaptic input , and therefore , changes in correlations are thought to imply changes in common input ( Komiyama et al . , 2010 ) . We calculated correlation coefficients between the spontaneous activity traces of pairs of neurons ( more than 4 s away from any stimuli ) , ( Figure 4A , dashed , blue curve ) . Interestingly , for trace conditioned learners , mean NC increased during the training session , with a peak at trial bin 16-20 ( 19 ± 2 , mean ± SEM ) . This peak precedes the peak in time-tuning ( trial bin 21–25 , Figure 3E ) and the averaged learning trial ( 24 ± 1 for imaged mice , trial 26 ± 2 , for all mice; mean ± SEM ) . This relationship was also observed at the level of individual mice ( Figure 4—figure supplement 1A , mean learning trial = 24 ± 1 , trial of peak reliability score = 25 ± 2 , peak NC trial = 19 ± 2 mean ± SEM; ANOVA p=0 . 03 ) . 10 . 7554/eLife . 01982 . 012Figure 4 . Area CA1 cell noise-correlations increase transiently during training . ( A ) Average neuron-pair noise correlations plotted as a function of training trials for task learner mice ( blue curves ) , pseudo-conditioned mice ( green curves ) and non-learners ( cyan curves ) . Solid lines represent average noise correlations between neurons that share similar time tuning ( same PT ) , whereas dashed lines indicate average noise correlations between random neuron pairs . Pair-wise noise correlations have been determined from spontaneous activity traces over five trial windows . Shaded regions indicate SEM . ( B ) Summary statistics comparing average noise correlations across early ( trials 1–10 ) , middle ( trials 21–30 ) and late ( trials 36–45 ) stages of training , between task learner mice ( blue bars ) and pseudo-conditioned mice ( green bars ) . Solid bars represent average noise correlations between neurons that share similar time tuning ( same PT ) whereas hatched bars represent average noise correlations between random neuron pairs . Error bars represent SEM . ( * indicates within condition [same time-tuning v/s random cell-pairs] , # indicates across conditions [trace learners v/s pseudo-conditioned] and @ indicates comparisons across stages of learning [early v/s middle v/s late] , p<0 . 01; n . s . indicates not significant ) . Figure 4—figure supplement 1A depicts the point in the session at which behavioral CR rates , CA1 cell timing reliability and spontaneous activity correlations reach their peaks , on an individual mouse basis . It also presents further characterization of the changes in NC during learning . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 01210 . 7554/eLife . 01982 . 013Figure 4—figure supplement 1 . Peaks of CR rate curve , CA1 activity timing reliability and NC; characterization of NC changes during learning . ( A ) Plots for individual mice , showing the trial number ( or middle of trial-bin ) of the training session when the learning trials , peaks in mean reliability scores of time-locked activity and mean spontaneous activity correlations occurred . For mean reliability scores and noise-correlations , the centers of the peak trial-bins have been plotted . Circles of a given color indicate values for a given mouse that learned the association . * indicates p<0 . 05 . ( B ) Noise correlations are higher for neurons with high activity timing reliability scores . Neurons were classified into two groups based on their activity timing reliability scores . The spontaneous activity correlations were calculated for pairs of cells sharing the same activity peak timing . Plotted here are the mean spontaneous activity correlations for high and low reliability score groups . * indicates p=0 . 004 . ( C ) Stimulus period activity sequences are not detectable in spontaneous activity . Spontaneous period activity traces for each neuron were given time offsets equal to the neurons peak timing in the stimulus period sequence . Thereafter , correlation coefficients were calculated for all neuron pairs . If similar sequences were re-capitulated during spontaneous activity , these correlations would be high . As a control , traces were given offsets randomly chosen from the existing peak timings and the same calculations carried out . The correlations of peak time offset activity traces were not significantly different from those for randomly offset traces ( n . s . –p=0 . 71 ) . ( D ) Mouse blink sizes remain stable over the training session . Blink size was estimated by calculating the area under the eyelid position curve during the tone-onset to puff-onset period for significant blink trials . The mean blink sizes before and after trial 25 of the training session were calculated across mice and compared and found not to be significantly different ( n . s . –p=0 . 62 ) . ( E ) Mouse blink peak latencies remain stable through the session . The latencies of the peak of the eyelid position trace from the time of tone onset were measured for significant blink trials . The mean blink latencies before and after trial 25 of the training session were calculated across mice and compared and found not to be significantly different ( n . s . –p=0 . 80 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 013 Furthermore , the NC calculated for cell-pairs where both cells had the same activity peak timing was significantly higher than for random cell-pairs ( same time-tuning = 0 . 22 ± 0 . 01 , random pairs = 0 . 14 ± 0 . 01 ( mean ± SEM ) , ANOVA followed by Tukey–Kramer h . s . d . , p<0 . 01 , n = 56 peak-timing points with an average of 25 cell-pairs in each group; Figure 4B ) . Towards the end of the session , NC declined , although , even at this stage , NC for neurons sharing the same peak times remained significantly higher than for randomly chosen cell pairs ( same PT = 0 . 16 ± 0 . 01 , random pairs = 0 . 10 ± 0 . 01; mean ± SEM; ANOVA followed by Tukey–Kramer h . s . d . , p<0 . 01; Figure 4B ) . Cells from pseudo-conditioned mice showed no such increase in NC ( same PT = 0 . 13 ± 0 . 02 , random pairs = 0 . 12 ± 0 . 01; mean ± SEM; ANOVA followed by Tukey–Kramer h . s . d . , p>0 . 01; n = 48 time-tuning points with an average of 25 cell-pairs in each; Figure 4B ) . We also checked if the spontaneous activity correlations were higher for cells with high activity peak timing reliability scores . Neurons with high reliability scores had significantly higher noise correlations than neurons with low reliability scores ( Figure 4—figure supplement 1B; high RS 0 . 24 ± 0 . 03 , low RS 0 . 14 ± 0 . 02; mean ± SEM; * indicates p<0 . 01 , two sample t test ) . Finally , we checked if activity sequences seen during the stimulus period were re-capitulated during spontaneous activity . We gave the spontaneous period activity trace of each neuron a time-offset equal to the relative timing of its activity peak timing within the stimulus period sequence . Then , we calculated the correlation coefficients between all such peak-time offset traces . If the same sequence indeed recurred during spontaneous activity , these correlations would be higher . As a control , we gave these spontaneous activity traces randomly chosen time-offsets from the set of peak-timings observed . Correlations for peak-timing offset traces were not significantly different from those for randomly offset traces ( Figure 4—figure supplement 1C; peak-time offset 0 . 048 ± 0 . 0012 , random offset 0 . 047 ± 0 . 0009 , means ± SE; two sample t test , p=0 . 71 ) . However , it must be pointed out that if sequence replay events are rare in comparison to spontaneous activity of area CA1 cells , then the correlation analysis described above , coupled with the limited size of our dataset ( to minimize calcium indicator dye photo-bleaching ) might not have the statistical power to detect them . Furthermore , spontaneous , sequential replay activity could still be occurring , but at an accelerated time-scale , as has been observed before ( Lee and Wilson , 2002 ) . Such fast replay events would not be resolvable , given our temporal resolution capability . Surprisingly , spontaneous activity correlations for trace learners were observed to drop towards pre-training levels in the second half of the session ( Figure 4A , B ) . This has been observed previously in the context of rats repeatedly exploring a novel track ( Cheng and Frank , 2008 ) . Here , the reduction in spontaneous activity correlations was ascribed to a reduction in the novelty of the particular congruence of stimuli related to the track being explored . Hence , the reduction in noise correlations we observe could also be a post-learning phenomenon , related to a reduction in the novelty of the specific congruence of the paired CS and US stimuli . To support this interpretation , we ruled out possible technical contributors to the reduction in correlations . First , the reduction in correlations were not due to a gradual decline in the quality of optical activity recordings , since we found that recordings remained stable through the training session , as discussed in ‘Results: Cells in hippocampal area CA1 imaged from awake , behaving mice exhibit temporal tuning’; Figure 2—figure supplement 1D , E . Second , this effect was not due to mice paying less attention to the task or fatiguing towards the end of the session . We tested this by measuring the sizes and peak latencies of the significant blinks before and after trial 25 , where the decline in spontaneous correlations was seen to begin . Mean significant blink sizes before and after trial 25 were not significantly different ( mean normalized early blink size = 1 . 03 ± 0 . 06 , late = 0 . 95 ± 0 . 03 , mean ± SEM; two sample t test p=0 . 62; Figure 4—figure supplement 1D ) . Additionally , mean significant blink peak latency from the time of tone onset did not change from the first to the second half of the session ( early latency = 565 . 3 ± 68 . 5 ms , late latency = 538 . 7 ± 92 . 3 ms; mean ± SEM; two sample t test p=0 . 80; Figure 4—figure supplement 1E ) . Hence , we concluded that blink responses remained stable across the two halves of the session , and that a lack of sustained attention was not likely to be the cause of the reduction in noise correlations . To summarize , spontaneous activity was seen to become more correlated as a result of trace conditioning , following which time-tuning peaked and mice learned the temporal association . Furthermore , neurons that eventually shared the same time-tuning , showed the greatest increases in NC . We next checked if area CA1 cells could be separated into groups sharing high within-group NC and low across-group NC . We used the previously described ‘meta k-means’ clustering algorithm to identify groups of neurons that showed highly correlated spontaneous activity ( Figure 5A–D; Ozden et al . , 2008; Dombeck et al . , 2009 ) . We observed that neurons clustered into distinct groups . As expected , within-cluster NC was higher than across cluster NC ( Figure 5B , D , two-sample t test , p<10−4 ) . Interestingly , in contrast to the groups of neurons sharing the same activity peak-timings ( Figure 3—figure supplement 1G , H ) , neurons belonging to the same correlation cluster were spatially clustered within the imaged fields of view ( examples of peak-timing groups and correlation clusters are shown in Figure 5E ) . The mean distance between neuron centroids for pairs of neurons belonging to the same cluster was significantly lower than for pairs of neurons taken from different clusters ( intra-cluster = 72 . 10 ± 1 . 00 µm , n = 1694 pairs , inter-cluster = 106 . 79 ± 0 . 90 µm , n = 3418 pairs , mean ± SEM; two-sample t test p<10−4 , Figure 5F ) . Furthermore , NC and cell–cell distance were also found to be significantly , inversely correlated ( correlation coefficient = −0 . 178 , p<10−4 , n = 24 , 522 cell pairs; Figure 5—figure supplement 1A ) . This indicates that neuronal NC are indeed spatially organized . 10 . 7554/eLife . 01982 . 014Figure 5 . Correlated neurons are spatially clustered . ( A ) Neurons within clusters show within-group , correlated activity . Spontaneous , ΔF/F calcium activity traces for neurons from an example dataset . Neurons are sorted as per clusters identified by meta k-means clustering . Cluster boundaries are indicated by red , dotted-lines . The color scale represents ΔF/F amplitude . A short , 6 s stretch of the traces outlined by the yellow box have been re-plotted on the right for clearer visibility . The white asterisk indicates a bout of correlated activity in a cluster . ( B ) Pair-wise noise correlation matrix for neurons sorted as per cluster identity , for the dataset in ( A ) . Color scale represents the pairwise correlation coefficient during spontaneous activity ( noise correlation ) . Boxes indicate within-cluster spontaneous activity correlations . ( C ) Masks of neuron ROIs for the dataset shown in A and B , color coded as per cluster identity . ( D ) Comparison of distributions of within cluster ( gray ) and across cluster ( black ) correlation-coefficients . The distributions are significantly different ( p<10−4 ) . ( E ) The same field of view , with neuron ROIs color-coded by timing of their activity peak in post-training trials ( top ) and by correlation cluster number ( bottom ) . The number of cells in the top panel is smaller as only 50% of the cells with the highest reliability scores were found to be time-tuned . ( F ) Correlated cell-clusters are spatially organized . Inter-cell distances or the distances between the centroids of pairs of neurons , when both belong to the same noise-correlation cluster ( Intra-Cluster ) or when they belong to different clusters ( Inter-Cluster ) . Error bars indicate SEM . ( * indicates p<10−4 ) . Figure 5—figure supplement 1 shows that the spontaneous activity correlations between pairs of cells fall off with increasing distance between the cells being considered . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 01410 . 7554/eLife . 01982 . 015Figure 5—figure supplement 1 . Spontaneous activity correlations fall with increasing inter-cell distance . Mean of the spontaneous activity correlations for all cell-pairs plotted against the binned distances between these cell-pairs . The shaded area represents SEM . Correlation coefficient = −0 . 178 , p<10−4 . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 015 To examine how these cell-groups or clusters changed over the session , we calculated a similarity score for clusters identified from different sections of the training session ( ‘Materials and methods: Data analysis’ ) . We considered spontaneous activity data from three segments of the session comprising early , middle and late blocks of trials . We then measured the similarity between pairs of matched clusters obtained using data from different trial blocks . The similarity score measured the mean fraction of neurons belonging to clusters matched across these blocks , normalized to the mean overlap between random cell-clusters of the same sizes . For example , a high similarity score for a cluster ( >> 1 ) would indicate a similarity much greater than chance , implying that this cluster is well preserved from one trial block to the next . Interestingly , we found highly reliable groupings ( i . e . , high similarity scores ) in the pre-training session datasets , indicating stable patterns of input from upstream neuronal populations ( example in Figure 6A ) prior to training . In contrast , cell-groupings changed drastically during the trace conditioning session ( Figure 6A , C ) , and were significantly less stable than those in the pre-training datasets . For pseudo-conditioned mice , cell groupings were reliable throughout the training session datasets ( ANOVA followed by Tukey Kramer h . s . d . , p<0 . 01; Figure 6B , C ) . Thus , spatially organized , correlated cell-clusters exist prior to and during learning . These correlated cell-groups remain stable in the absence of training , but undergo continuing re-organization during the process of learning the trace conditioning task . 10 . 7554/eLife . 01982 . 016Figure 6 . Correlated neuron-groups are spatially re-distributed during trace conditioning . ( A and B ) Sample trace-conditioned ( A ) and pseudo-conditioned ( B ) mouse neuron-clusters , identified in early ( left ) , middle ( center ) and late ( right ) trials , for pre-training ( top ) as well as training ( bottom ) sessions . Neurons have been color-coded as per the noise-correlation cluster they belong to , and cluster numbers have been sorted for maximum cluster member overlap across trial-groups . ( C ) Spatial organization of correlated cell-clusters changes during learning . Summarized statistical testing of stability of clusters over the training session . A cluster similarity score ( SS ) was computed to measure the similarity between clusters from different parts of the session . Mean similarity scores for clusters of neurons from trace conditioned mice are shown in blue , pseudo-conditioned mice in green , and non-learners in cyan . Pre-training session scores are displayed as hatched bars and training session scores , solid bars . Error-bars denote SEM . The black , dotted line denotes chance level overlap score , and the red , dotted line indicates mean overlap score with perfectly overlapping clusters ( * indicates p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 016 Are groups of time-tuned neurons organized into these correlation clusters ? We calculated a cluster similarity score for neurons grouped by correlated activity and neurons grouped by time-tuning , during the tone to puff period . We found that this similarity was close to chance ( mean overlap score = 2 . 08 , SD = 0 . 6 , in multiples of similarity expected by chance; ‘Materials and methods: Data analysis’ ) .
The objective of this study was to shed light on the role played by the hippocampus in the association of stimuli separated in time , specifically during trace eyeblink conditioning . Previous work had identified neurons that were primarily air-puff responsive early in training , but became progressively more tone responsive as training proceeded . This process was thought to ascribe US valence to the CS ( McEchron and Disterhoft , 1999 ) . The progressive changes in neuronal activity timing we observed are consistent with this finding , as a small subset of neurons did indeed shift from being tone to puff responsive ( Figure 2E , F , Figure 3G–I ) . However , the predominant learning-related change we observed was the emergence of sequentially timed activity of groups of neurons ( Figure 3D , E , J , Figure 3—figure supplement 1D–F ) . Here , we note that calcium response recordings typically have a time resolution of about 100 ms , a time window within which entire sequences of neuronal activation have been known to occur ( Foster and Wilson , 2006 ) . Hence , we interpret our data to be indicative of sequential activity with respect to time-windows comparable to the average frame time ( 74 ms ) of our recordings . Even at our lower time resolution , the total space of possible sequences is very large . Our study thus indicates that a very small fraction of all possible sequences is important for successful trace conditioning . Grouped by activity peaks within ∼70 ms time-windows , neurons might show finer-scale sequences or even fluctuations that would not be visible in our recordings . However , at the time-scale of a few hundred milliseconds relevant to trace conditioning , area CA1 cells show progressively emerging activity sequences after network re-organization during learning . Another suggested model for stimulus trace representation involves sustained neuronal activity ( Solomon et al . , 1986 ) . We saw no evidence to support this model . Our observation of the progressive formation of cell activation sequences ( Figure 2C , D , 3E , J ) significantly extends our understanding by revealing how the representation of the tone stimulus is maintained over the tone-puff interval in the trace conditioning task . Time-tuned firing has previously been interpreted to imply that neurons encode time ( MacDonald et al . , 2011 ) or that they encode a triggering stimulus in the form of a well-timed sequence , effectively bridging two temporally separated stimuli ( Itskov et al . , 2011; MacDonald et al . , 2013 ) . Stimulus representation by sequential activity would also allow for the association of two temporally separated stimuli , by maintaining a representation of the first stimulus until the arrival of the second one . This would then allow Hebbian plasticity to strengthen synapses linking the neurons associated with the non-stationary , sequential stimulus representation . This stimulus maintenance is necessary since Hebbian plasticity cannot occur with stimuli separated by over ∼100 ms ( Levy and Steward , 1983 ) . Simulation studies of the hippocampal CA3 network , have shown that stimuli separated in time can be ‘bridged’ ( Wallenstein et al . , 1998 ) . Simple recurrent network models of the hippocampal CA3 region have also been trained on a close proxy of the trace conditioning paradigm ( Howe and Levy , 2007; Levy et al . , 2005 ) . Notably , in these studies , the tone stimulus triggers a well-timed relay of CA3 cell activation that culminates in the activation of US representing neurons . Our experimental observations are in agreement with this model , if we assume that qualitatively , grouping in area CA1 cell activity reflects grouping of upstream CA3 activity ( Brivanlou et al . , 2004 ) . We suggest that the progressive emergence of sequentially timed activity we observe in our study is a signature of such network changes . Sequential activity has been observed before , in subjects that have already learnt a temporal memory task ( Lee and Wilson , 2002; Pastalkova et al . , 2008; MacDonald et al . , 2011 ) . While sequentially-timed activity has also been observed in un-trained mice ( Luczak et al . , 2009; Dragoi and Tonegawa , 2011 ) , these sequences of activation were very short , only 100–200 ms long . Post training , we observed sequential activation over an 800 ms period , which is a task relevant timescale ( Figure 2D ) . Importantly , to our knowledge , our study is the first to monitor the emergence of these sequences over the entire process of learning a temporal memory task ( schematic in Figure 7A ) . As we discuss below , this has allowed us to identify transient , learning-related network changes in spontaneous activity correlations that could underlie the emergence of the longer lasting sequential activity . 10 . 7554/eLife . 01982 . 017Figure 7 . Summary and model schematics . ( A ) Schematic summarizing findings: early in the session ( top ) , neuron activity peaks are tuned to stimuli . Spontaneous activity is largely un-correlated . By the middle of the session ( center ) , time-tuned activity during the stimulus period begins to emerge , after spontaneous activity correlations have risen . At the end of the session ( bottom ) , stimulus–period activity is time-tuned , but spontaneous correlations have fallen back towards baseline levels . ( B ) Cartoon model of possible network changes occurring at the synaptic level: early in training ( top ) , CA3 to CA1 synapses ( gray circles ) are at baseline strengths . As training progresses , specific synapses are strengthened ( larger gray circles ) , towards the middle of the session ( middle ) . At this point , neurons that share time tuning also begin to display increased noise-correlations as they receive more common input . Late in the training session ( bottom ) , after repeated re-organization of the network , synapses undergo homeostatic normalization ( gray circles reduced in size ) , thus causing total common input and spontaneous activity correlations to drop . ( C ) Schematic diagram depicting network changes at the level of groups of cells . CA1 cell clusters ( horizontal row at bottom of each panel; clusters are color-coded ) have high correlations in spontaneous activity , receiving shared input ( direction indicated with black arrows ) from groups of CA3 cells that are also correlated in their spontaneous activity . These correlation-groups of cells are spatially clustered . As learning progresses , the clusters of correlated CA3 cells change , resulting in changes in the groups of correlated CA1 cells . These new groups are also spatially organized , but differ significantly from pre-training groups . DOI: http://dx . doi . org/10 . 7554/eLife . 01982 . 017 As discussed earlier ( ‘Results: Mice learn a trace eyeblink conditioning task within a single session’ ) , area CA1 activity data pooled from non-learners is probably confounded by contrary effects due to the uncertain state of learning of each mouse . However , for the sake of completeness , we have included data from non-learners in some key figures . These panels show that the data from non-learners is consistent with our interpretations and that trace learners are significantly different from non-learners as well as from pseudo-conditioned mice , in all cases . Correlations between neurons during periods of spontaneous activity or noise correlations ( NC ) have been thought to be indicative either of shared inputs , or of direct or indirect connectivity ( Ts’o et al . , 1986; Bair et al . , 2001; Fujisawa et al . , 2008 ) . While high temporal resolution measurements of neuronal activity are required to infer direct , synaptic connectivity , we , in our study , infer changes in NC to be indicative of changes in clustering of inputs to CA1 cells from upstream CA3 cells ( Komiyama et al . , 2010 ) . This view is consistent with earlier findings where the correlations between place cell pairs were seen to increase as rats explored a novel track ( Cheng and Frank , 2008 ) . Importantly , this increase in correlations was dependent on the expression of N-methyl-D-aspartate receptors and , therefore , the capacity for synaptic plasticity in area CA3 ( Dragoi and Tonegawa , 2013 ) . We found that cell-pairs that eventually share the same time-tuning show significantly higher increases in NC than do random cell-pairs , though on average , all cell-pairs show a significant increase in pair-wise correlations ( Figure 4A , B ) . We interpret this to be due to increased common input from changing groupings of inter-connected CA3 neurons . However , this increase is temporary as average NC declines near the end of the session , although neurons sharing the same time-tuning still maintain significantly elevated NC . This is consistent with a previous observation that novelty-induced increases in spontaneous correlations gradually reduce with repeated exposure to the novel environment ( Cheng and Frank , 2008 ) . The transient , global increase in NC may arise from a network-wide increase in synaptic weights during learning , and the subsequent decline may reflect weakening of mean synaptic weights due to synaptic re-normalization ( schematic in Figure 7B ) . Consistent with this interpretation , we observed that mean NC peaked just prior to the peaks in behavioral responsiveness and reliability of sequential activity , even on an individual mouse basis ( Figure 4—figure supplement 1A ) . Similar results have also been observed in previous studies . Cortical neurons sharing task-specific tuning were seen to increase their spontaneous correlations as learning progressed on a sensory association task ( Komiyama et al . , 2010 ) . In recordings from rodent CA1 neurons , place cells were seen to show gradually increasing spontaneous correlations as a novel environment was explored ( Cheng and Frank , 2008; Dragoi and Tonegawa , 2013 ) . Furthermore , over extended exposure to the same novel track , these elevated correlations gradually reduced ( Cheng and Frank , 2008 ) . Interestingly , when neurons were divided into clusters that shared high NC , these clusters turned out to be spatially organized even prior to training ( Figure 5E , F ) . This might indicate that area CA1 cells receive spatially segregated input from bundles of area CA3 cells . This is borne out by the observation of spatially segregated connectivity from CA3 to CA1 cells in hippocampal slices ( Brivanlou et al . , 2004 ) . Importantly , we observed that these spatially segregated clusters changed only during learning , while in pseudo-conditioned mice and in data collected prior to training , NC clusters remained relatively constant ( Figure 6C ) . This indicates that during learning , area CA1 cells ( and probably also , by extension , CA3 cells ) enter a dynamic state , where neurons change from one spatially-organized , correlated grouping to another ( schematic in Figure 7C ) . In our view , these fluctuating , spatial groupings reflect network changes that underlie the emergence of spatially un-organized , behavior-related sequentially timed activity . We propose that when learning the trace eyeblink conditioning task , and perhaps more generally when associating temporally separated stimuli , the hippocampus is recruited as follows: Through a network-wide , transient increase in specific synaptic strengths within the CA3 and in the CA3 to CA1 network , changing groups of CA3 cells remodel the strengths of their connections to CA1 cells ( Figure 7B , C ) . This is reflected in changing clusters of area CA1 cells showing increased noise-correlations . Progressively , these groups stabilize into a sequence of stimulus-activated cell-groups , after which global synaptic re-normalization causes average noise correlations to move back towards baseline levels ( Figure 7C ) . Only neurons sharing the same timing in the sequence maintain significantly elevated noise correlations ( Figure 4B , schematic in Figure 7B ) . Importantly , these spontaneously correlated clusters of cells are also spatially clustered and changes in these clusters underlie the emergence of sequentially activated cell-groups . These sequential groups allow CS-representing neurons to be indirectly connected to US-representing neurons , thereby forming an association between stimuli separated in time .
All experimental procedures were approved by the National Centre for Biological Sciences Institutional Animal Ethics Committee ( Protocol number USB–19–1/2011 ) , in accordance with the guidelines of the Government of India ( animal facility CPCSEA registration number 109/1999/CPCSEA ) and equivalent guidelines of the Society for Neuroscience . All recordings and behavioral experiments were carried out on male , 30 to 45 day old C57BL/6 mice . Data acquired from a total of 35 mice is included here . A total of 18 mice were trace conditioned , and 17 mice pseudo-conditioned . Of the 18 trace conditioned mice , 9 learned the task to criterion and the other 9 failed to learn . Imaging data was acquired for 14 of the trace conditioned mice . Of these , 6 were learners and 8 were non-learners . Of the 17 pseudo-conditioned mice , imaging data was acquired for 6 mice ( Figure 2—figure supplement 1B ) . 1 . 5 to 2 hr after recovery from anesthesia , mice were head-fixed on a custom-designed clamp ( gift from the Albeanu Lab , Cold Spring Habor Laboratory , New York , NY , USA ) and mounted onto the stage of a custom-built , two-photon microscope . Left eyelid blink responses were recorded using a custom-made magnetometer ( Koekkoek et al . , 2002 ) positioned close to ( 1–3 mm away ) the neodymium magnet fragment glued to the eyelid and was left un-disturbed throughout the experiment . The magnetometer was constructed using a magneto-resistive sensor ( HMC1051Z; Honeywell , Linden , NJ , USA ) , whose output was further amplified ( SR560; Stanford Research Systems ) and digitized at 10 KHz using a data acquisition card ( PCI-6221; National Instruments , Austin , TX , USA ) . The tone , conditioned stimulus was delivered using a speaker positioned to the left of the mouse . The stimulus was a pure tone of frequency 9 . 5 KHz , delivered at an intensity of 80–90 dB at the microscope stage and of 350 ms duration . In a pre-training session , only tones were delivered for 30 trials . Blink responses were acquired for 5 s prior to tone onset and terminated 10 s after tone onset . The inter-trial interval was randomized between 15 and 30 s , making the total duration between tone-presentations 30 to 45 s . Thereafter , a single trace conditioning session of 50 trials was carried out . On training trials a 100 ms air-puff ( aversive , un-conditioned stimulus ) , was delivered 250 ms ( trace interval ) after the end of the tone . Pressurized air at 5 psi was passed through a nozzle of tip-diameter 1–1 . 5 mm positioned ∼1–2 cm away and aimed at the mouse’s left eye . Air-flow was switched on and off with a computer-controlled electronic , solenoid valve ( EV mouse valve , Clippard , Cinncinati , OH , USA ) . Every fifth trial was a probe trial ( on which no puff was delivered ) however as these amounted to only 10 probe trials per mouse , data from these trials have not been used for any specific analysis . Only one training session of 50 trials was carried out per mouse . Longer sessions were seen to cause blinking fatigue and discomfort to the mice . Additional sessions over multiple days were not possible due to the acute nature of the imaging preparation ( ‘Materials and methods: Mice and surgical procedure’ ) . All trained mice were imaged , irrespective of usability of imaging data , while training was carried out . Microscope scan-mirrors continued to oscillate during the inter-trial interval , to remove auditory cues signaling the beginning or end of a trial . Hippocampal area CA1 cells were imaged in parallel with eye-blink acquisition and training . We used a custom-built two-photon microscope , built around a Tsunami , Ti-Sapphire , 80 MHz pulsed laser ( Spectra Physics , Mountain View , CA , USA ) tuned to 810 nm for excitation with a water dipping objective lens ( 5CFI75 LWD 16x , NA 0 . 8 , Nikon , Japan ) . Emitted fluorescent light was detected using an analog GaAsP PMT ( H7422P-40; Hamamatsu , Japan ) . The amplified signal was binned over 2 µs pixel times . Imaging and training were carried out in complete darkness . All analysis was carried out using custom-written programs ( Matlab , Mathworks , Natick , MA , USA ) , unless otherwise mentioned . Matlab code for the behavioral learning curves analysis ( Wirth et al . , 2003; Smith et al . , 2004 ) and meta k-means analysis ( Ozden et al . , 2008; Dombeck et al . , 2009 ) is available with the authors of the studies to first use it . Code for the main analyses carried out for this study is available for download at https://github . com/mehrabmodi/CA1-sequences-data-analysis-code . git . | Ivan Pavlov famously discovered that dogs would salivate upon hearing a bell that had previously been used to signal food , even when there was no food present . This ability to connect events that occur close together in time is known as associative learning . But how is it supported within the brain ? In the late 1940s , neuroscientist Donald Hebb proposed that if one neuron persistently and repeatedly takes part in firing a second neuron , the connection between the two neurons will be strengthened . Thus , if neurons that encode the sound of a bell are active at the same time as neurons that encode receiving food , connections between the two groups will be strengthened , and this might enable the dogs to associate the two events . However , animals can also learn to associate events that do not overlap in time . For example , we can associate a bout of food poisoning with a meal we consumed several hours earlier . In rodents , this type of learning is often studied using a task known as trace eyeblink conditioning , in which a tone signals the delivery of a puff of air to the eye after a short delay . Rodents eventually begin to blink in response to the tone , even thought the tone and the air puff are never presented simultaneously . Two possibilities have been proposed for how this might occur: either the neurons that encode the tone remain active until delivery of the air puff , or different groups of neurons are successively activated in a relay that spans the interval between the tone and the air puff . Now , Modi et al . have used in vivo imaging in awake mice to obtain evidence in favour of the second option . Mice were trained on the conditioning task while imaging was used to follow the activity of neurons in a region of the brain known as the hippocampus . As animals learned the task , neurons in part of the hippocampus called CA1 began to reorganize their firing patterns so that distinct groups of cells were active at each time point in the interval between the tone and the air puff . By contrast , hardly any neurons were active across the entire delay . The organized firing became particularly apparent at the same time as the mice first began to blink in response to the tone , and was only ever seen in animals that learned the task successfully . As well as providing evidence to distinguish between competing theories of associative learning across a delay , this study is the first to follow in real-time the reorganization of networks of neurons within the hippocampus during this common type of learning . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | CA1 cell activity sequences emerge after reorganization of network correlation structure during associative learning |
The protective immunity afforded by CD8+ T cells against blood-stage malaria remains controversial because no MHC class I molecules are displayed on parasite-infected human erythrocytes . We recently reported that rodent malaria parasites infect erythroblasts that express major histocompatibility complex ( MHC ) class I antigens , which are recognized by CD8+ T cells . In this study , we demonstrate that the cytotoxic activity of CD8+ T cells contributes to the protection of mice against blood-stage malaria in a Fas ligand ( FasL ) -dependent manner . Erythroblasts infected with malarial parasites express the death receptor Fas . CD8+ T cells induce the externalization of phosphatidylserine ( PS ) on the infected erythroblasts in a cell-to-cell contact-dependent manner . PS enhances the engulfment of the infected erythroid cells by phagocytes . As a PS receptor , T-cell immunoglobulin-domain and mucin-domain-containing molecule 4 ( Tim-4 ) contributes to the phagocytosis of malaria-parasite-infected cells . Our findings provide insight into the molecular mechanisms underlying the protective immunity exerted by CD8+ T cells in collaboration with phagocytes .
Malaria is one of the world's three major infectious diseases , together with AIDS and tuberculosis , accounting for approximately 200 million cases annually , with 600 , 000 deaths ( Snow et al . , 2005; Murray et al . , 2012 ) . With the spread of drug-resistant parasites and the lack of effective vaccines , malaria is a serious global health problem , especially in developing countries . To develop malarial vaccines , it is necessary to understand the protective immune response against malaria . However , because the malaria parasite successfully evades the host immune responses ( Hisaeda et al . , 2004 ) , it is difficult to identify the truly important immune responses , hindering the development of a malarial vaccine ( Good and Engwerda , 2011 ) . Antibodies play a major role in the protective immunity directed against the blood-stage malaria parasite . CD4+ T cells contribute to protection against blood-stage malaria though induction of antibody production and macrophage activation ( Good and Doolan , 1999; Marsh and Kinyanjui , 2006; Jafarshad et al . , 2007; Langhorne et al . , 2008 ) . However , the contribution of CD8+ T cells to this protection remains controversial because there are no major histocompatibility complex ( MHC ) class I antigens on human erythrocytes infected with the malaria parasite . Some studies have shown that infection of BALB/c mice with non-lethal Plasmodium yoelii was controlled even after depletion of CD8+ T cells comparable to control mice ( Vinetz et al . , 1990 ) . Moreover , MHC class I null mice ( beta 2-microglobulin-deficient mice ) recovered from infection with Plasmodium chabaudi chabaudi AS or Plasmodium chabaudi adami ( van der Heyde et al . , 1993b ) . Other studies have reported that depletion of CD8+ T cells from mice infected with P . chabaudi attenuated their protection , confirming the importance of CD8+ T cells ( Suss et al . , 1988; Podoba and Stevenson , 1991; van der Heyde et al . , 1993a; Horne-Debets et al . , 2013 ) . However , these studies did not show the effector mechanism of CD8+ T cells against blood-stage malaria protection . We have conclusively demonstrated the protective roles of CD8+ T cells using prime–boost live vaccination with the non-lethal rodent parasite P . yoelii 17XNL ( PyNL ) against challenge with the lethal P . yoelii 17XL ( PyL ) strain ( Imai et al . , 2010 ) . The transfer of CD8+ T cells from mice cured of PyNL infection into Rag2−/− or irradiated recipients , followed by two boosts with PyL , conferred protection against PyL . The major protective mechanism of CD8+ T cells is the interferon γ ( IFN-γ ) -dependent activation of phagocytes , resulting in the enhanced phagocytosis of parasitized red blood cells ( pRBCs ) . The cytotoxic activity of CD8+ T cells also contributes to protecting the host against blood-stage malaria . However , the target cells of this cytotoxicity and how this cytotoxicity acts against blood-stage malaria are as yet unknown . Although recent reports have demonstrated that the human malaria parasites Plasmodium falciparum and Plasmodium vivax parasitize erythroblasts ( Ru et al . , 2009; Tamez et al . , 2009 ) , the host response and protective immunity against these parasitized erythroblasts are unclear . We have reported that PyNL parasites also infect erythroblasts that express MHC class I molecules on their surfaces and that CD8+ T cells produce IFN-γ in response to parasitized erythroblasts in an antigen-specific manner . These results suggest that parasitized erythroblasts are the targets of CD8+ T cells . In this study , we investigated the effector mechanism of CD8+ T cells against blood-stage malaria in detail . Splenic CD8+ T cells activated during malaria express Fas ligand ( FasL ) and interact with Fas-expressing parasitized erythroblasts . As a result , phosphatidylserine ( PS ) is externalized to the outer leaflet of the cell membrane , leading to enhanced phagocytosis of the parasitized cells . Thus , CD8+ T cells expressing FasL contribute to the immune response to blood-stage malaria by making parasitized cells susceptible to phagocytosis .
C57BL/6 mice infected with PyNL exhibited peak parasitemia of up to 30% and recovered from the infection . However , those depleted of CD8+ T cells showed significantly greater parasitemia and died from the infection ( Figure 1A , Figure 1—figure supplement 1C ) . BALB/c mice depleted of CD8+ T cells showed similar results ( Figure 1—figure supplement 1D ) , indicating that CD8+ T cells play an essential role in protecting mice against blood-stage malaria . CD4+ T cells are known to be important in the protective immune response to blood-stage malaria ( Suss et al . , 1988; Kumar et al . , 1989; Podoba and Stevenson , 1991; Good and Doolan , 1999 ) , and we confirmed that CD4+-T-cell-depleted mice displayed greater parasitemia and a higher mortality rate ( Figure 1A ) . However , the course of infection clearly differed in CD8+-T-cell-depleted and CD4+-T-cell-depleted mice . Although mice depleted of CD8+ T cells suffered from much greater parasitemia from the early phase to its peak , the survivors eliminated the parasites similar to the control mice , whereas the CD4+-T-cell-depleted survivors took longer to recover from infection . This suggests that CD4+ and CD8+ T cells have different effector mechanisms for parasite clearance , and that the protective immunity mediated by CD8+ T cells is important in controlling infection during the early phase , within the period of peak parasitemia . Therefore , the following analyses were conducted 7–8 days after infection , when the CD8+ T cells might be activated in response to the parasite , and 16–18 days after infection , when the parasites begin to be eliminated . 10 . 7554/eLife . 04232 . 003Figure 1 . CD8+ T cells and FasL protect against infection with P . yoelii NL ( PyNL ) . ( A ) Daily parasitemia and survival rates of C57BL/6 mice depleted of CD8+ or CD4+ T cells after infection with PyNL . Parasitemia was estimated from microscopic observation of Giemsa-stained blood films . Parasitemia values are means ± SE of three pooled experiments ( control: N = 20; CD4 depletion: N = 20; CD8 depletion: N = 23 ) . *p < 0 . 05 , **p < 0 . 01 , and ***p < 0 . 001 , Mann–Whitney U-test . Survival rate was calculated from three pooled individual experiments , as described above . p values for the Kaplan–Meier log rank test are shown . ( B ) Flow-cytometric analyses of splenic CD8+ T cells were performed 7 days after infection with PyNL . Single-cell suspensions from spleens were stained with fluorescence-labeled anti-CD8β antibody . CD8+ T cells were analyzed for the expression of the indicated molecules . Shaded areas and lines in the histograms represent their expression in uninfected and infected mice . Bar graph indicates percentages of CD8+ T cells expressing the molecules as means ± SD of five mice from one of three experiments . *p < 0 . 05 , **p < 0 . 01 , Mann–Whitney U-test . ( C ) Daily parasitemia ( upper panel ) and survival rates ( bottom panel ) of wild type mice ( WT ) or gld mice infected with PyNL , monitored as in Figure 1A . WT , N = 6; gld , N = 9 . Parasitemia values are means ± SD from one of five experiments . *p < 0 . 05 and ***p < 0 . 001 , Mann–Whitney U-test . Survival rates are from five pooled individual experiments ( WT , N = 28; gld , N = 25 ) . ( D ) Contribution of FasL expressed on CD8+ T cells to the protective effects against blood-stage malaria . Expression of FasL on splenic CD4+ T cells was evaluated . *p < 0 . 05 , Mann–Whitney U-test . Data of FasL on CD8 are the same experiment as Figure 1B . ( E ) Experimental protocol for the adaptive transfer of cells after the prime–boost PyNL vaccine regime against lethal PyL infection . WT and gld mice were infected with PyNL , and then boosted twice with PyL . CD4+ and CD8+ T cells isolated from the vaccinated donors were transferred into irradiated recipients . Note that although some gld mice died from the PyNL infection , the survivors were as resistant to PyL infection as the WT mice . ( F ) Parasitemia was monitored in the recipients of the indicated cells . Each symbol indicates means ± SD . Each group contained five mice . The final survival rate of each group is also indicated . The results are from one experiment , representative of the two performed . Dagger indicates death . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 00310 . 7554/eLife . 04232 . 004Figure 1—figure supplement 1 . CD8+ T cells play protective roles in C57BL/6 mice and BALB/c mice infected with PyNL . ( A ) Successful depletion of CD8 T cells in mice treated with anti-CD8α or anti-CD8β antibody was evaluated in peripheral blood 24 hr after inoculation . ( B ) Spleen cells isolated from the indicated mice 3 days after infection were stained with anti-CD8 antibody ( clone: 53 . 6 . 7 ) with a different specificity from the depleting antibody . ( C ) Daily parasitemia and survival rates of C57BL/6 mice depleted of CD8α+ or CD8β+ cells after infection with PyNL-GFP . Parasitemia values are shown as the means ± SD ( control , N = 18; CD8α depletion , N = 9; CD8β depletion , N = 18 ) from pooled two experiments . Survival rate was calculated from two pooled individual experiments as above . ( D ) Daily parasitemia and survival rates of BALB/c mice depleted of CD8+ T cells after infection with PyNL-GFP . Parasitemia values are shown as the means ± SD of eight mice in one experiment , which is representative of the two experiments performed . Survival rate was calculated from two pooled individual experiments . Control , N = 16; CD8 depleted , N = 16 . It should be noted that 60% of C57BL/6 mice treated with anti-CD8α survived ( C ) , contrasting to the results in Figure 1A where only 20% mice survived . This difference might be due to the environmental influences as these results were obtained from experiments conducted in different animal facilities ( both are specific pathogen-free ) using mice from different a supplier . For instance , alterations in microbiota due to environmental differences are known to affect immune responses ( Furusawa et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 00410 . 7554/eLife . 04232 . 005Figure 1—figure supplement 2 . Confirmation that CD8+ T cells are responsible for transferring protection to Rag2−/− mice . ( A ) The adaptive transfer experiments were performed as in Figure 1E except for using Rag2−/− mice ( CD45 . 2 ) as recipients and B6SJL mice ( CD45 . 1 ) as donors to discriminate the transferred cells . ( B ) Spleen cells from donors ( left panel ) and sorted CD8+ T cells used for transfer ( right panel ) were analyzed for CD4 and CD8 . ( C ) The transferred cells expressing CD45 . 1 ( left panels ) were analyzed for contaminating CD4+ T cells ( right panels ) in spleens of Rag2−/− recipients 5 and 10 days after PyL challenge . ( D ) Parasitemia was monitored in the Rag2−/− recipients transferred with the indicated cells . Each line indicates an individual mouse . Daggers indicate death . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 005 First , we evaluated whether the activation of CD8+ T cells occurs during infection with PyNL . PyNL infection increased the proportion of CD8+ T cells that expressed activation markers such as CD25 and CD69 ( Figure 1B ) , and the CD8+ T cells started to express the cytotoxicity-related molecules FasL ( Krueger et al . , 2003 ) and lysosome-associated membrane protein 1 ( LAMP1 ) ( Wolint et al . , 2004 ) ( Figure 1B ) . These results indicate that CD8+ T cells contribute to the protective response to blood-stage malaria . The proportion of CD8+ T cells that express FasL increased after infection , suggesting that this molecule is involved in the immune response . To investigate this possibility , FasL-mutant gld mice were infected with PyNL . The course of infection in the gld mice resembled that in mice depleted of CD8+ T cells , insofar as parasitemia was exacerbated before peak parasitemia and the survival rate was lower than in wild-type ( WT ) mice ( Figure 1C ) . Thus , FasL is important in controlling blood-stage malaria . Although we hypothesized that FasL expressed on CD8+ T cells is crucial , the FasL expressed on CD4+ T cells ( el-Khatib et al . , 1995; Hahn et al . , 1995 ) may also play a protective role . However , this is unlikely because infection did not increase the expression of FasL on CD4+ T cells , in contrast to CD8+ T cells ( Figure 1D ) . To confirm these inferences , we used cell transfer experiments combined with a prime–boost live vaccination system in which CD8+ and CD4+ T cells isolated from mice that had recovered from PyNL infection after two homologous boosts with PyL transferred protection from an otherwise lethal infection with PyL to the recipient mice ( Figure 1E , F ) ( Imai et al . , 2010 ) . Mice that had received gld immune CD8+ T cells exhibited higher parasitemia at an early stage of infection , and some of them failed to control the challenge infection and died ( Figure 1F , left panel ) . In contrast , CD4+ T cells from gld donors protected the recipients from challenge with PyL , similar to the protection afforded mice by CD4+ T cells from WT donors ( Figure 1F , right panel ) . Therefore , FasL plays a crucial role in CD8+-T-cell-mediated protective immunity against blood-stage malaria . To confirm the roles of CD8+ T cells responsible for resistance , we excluded the possibility that contaminants in transferred cells play a role using Rag2−/− mice as recipients ( Figure 1—figure supplement 2A ) . CD8+ T cells obtained from ‘immune’ CD45 . 1-bearing C57BL/6 mice could transfer protection to Rag2−/− mice ( CD45 . 2 ) , although parasitemia was greater compared with when the irradiated mice were used as hosts ( Figure 1—figure supplement 2D , Imai et al . , 2010 ) . In such lymphopenic recipients , a very few cells proliferate intensively and acquire effector capacities . Thus , the homeostatic proliferation of contaminants other than CD8+ T cells was evaluated . The ratio of the contaminated CD4+ cells in sorted CD8+ cells used for the transfer was less than 0 . 1% ( Figure 1—figure supplement 2B ) . That in donor-derived CD45 . 1+ cells recovered from the recipients did not exceed 0 . 8% , and CD8+ T cells constantly occupied more than 96% even after infection with PyNL ( Figure 1—figure supplement 2C ) . To further exclude the possibility , we determined how many CD4+ T cells are required to protect the recipient . The recipients transferred with 1 × 107 CD4+ T cells could control the challenge infection , but those with 1 × 106 CD4+ T cells could not ( Figure 1—figure supplement 2D ) . Thus , 0 . 1% contaminated CD4+ T cells corresponding to 1 × 104 in transferred 107 CD8+ T cells seem to have no protective ability . Based on these findings , we concluded that CD8+ T cells are responsible for the transferred protection . We next examined the cell types targeted by FasL-dependent immunity . FasL interacts with Fas expressed on target cells , inducing the apoptosis of the Fas-expressing cells ( Nagata and Golstein , 1995 ) . Recently , erythroid cells have been reported to express Fas ( De Maria et al . , 1999; Tsushima et al . , 1999; Mandal et al . , 2005; Liu et al . , 2006 ) . Based on our previous finding that malaria parasites infect erythroblasts ( Imai et al . , 2013 ) . We postulated that infected erythroid cells are the targets of FasL-expressing CD8+ T cells . Therefore , we analyzed the expression of Fas on infected erythroid cells in the spleens and peripheral blood of mice infected with PyNL–green fluorescent protein ( GFP ) . Very few TER119+ erythroid cells expressed Fas in the peripheral blood , even among the infected GFP+ cells ( Figure 2 ) . In contrast , a number of infected GFP+ cells expressing Fas were present in the spleen , and the frequency of those cells among the parasitized cells reached 50% before peak parasitemia ( Figure 2A , B ) . To identify the erythroid cells that express Fas in the spleen , we examined the expression of MHC class I molecules on the infected cells because erythroblasts are distinguished from reticulocytes and mature RBCs by their high-level expression of MHC class I antigens ( Imai et al . , 2013 ) . Almost all Fas-expressing cells , both infected and uninfected , were MHC class Ihi ( Figure 2C ) , indicating that the infected Fas+ cells were erythroblasts . As those cells present antigens in conjunction with MHC class I molecules and are recognized antigen-specifically by CD8+ T cells ( Imai et al . , 2013 ) , it is possible that FasL-bearing CD8+ T cells affect infected erythroblasts expressing Fas . Notably , the infection of erythroblasts with PyNL may induce their expression of Fas , because Fas− erythroblasts were markedly reduced in the infected cells relative to their numbers in uninfected cells ( 41% and 14% , respectively; Figure 2C ) . Moreover , the intensity of Fas expression was much higher on parasitized erythroblasts than in uninfected erythroblasts . 10 . 7554/eLife . 04232 . 006Figure 2 . Fas is expressed on erythroid cells infected with PyNL . ( A ) Spleen cells and peripheral blood cells obtained from mice infected with PyNL–GFP were stained with anti-TER119 , anti-Fas , and anti-MHC class I antibodies . TER119+ GFP+ infected or TER119+ GFP− uninfected cells were analyzed for their expression of Fas . Numbers on the histograms indicate the percentages of Fas+ cells in the gated cells . ( B ) Percentages of Fas+ cells in parasitized cells ( TER119+ GFP+ Fas+/TER119+ GFP+ ) are shown as means ± SD from one experiment ( N = 4 ) , representative of the three performed . ( C ) Splenic TER119+ cells infected ( right panel ) or uninfected ( left panel ) in mice infected with PyNL–GFP were separated into MHC class Ihi erythroblasts ( fluorescence intensity > 102 ) , class Ilo-neg reticulocytes , and mature RBCs and analyzed for their Fas expression . Numbers indicate the percentages of the gated cells in each quadrant . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 006 As a consequence of the interaction between FasL and Fas , Fas-expressing cells undergo apoptosis ( Nagata , 1996a , 1996b ) , which is characterized by the fragmentation of their nuclei and the exposure of PS on the surface of the cell ( Yoshida et al . , 2005 ) . PS-displaying infected RBCs are more susceptible to phagocytosis , and this phenomenon is involved in the protection of the host from malaria . Therefore , we investigated whether PS is exposed on erythroid cells in response to the FasL–Fas interaction during malaria ( Figure 3 ) . PS+ cells were significantly increased in splenic infected TER119+ cells ( Figure 3A ) . CD8+-T-cell-depleted or gld mice had much fewer PS+ cells than the control mice ( Figure 3B , C ) . Moreover , the majority of infected Fas+ splenic erythroblasts displayed PS ( Figure 3D ) , suggesting that CD8+ T cells and FasL are involved in increasing the exposure of PS on infected cells in the spleen . In contrast , the number of PS+ cells among the infected RBCs was only slightly increased in the peripheral blood . Because the gld and CD8+-T-cell-depleted mice contained fewer PS+ infected RBCs , the increase in PS+ cells seemed to be dependent on FasL and CD8+ T cells , despite the absence of Fas+ cells in the peripheral blood . 10 . 7554/eLife . 04232 . 007Figure 3 . Infection with PyNL induces externalization of phosphatidylserine ( PS ) on parasitized cells . ( A ) Spleen cells and peripheral blood obtained from the indicated mice 8 days after infection with PyNL–GFP were stained with anti-TER119 antibody and annexin V . Infected GFP+ or uninfected GFP− TER119+ cells were analyzed for the expression of PS . ( B ) Percentages of TER119+ GFP+ PS+ cells in the TER119+GFP+ cells in the control ( open symbols ) and CD8+-depleted mice ( closed symbols ) are shown as means ± SD from one experiment ( N = 4 ) , representative of the three performed . ( C ) Those in the gld mice were also analyzed . **p < 0 . 01 , Mann–Whitney U-test . ( D ) Splenic TER119+ cells infected ( right panel ) or uninfected ( left panel ) obtained from mice 8 days after infection with PyNL–GFP were analyzed for the expression of PS and Fas . Numbers indicate the percentages of the gated cells in each quadrant . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 007 To further investigate the involvement of CD8+ T cells in PS exposure , splenic TER119+ cells isolated from gld mice , which contained fewer PS+ cells despite similar numbers of Fas+ cells ( Figures 2B , 3C ) , were cocultured with CD8+ T cells of various origins ( Figure 4A ) . CD8+ T cells from infected WT mice efficiently induced PS exposure in a dose-dependent manner , whereas those from uninfected WT mice did not ( Figure 4B ) . Exposure of PS was only observed in infected GFP+ cells , and not in uninfected cells ( Figure 4C ) . Importantly , CD8+ T cells from infected gld mice induced PS exposure in only a few infected GFP+ cells , as did the CD8+ T cells from uninfected WT mice ( Figure 4B , C ) , indicating that FasL expressed by CD8+ T cells is responsible for PS exposure . 10 . 7554/eLife . 04232 . 008Figure 4 . Exposure of PS is dependent on CD8+ T cells and FasL . ( A ) Experimental protocol for the evaluation of CD8+-T-cell-dependent PS externalization in parasitized cells in vitro . Splenic TER119+ cells containing RBC , pRBC , erythroblasts ( EB ) and pEB ( 3 × 105 ) isolated from gld mice 17 days after infection with PyNL–GFP were cultured for 4 hr with CD8+ T cells from WT or gld mice 17 days after PyNL infection , at the indicated ratios . ( B ) Cultured TER119+ cells with CD8+ T cells from the indicated mice were stained with annexin V , and GFP+ cells were analyzed for PS expression . Numbers in histograms indicate percentages of annexin V+ cells in the gated cells . ( C ) Values are means ± SD from triplicate cultures in one experiment , representative of the four performed . **p < 0 . 01 , Mann–Whitney U-test . ( D ) Experimental protocol for the evaluation of FasL-dependent PS externalization in parasitized cells in vitro . TER119+ cells isolated from spleens and peripheral blood of gld mice 7 days after infection with PyNL–GFP were cultured for 4 hr with the indicated amounts of FasL–Strep or bovine serum albumin ( negative control ) . ( E ) Cultured cells were collected and stained with annexin V , and annexin V+ cells among the GFP+ parasitized cells were quantified . Values are means ± SD of triplicate cultures in one experiment , representative of the four performed . *p < 0 . 05 and **p < 0 . 01 , Mann–Whitney U-test . ( F ) Annexin V-positive or -negative GFP+ parasitized cells were analyzed for the expression of MHC class I antigens , as in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 008 Next , we examined FasL-dependent PS exposure in vitro using FasL–Strep . Splenic TER119+ cells collected from gld mice infected with PyNL–GFP were cultured with FasL–Strep ( Figure 4D ) . The addition of FasL–Strep induced the externalization of PS in infected GFP+ cells in a dose-dependent manner ( Figure 4E ) . Because the Fas-expressing cells in the spleen were erythroblasts ( Figure 2 ) , we confirmed that FasL–Strep induced erythroblasts to expose PS . The PS+ cells were all MHC class Ihi erythroblasts , even in the absence of FasL–Strep ( Figure 4F ) . Furthermore , FasL–Strep reduced the proportion of MHC class Ihi erythroblasts in PS−cells ( from 64% to 50% ) , indicating that PS− Fas-expressing erythroblasts expose PS in the presence of FasL–Strep ( Figure 4F ) . In contrast , PS exposure was not observed on infected cells from the peripheral blood , even in the presence of CD8+ T cells or FasL–Strep ( Figure 5 ) , presumably because Fas was not expressed on these cells ( Figure 2 ) . FasL is known to localize to the cell surface , but it is also secreted ( Morello et al . , 2013 ) . Therefore , we determined whether FasL-dependent PS exposure requires cell contact or a soluble factor . Transwell experiments revealed that PS was only displayed on infected erythroid cells when the erythrocytes contacted CD8+ T cells , and was not affected by soluble factors secreted from CD8+ T cells ( Figure 6 ) . These results indicate that activated CD8+ T cells can induce PS exposure on infected splenic cells ( parasitized erythroblasts ) in a FasL- and contact-dependent manner during blood-stage malaria , although the involvement of exosomes bearing pro-apoptotic membranous FasL could not be ruled out ( Andreola et al . , 2002 ) . 10 . 7554/eLife . 04232 . 009Figure 5 . Externalization of PS in pRBCs was not induced in vitro . Peripheral blood cells obtained from gld mice infected with PyNL–GFP were cultured with CD8+ T cells ( A ) or FasL–Strep ( B ) and analyzed as in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 00910 . 7554/eLife . 04232 . 010Figure 6 . Externalization of PS in parasitized cells requires contact with CD8+ T cells . ( A ) Protocol of the contact dependence assay using Transwell cultures . Splenic TER119+ cells from gld mice infected with PyNL–GFP and CD8+ T cells from WT mice infected with PyNL were placed into the upper and/or lower wells and cultured for 6 hr . GFP+ parasitized cells were analyzed for PS expression , as in Figure 4B . The ratio of the percentages of PS+ cells in the GFP+ cells in the upper ( B ) and lower wells ( C ) was calculated as ( % PS+ GFP+ of GFP+ cells in each test ) / ( % PS+GFP+ in GFP+ cells in the absence of cell components in the lower well ) in ( B ) , and as ( % PS+ GFP+ in GFP+ cells in the presence of CD8+ T cells ) / ( % PS+ GFP+ in GFP+ cells in the absence of CD8+ T cells in the lower well ) in ( C ) . Values shown are the means ± SD of triplicate cultures in one experiment , representative of the three performed . **p < 0 . 01 , Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 010 In line with previous reports from ourselves and others on the importance of the phagocytosis of infected RBCs in the protective response to blood-stage malaria ( Zhang et al . , 1999; Couper et al . , 2007; Imai et al . , 2010; Matsuzaki-Moriya et al . , 2011; Duan et al . , 2013 ) , mice depleted of macrophages by administration with clodronate/liposome ( C/L ) quickly died in association with high parasitemia ( Figure 7A , B ) . It has also been reported that PS-exposing cells are highly susceptible to phagocytosis by macrophages ( Fadok et al . , 1998; van den Eijnde et al . , 1998 ) . These findings led us to hypothesize that PS exposure on infected cells , induced by CD8+ T cells , accelerates their engulfment by phagocytes such as macrophages . To test this hypothesis in vitro , infected cells were isolated from CD8+-T-cell-deleted , gld , and WT control mice infected with PyNL . The degree of PS exposure was different in each preparation . Those cells were then labeled and cocultured with macrophages from uninfected mice ( Figure 7C ) . Only a few macrophages phagocytosed the uninfected cells , whereas up to 30% of macrophages engulfed the infected cells from the WT mice . When cocultured with infected cells from CD8+-T-cell-depleted WT or gld mice , the number of phagocytotic macrophages was significantly reduced ( Figure 7D ) . A positive correlation between the degree of PS exposure on the infected cells and the numbers of phagocytotic macrophages was observed , indicating that PS+ cells are readily phagocytosed ( Figure 7E ) . 10 . 7554/eLife . 04232 . 011Figure 7 . Phagocytosis of parasitized RBCs ( pRBCs ) by macrophages correlates with RBC PS expression in vitro . ( A ) Experimental protocol for depleting macrophage with clodronate/liposomes ( C/L ) . ( B ) Parasitemia ( left panel ) and survival rate ( right panel ) were evaluated from two pooled separate experiments . Control: N = 17; C/L: N = 10 . ***p < 0 . 001 , Mann–Whitney U-test . ( C ) Protocol used to evaluate the phagocytosis of pRBCs . pRBCs obtained from WT , CD8+-depleted , or gld mice were labeled with CFSE , and then cocultured for 4 hr with CD11b+ macrophages obtained from uninfected WT mice , which had been labeled with PKH26 fluorescence . RBC from uninfected WT mice was also tested . The ratio of macrophages to pRBCs or RBC was 1:30 . ( D ) Phagocytosis was evaluated by detecting the PKH+ CFSE+ macrophages after culture with pRBCs isolated from the indicated mice . Numbers in the upper panels indicate the percentage of phagocytic macrophages in the squares in the total macrophages ( % phagocytosis = PKH+ CFSE+/PKH+ ) . Values in the bar graph are means ± SD from triplicate cultures in one experiment , representative of the two experiments performed . **p < 0 . 01 , Mann–Whitney U-test . ( E ) PS exposure correlates with the degree of phagocytosis . The percentage of PS+ cells in each pRBC preparation and the percentage of phagocytic macrophages when each preparation was used are plotted . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 011 The phagocytosis of the infected cells was analyzed in vivo using PyNL–GFP . Spleen and peripheral blood cells were stained with CD11b and separated into CD11b+ GFP+ cells and CD11b− GFP+ cells , as infected cells phagocytosed by macrophages or infected cells , respectively ( Figure 8A , B ) . Macrophages phagocytosing infected GFP+ cells were observed under a microscope ( Figure 8C ) . There were more infected CD11b− GFP+ cells in the peripheral blood of CD8+-T-cell-depleted mice than in the peripheral blood of the control mice ( Figure 8A ) , reflecting the higher parasitemia in the CD8+-T-cell-depleted mice ( Figure 1A ) . Substantial numbers of infected cells were phagocytosed in the spleen but not in the peripheral blood , indicating the importance of this organ in the elimination of the malaria parasite ( Figure 8A ) . The proportion of phagocytosed infected cells in the total infected cells ( CD11b+ GFP+/GFP+ ) in the spleens of CD8+-T-cell-depleted mice was significantly lower than the proportion in the control mice ( Figure 8A , B , D ) . Furthermore , gld mice showed similar results to those of CD8+-T-cell-depleted mice ( Figure 8E ) . Finally , we analyzed macrophage subsets and found that F4/80+ red pulp macrophages are responsible for the ingestion of parasites . SIGNR1+ marginal zone macrophages , CD169+ marginal metallophilic macrophages , and CD68+ tingible-body macrophages appeared not to be involved in phagocytosis ( Figure 8F ) . Although depletion of CD8+ T cells did not affect the numbers of each macrophage subset ( data not shown ) , it dramatically reduced the number of phagocytic F4/80 macrophages . 10 . 7554/eLife . 04232 . 012Figure 8 . Phagocytosis of parasitized cells by macrophages in vivo . Spleen cells and peripheral blood were isolated seven or 17 day after mice depleted of CD8+ T cells were infected with PyNL–GFP . ( A ) Those cells were then stained with anti-CD11b antibody and separated into free parasitized cells ( GFP+ CD11b− ) and phagocytosed cells ( GFP+ CD11b+ ) . The numbers represent the percentages of cells in each quadrant . ( B ) Histograms indicate CD11b expression in GFP+ gated cells . ( C ) Images of phagocytosed parasitized cells are shown . Hemozoin-containing adherent macrophages were isolated from spleens with magnetic sorting and were observed microscopically . Scale bars represent 10 μm . Fractions of phagocytosed GFP+ parasitized cells were quantified as CD11b+ cells from mice depleted of CD8+ T cells ( D ) or from gld mice ( E ) . ( F ) Macrophage subsets expressing the indicated macrophage markers were also calculated in control and CD8+ T cell-depleted mice . Values are means ± SD of 5–7 mice from three pooled individual experiments . *p < 0 . 05 , Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 01210 . 7554/eLife . 04232 . 013Figure 8—figure supplement 1 . Dendritic cells also phagocytose parasitized cells , presumably in response to PS exposure . The phagocytosis of parasitized cells by dendritic cells was analyzed as in Figure 8 , except that CD11c was used as the dendritic-cell marker instead of CD11b . Spleen cells and peripheral blood were isolated seven or 17 day after mice depleted of CD8+ T cells were infected with PyNL–GFP . ( A ) Those cells were then stained with anti-CD11c antibody and separated into free parasitized cells ( GFP+ CD11c− ) and phagocytosed cells ( GFP+ CD11c+ ) . The numbers represent the percentages of cells in each quadrant . ( B ) Histograms indicate CD11c expression in GFP+ gated cells . ( C ) Fractions of phagocytosed GFP+ parasitized cells were quantified as CD11c+ cells from mice depleted of CD8+ T cells ( D ) or from gld mice . Values are means ± SD of 5–7 mice from three pooled individual experiments . *p < 0 . 05 , Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 013 As the macrophages in the CD8+-T-cell-depleted mice were activated to a similar degree as those in the control mice during malaria ( Figure 9 ) , the proportion of cells exposing PS may correspond to this difference in the number of phagocytosing macrophages . These results indicate that the phagocytosis of infected cells occurs in the spleen and correlates with the exposure of PS on the infected cells , which is dependent on CD8+ T cells and FasL . We obtained the same results using dendritic cells instead of macrophages ( Figure 8—figure supplement 1 ) . 10 . 7554/eLife . 04232 . 014Figure 9 . Depletion of CD8+ cells did not affect the activation of macrophages . Spleen cells collected from the indicated mice 17 days after infection with PyNL were stained with anti-CD11b , anti-MHC class I , and anti-MHC class II antibodies . CD11b+ cells were analyzed for their expression of class I ( A ) and class II molecules ( B ) . Values shown are the means ± SD of 3–5 mice in one experiment , representative of the two performed . *p < 0 . 05 , Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 014 Recently , T-cell immunoglobulin- and mucin-domain-containing molecule ( Tim-4; also known as Timd4 ) was identified as a PS receptor ( Miyanishi et al . , 2007 ) . In this study , the phagocytosis of PS-exposing infected erythroid cells was observed . Therefore , we investigated the involvement of Tim-4 as a novel receptor in the protective immune response against malaria . The expression of Tim-4 on splenic macrophages was upregulated , and the number of Tim-4+ macrophages increased in response to infection with PyNL ( Figure 10A ) . The phagocytosis by macrophages of infected RBCs isolated from infected WT mice was dose-dependently inhibited by the presence of antibodies directed against Tim-4 ( Figure 10B , C ) . These results indicate that Tim-4 contributes to the phagocytosis of infected RBCs . 10 . 7554/eLife . 04232 . 015Figure 10 . Tim-4 expressed on macrophages contributes to the phagocytosis of pRBCs . ( A ) Infection with PyNL induced the expression of Tim-4 on macrophages . Spleen cells obtained from mice 17 days after infection were stained with anti-CD11b and anti-Tim-4 antibodies , and the CD11b+ cells were analyzed for Tim-4 expression . The expression levels of Tim-4 are shown in a histogram . Values in the bar graph are the means ± SD of six mice in two pooled individual experiments . *p < 0 . 05 , Mann–Whitney U-test . ( B ) Addition of anti-Tim-4 antibody suppressed the phagocytosis of pRBCs by macrophages . PKH-labeled macrophages from uninfected mice were cultured with pRBCs isolated from WT mice 17 days after infection in the presence of the indicted concentrations of anti-Tim-4 antibody . The phagocytic macrophages were evaluated as in Figure 7D . ( C ) Inhibitory effects of anti-Tim-4 antibody on phagocytic cells were quantified as ( % phagocytic macrophages in the presence of the antibody ) / ( % phagocytic macrophages in the absence of the antibody ) . Values are the means ± SD of triplicate cultures in one experiment , representative of the four performed . *p < 0 . 05 and **p < 0 . 01 , Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04232 . 015
Here , we have demonstrated a novel protective mechanism against blood-stage malaria conferred by CD8+ T cells . CD8+ T cells interact with infected erythroblasts and induce them to display PS in a FasL-dependent manner . In turn , PS exposure enhances the susceptibility of infected cells to phagocytosis , which contributes to the elimination of the parasite . Our proposal may resolve the controversial protective roles of CD8+ T cells against infected erythroid cells . Vinetz et al . had reported that CD8+ T cells are not contributed to protection against blood-stage murine malaria ( Vinetz et al . , 1990 ) . They used P . yoelii 17X clone 1 . 1 , which results in an obviously different course of infection from ours . The PyNL clone that we used appears more virulent than the 17× clone 1 . 1 as judged by the higher peak parasitemia ( 30–40% vs 10% ) and prolonged period for parasite elimination ( 30 days vs 15 days ) , suggesting that the difference in virulence may cause the different results when mice were depleted of CD8+ T cells . It is quite possible that CD8+ T cells target erythroblasts that strongly express MHC class I antigens . However , we previously reported the contribution of macrophages to CD8+-T-cell-mediated protection against malaria ( Imai et al . , 2010 ) . Those findings , together with the present study , suggest that CD8+ T cells enhance not only the phagocytotic capacity of macrophages but also the susceptibility of infected erythroblasts to phagocytosis through their display of PS . Thus , this FasL-dependent effect of CD8+ T cells on infected erythroblasts might be essential for the protective immune response to blood-stage malaria by supporting enhanced phagocytosis . Thus , CD8+ cells collaborate with macrophages to completely eradicate the parasites . The CD8+-T-cell-mediated protection that targets parasitized erythroblasts may operate in the early phase of infection , as inferred from the course of infection in mice depleted of CD8+ T cells . We have previously shown that the proportion of infected erythroblasts is constant during the course of infection , unlike the proportion of infected RBCs , which increases dramatically in the later stages of infection ( Imai et al . , 2013 ) . This means that there are relatively more infected erythroblasts in the early stage of infection . Therefore , the reduction of infected erythroblasts by CD8+ T cells in the early phase would efficiently control blood-stage malaria . From this perspective , this protective mechanism might effectively control malaria parasites in humans , in which parasitemia develops to a lower level than that observed in animal models . Indeed , parasitized erythroblasts were found within the bone marrow of patients with vivax malaria ( Ru et al . , 2009 ) , and P . falciparum parasites ( Tamez et al . , 2009 ) can infect erythroblasts in vitro . Therefore , these cells might be targets of CD8+ T cells in humans . It will be very interesting to evaluate whether parasitized erythroblasts are phagocytosed in the human bone marrow or spleen ( although this will be difficult to demonstrate experimentally ) . Notably , we have demonstrated the expression of Fas on erythroblasts infected with the malaria parasite ( Figure 2C ) . There are two possible explanations for the expression of Fas on infected erythroblasts . One is that the infection of Fas− erythroblasts with the malaria parasite induces the expression of Fas . Our findings support this because Fas− erythroblasts were markedly reduced in infected cells compared with their numbers in uninfected cells , indicating a transition from Fas− to Fas+ cells upon infection . However , the precise mechanism of this induction of Fas remains to be clarified . The other possible explanation is that the malaria parasite infects Fas+ erythroblasts . Erythroblasts are known to express Fas under physiological conditions , and Fas is considered to be involved in the negative regulation of erythropoiesis ( De Maria et al . , 1999; Liu et al . , 2006 ) . Furthermore , uninfected Fas+ erythroblasts were found in mice infected with PyNL . Activated CD8+ T cells expressing FasL might interact with uninfected erythroblasts expressing Fas and induce bystander cell damage during infection , and this system may underlie the pathogenicity of malarial anemia . The Fas/FasL-dependent , CD8+ T cell-induced PS externalization by parasitized erythroblasts required cell-to-cell contact ( Figure 6 ) . Do CD8+ T cells contact their target cells in the spleen ? As CD8+ T cells and erythroblasts occur in the splenic white pulp and red pulp , respectively ( Mebius and Kraal , 2005 ) , the opportunity for contact between them might be limited . However , infection with the malaria parasite changes the structure of the spleen and makes the white pulp indistinguishable from the red pulp ( Del Portillo et al . , 2012 ) . Thus , it is possible that CD8+ T cells contact erythroblasts in the spleens of mice infected with PyNL . In vivo imaging would be useful in confirming this , because imaging has recently shown that CD8+ T cells accumulate in the liver after sporozoite infection ( Kimura et al . , 2013 ) . We did not investigate whether erythroblasts undergo apoptosis after the ligation of Fas , as in normal cells , and whether apoptosis suppresses parasite growth . Unlike viruses , malaria parasites can multiply inside RBCs but may not require any cellular machinery for their replication , suggesting that the apoptosis of the host cells may not influence parasite growth . Indeed , a related protozoan , Toxoplasma gondii , can survive inside damaged host cells ( Yamashita et al . , 1998 ) . Instead , the elimination of the malaria parasite may require the phagocytosis of the infected cells by the phagocytes of the reticuloendothelial system , and the externalization of PS on parasitized erythroblasts via Fas/FasL plays an important role in this process . PS acts as an ‘eat me’ signal for phagocytes ( Savill and Gregory , 2007 ) and contributes to the rapid removal of infected erythroblasts and apoptotic cells . Erythroblasts are distributed in a specific region called the ‘erythroblastic island’ in the splenic red pulp . Macrophages are located in the center of the erythroblastic island and rapidly phagocytose the nuclei of erythroblasts after their enucleation under physiological conditions ( Chasis and Mohandas , 2008 ) . These macrophages may rapidly engulf infected erythroblasts as soon as PS is exposed after their interaction with CD8+ T cells . Not only the erythroblasts in the spleen , but also the infected RBCs in the peripheral blood , expose PS in response to CD8+ T cells and FasL ( Figure 3 ) . Although PS exposure on infected RBCs induced by Fas stimulation could not be reproduced in vitro coincident with the absence of Fas+ cells in the peripheral blood , we observed a substantial number of infected PS+ RBCs in the peripheral blood . One possible explanation for FasL-dependent PS exposure on infected Fas− RBCs is that infected erythroblasts exposing PS develop into RBCs after enucleation , which is associated with the shedding of MHC class I molecules . PS exposure on infected RBCs has been reported in response to several stressors during malaria ( Foller et al . , 2009 ) , and the FasL- and CD8+-T-cell-dependent system might be one cause of this PS exposure . PS exposure on infected RBCs might be part of the CD8+-T-cell-mediated protective mechanism against blood-stage malaria . We proposed that Tim-4 is a novel phagocytic receptor for infected cells . The rate at which an anti-Tim-4 antibody inhibited the phagocytosis of infected RBCs ( up to 20% ) seems appropriate because 15–20% of the macrophages used here ( obtained from uninfected mice ) expressed Tim-4 ( Figure 10 ) . However , infection with PyNL induced the expression of Tim-4 on macrophages , which may play a major role in the phagocytosis of infected cells during malarial infection . Our results also indicate that other molecules that are known PS receptors , such as PS receptor ( Hoffmann et al . , 2001 ) and developmental endothelial locus 1 ( Del-1 ) ( Hanayama et al . , 2004 ) , might be involved in the phagocytosis of infected cells . In summary , we have clearly demonstrated the protective mechanisms of CD8+ T cells against blood-stage malaria . Our findings should provide novel strategies for the development of a blood-stage vaccine based on the activation of CD8+ T cells , distinct from those strategies based on the induction of antibodies . Antigens recognized by antibodies must be expressed on the parasite's surface . Such molecules are exposed to immune pressure and acquire polymorphisms , allowing them to evade antibody recognition and causing ‘strain-specific immunity’ , which hampers the development of effective vaccines . In contrast , antigens recognized by CD8+ T cells are not restricted in their locations , and conserved intracellular molecules could be recognized after antigen presentation . Therefore , the development of malaria vaccines that activate protective CD8+ T cells against blood-stage malaria might be useful and have wide applications .
C57BL/6 ( B6 ) mice , C57BL/6JSlc-gld ( gld: generalized lymphoproliferative disease ) mice and BALB/c mice were obtained from SLC ( Hamamatsu , Japan ) or Kyudo ( Tosu , Japan ) . Rag2−/− mice were obtained from Central Laboratory of Experimental Animals ( Kawasaki , Japan ) . All mice were maintained under specific-pathogen-free conditions . Experiments were generally performed in mice aged 8–12 weeks . All mouse experiments were approved by the Committee for Ethics on Animal Experiments in the Faculty of Medicine , and performed under the control of the Guidelines for Animal Experiments in the Faculty of Medicine , Gunma University and Kyushu University , according to Japanese law ( no . 105 ) and notification ( no . 6 ) of the Government of Japan . No human samples were used in these experiments . The clonal lines of blood-stage P . yoelii 17XNL ( PyNL ) and 17XL ( PyL ) parasites originated from Middlesex Hospital Medical School , University of London , 1984 , are generous gifts from Dr M Torii ( Ehime University ) , and the generation of PyNL–GFP has been described previously ( Imai et al . , 2013 ) . Blood-stage parasites were obtained after the fresh passage of frozen stock through a donor mouse , 4–5 days after inoculation . Mice were infected intraperitoneally with 25 , 000 parasitized red blood cells ( pRBCs ) , unless otherwise indicated . All antibodies were purchased from BD Pharmingen ( Franklin Lakes , NJ , USA ) , eBioscience ( San Diego , CA , USA ) , or BioLegend ( San Diego , CA , USA ) . Fluorescein isothiocyanate ( FITC ) - and allophycocyanin ( APC ) -conjugated anti-CD3 , FITC- , phycoerythrin ( PE ) –Cy7- , and APC-conjugated anti-CD8α or β , FITC- , PE–Cy7- , and APC-conjugated anti-CD4 , FITC- , and PE- conjugated anti-CD62L , FITC- and PE-conjugated anti-CD44 , PE-conjugated anti-CD25 , PE-conjugated anti-CD69 , PE- , and biotin-conjugated anti-FasL , PE-conjugated anti-LAMP1 , PE- , PE-Cy7 , biotin-conjugated anti-Fas , FITC- , PE- , PE-Cy7- , and APC-conjugated anti-TER119 , PE- , PE-Cy7- , and APC-conjugated anti-MHC class I , PE- and PE-Cy7-conjugated anti-MHC class II , PE- and PE-Cy7-conjugated anti-CD11b , PE- and PE-Cy7-conjugated anti-F4/80 , PE- and PE-Cy7-conjugated anti-CD11c , PE- and PE-Cy7-conjugated anti-CD11b , PE- and PE-Cy7-conjugated anti-CD169 , PE- and PE-Cy7-conjugated anti-CD68 , PE- and PE-Cy7-conjugated anti-SIGNR1 , PE-conjugated anti-PanNK , and PE-conjugated anti-Tim4 antibodies were used for flow cytometry . Purified anti-CD16/32 antibodies were used for blocking . Propidium iodide ( Sigma , St . Louis , MO , USA ) or 7-amino-actinomycin D ( BioLegend ) were used for dead cell staining , when in some experiments , dead cells were excluded from the analysis . Annexin V ( BD Pharmingen ) was used to stain PS . Anti-PE- , anti-FITC- , anti-APC- , anti-CD8- , and anti-TER119 microbeads and CD8α+ T cell isolation kit ( Miltenyi Biotech , Bergisch Gladbach , Germany ) were used for MACS cell purification ( Miltenyi Biotech ) . The PKH26 Red Fluorescent Cell Linker Kit for General Cell Membrane Labeling was from Sigma–Aldrich . The culture medium was RPMI 1640 ( Sigma ) containing 10% fetal bovine serum , 2 mM l-glutamine , 1 mM sodium pyruvate , 0 . 1 mM nonessential amino acids , penicillin–streptomycin , and 2-mercaptoethanol . To induce FasL-dependent apoptosis , FasL–Strep and Strep-Tactin microtiter plates were purchased form IBA ( St . Louis , MO , USA ) . For the analysis of white blood cells ( WBCs ) , but not erythroid cells , and in vivo phagocytosis , spleen cells were added to ACK lysing buffer ( 8024 mg/l NH4Cl , 1001 mg/l KHCO3 , 3 . 722 mg/l EDTA Na2⋅2H2O ) to remove the RBCs . Cell suspensions were incubated with anti-CD16/32 antibody ( Fc block ) and stained with fluorochrome-labeled antibodies . Isotype control antibodies were also used to evaluate specific staining . Propidium iodide ( Sigma ) or 7-amino-actinomycin D ( BioLegend ) were used to stain dead cells , because dead cells were excluded from the analysis in some experiments . Cells were analyzed with a FACSCalibur or FACSAria II flow cytometer ( Becton Dickinson , San Jose , CA , USA ) , and the data were analyzed with the FlowJo software ( Treestar , Ashland , OR , USA ) . Samples were analyzed using a Biorevo BZ-9000 microscope ( Keyence , Osaka , Japan ) . The data were analyzed with the BZ-II software ( Keyence ) . The depletion of CD4+ or CD8+ T cells was performed as previously described ( Imai et al . , 2008 ) . Briefly , mice were intraperitoneally administered 0 . 5 mg of anti-CD4 ( clone: GK1 . 5 ) , anti- CD8α ( 2 . 43 ) , or anti- CD8β ( YTS156 . 7 . 7 ) antibodies 1 day before and 14 and 28 days after PyNL infection . The depletion of each T-cell subset was checked by flow cytometry , and >99% of the appropriate cell subset was depleted in the peripheral blood by 24 hr after inoculation ( Figure 1—figure supplement 1A ) . The depletion of splenic CD8+ T cells in malaria-infected mice is shown in Figure 1—figure supplement 1B . The protocols for the prime–boost live vaccination and cell transfer are shown in Figure 1D . CD8+ T cells were isolated from WT and gld mice infected with PyNL ( 25 , 000 pRBC ) after two boosts with PyL ( 50 , 000 pRBC ) at 6 and 9 weeks after the primary PyNL infection . Then , 1 × 107 purified cells were transferred to recipient x-ray-irradiated ( 5 . 5 Gy ) mice or Rag2−/− mice . The recipients were infected with PyL ( 50 , 000 pRBC ) 1 week after cell transfer . The collected RBCs or pRBCs were washed twice with medium . The cells ( 2 × 107 cells/ml ) were stained with 250 nM carboxyfluorescein succinimidyl ester ( CFSE: Molecular Probes; Life Technologies , Carlsbad , CA , USA ) for 1 min . Staining was stopped by the addition of fetal calf serum , and the cells were washed three times with medium . Splenic CD11b+ macrophages from uninfected mice were sorted with the MACS cell separation system , and the labeled with PKH26 , according to the manufacturer's instructions . Splenic CD11b+ macrophages ( 1 × 105 cells ) were cocultured with CFSE-labeled pRBCs or normal RBCs in a 1:30 ratio , at a final volume of 200 μl for 4 hr at 37°C in a CO2 incubator with culture medium . Following coculture , the noningested RBCs were removed with ACK lysing buffer . The capacity of macrophages to phagocytize CFSE-labeled pRBCs or normal RBCs was analyzed with a FACSCalibur flow cytometer . For the in vitro phagocytosis inhibition assay , anti-Tim-4 antibody and its isotype control antibody were added to the test sample . Sorted erythroid cells ( 3 × 105 cells/well ) from gld mice 17 days after infection with PyNL–GFP were cocultured with CD8+ T cells from WT or gld mice 17 days after PyNL infection or from uninfected WT mice at 37°C for 4 hr in a CO2 incubator with culture medium . Effector ( CD8 ) : The target ( erythroid ) ratio was 0:1–10:1 . The cells were Fc-blocked and stained with PE-Cy7-conjugated anti-TER119 antibody . PS was then stained with PE-conjugated annexin V in calcium-containing annexin V binding buffer ( BD Pharmingen ) . The parasitized cells ( TER119+ GFP+ ) or unparasitized cells ( TER119+ GFP− ) were analyzed for PS expression with flow cytometry . FasL–Strep ( 0–100 ng/ml ) was added to a Strep-Tactin microtiter plate , and incubated for 20 min at 37°C . Sorted erythroid cells from PyNL–GFP-infected gld mice ( 2 × 105 cells/well ) were cultured for 4 hr at 37°C in the abovementioned plate . The PS was then surface stained with PE-conjugated annexin V in annexin V binding buffer . In some cases , cells were Fc-blocked and stained with APC- or PE-Cy7-conjugated anti-TER119 antibody and PE-Cy7-conjugated anti-MHC class I antibody , and then analyzed with flow cytometry . Macrophage depletion methods have been previously described ( Van Rooijen and Sanders , 1994; Ishida et al . , 2013 ) . Mice were intravenously injected with clodronate ( Sigma ) liposome ( C/L: 1 . 5 mg clodronate/300-μl liposome suspension ) 3 and 9 days after PyNL infection . Two sets of data ( control vs experimental group ) were compared and Mann–Whitney U-test was used for statistical analysis . A p-value of p < 0 . 05 was considered to be statistically significant . Significant differences in survival were tested with a log-rank test using Kaplan–Meier survival curves . | The immune system consists of several different types of cell that work together to prevent infection and disease . For example , immune cells called cytotoxic CD8+ T cells kill tumor cells or other cells that are infected . To do so , the CD8+ T cells must recognize certain molecules on the surface of the tumor or infected cells and bind to them . Malaria is an infectious disease caused by the Plasmodium parasite , which is transferred between individuals by mosquitoes . The parasite is able to evade the immune system—so much so that it is not well understood how the immune system tries to respond to stop the infection . This has made it difficult to develop a vaccine that protects against malaria . During the latter stages of a malaria infection , the parasite infects the host's red blood cells . It was long believed that CD8+ T cells did not help to eliminate the red blood cells that had been infected by Plasmodium . However , recent work in mice suggested that CD8+ T cells do respond to infected erythroblasts—precursor cells that develop into red blood cells—and that CD8+ T cells help protect mice against blood-stage malaria . Now , Imai et al . describe how the CD8+ T cells in mice help to kill erythroblasts infected with Plasmodium yoelli , a species of the parasite used to study malaria in mice . The infected cells display a protein called Fas on their surface . Imai et al . found that , during a malaria infection , the CD8+ T cells produce a protein that can interact with Fas . This interaction causes the infected cell to move a signaling molecule to its outside surface , which encourages another type of immune cell to engulf and destroy the infected cell . This knowledge of how CD8+ T cells fight Plasmodium parasites in the bloodstream could now help to develop new types of blood-stage vaccine for malaria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2015 | Cytotoxic activities of CD8+ T cells collaborate with macrophages to protect against blood-stage murine malaria |
An ‘interactome’ screen of all Drosophila cell-surface and secreted proteins containing immunoglobulin superfamily ( IgSF ) domains discovered a network formed by paralogs of Beaten Path ( Beat ) and Sidestep ( Side ) , a ligand-receptor pair that is central to motor axon guidance . Here we describe a new method for interactome screening , the Bio-Plex Interactome Assay ( BPIA ) , which allows identification of many interactions in a single sample . Using the BPIA , we ‘deorphanized’ four more members of the Beat-Side network . We confirmed interactions using surface plasmon resonance . The expression patterns of beat and side genes suggest that Beats are neuronal receptors for Sides expressed on peripheral tissues . side-VI is expressed in muscle fibers targeted by the ISNb nerve , as well as at growth cone choice points and synaptic targets for the ISN and TN nerves . beat-V genes , encoding Side-VI receptors , are expressed in ISNb and ISN motor neurons .
Protein-protein interactions ( PPIs ) control a vast array of processes in metazoans , ranging from signal transduction and gene regulation within cells to signaling between cells via cell surface and secreted proteins ( CSSPs ) . The strength of PPIs varies widely , from high-affinity interactions that create stable protein complexes to weak transient interactions ( Nooren and Thornton , 2003 ) . Defining global PPI patterns ( ‘interactomes’ ) has been the focus of much recent research . Progress has been made in generating high-throughput protein interaction data for a variety of organisms , including S . cerevisiae ( Tarassov et al . , 2008 ) , C . elegans ( Li et al . , 2004; Simonis et al . , 2009 ) and D . melanogaster ( Guruharsha et al . , 2011; Giot et al . , 2003 ) . Methods used to create interactomes include affinity purification/mass spectrometry ( AP-MS ) and the yeast two-hybrid assay ( Y2H ) . It is estimated that up to one sixth of human genes encode CSSPs ( Bushell et al . , 2008 ) . CSSPs control signaling from the extracellular milieu to cells and the flow of information between cells . Due to their importance and accessibility , CSSPs are often the targets for therapeutic agents , including humanized monoclonal antibody drugs such as checkpoint inhibitors ( e . g . , Yervoy and Keytruda ) , the non-Hodgkin’s lymphoma drug Rituxan , and the breast cancer drug Herceptin . However , the biochemical properties of many CSSP interactions prevent them from being detected by commonly used techniques employed in high-throughput PPI screens , and CSSPs are underrepresented in global interactomes ( Guruharsha et al . , 2011; Braun et al . , 2009; Miller et al . , 2005 ) . There are several reasons for this . First , these proteins are usually glycosylated and have disulfide bonds , so they need to be expressed in the extracellular compartment ( Wright et al . , 2010 ) . CSSP interactions between monomers are also often weak , with KDs in the μM range ( van der Merwe et al . , 1994 ) , making them difficult to capture due to their short half-lives . Lastly , insoluble transmembrane domains on cell surface proteins preclude their purification with standard biochemical techniques , which makes them difficult to study using methods such as AP-MS ( Wright , 2009 ) . Despite these difficulties , recent advances have been made in the study of global CSSP interaction patterns . Interactions among cell-surface proteins ( CSPs ) often occur between clusters of proteins on cell surfaces , and avidity effects ( stronger binding due to clustering ) can make these cell-cell interactions stable even when monomers bind only weakly . To facilitate detection of interactions among CSSP extracellular domains ( ECDs ) in vitro , several groups have taken advantage of avidity effects by attaching ECDs to protein multimerization domains and expressing ECD fusions as soluble secreted proteins ( Bushell et al . , 2008; Wojtowicz et al . , 2007; Söllner and Wright , 2009; Ramani et al . , 2012 ) . These methods have been shown to be effective , allowing detection of interactions that otherwise would not have been observable . Özkan et al . scaled up the avidity-based approach , developing a high-throughput ELISA-like screening method , the Extracellular Interactome Assay ( ECIA ) . The ECIA was used to define interactions among 202 Drosophila CSSPs , comprising all CSSPs within three ECD families . These were the immunoglobulin superfamily ( IgSF ) , fibronectin type III ( FNIII ) and leucine-rich repeat ( LRR ) families . The ECIA utilized dimers as ‘bait’ and pentamers as ‘prey’ . It detected 106 interactions , 83 of which were previously unknown ( Özkan et al . , 2013 ) . The most striking finding from the ECIA interactome was that a subfamily of 21 2-IgSF domain CSPs , the Dprs , selectively interacts with a subfamily of 9 3-IgSF domain CSPs , the DIPs , forming a network called the ‘Dpr-ome’ ( Özkan et al . , 2013 ) . Each Dpr and DIP that has been examined is expressed by a small and unique subset of neurons at each stage of development . One Dpr-DIP pair is required for normal synaptogenesis and influences neuronal cell fate . In the pupal optic lobe , neurons expressing a Dpr are often presynaptic to neurons expressing a DIP to which that Dpr binds in vitro ( Carrillo et al . , 2015; Tan et al . , 2015 ) . The Dpr-ome initially defined by the global interactome contained several ‘orphans’ , proteins with no binding partner ( Özkan et al . , 2013 ) . By expressing new versions of Dprs and DIPs , including chimeras , and using these to conduct a ‘mini-interactome’ analysis of the Dpr-ome , we were able to find partners for all but one orphan . That protein , Dpr18 , has changes to conserved binding interface residues and may lack the capacity to bind to any DIPs ( Carrillo et al . , 2015 ) . The ECIA also identified a second IgSF network , formed among members of the Beaten Path ( Beat ) and Sidestep ( Side ) protein subfamilies . Beat-Ia and Side were identified by genetic screens for motor axon defects , and were later shown to have a ligand-receptor relationship . They control the projection of motor axons to muscle targets ( Fambrough and Goodman , 1996; Sink et al . , 2001; de Jong et al . , 2005; Siebert et al . , 2009 ) . Beat-Ia is expressed on motor axons , where it binds to Side , which is expressed on muscles . This binding causes motor axons to decrease their adhesion to each other , allowing them to leave their bundles and turn onto the muscle fibers . beat-Ia and side have strong motor axon phenotypes . In the absence of either protein , motor axons often remain in their fascicles and never leave to arborize on their target muscles ( Siebert et al . , 2009; Aberle , 2009 ) . The ECIA detected the known Beat-Ia::Side interaction , and also uncovered other interactions between members of the Beat and Side subfamilies ( Özkan et al . , 2013 ) . Seven of the 14 Beats were found to bind to four of the eight Sides . The remaining Beats and Sides were still orphans with no binding partners in the other subfamily . The functions of the newly defined interactions between Beats and Sides were unknown . Most beat genes are expressed in embryonic neurons . Some Beats were genetically characterized using deletion mutations and RNAi , but loss of these Beats did not cause strong motor axon phenotypes ( Pipes et al . , 2001 ) . None of the other Side subfamily members had been examined . This paper describes the development of a new method for interactome screening , which we call the BPIA ( Bio-Plex Interactome Assay ) . This method uses the ‘Bio-Plex’ system , which employs Luminex xMAP technology ( Houser , 2012 ) . Our method detects binding of a prey protein to many bait proteins , each conjugated to a bead of a different color , in each assay well . For the ECIA , the number of assays required for the interactome screen was the square of the number of proteins examined , while with the Bio-Plex the number of assays could be equal to the number of proteins . In principle , then , the Bio-Plex might greatly speed up interactome screening , and might also be more sensitive , since the available signal-to-background ratio is much greater for the Bio-Plex than for the ECIA . As a test of the method , we used a Bio-Plex 200 to do a mini-interactome screen of the Beat-Side network . Based on the the fact that the Dprs and DIPs that were initially orphans ( Özkan et al . , 2013 ) were later shown to have binding partners ( Carrillo et al . , 2015 ) , we hypothesized that some of the orphan Beats should have Side partners , and vice versa . Consistent with this hypothesis , we were able to deorphanize two more Beats and two Sides using the BPIA . To further our understanding of Beat and Side function during embryonic development , we characterized expression of several Beats and Sides , focusing primarily on Side-VI and the three Beat-Vs , which were the strongest interactors in both the ECIA and BPIA screens . The three beat-Vs exhibit differential expression in identified motor neurons , while side-VI is expressed at motor axon choice points and in a subset of target muscle fibers .
Gene duplication , a key phenomenon in the expansion of gene families , provides opportunity for the fine-tuning or innovation of protein interactions and functions ( Ohno , 1970 ) . The duplication of genes encoding receptors or ligands that have multiple binding partners can lead to partitioning of the interactions among the paralogs . Relaxed constraints due to redundancy between duplicated genes can result in the exploration of new functions . In these ways , members of ligand and receptor families can establish a complex interaction network in which each binding pair has a distinct expression pattern and function . The Beat IgSF subfamily was previously characterized in Drosophila melanogaster ( Pipes et al . , 2001 ) . Here we show that orthologs of each Beat are found in most of the other 12 sequenced Drosophilid species ( Figure 1A ) . Beats have two IgSF domains , and there are both secreted and membrane-bound isoforms . There are 14 Beat proteins , divided into seven clusters based on their phylogenetic relationships . Beats that are most closely related to each other ( e . g . , Beats Ia , Ib , and Ic ) are encoded by clustered genes and denoted by a Roman numeral followed by a letter . While divergence rates within the beat family phylogeny are highly asymmetric following the earliest duplications , groups of beats within each of the clusters of paralogs have similar divergence levels ( e . g . , Beat-IIa and Beat-IIb present similar rates of evolution ) . Beats encoded within clusters also have similar ( but not identical ) embryonic expression patterns ( Pipes et al . , 2001 ) . These observations suggest that beats have undergone two levels of specialization: functional specialization after duplication and emergence of the seven major Beat branches , followed by individuation of expression patterns and binding specificities for members of the four subclusters ( I , II , III , V ) . Proteins in the Side subfamily are all transmembrane proteins , and none are encoded by clustered genes ( Sink et al . , 2001; Aberle , 2009 ) . An exhaustive analysis using SMART ( Schultz et al . , 1998 ) , HMMER ( Eddy , 2011 ) and DOUT-finder ( Novatchkova et al . , 2006 ) to identify outlier homologs of structural domains reveals that the Side family of paralogs has an invariant ECD architecture composed of five IgSF domains followed by an FNIII domain ( Figure 1B ) . In addition to the protein domain-based composition , phylogenetic inferences provide evidence for a cohesive subfamily of Side proteins ( Figure 1C ) . We refer to Sidestep as Side , and have designated names for the other seven Side paralogs based on their evolutionary distance from Side . All Side paralogs seem to present similar or comparable levels of inter-species ( intra-paralog ) divergence , indicating that different Side paralogs have undergone similar selective constraints . The presence of Side paralogs in most of the 12 sequenced Drosophilids , and the presence of orthologs of some of these paralogs in the mosquito Anopheles gambiae , clearly indicates the origin of the Side family through successive duplication events that pre-dated Drosophilid speciation . We could not identify orthologs in all 12 Drosophilids for all Sides , likely due to incomplete genomic sequence ( see Materials and methods ) . The most likely scenario given our phylogenetic trees is that the ancestral Side subfamily gene duplicated successively , followed by a rapid sequence and functional divergence predating Drosophilid speciation . Indeed , rooted trees for the Side family show a dynamic history of gene duplication and divergence , with asymmetric clusters of duplicates resulting from faster evolution of one gene copy compared to its sister , indicating possible functional divergence and specialization after gene duplication . Our rooted phylogeny of the Side paralogs differs from a previous unrooted one ( Aberle , 2009 ) . The low bootstrap support values ( p<60% ) for some of the internal tree branches indicate rapid successive duplication events . The long branches post-dating duplications but predating speciation support enormous divergence between the duplicates at the sequence level , followed by strong purifying selection after speciation . To attempt to deorphanize more members of the Beat and Side superfamilies , and to develop new assays that might eventually streamline the process of creating global interactomes , we investigated technologies that have the potential to provide higher throughput and greater sensitivity . The Bio-Rad Bio-Plex system is based on the principles of flow cytometry and can be used for a variety of high-throughput , multiplexed assays . It uses magnetic polystyrene beads impregnated with different ratios of fluorescent dyes ( each variant is called a ‘bead region’ ) , rendering them spectrally distinct when excited by a laser . The Bio-Plex 200 used in our experiments employs two dyes and has 100 different bead regions , allowing for the simultaneous analysis of up to 100 distinct bead-bound analytes , while the Bio-Plex 3D has 500 bead regions . The beads can be conjugated to lysine residues on proteins through carboxyl groups on their surfaces . Protein-conjugated bead regions are mixed and incubated with soluble proteins . Binding between soluble and bead-bound proteins can be detected using phycoerythrin ( PE ) -coupled secondary antibodies or other fluorescent reagents . Beads flow through the machine in single file , and are interrogated by two lasers: one to discern the identity of the bead region , and the other to detect the PE signal , representing the amount of bound binding protein . This assay has a high signal-to-background ratio , because strong binding of the bead-bound analyte can generate readings of >20 , 000 , vs . <100 for unconjugated beads . Luminex xMAP technology has been used by other groups to assay interactions between proteins ( Blazer et al . , 2001; Rimmele et al . , 2010; Blazer et al . , 2011 ) . For example , Blazer et al . used avidin-coupled bead regions to capture Regulator of G protein Signaling ( RGS ) proteins . These were then incubated with fluorescently labeled Gαo to measure RGS-G protein interactions and identify compounds that could inhibit these interactions . In developing a Bio-Plex-based assay for CSSP interactions , we took advantage of avidity , utilizing the same dimer and pentamer constructs employed for the ECIA , but reversing the bait and prey roles ( Özkan et al . , 2013 ) . Bait proteins were alkaline phosphatase ( AP ) fusion proteins of ECDs pentamerized using a sequence from cartilage oligomeric matrix protein ( COMP ) ( Bushell et al . , 2008; Voulgaraki et al . , 2005 ) . Prey constructs were fusions of ECDs to human Fc , which is a dimer . Preys contain a C-terminal V5 epitope tag , so a V5 antibody was used to detect binding , followed by a secondary antibody conjugated to PE . We developed an affinity-capture method to attach bait proteins to the beads that avoided the necessity to purify baits ( Figure 2A ) . To accomplish this , we added a sequence encoding an Avitag , a 15 amino acid sequence that is recognized by the enzyme biotin ligase ( BirA ) , at the C-terminal end of each bait protein coding region . BirA adds one biotin molecule to the tag ( Ashraf et al . , 2004; Sung et al . , 2011; Wang et al . , 2013 ) . To perform in vivo biotinylation , we co-transfected the bait constructs with an endoplasmic reticulum ( ER ) -localized BirA construct optimized for expression in S2 cells ( Tykvart et al . , 2012 ) . To capture bait proteins , we coupled each bead region to streptavidin , and incubated each with media containing a different biotinylated bait protein , thus bypassing the purification step . Each Fc-tagged prey protein was also expressed in S2 cells , and purified with Ni-NTA resin . The bait-coated beads were then mixed and aliquoted and a different Fc prey protein added to each tube . The reactions were then washed and incubated with anti-V5 antibody , followed by PE-conjugated secondary antibody , before being transferred to a 96-well plate and read with the Bio-Plex . This is the BPIA assay . The biotin-streptavidin interaction is one of the strongest non-covalent interactions known in nature , with a KD on the order of ~10−14 mol/L ( Hendrickson et al . , 1989 ) , so we expected that bait proteins should not be able to ‘jump’ to other beads after bead regions are mixed . To test this , we incubated multiple bead regions with bound baits overnight together with a streptavidin-coupled bead region lacking a bait . We observed no jumping of baits to the bead region without a bait ( see Materials and methods for details ) . We performed a Beat-Side mini-interactome for the 22 Beat and Side subfamily members plus the IgSF protein CG17839 , which binds to Side-VII ( Özkan et al . , 2013 ) , using the Bio-Plex system . The purposes of this experiment were to: ( 1 ) demonstrate that we could screen for binding of all baits to a prey within a single well , and therefore that we could do the entire assay with 23 wells plus controls ( vs . the 529 wells that would have been required by the original ECIA ) , and , ( 2 ) determine if we could observe all of the interactions found by the ECIA . We also hoped that the BPIA might be more sensitive than the ECIA , due to its high signal-to-background ratio , and therefore might uncover previously unknown interactions . 23 bait constructs were co-transfected with the BirA plasmid in S2 cells , and AP bait protein-containing media harvested . The bait proteins were then captured directly from media using streptavidin-coupled beads , and the beads mixed together . As prey , 23 different His-tagged ECD-Fc constructs were transfected into S2 cells , and the fusion proteins purified with Ni-NTA resin . We analyzed each potential interaction pair in both orientations , with protein A as bait and protein B as prey , and vice versa . The pooled beads were incubated with a prey protein overnight , washed , and then incubated with anti-V5 antibody followed by PE-conjugated secondary antibody , transferred to a 96-well plate , and analyzed with the Bio-Plex . Figure 2B shows a heat map of the raw interaction signals for the 23 × 23 matrix . Direct protein capture with streptavidin-coupled beads enabled us to bypass protein purification for bait proteins . We were interested to see if the assay could also be performed using unpurified prey proteins , which would further reduce the workload involved . To test this , we performed the BPIA using a subset of the Beat and Side subfamily proteins . Bait proteins were expressed and captured as described above . The prey proteins were expressed in S2 cells grown in Sf-900 III , a serum-free media optimized for protein expression in insect cells . We chose serum-free media due to the fact that the high concentration of extraneous proteins present in normal ( serum-containing ) S2 media lowers the signal to background ratio of the assay ( data not shown ) . Using this method , we were able to find all of the interactions seen with purified prey , except for Beat-Ic::Side ( perhaps due to low expression of Beat-Ic-AP bait in this experiment ) ( Figure 2C ) . These results show that the BPIA is also compatible with the use of unpurified prey proteins . To analyze the Beat-Side mini-interactome , we utilized methods based on those of Özkan et al , who used a Z score system to classify interactions ( Özkan et al . , 2013 ) . They used a cutoff to eliminate outliers , but we could not do this because our data set is much smaller and a large fraction of the proteins interact ( see Materials and methods ) . Accordingly , to process our data , we utilized bootstrapping of the median for each row and each column . Briefly , for every row and column , 23 numbers were chosen randomly with replacement , and the median calculated . After n cycles , a histogram of the median was generated and the mean and SD for that row or column calculated . A Z score was then calculated for each number in the row or column based on the generated mean and SD . In this manner , for each number in the matrix , two different Z scores were generated ( one based on row , the other on column ) . The two Z scores were then averaged . Each Beat-Side protein pair appears twice in the matrix , since each protein is used as both bait and prey . The two averaged Z-scores for each pair represent interactions in opposite orientations . It was expected that these values would be discrepant , as in the ECIA , due to differences in protein expression , binding geometries , and other factors . To incorporate interactions in both directions into the analysis , we calculated the geometric mean ( square root of the product ) of the two Z scores ( note that geometric mean can only be calculated if both scores are >0 ) . If the geometric mean was greater than five , the Beat-Side pair was scored as a genuine interactor . Figure 3A graphically displays these results . It is a quantized heat map generated from the geometric means of Z scores for each protein pair . Each Z score was assigned to one of three categories: high ( dark blue ) , mid ( light blue ) , and low ( white ) , which were determined using cutoffs of 80% and 90% . These cutoffs were chosen so that the ‘high’ Z-score category corresponded to the pairs we scored as genuine interactors , having a geometric mean of Z-scores that was >5 . For comparison , Figure 3B shows a heat map based on the ECIA data ( Özkan et al . , 2013 ) . All of the hits found with the ECIA were in the high category , with the exception of Beat-Ia::Side . This is likely due to poor expression of Beat-Ia-AP bait , resulting in one of the Z scores being zero . Note that in Figure 2B a Beat-Ia::Side interaction is detected when Beat-Ia is the Fc prey and Side is the AP bait . Beat-VI::Side-II , Beat-Ic::Side-III , and Beat-Ic::Side are in the high category in our heat map , and we concluded that these are new interacting pairs . The ‘revised’ Beat-Side subfamily interaction network is shown in Figure 3C , with new interactions identified by the BPIA indicated by red lines . Interestingly , we observe a ‘phylogenetic mirroring’ between Sides and Beats when we compare the network diagram ( Figure 3C ) with the evolutionary trees of Figure 1 . Sides closer to the root of the tree tend interact with Beats also close to the root of the tree and vice versa . Side members separated by more branchpoints from the root of the tree may have become functionally specialized to interact with the more recently duplicated Beats . To confirm the interactions between Beat and Side subfamily proteins observed in the ECIA and BPIA and determine their affinities , we used surface plasmon resonance ( SPR ) . We purified monomeric ECDs from proteins expressed using the baculovirus system . For the Beat-V::Side-VI interactions found in the ECIA , we flowed Beat-V ECDs over the surface of Biacore chips layered with Side-VI to determine their binding affinities and the kinetics of the interactions . Binding data show that dissociation kinetics are too fast to measure ( koff ≥0 . 5 s−1 ) . Therefore , SPR responses are only fitted at equilibrium to a binding isotherm and their fit is indicative of specific interaction . Affinities ( KDs ) for binding of the three Beat-Vs to Side-VI are in the µM range ( 0 . 76 µM , 2 . 3 µM and 9 . 4 µM for Beat-Va , Beat-Vb and Beat-Vc , respectively; Figure 4A–C ) . These dissociation constants are in the same range as those we have previously described for interactions between Beat-Ia and Side ( Özkan et al . , 2013 ) and are typical for interactions of cell adhesion molecules ( van der Merwe et al . , 1994 ) . We also examined the three new Beat/Side interactions discovered with the BPIA: Beat-VI::Side-II , Beat-Ic::Side-III , and Beat-Ic::Side . To verify these interactions , we also measured them using SPR . Beat-VI and Beat-Ic ECDs were captured on Biacore chips and Side-II and Side-III were flowed over the chips . Interactions were observed between Beat-VI and Side-II ( KD: 2 . 78 μM ) and Beat-Ic and Side-III ( KD: 63 . 5 μM ) ( Figure 4D , E ) . Binding was also observed between Beat-Ic and Side , although Side ECD precipitation precluded the collection of a titration series . The strongest interactions observed in both the ECIA and BPIA were those between Beat-Vs and Side-VI ( Figures 2B and 3 ) . Interestingly , with both assay the observed signals were related to the measured KDs for these interactions , with Beat-Va ( tightest binder ) >Beat-Vb>Beat Vc . To further characterize these interactions , we expressed Beat-Va , Beat-Vb , and Beat-Ia on the surfaces of S2 cells , and evaluated their binding to Side-VI-AP . We observed binding for both Beat-Va and Vb , but not for Beat-Ia ( Figure 5—figure supplement 1 ) . We then determined if Side-VI could bind to Beat-Vb in embryos by live staining with Side-VI-AP ( Özkan et al . , 2013; Lee et al . , 2009; Fox and Zinn , 2005 ) . To do this experiment , we expressed Beat-Vb from a UAS construct using a strong pan-cellular GAL4 driver , Tub-GAL4 . Figure 5 shows that Side-VI-AP strongly stains muscle fibers in Tub >Beat Vb embryos , but not in wild-type embryos . In wild-type embryos , punctate Side-VI-AP staining is observed on motor axons ( Figure 5—figure supplement 2 ) , consistent with the fact that Beat-Vs are expressed by motor neurons ( see below ) . The expression pattern of Side protein is dynamic in space and time . Its expression pattern changes as motor axons grow toward their targets , so that Side marks the cells over which growth cones travel during each stage of development ( Sink et al . , 2001; Siebert et al . , 2009 ) . At stage 12 it is expressed in cells in a belt-like pattern within the CNS and slightly later in a cluster of cells with a triangular pattern that are contacted by pioneer motor axons in the intersegmental nerve ( ISN ) on their way to the dorsal muscle field . At later stages Side is expressed in sensory afferents , where it is downregulated following contact with Beat-Ia expressing motor axon growth cones , and Side subsequently appears on the muscle fibers . Thus , Side labels substrates followed by ISN axons at each stage of their growth toward their muscle targets ( Siebert et al . , 2009 ) . We reasoned that expression of side paralogs at guideposts or choice points along these nerve tracts would be an indicator of other Beat-Side interactions that might be important for motor axon guidance . Therefore , we examined the embryonic expression of side-II , side-III , side-IV , side-VI , side-VII and side-VIII by fluorescent in situ mRNA hybridization ( FISH ) and labeled all motor axons with the anti-fasciclin-II antibody mAb 1D4 , to assess the coordinates of side paralog expression relative to motor axon trajectories . All of these side genes , with the exception of side-VIII , are expressed in peripheral tissues traversed by motor and sensory axons . side-II is transcribed broadly in the CNS and to a lesser extent in the developing musculature at stage 15 ( Figure 6A ) . side-III is initially expressed at high levels in the mesoderm and muscle primordia and broad transcription in the CNS increases as embryonic development progresses . By stage 14 peripheral side-III expression is strongest in the developing trachea and in stripes in the ectoderm along the parasegmental furrows ( Figure 6B , N ) . The tracheal branches are known intermediate targets of the ISN and sensory axons ( Younossi-Hartenstein and Hartenstein , 1993; Harris and Whitington , 2001 ) . side-IV is expressed in ventral muscle precursors at stage 12 , and by stage 14 it is localized to ventral muscles ( muscles 15 , 16 , and 17 ) and to lateral muscles 5 and 8 ( Figure 6C , D , N ) . side-VII shows broad expression in the CNS at stage 14 , and is also detected in the dorsal tracheal trunk ( Figure 6K , N ) . side-VIII expression is quite different from these other side genes . There is no detectable expression outside the CNS ( data not shown ) , and at stage 16 expression is restricted to a subset of CNS neurons , including the RP1 , 3 , 4 , and five motor neurons and the pCC interneuron ( Figure 6L , M ) . The pattern of expression of side-VI is of particular interest , because it is expressed at key choice points for motor axons in the ISN , intersegmental nerve b ( ISNb ) nerves , and transverse ( TN ) nerves . It is broadly expressed in the CNS ( Figure 6E , I ) . At stages 15 and 16 it is expressed in subsets of muscle fibers , including ventrolateral muscles 12 and 13 , which are the targets of the RP4 and RP5 ISNb motor axons ( Figure 6J , N ) . side-VI is also transcribed in cells whose surfaces are explored by the ISN tip , such as the dorsal cluster of Lim3-positive sensory neurons that fasciculate with the ISN ( Figure 6F , N ) , and in a ‘persistent Twist expressing cell’ ( PT cell ) which coincides with the first branch point of the ISN within the dorsal musculature ( Bate et al . , 1991 ) ( Figure 6G , N ) . side-VI is also expressed in certain targets of the TN that are known to be essential for its guidance , including the lateral bi-dendritic neuron ( LBD ) ( Figure 6H , N ) and the dorsal median cell ( DMC ) ( Chiang et al . , 1994; Gorczyca et al . , 1994 ) ( Figure 6I , N ) . Overall , the expression patterns of the side-III , -IV , -VI , and -VII genes in the periphery are consistent with the idea that they could have roles like that of side , encoding guidance cues for motor or sensory axons . In particular , the dynamic nature of side-VI expression and the fact that it is expressed at intermediate and final targets of the ISN , ISNb , and TN suggests that it may play a role in guiding motor axons towards their targets through receptors expressed on these nerves . However , our phenotypic analysis ( see Discussion ) suggests that it has redundant functions with other guidance cues , since guidance defects at the Side-VI-expressing choice points are not observed in most segments of side-VI mutant embryos . The embryonic in situ hybridization data described above show that side-VI is expressed in the CNS , but do not indicate whether it is transcribed in motor neurons . We thus used a side-VI-T2A-GAL4 line ( Diao et al . , 2015 ) generated from a MiMIC insertion ( Venken et al . , 2011 ) to evaluate expression of the gene . This driver did not express in embryos , but in third instar larvae the GFP reporter driven by the GAL4 is present in all type 1b and 1 s NMJs , indicating that at this stage of development the gene is expressed in all glutamatergic motor neurons ( Figure 6—figure supplement 1 ) . We also observed expression in subsets of sensory neurons and in the ventral nerve cord ( data not shown ) . Pipes et al . showed that expression of all beat genes except beat-IIs is restricted to the CNS at stage 16 ( Pipes et al . , 2001 ) . We focused on five of the six beat-V and beat-I genes , for which we could readily detect expression in single cells by in situ hybridization , in order to determine whether the individual genes within these clusters had acquired unique expression patterns , as one might predict based on the fact that these duplicated genes are maintained in most or all Drosophilid species . Having shown that side subfamily genes are expressed in cells targeted by motor axons , we wished to determine if the genes encoding their Beat receptors were expressed in motor neurons and , if so , to identify those motor neurons . We thus performed in situ mRNA hybridization combined with simultaneous immunohistochemistry and confocal imaging using two marker lines ( RN2-Gal4 and Lim3A-tau-myc [Thor et al . , 1999; Fujioka et al . , 2003] ) , to specifically identify the ISN neurons aCC and RP2 , which innervate dorsal muscles ( Fujioka et al . , 2003 ) , and the ISNb neurons RP1 , 3 , 4 and 5 , which innervate ventrolateral muscles . Within the beat-I subfamily , beat-Ia is expressed in both the aCC and RP2 motor neurons of the ISN ( Figure 7A , F ) where its transcription is dependent on eve ( [Zarin et al . , 2014] and data not shown ) and in RP1 , 3 , 4 and 5 ( [Pipes et al . , 2001] and data not shown ) . beat-Ic is expressed in aCC and RP2 , as well as in the pCC interneuron , which also expresses RN2-GAL4 , but not in RP1 , 3 , 4 , and 5 ( Figure 7D , E , F ) . beat-Ib is expressed at low levels in aCC and pCC ( dots in Figure 6C ) , but not detectably in RP1 , 3 , 4 , and 5 ( Figure 6B ) . beat-Va and beat-Vb are differentially expressed in ISNb and ISN motor neurons . beat-Va is absent from RP1 , 3 , 4 and 5 ( Figure 6F , G ) , but is expressed at high levels in RP2 and at lower levels in aCC ( Figure 6F , H ) . By contrast , beat-Vb is expressed at high levels in RP1 , 3 , 4 and 5 ( Figure 6F , I ) , but at lower levels in aCC and RP2 ( Figure 6F , J ) . These results show that embryonic expression patterns within the CNS have diversified between clustered beat-I and beat-V paralogs .
We examined the expression patterns of side and beat genes in order to obtain insights into their possible functions . Most sides are expressed in cells in the periphery as well as in the CNS , while most beats are expressed only by CNS neurons , including motor neurons ( Figures 6 and 7 ) ( Pipes et al . , 2001 ) . Beat-Ia::Side interactions are required for normal motor axon guidance , and highly penetrant motor axon defects in which muscles remain uninnervated are observed in mutants lacking either protein ( Fambrough and Goodman , 1996; Sink et al . , 2001; Siebert et al . , 2009 ) . By contrast , partial loss of function of beat-Ib , beat-Ic , both beat-IIs , or beat-VI causes motor axon defects with less than 20% penetrance ( Pipes et al . , 2001 ) . Genetic redundancy is a common theme in motor axon guidance ( see ref . [Zarin et al . , 2014] ) , so it is not surprising that only low-penetrance defects are observed when Beat paralogs are not expressed . Given that Beat-Ia and Side both interact with other partners ( Figure 3 ) , it is perhaps remarkable that beat-1a and side have such strong phenotypes as single mutants . We found that Beat-V and Side-VI also have redundant functions in motor axon guidance . side-VI is expressed in motor axon targets , including muscles 12 and 13 ( Figure 6 ) and interacts with the three Beat-Vs , at least two of which are expressed in subsets of motor neurons ( Figure 7 , Figure 5—figure supplement 2 ) . Beat-V::Side-VI interactions produced the strongest signals in both the ECIA and BPIA ( Figures 2B and 3 ) . We observed low-penetrance ( ~1/5 of stage 17 embryonic hemisegments affected ) muscle 12 innervation defects in side-VI insertion mutants or in deletion mutants lacking all three beat-V genes ( unpublished results ) . There were also low-penetrance ISN guidance defects in both mutants . The fact that most muscle 12 s are innervated normally in beat-V or side-VI mutants indicates that , while Beat-V::Side-VI interactions may contribute to correct targeting of the RP5 axon to muscle 12 , other cues must also be involved . Muscles 12 and/or 13 also express Wnt-4 ( a repulsive ligand ) and the LRR protein Capricious ( Caps; probably an adhesion molecule ) , and low-penetrance RP5 targeting defects are observed in Wnt-4 ( Inaki et al . , 2007 ) and caps mutants ( Kurusu et al . , 2008 ) . Perhaps muscle 12 is distinguished from other nearby muscles by a set of partially redundant cues , so that strong targeting phenotypes are not observed in any single mutant . Although Beat and Side paralogs may not be central to motor axon guidance , their expression patterns suggest that they could be important for determining synaptic connections within the CNS . side-VIII , encoding an orphan Side , is expressed in a small subset of embryonic CNS neurons ( Figure 6 ) . In the optic lobe of the pupal brain , an RNAseq analysis of two photoreceptors ( R7 and R8 ) and five types of lamina neurons ( L1-L5 ) revealed that beats and sides have highly specific expression patterns ( Tan et al . , 2015 ) . For example , beat-VII is specific to L2 , beat-VI to L5 , beat-IIa to L3 ( with lower levels in L4 ) , and beat-IIIc to R8 , being expressed at much higher levels in those cells relative to all other cells . side and side-III are specific to L3 , side-II is specific to L1 , side-IV is specific to L2 , and side-V is specific to L5 . R7 , R8 and each of the L neuron types synapse with different sets of neurons in the medulla , a ten-layered structure that processes visual information from the retina and lamina . It has been observed that R and L neurons expressing specific Dprs often form synapses on medulla neurons expressing DIPs to which those Dprs bind in vitro ( Carrillo et al . , 2015; Tan et al . , 2015 ) . In a similar manner , perhaps some of the medulla neurons that are postsynaptic to L or R neurons expressing specific Sides or Beats express their in vitro binding partners , and these Beat-Side interactions might be important for synapse formation or maintenance .
Orthologs for the beat and side genes in the 12 sequenced Drosophila species were established using a reciprocal BLAST approach , first against the annotated predicted transcript databases ( Clark et al . , 2007 ) . Where full length orthologous coding sequence had not been predicted in the public databases , coding sequences of the N terminal ectodomains were inferred and annotated by aligning the full length orthologs from the closest related species against the genome assembly , and other available predicted transcripts in the host . Protein domains were inferred using the online implementations of SMART ( Schultz et al . , 1998 ) , HMMER ( Eddy , 2011 ) and DOUT-finder ( Novatchkova et al . , 2006 ) . Multiple sequence alignments were carried out using the Muscle , t-coffee ( Notredame et al . , 2000 ) and clustal-Ω ( Sievers et al . , 2011 ) algorithms . Alignments were manually edited in SeaView ( Gouy et al . , 2010 ) and UGENE ( Okonechnikov et al . , 2012 ) ; poorly aligning sequences were removed . Maximum likelihood protein phylogenies and bootstrap analyses were performed using RaxML source code ( Stamatakis , 2006 ) and RaxML via the CIPRES Science Gateway and visualised and edited in SeaView . We could not identify orthologs in all 12 Drosophilids for all Sides , likely due to incomplete genomic sequence rather than to stochastic loss of some non-functionalized paralogs after gene duplication as predicted by Ohno’s theory ( Ohno , 1970 ) . The missing orthologs within Side clusters are likely due to the limitations of the methods used to identify them because: ( a ) Side paralog clusters containing low numbers of orthologs present similar inter-species divergence levels as those containing high numbers of orthologs , hence equal selective constraints; ( b ) Evolutionary instability of functionally redundant gene copies , which would lead to the non-functionalization and erosion of redundant paralogs , is not a plausible evolutionary explanation for missing orthologs since the large inter-Side divergence levels imply that paralogs diverged functionally after gene duplication , and thus were not functionally redundant , and ( c ) the loss of redundant paralogs is expected soon after duplication ( Lynch and Conery , 2000 ) , likely pre-dating speciation . Bait expression vectors were modified from the pECIA14 vector ( Özkan et al . , 2013 ) . An Avitag ( Avidity ) was added between the hexahistidine and FLAG tags at the C-terminus of the vector with standard cloning procedures to make a new Gateway ( Thermo Fisher , Waltham , MA ) destination vector . ECD sequences were moved from entry vectors for Beats and Sides , described in ( Özkan et al . , 2013 ) , into the modified pECIA14 vector using LR Clonase II ( Thermo Fisher ) . Prey proteins were expressed from the pECIA2 vector ( Özkan et al . , 2013 ) . All proteins , excepting the unpurified prey , were expressed in Drosophila Schneider 2 cells grown in S2 media with 10% fetal bovine serum , 50 units/mL penicillin and 50 μg/mL streptomycin . The unpurified prey proteins were expressed in Sf-900 III media ( Thermo Fisher ) . Proteins were transfected using Effectene ( Qiagen , Hilden , Germany ) , following manufacturer’s instructions . Copper ( 0 . 5 mM CuSO4 ) was added the day after transfection to induce expression of protein . For the baits , 3 mM biotin was also added to the media to facilitate in vivo biotinylation . Prey proteins were purified using Ni-NTA resin , following standard procedures . We first explored direct coupling by conjugating purified AP bait proteins to bead regions . Anti-AP antibody was added to the coupled beads , followed by PE-conjugated secondary antibody , and the bead mixture was run on the Bio-Plex machine to evaluate coupling efficiency . We found that this direct coupling method was not optimal , as different bait proteins coupled to the beads with very different efficiencies ( data not shown ) , and a great deal of protein was lost during the purification steps . For the affinity capture method , Bio-Plex Pro Magnetic COOH Beads ( Bio-Rad , Hercules , CA ) were coupled to streptavidin following the manufacturer’s instructions , and beads were blocked with 1% i-Block ( Tropix , Bedford , MA ) in PBS . Bait protein constructs were transfected into 10 cm dishes of S2 cells , and the medium from each dish was concentrated in Amicon centrifugal filters to 1 mL . The baits were then captured directly from concentrated media with 4–8 μl of beads ( corresponding to 50 , 000–100 , 000 beads per region ) . Purified ( or unpurified ) prey was added to the bead mix and incubated overnight at 4° C . For most purified prey proteins , we added 1 μg protein in a 100 μl reaction . For some preys , because of differential stickiness and low expression levels , different amounts were added . These ranged from 0 . 05 ug to 4 μg . For unpurified protein , 50 μl of protein in media was added to an overall reaction volume of 100 μl . The next day , beads were washed with PBST containing 0 . 02% i-Block and incubated with anti-V5 antibody ( Invitrogen ) at 2 μg/mL . The beads were then washed again and incubated with PE-conjugated goat anti-mouse IgG ( Santa Cruz Biotechnology . Dallas , TX ) . The beads were washed again , transferred into a 96-well plate and run on the Bio-Plex 200 . We tried to at use at least 1000 beads per region , but because of differential bead loss during the various incubation and wash steps , different numbers of beads were counted for each region . For our analysis , we always counted at least 35 beads per bead region . Each reaction was run in duplicate . To test whether jumping occurs , we coupled four different bead regions to streptavidin . Three of the streptavidin-coupled bead regions were used to capture three different pentamerized , biotinylated proteins . The beads were then mixed together and incubated overnight with anti-AP antibody , followed by PE-conjugated secondary antibody , and run on the Bio-Plex . Strong PE signal was detected for the bead regions with captured bait proteins , while streptavidin beads with no bait protein had no detectable signal over background ( data not shown ) . These results show that there is no jumping of proteins between different bead regions . Before generating Z scores , Özkan et al . ( Özkan et al . , 2013 ) used a cutoff to eliminate outliers ( high values , probably due to binding ) , as these would artificially inflate the mean and standard deviation ( SD ) . Since our data set is relatively small ( a 23 × 23 matrix ) , and a large fraction of the proteins interact with each other ( since we are using preselected proteins that are already known to be part of a network ) we could not exclude signals due to genuine binding as outliers , as that would eliminate much of the data . By contrast , in the 202 × 202 matrix of the global interactome , the probability that any randomly chosen pair of proteins actually bind to each other is very low . To process our data , then , we utilized bootstrapping of the median for each row and each column . We construct an N x N matrix X with the rows and columns containing the N proteins in the same order . The rows denote the prey and the columns denote the bait . Thus , the ith prey interaction with jth bait is quantified by X ( i , j ) , and the jth prey interaction with the ith bait is quantified by X ( j , i ) . We then selected with replacement N random samples from the ith column of the matrix X . The process was repeated B times ( B = 300 was used ) to obtain N B-dimensional vectors . Similarly , we selected with replacement N random samples from the ith row of X , and the process was repeated to obtain N B-dimensional vectors . The mean and standard deviation of each of the N rows and N columns were calculated and each component in X was Z-scored with respect to the column and row statistics to obtain two N x N matrices Xzc and Xzr , respectively . A matrix Xzrc was formed via the element-by-element computation Xzrc ( i , j ) = ( Xzr ( i , j ) +Xzc ( i , j ) ) /2 . In the scenario of both Xzrc ( i , j ) and Xzrc ( j , i ) being positive , the geometric mean of Xzrc ( i , j ) and Xzrc ( j , i ) were computed . If the geometric mean exceeded the threshold of five , then the i and j pair were labeled as an interaction . We were able to recapitulate all interactions found in the ECIA except for BeatIa::Side , which was clearly observed in one orientation ( Figure 2B ) but not detected in the other due to failure of expression of Beat-Ia AP bait . This interaction was not scored because the geometric mean cannot be calculated if one of the Z scores is zero , which was the case for Beat-Ia-AP bait and Side prey . The interaction between Side-II and Side-III was very strong in one orientation but just below our cutoff in the opposite orientation , so we have preserved this interaction , seen in the original ECIA . All Beat and Side extracellular domains with C-terminal hexahistidine tags were expressed in and secreted from Trichoplusia ni High Five Cells using the baculovirus system . Proteins were first purified with Ni-NTA agarose resin , followed by size exclusion chromatography using Superdex 75 or 200 10/300 columns ( GE Healthcare ) . For capturing on Surface Plasmon Resonance chips , Side-VI ( CG34114 ) , Beat-Ic and Beat-VI expression constructs also included a biotin acceptor peptide sequence , which was biotinylated using E . coli BirA biotin ligase , and this allowed proteins to be captured on SA ( streptavidin ) Biacore chips ( GE Healthcare ) . Beat-Va , Vb , Vc , Side-II , and Side-III were titrated in the mobile phase over the SA chips . Side-VI , Beat-Ic , Beat-VI and Beat-Va , Vb , and Vc expression constructs included complete ectodomains . Due to problems with expression and/or purification for full-ectodomain constructs of Side , Side-II , and Side-III , shorter fragments of these Side ectodomains were used for SPR , based on the knowledge that the first IgSF domains of Sides are sufficient for Beat-Side interactions ( unpublished data ) . The following constructs were used during SPR experiments: N-terminal two IgSF domains of Side , N-terminal single-IgSF domains of Side-II and Side-III . Surface Plasmon Resonance ( SPR ) experiments for Side-VI against Beat-Va , -Vb and -Vc were performed on a Biacore T100 ( GE Healthcare ) , and for Beat-Ic and Beat-VI against Side-II and Side-III were performed on a Biacore 3000 . Unless noted , all SPR binding measurements are done in HBSp+ ( GE Healthcare ) , which includes 10 mM HEPES pH 7 . 2 , 150 mM NaCl , and 0 . 05% surfactant Polysorbate 20 . To prevent non-specific binding to Biacore chip surfaces , Beat-Va and Vb binding experiments were performed with HBSp +containing 500 mM NaCl and 15% Glycerol . For similar reasons , Side-II and Side-III binding was performed in the buffer HBSp +and 1% ( w/v ) bovine serum albumin ( BSA ) . Binding between Sidestep ( mobile phase ) and Beat-Ic ( stationary phase ) could also be observed , but precipitation of Sidestep prevented us from collecting a titration series . AP-fusion constructs were generated by Gateway recombination into a destination vector , pUAS-LPGWAP , containing a metallothionein promoter N- terminal leader peptide and C-terminal AP . Secreted AP-ectodomains were produced in Drosophila S2 cells by co-transfecting the pUAS-prey-AP and pAct-Gal4 plasmids using FuGENE HD transfection reagent ( Promega , Madison , WI ) . Cell surface binding assays were adapted from those previously described ( Cheng and Flanagan , 1994 ) . Briefly , 106 Drosophila S2 cells were seeded in 6-well plates , transfected with cell surface bait or control constructs , expression was induced and cells were harvested by centrifugation . Cells were washed and incubated with 0 . 5 nM Prey-AP or LP-AP ( control ) conditioned S2 media for 90 min at room temperature , washed , and bound AP activity was measured . UAS-Beat-Vb x Tub-GAL4 or wild-type embryos ( Figure 5 ) were collected , dissected , and stained following procedures described in Lee et al ( Lee et al . , 2009 ) . Similar methods were used for wild-type embryos in Figure 5—figure supplement 2 . Dissected embryos were stained with Side-VI-AP ( in S2 media ) , followed by primary antibodies rabbit anti-AP ( Serotec ) and mAb 1D4 . Secondary antibodies used were Alexa-Fluor 568 anti-rabbit and Alex-Fluor 488 anti-mouse ( Invitrogen ) at a 1:1000 dilution . Images were collected on a Zeiss LSM 710 using a 40X objective . In situ mRNA hybridization was performed as previously described ( Zarin et al . , 2012 ) . Probes were generated from cDNA vectors ( Drosophila Genomics Resource Centre; beat-Ia cDNA kindly provided by H . Aberle ) for the genes of interest and specific motor neurons were labeled in the following stocks: RN2-Gal4::UAS-tau-myc-GFP , RN2-LacZ ( Fujioka et al . , 2003 ) , Lim3A-tau-myc ( Thor et al . , 1999 ) . Motor axon staining to define phenotypes was done as described in Patel ( 1994 ) . Third instar larvae of Side-VI and Beat-Va T2A-GAL4 lines driving GFP were dissected following procedures described in ( Menon et al . , 2009 ) . Dissected larvae were stained with rabbit anti-GFP ( Invitrogen ) at 1:500 , followed by rhodamine-conjugated anti-HRP ( Jackson ImmunoResearch ) at 1:50 and Alexa-Fluor 488 anti-rabbit ( Invitrogen ) at 1:1000 . Samples were imaged with a Zeiss LSM 710 with a 40X objective . Images were processed with ImageJ and Adobe Photoshop . | Within every organ of the body , cells must be able to recognise and communicate with one another in order to work together to perform a particular role . Each cell has a specific protein on its surface that acts like a molecular identity card , and which can form weak bonds with a complementary protein on another cell . There are thousands of different cell surface proteins , and the interactions between them – known collectively as the interactome – dictate the how cells interact with one another . Many cell surface proteins are similar across species . Humans and fruit flies , for example , both possess a family of cell surface proteins that contain a region called the Immunoglobulin Superfamily domain . This family can be further divided into subfamilies , two of which are known as “Beats” and “Sides” for short . As the nervous system develops , nerve cells carrying a particular Beat protein interact with nerve or muscle cells carrying a corresponding Side protein . Yet while experiments have matched up many Beats and Sides , the partners of others remain unknown . Li et al . have now developed a new technique called the Bio-Plex Interactome Assay to rapidly screen for interactions between multiple cell surface proteins in a single sample . Applying the technique to cells from fruit flies revealed new binding partners within the Beats and the Sides . After verifying several of these interactions , Li et al . explored the role of various Beats and Sides in the developing nervous system of fruit fly embryos by mapping the cells that display them on their surfaces . This increased knowledge of the Beat-Side binding network should provide further insights into how connections form between nerve cells . The new screening technique could also eventually be used to map the cell surface protein interactome in humans . A number of key drugs , including the breast cancer drug Herceptin , target cell surface proteins . Identifying interactions among cell surface proteins could thus provide additional leads for developing new therapies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Deconstruction of the beaten Path-Sidestep interaction network provides insights into neuromuscular system development |
Many species perform rapid limb movements as part of their elaborate courtship displays . However , because muscle performance is constrained by trade-offs between contraction speed and force , it is unclear how animals evolve the ability to produce both unusually fast appendage movement and limb force needed for locomotion . To address this issue , we compare the twitch speeds of forelimb muscles in a group of volant passerine birds , which produce different courtship displays . Our results show that the two taxa that perform exceptionally fast wing displays have evolved 'superfast' contractile kinetics in their main humeral retractor muscle . By contrast , the two muscles that generate the majority of aerodynamic force for flight show unmodified contractile kinetics . Altogether , these results suggest that muscle-specific adaptations in contractile speed allow certain birds to circumvent the intrinsic trade-off between muscular speed and force , and thereby use their forelimbs for both rapid gestural displays and powered locomotion .
Signal evolution in animals has produced a variety of unique and unusual displays , including the elaborate physical routines performed for social communication and advertisement ( Beehler and Pruett-Jones , 1983; Prum , 1990; Masonjones and Lewis , 1996; Voigt et al . , 2001; Hogg and Forbes , 1997 ) . Little is understood about how these displays evolve , and their emergence represents a functional feat in animal design ( Fuxjager and Schlinger , 2015; Irschick et al . , 2007 ) . Most physical displays , for example , involve rapid limb movements , which are controlled by the same neuro-motor architecture that governs the otherwise 'normal' movements used to power locomotion . However , because motor performance is constrained by a trade-off between muscle contraction force and speed ( Rome et al . , 1999; Young and Rome , 2001 ) , limb muscles should in theory be unable to generate both the swift appendage movements incorporated in showy physical displays and the force needed to drive locomotion ( Young and Rome , 2001; Rome and Lindstedt , 1998 ) . How , then , do limb motor systems and their underlying musculature control extraordinarily fast limb movements necessary for the production of adaptive behavioral displays ? We study this issue in a group of volant passerine birds , which produce different types of elaborate courtship behavior ( Figure 1A ) . Trade-offs between muscle force and contraction speed clearly exist in the wing muscles of birds within this size range ( Biewener , 2011; Biewener and Roberts , 2000; Biewener et al . , 1992; Dial and Biewener , 1993; Hedrick et al . , 2003 ) , and evidence suggests that such muscles maintain sufficient force-generating abilities to power flight and regulate flying speeds ( Biewener et al . , 1992; Dial and Biewener , 1993; Hedrick et al . , 2003 ) . Of the species we study , however , both golden-collared ( Manacus vitellinus ) and red-capped manakins ( Ceratopipra mentalis ) produce exceptionally rapid wing movements as part of their acrobatic courtship displays ( Bostwick and Prum , 2003; Fusani et al . , 2007; Fuxjager et al . , 2013 ) . For example , male golden-collared manakins perform roll-snaps , whereby they hit their wings together above their backs at ≈60 Hz to produce a loud mechanical sonation ( Fusani et al . , 2007; Fuxjager et al . , 2013 ) . Kinematic analyses suggest that this behavior is performed by elevating the opened ( extended ) wings and then by repeatedly retracting the wings to force the wrists to collide in quick succession ( Bostwick and Prum , 2003; Fusani et al . , 2007; Fuxjager et al . , 2013 ) . Likewise , male red-capped manakins produce a similar wing sonation called a clap , in which the wings are laterally extended slightly above the body and then immediately retracted back to the sides ( Bostwick and Prum , 2003 ) . Claps are also produced in rapid succession , with the wing-extension and wing-retraction phases occurring at ≈45 Hz ( Bostwick and Prum , 2003 ) . In both species , the wing oscillation frequencies used for courtship displays exceed those that similarly sized birds use for flight ( ≈25–30 Hz ) ( Donovan et al . , 2013; Pennycuick , 2001 ) ; thus , the forelimb motor system is likely modified to actuate courtship movements , while preserving the force-generating ability needed to drive powered locomotion . However , the nature of these modifications has never been explored . 10 . 7554/eLife . 13544 . 003Figure 1 . Species and muscles examined in this study . ( A ) Species included in our study , with common names in boldface typesetting and scientific names in italic typesetting . A brief description of each species’ display and reason for inclusion in the study is described . Photos with permission from Nick Athanas . ( B ) Illustration of the three main wing muscles in a golden-collared manakin that are involved in the production of the roll-snap . These include ( i ) the supracoracoideus ( SC ) , which raises the wing by elevating the humerus; ( ii ) the pectoralis ( PEC ) , which lowers the wing by depressing the humerus , and ( iii ) the scapulohumeralis caudalis ( SH ) , which retracts the wing via the humerus ( Biewener , 2011; Dial , 1992; Dial et al . , 1991 ) . Note that the SC is a darker shade of pink , compared to the PEC and SH , because the SC lies deep to the PEC . Scientific illustrations of these muscles can be found elsewhere ( Welch and Altshuler , 2009; George and Berger , 1966 ) . This schematic is modified with permission from Schlinger , et al . ( Schlinger et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 003 We hypothesize that , in both golden-collared and red-capped manakins , the main muscles involved in humeral retraction have evolved rapid contractile kinetics to support display maneuvering . This idea is rooted in the kinematic studies that suggest that retraction of the wings is a major feature of the roll-snap and clap . Building on this notion , we expect that the other primary wing muscles that supply the main force for forelimb movement during flight are unmodified . To test our hypothesis , we therefore measure twitch contraction frequencies in situ from the three main wing muscles: ( i ) the supracoracoideus ( SC ) , which raises the wing by elevating the humerus; ( ii ) the pectoralis ( PEC ) , which lowers the wing by depressing the humerus , and ( iii ) the scapulohumeralis caudalis ( SH ) , which retracts the wing via the humerus [Figure 1B; Biewener , 2011 , Dial , 1992 , Dial et al . , 1991] . Because the SH acts as the main wing retractor and generates the least aerodynamic force necessary for powered flight , we predict that its contractile kinetics are significantly increased in golden-collared and red-capped manakins , compared to the SC and PEC . In this vein , we predict that these latter two muscles show similar contractile kinetics , as they appear to contribute less to the rapid wing movements that are incorporated into these two birds’ displays . Inasmuch , these muscles are likely more constrained by natural selection to generate aerodynamic force for powered flight ( Biewener et al . , 1992; Dial and Biewener , 1993; Hedrick et al . , 2003 ) . To further examine whether unusually rapid muscle kinetics co-evolve with fast-moving wing displays , we compare muscle twitch frequencies from wild-caught golden-collared and red-capped manakins to those of three other related ( wild-caught ) species ( Figure 1A ) . The first of these is the blue-crowned manakin ( Lepidothrix coronata ) , which is a close relative of the aforementioned manakins and produces an elaborate physical display without rapid wing-snaps or wing-flicks ( Durães , 2009 ) . The second species is the dusky antbird ( Cercomarca tyrannica ) , which inhabits the same niche as the manakins described above but produces elaborate vocalizations in lieu of a physical display ( Morton and Derrickson , 1996; Morton , 2000 ) . The last bird is the house wren ( Troglodytes aedon ) , a more distant relative within the same avian order that sings to attract mates and does not produce a robust physical display ( Johnson and Kermott , 1991 ) . These comparisons collectively provide a framework for evaluating how the muscular properties that control twitch frequencies vary across species , and whether rapid wing maneuvering used for courtship is related to this variation . Therefore , based on our hypothesis described above , we predict that rapid kinetics of a humeral retractor muscle will be absent from blue-crowned manakins , dusky antbirds , and house wrens , as these species do not produce rapid wing maneuvers as part of their courtship displays .
We measured mean levels of muscle relaxation in response to electrical stimulation at different frequencies ( Figures 2A , B ) . Therefore , our first aim was to verify that the repeated stimulations administered to a given muscle did not exhaust or damage the tissue and thereby confound our results . To do this , we compared muscle recovery in response to the 20 Hz stimulation given at the beginning of our experimental series to muscle recovery in response to a second 20 Hz stimulation given at the end of our experimental series ( Figure 2A ) . Overall , we found no significant difference in percent relaxation values between these two separate stimulation events ( Table 1; all p values ≥0 . 18 ) . These results therefore indicate that the muscles were functionally intact throughout the twitch frequency recording sessions . 10 . 7554/eLife . 13544 . 004Figure 2 . Experimental design . ( A ) Schematic of the work flow and procedural design . Muscles were prepared in situ ( see methods ) and stimulated at frequencies ranging from 10 Hz to 100 Hz , increasing at increments of 10 Hz . Stimulation trains were spaced 1 min apart . After the 100 Hz stimulation train , we administered a second 20 Hz stimulation train ( shown underlined and in boldface typesetting ) . Percent recoveries were compared between this 20 Hz stimulation train and the first 20 Hz stimulation train to validate that the procedure did not exhaust/damage muscle . ( B ) Representative twitch recordings from a red-capped manakin ( 10 Hz from the SC and 100 Hz from the SH; note the differences in time scale ) . For each individual , percent recovery at a given stimulation frequency was calculated by averaging the percent recoveries of the first eight stimulations within the administered train . This corresponds to reasonable numbers of wing oscillations that golden-collared and red-capped manakins incorporate into their respective roll-snap or clap displays . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 00410 . 7554/eLife . 13544 . 005Table 1 . Mean ( ± 1 SEM ) percent muscle relaxation in response to a 20 Hz stimulation at the beginning ( first ) and end ( second ) of the stimulation series . GCM = golden-collared manakin; RCM = red-capped manakin; BCM = blue-crowned manakin; DAB = dusky antbird; HW = house wren . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 005SpeciesMuscleFirst 20 Hz stimulationSecond 20 Hz stimulationt Statisticp Value*GCMPEC99 . 06 ( 0 . 94 ) 99 . 95 ( 0 . 05 ) -1 . 00 . 42SC99 . 44 ( 0 . 49 ) 99 . 61 ( 0 . 35 ) -2 . 030 . 18SH100 ( 0 . 0 ) 95 . 27 ( 4 . 73 ) -1 . 090 . 39RCM‡PEC100 ( 0 . 0 ) 100 ( 0 . 0 ) -NSSC100 ( 0 . 0 ) 100 ( 0 . 0 ) -NSSH100 ( 0 . 0 ) 100 ( 0 . 0 ) -NSBCMPEC85 . 01 ( 14 . 98 ) 100 ( 0 . 0 ) -1 . 00 . 42SC100 ( 0 . 0 ) 99 . 90 ( 0 . 097 ) 1 . 00 . 42SH98 . 55 ( 1 . 45 ) 99 . 86 ( 0 . 14 ) 1 . 00 . 50DABPEC95 . 93 ( 1 . 71 ) 98 . 14 ( 0 . 92 ) -1 . 560 . 26SC97 . 97 ( 1 . 30 ) 99 . 63 ( 0 . 37 ) -1 . 290 . 29SH100 ( 0 . 0 ) 99 . 75 ( 0 . 25 ) 1 . 00 . 39HWPEC95 . 09 ( 2 . 63 ) 95 . 43 ( 2 . 55 ) -0 . 0830 . 94SC98 . 86 ( 1 . 14 ) 92 . 54 ( 7 . 46 ) 1 . 00 . 42SH86 . 38 ( 6 . 76 ) 93 . 99 ( 3 . 02 ) -1 . 990 . 19*p values are derived from paired sample t tests . ‡In the RCM , muscle relaxation in both 20 Hz frequency groups was 100% . Therefore , t statistics cannot be computed because the standard error difference is 0 and the groups are assumed to be indistinguishable ( NS = not significant ) . A benefit of collecting muscle twitch recordings in situ is that we can evaluate whether the muscle produces movement in response to stimulation under normal load conditions ( i . e . , whether stimulated contractions can move the wing and thus overcome its weight and inertia ) . Indeed , we found that this was the case , as the muscle stimulations that we administered actuated wing movement . This included instances when we administered high stimulation frequencies that did not induce complete muscle fusion , such as when we gave 90 Hz and 100 Hz stimulation to the golden-collared manakin’s SH . These data indicate that high contraction-relaxation cycling rates translate to corresponding high frequency wing oscillations . Next , to evaluate species differences in in situ contractile dynamics of the PEC , SC , and SH , we compared non-linear regression models that characterize muscle contraction speeds in response to varying stimulation frequencies ( Figures 3A , B and C ) . For each muscle , models differed significantly among taxa ( Figure 3A , B and C; PEC: F16 , 130 = 2 . 50 , p =0 . 0023; SC: F16 , 140 = 5 . 50 , p<0 . 001 . ; SH: SC: F16 , 134 = 15 . 68 , p<0 . 001 ) . This suggests that there is performance variability in these muscles , and we suspect that such results underlie functional differences in the way in which the forelimb musculature controls rates of wing flapping . 10 . 7554/eLife . 13544 . 006Figure 3 . Muscle twitch speed dynamics in the ( A , D ) pectoralis , ( B , E ) supracoracoideus , and ( C , F ) scapulohumeralis of the five avian species included in our study . ( A–C ) Non-linear models generated to depict the relationship between mean muscle relaxations in response to different muscle stimulation frequencies . In each graph , muscle relaxation at a given stimulation frequency represents the mean ( ± 1 SEM ) among individuals of a given species . ( D–F ) Half-relaxation frequencies of the different forelimb muscles across the five species in our analysis . Data represent mean ( ± 1 SEM ) half-relaxation frequency for the given species . ( F ) For the SH , differences in the letters atop each bar denote statistically significant differences in mean half-relaxation values , according to post-hoc analyses ( SNK tests , p<0 . 05 ) . In all graphs , GCM = golden-collared manakin ( orange ) ; RCM = red-capped manakin ( red ) ; BCM = blue-crowned manakin ( blue ) ; DAB = dusky antbird ( black ) ; and HW = house wren ( brown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 00610 . 7554/eLife . 13544 . 007Figure 3—source data 1 . Mean percent recoveries of the main wing muscles at different stimulation frequencies ( see Materials and methods ) across all species . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 007 Building on the analyses above , we investigated species differences in muscle contraction speeds that might explain the ability of certain taxa to rapidly move their limbs during display performances . For each muscle , we therefore compared the half-relaxation frequency among species ( Figures 3D , E and F ) . This measure is the regression models’ estimate of intrinsic contractile behavior , and it is defined as the maximum stimulation frequency at which the muscle’s percent relaxation is half of its predicted functional range ( see Methods for further details ) . We used this measure because it provides a consistent and unbiased metric by which we can compare muscle contraction speeds across birds . For both the PEC and the SC , half-relaxation frequencies averaged ≈50 Hz and were indistinguishable among species ( Figure 3D , PEC: F4 , 10 = 0 . 85 , p = 0 . 53; Figure 3E , SC: F4 , 11 = 1 . 74 , p = 0 . 39 ) . By contrast , in the SH , the half-relaxation frequencies differed significantly among taxa ( Figure 3F , F4 , 11 = 10 . 09 , p = 0 . 0011 ) , and post-hoc analysis of this effect showed these frequency measures were significantly higher in both golden-collared and red-capped manakins compared to the other three species in the study ( Figure 3G and Table 2; golden-collared: SNK post-hoc tests , q≥6 . 17 , p<0 . 01; red-capped: SNK post-hoc tests , q≥3 . 64 , p<0 . 05 ) . Estimates of these two birds' SH half-relaxation frequencies were exceptionally high , coming in at ≈100 Hz and ≈80 Hz , respectively . Moreover , the SH half-relaxation frequencies between golden-collared and red-capped manakins were statistically indistinguishable ( Figure 3F and Table 2; SNK post-hoc tests , q = 2 . 52 , p>0 . 05 ) . 10 . 7554/eLife . 13544 . 008Table 2 . Summary of SNK post-hoc results comparing SH half-relaxation frequencies between species , with statistically significant differences are shown in boldface typesetting . GCM = golden-collared manakin; RCM = red-capped manakins; BCM = blue-crowned manakin; DAB = dusky antbird; HW = house wren . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 008Pairwise comparisonq statisticDFp valueGCM vs . RCM2 . 5211>0 . 05GCM vs . BCM6 . 1711<0 . 01GCM vs . DAB6 . 6511<0 . 01GCM vs . HW7 . 5811<0 . 01RCM vs . BCM3 . 6411<0 . 05RCM vs . DAB3 . 9611<0 . 05RCM vs . HW5 . 0611<0 . 05BCM vs . DAB0 . 06211>0 . 05BCM vs . HW1 . 4211>0 . 05DAB vs . HW1 . 4511>0 . 05 Finally , we used our models to evaluate whether twitch speeds obtained from the golden-collared and red-capped manakin SH are in theory sufficient to drive wing oscillations necessary for each species’ display . In the case of the golden-collared manakin , we found that the SH achieves on average ≈85% relaxation in response to stimulation frequencies of around 95 Hz . In the red-capped manakin , the SH achieves the same amount of mean relaxation in response to stimulation frequencies of roughly 70 Hz . The observation of concurrent wing movements at these stimulation frequencies ( see above ) shows that these changes in muscle length can actuate humeral retraction; thus , our results suggest that the SH is more than capable of driving the natural wing oscillations that make up each species’ wing display ( Bostwick and Prum , 2003; Fusani et al . , 2007; Fuxjager et al . , 2013 ) . Such rapid kinetics of the SH suggest that this muscle features specializations that are similar to those of so-called 'superfast' muscles ( Rome et al . , 1996; Elemans et al . , 2008; Elemans et al . , 2004 ) .
We hypothesize that the SH is one of the main actuators of the rapid forearm movements that are incorporated into golden-collared and red-capped manakin displays . Kinematic studies suggest that these species generate their wing sonations by moving their forelimbs at frequencies of 55–63 Hz and 45 Hz , respectively ( Bostwick and Prum , 2003; Fusani et al . , 2007; Fuxjager et al . , 2013 ) . Our results show that these rates fall well within the range of stimulation frequencies at which significant recovery of the SH is observed ( at least 85% ) , and we find that individual stimulations are accompanied by observable wing movements . In addition , these same kinematic studies suggest that the bulk of the rapid forearm movement is driven by repetitive humeral retraction , and such movement of the wing is fully consistent with the biomechanical role of the SH ( Dial , 1992; Dial et al . , 1991 ) . Thus , these data collectively suggest that the SH is the main muscle driving display production in both animals . Since the elaborate displays of manakins are used for courtship and male-male rivalry ( Bostwick and Prum , 2003; McDonald et al . , 2001; Barske et al . , 2011 ) , sexual selection likely favored the emergence of rapid contractility in the SH . To this end , these superfast kinetics make the SH in these two birds the fastest vertebrate limb muscles currently known , at least with respect to measures of twitch contraction . Moreover , we know that these frequencies markedly exceed the typical wing-beat frequencies during flight ( 25–30 Hz ) for birds of similar size ( Donovan et al . , 2013; Pennycuick , 2001 ) , which again implies that sexual selection has positvely favored the emergence of an extreme muscle phenotype to accommodate the evolution of an equally extreme behavior . With this in mind , it is interesting to speculate about why sexual selection has not forced faster display maneuvering , given that the SH appears capable of more rapid contractile kinetics than roll-snapping or clapping demands . Studies in golden-collared manakins suggest that females preferentially mate with males that perform certain display maneuvers at fractions-of-a-second faster speeds ( Barske et al . , 2011; 2015 ) . Therefore , one intriguing possibility is that other wing muscles , like the 'slower' SC and PEC , also contribute to display production by helping position the wings ( Fusani et al . , 2014; Schlinger et al . , 2013 ) . In doing so , these muscles may limit the overall speed with which a given maneuver can be performed , owing to the effects of natural selection that presumably preserve these tissues’ force-generating properties for powered flight ( see below ) . If this is the case , then display speed itself may be constrained by neural circuits that integrate and/or coordinate the activation of numerous forearm motor units , as opposed to only the speed of the SH . Our results uncover a possible motor design that enables birds to produce rapid wing movements for a display , while preserving their ability to generate aerodynamic force needed to sustain powered flight . This , in effect , is rooted in our finding that muscle twitch speeds are modified in the SH , but not the SC or PEC . Rather , the latter two muscles maintain half-relaxation rates of ≈50 Hz , which are similar to the half-relaxation frequencies of all three muscles in the non-wing-displaying blue-crowned manakin , dusky antbird , and house wren . Thus , in golden-collared and red-capped manakins , the SC and PEC twitch speeds are largely conserved , which means that the ability of these muscles to generate aerodynamic force needed for powered flight is not compromised by the mutually exclusive ability to contract rapidly ( Rome et al . , 1999; Young and Rome , 2001 ) . Flight in these two manakins is therefore unencumbered ( Moore et al . , 2008 ) . To this end , this model is consistent with the biomechanical role of the SH , which contributes to aerodynamic force by adjusting the wings’ angle of attack , area , and/or camber ( Dial , 1992; Dial et al . , 1991 ) . The SH therefore likely contributes least to the generation of power for flight , and thus presents more opportunity for evolutionary modification without severely compromising flight ability . To our knowledge , this is the first glimpse into how evolution might circumvent trade-offs between muscle contraction speed and force generation to favor the emergence of rapid appendage displays , while maintaining normal locomotion . Additional work is clearly needed to examine these ideas further , especially since the physiological and biomechanical mechanisms that underlie animal flight vary significantly among taxa . In vivo studies of muscle performance during display production would be especially helpful , as well as a better understanding of the force-velocity relationships of these birds’ wing muscles . The muscular design described above may impose a 'cost' on flight that has not yet been identified . Past work shows that both golden-collared and red-capped manakins are strong fliers , relative to other similarly sized tropical birds ( Moore et al . , 2008 ) . This result indicates that dramatically enhancing the contraction speed of the SH does not altogether impede or diminish locomotion . However , studies have noted that both of these species produce a muffled fluttering sound when they fly ( Bostwick and Prum , 2003 ) . It is unclear if this sound , like the other sonations these birds produce , is a sexual signal , or whether it is a by-product of physiological and/or biomechanical adaptations that enable elaborate courtship maneuvering . If the latter is true , then our results might provide a mechanism for such action , as the ability of the SH to appropriately control wing positioning ( Dial , 1992 ) might be altered by this muscle’s performance shift . With this in mind , it will be interesting to pursue putative 'costs' of sacrificing force for speed in the SH , but not the SC and PEC . Our results provide the first insight into how the complex motor system for flight is designed to support wing-based acrobatic and elaborate behavioral displays . Specifically , in two species that use rapid wing movements as part of an adaptive behavioral display , we show that one of the muscles actuating wing movement has evolved 'superfast' contraction kinetics , while the kinetics of other muscles that provide more power for flight are functionally conserved . Thus , our study provides evidence not only for the emergence of the fastest known vertebrate limb muscle , but also a unique evolutionary design of the forelimb muscular system that enables both rapid movement for displaying and force-generating movement for locomotion .
Adult male manakins were captured via passive mist netting at their respective breeding leks ( golden-collared manakins , n = 3 , red-capped manakins , n = 3 , blue crowned manakins , n = 3 ) . At the same time , adult male dusky antbirds ( n = 4 ) and house wrens ( n = 3 ) were captured from their breeding territories by luring individuals into the net with conspecific playback . Birds were immediately transported to our nearby laboratory for muscle recordings ( see below ) , after which they were euthanized with an overdose of isoflurane and their tissues were removed for another study . We inspected the testes of each specimen to verify that they were enlarged and that the individual was in reproductive condition . The work was conducted from March to April in Gamboa , Panama at the Smithsonian Tropical Research Institute ( STRI ) . All appropriate institution and governmental authorities approved the methods described herein , including the relevant Institutional Animal Care and Use Committees ( IACUC ) from STRI and Wake Forest University . Using a similar approach described elsewhere ( Elemans et al . , 2004; 2008 ) , we assessed the twitch speed dynamics of the PEC , SC , and SH ( Figure 1B ) in each individual collected from the field . Contraction-relaxation speed within a stimulus period was assessed by measuring the degree to which each muscle relaxed relative to its un-stimulated length ( i . e . , percent recovery ) in response to different frequencies of electrical stimulation ( see Figure 2A ) . All muscles were subjected to the following stimulation frequencies: 10 Hz , 20 Hz , 30 Hz , 40 Hz , 50 Hz , 60 Hz , 70 Hz , 80 Hz , 90 Hz , and 100 Hz . Each stimulation train consisted of 10 separate pulses . Pulse duration was set to 1 ms , and the pulse current was set between 0 . 5–0 . 8 mA . Stimulation trains were always delivered from low frequency ( 10 Hz ) to high frequency ( 100 Hz ) . We space subsequent stimulation trains 1 min apart from each other to allow the muscle to temporarily rest . To confirm that our results were not confounded by damage or exhaustion of the muscle in response to repeated stimulations , we delivered a second 20 Hz stimulation train 1 min after the series’ final 100 Hz stimulation ( Figure 2A ) . We then compared the muscle recoveries in response to these two ( temporally spaced ) 20 Hz stimulations . In this study , every muscle was subjected to a single experimental series ( i . e . , 10 Hz to 100 Hz pulse as described above ) , and thus each muscle was also subjected to this validation procedure . Muscle twitch speeds were recorded in situ ( Figure 4 ) . For all preparations , birds were restrained on a soft foam pad that was securely attached to the surgical bench . Birds were then anesthetized with isoflurane ( 2–4% in O2 ) . To prepare the PEC and SH , we first cut a small ( 1 cm ) incision in the skin . We then exposed the moist surface of the muscle by gently pushing apart the skin . Once the muscle was in clear view , we implanted it with the stripped ends ( 1–2 mm ) of two insulated silver wire electrodes ( diameter: 0 . 14 mm ) . These electrodes were then connected to a nearby stimulator ( Model 2100 , A-M Systems , WA , USA ) . We next fastened a stainless steel hook ( 0 . 1 mm diameter ) to the muscle directly adjacent to the implanted electrodes . This hook was connected through a monofilament line to a force transducer ( Model FT03 , Grass Technology , RI , USA ) , which was tightly clamped to a heavy stand ( ≈6 kg ) . As soon as this preparation was completed , we placed a drop of normal avian saline ( 0 . 9% ) over the exposed muscles to prevent tissues desiccation during the recording session . We adjusted the slack in the line to the force transducer by slightly moving the heavy stand; this allowed us to maintain tension in the line between the muscle and force transducer , and thus optimize the sensitivity of muscle twitch recordings without overloading the force transducer signal . When the recordings were completed , the electrodes and monofilament line from the force transducer were removed from the muscle , and the incision was closed using Vetbond tissue adhesive . All surgical preparations occurred at room temperature , which was similar to the outside ambient temperature ( ≈30°C ) . 10 . 7554/eLife . 13544 . 009Figure 4 . Schematic representation of the experimental set up used to record muscle twitch in situ . Birds were deeply anesthetized with isoflurane; their muscles were exposed , attached to the force transducer , and implanted with silver electrode wires . Data were collected on a nearby laptop computer . Note that the elements in this figure are not drawn to scale . DOI: http://dx . doi . org/10 . 7554/eLife . 13544 . 009 We prepared the SC in a similar manner to the PEC and SH; however , we made a few modifications to this surgery , given that the SC lies deep to the PEC and is more difficult to access ( Figure 1 ) . First , we cut a small ( 1 . 5 cm ) incision in the skin above the furcula . We then gently moved aside the underlying fat-pad and exposed the inter-clavicular air-sac . To gain access to the SC , we carefully moved the membrane of this air-sac to the side , taking great caution not to puncture it . From this angle , we could then see the SC positioned above the keel and below the PEC . We quickly implanted the SC muscle with the stripped ends of the silver electrodes , and allowed the inter-clavicular air-sac to fall back into position . Finally , we gently moved the fat pad back over the furcula to prevent desiccation during the recordings and placed a small drop of normal avian saline over this tissue . Given the deep ( and difficult ) position of the SC , it was not possible to fasten the hook from the force transducer directly to the muscle for twitch speed recordings . Instead , we attached the hook to the bird’s elbow , and recorded the action of the stimulated SC as movement of the wing . When recordings were completed , we lightly pulled the electrodes from the muscle , removed the force transducer line from the elbow , and closed the dissection with Vetbond glue . Both the stimulator and force transducer were connected to a laptop through an A-D converter ( Model NI USB-6212 , National Instruments , TX , USA ) . The signal from the force transducer was first amplified ( 5K–10K ) and low-pass filtered ( 3000 Hz ) using an AC/DC strain gage amplifier ( Model P122 , Grass Technologies ) . We collected all recordings in AviSoft-RECORDER ( v . 4 . 2 . 22 ) and measured the data in Praat software ( v . 5 . 4 . 21 , P . Boersma and D . Weenink ) . This was accomplished by measuring the percentage at which muscles relax after each stimulatory pulse within a given stimulation train . Complete muscle relaxation ( 100% relaxation ) occurred when the extra tension detected by the force transducer in response to stimulation was fully relieved ( i . e . , the signal returned to its baseline level ) . Partial muscle relaxation occurred when the extra tension placed on the force transducer in response to stimulation was less than fully relieved ( i . e . , summation occurred ) . Inasmuch , the values of partial relaxation ranged between 99% and 0% ( full fusion ) and were calculated by dividing the actual amount of relaxation by the amount of relaxation that would otherwise be necessary for full recovery . For each stimulation train , we averaged the percent recovery values for the first eight stimulations . This corresponds to similar numbers of wing-oscillations that golden-collared and red-capped manakins use in their forelimb displays ( Bostwick and Prum , 2003; Fuxjager et al . , 2013 ) . Notably , in our calculations , we used baseline as the measure of a fully relaxed ( un-stimulated ) muscle , even though summation changes peak force production over the course of the stimulation train . As such , saturation levels may not always reflect 0% recovery . For each muscle , we compared the percent recoveries obtained from 20 Hz stimulations administered at the onset and after the experimental series using paired t-tests . One exception to this analysis occurred in red-capped manakins , as 20 Hz stimulations elicited 100% recovery in all cases . Variation across biological replicates is therefore 0 . 0 , which means that statistics cannot be performed . We interpret these results to indicate no significant difference in muscle recovery in response to 20 Hz stimulations before or after the experimental series , which is entirely consistent with the results of the paired t-tests obtained from the other species ( see Table 1 ) . Next , we examined muscle twitch dynamics by fitting the data with a four-parameter logistic non-linear regression model . This type of model characterizes data that form a reverse sigmoidal curve , and it is widely used to assess a variety of biological functions [e . g . , Kent et al . , 1972 , Johnson et al . , 2003] . We constrained model thresholds at or below 100% ( maximum possible percent relaxation ) and model saturations at or above 0% ( minimum possible percent recovery ) . Each model produced an estimated inflection point ( ± 1 SEM ) , which represents the muscle’s half-relaxation frequency . This metric is defined as the stimulation frequency at which the muscle’s average recovery equals half of its predicted functional range . The half-relaxation point is biologically significant , in that it represents a change in muscle length that is still highly effective in actuating movement of the appendage ( wing ) . This was validated in all species by watching the wings move in response to such stimulations ( also see Results ) . Thus , at the level of muscle fusions associated with the half-relaxation frequency , generation of the display behavior is still highly feasible . To this end , we statistically compared models and their estimates of half-relaxation frequency only within muscle type , considering that differences in the preparations of the separate muscles might affect model parameters . Therefore , we used extra sum-of-squares tests to compare models across species and one-way ANOVAs to compare half-relaxation frequencies across species . Significant effects from ANOVAs were followed by Student-Newman-Keuls ( SNK ) post-hoc comparisons to control for multiple pairwise contrasts . Data used for our analyses are included as a supplemental file ( Figure 3—source data 1 ) . | Many animals court mates and fight with rivals by performing physically elaborate and showy displays . From male fiddler crabs waving their claws to attract females , to the leaping dances of whooping cranes , these displays often involve remarkably fast limb movements . However , in many cases it is puzzling how animals can perform these behaviors , because the muscles that move the limbs are often geared to produce strength for walking , running or flying , and not speed . Indeed , decades of research in animal physiology has confirmed that limb-moving muscles can contract with either great strength or great speed , but never both . A small group of tropical birds called manakins produce different types of courtship displays , including some in which the wings are moved extremely rapidly . To date , nobody has examined if or how the limb muscles can generate such superfast movements . Fuxjager et al . now show that , in two species of manakins that produce rapid wing movements as part of their courtship displays , one of the main wing muscles has evolved to move the wings at superfast speeds . In fact , this muscle can move the wing at speeds that are more than twice as fast as those required for these birds to fly , and appears to be the fastest limb muscle on record for any animal with a backbone . Fuxjager et al . also show that the manakins’ other wing muscles are no different from other birds , and suggest that these muscles are preserved to produce the strength needed for flying . Further studies could now explore how this one muscle can create such superfast wing movements and whether male hormones , like testosterone , play a role in regulating the muscle’s speed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"neuroscience"
] | 2016 | Select forelimb muscles have evolved superfast contractile speed to support acrobatic social displays |
Microbial community structure and function rely on complex interactions whose underlying molecular mechanisms are poorly understood . To investigate these interactions in a simple microbiome , we introduced E . coli into an experimental community based on a cheese rind and identified the differences in E . coli’s genetic requirements for growth in interactive and non-interactive contexts using Random Barcode Transposon Sequencing ( RB-TnSeq ) and RNASeq . Genetic requirements varied among pairwise growth conditions and between pairwise and community conditions . Our analysis points to mechanisms by which growth conditions change as a result of increasing community complexity and suggests that growth within a community relies on a combination of pairwise and higher-order interactions . Our work provides a framework for using the model organism E . coli as a readout to investigate microbial interactions regardless of the genetic tractability of members of the studied ecosystem .
Microorganisms rarely grow as single isolated species but rather as part of diverse microbial communities . In these communities , bacteria , archaea , protists , viruses and fungi can coexist and perform complex functions impacting biogeochemical cycles and human health ( Falkowski et al . , 2008; Flint et al . , 2012 ) . Deciphering microbial growth principles within a community is challenging due to the intricate interactions between microorganisms , and between microorganisms and their environment . While interest in microbial communities has dramatically increased , our understanding of microbial interactions within communities is lagging significantly behind our ability to describe the composition of a given community . Approaches relying on 16S rDNA sequencing analyses of microbial communities can be used to reconstruct ecosystem networks and detect patterns of co-occurrence to infer general interactions such as competition , mutualism and commensalism ( Faust and Raes , 2012 ) . However , the molecular mechanisms underlying these interactions remain largely uncharacterized . Further , the way in which these interactions are organized within a community , such as whether they consist of predominantly pairwise or higher-order interactions , is even less clear . A more precise understanding of microbial interactions , their underlying mechanisms , and how these interactions are structured within a community , are all necessary to elucidate the principles by which a community is shaped . In this study , we combine genome-scale genetic and transcriptomic approaches within an experimentally tractable model microbial community to begin to address these questions . Genome-scale approaches , such as transposon mutagenesis coupled to next-generation sequencing ( TnSeq approaches ) have been successfully used to quantify the contribution and thus the importance of individual genes to a given phenotype ( van Opijnen and Camilli , 2013 ) . These techniques use a pooled library of transposon insertion mutants whose frequency is measured to identify genes important for growth in a given condition . Recently , the generation and introduction of unique random barcodes into transposon mutant libraries made this approach more high-throughput , enabling screens of important genes across hundreds of conditions and for numerous genetically tractable microorganisms ( Wetmore et al . , 2015; Price et al . , 2018 ) . To investigate the genetic basis of microbial interactions , we have adapted this approach to identify and compare genetic requirements in single-species ( non-interactive ) and multi-species ( interactive ) conditions . We used a large and diverse transposon library previously generated in the genetically-tractable model bacterium E . coli K12 ( Wetmore et al . , 2015 ) to characterize the genetic requirements of interactions within a model community based on the rind of cheese ( Wolfe et al . , 2014 ) . The fact that the E . coli genome has undergone extensive characterization can help more effectively interpret the genetic requirements introduced by growth within communities . Although E . coli K12 is not a typical endogenous species of this particular microbiome , non-pathogenic E . coli strains can be found in raw milk and raw-milk cheese ( Trmčić et al . , 2016 ) . Shiga-toxin-producing E . coli 0157:H7 and non-0157 pathogenic E . coli species are common invaders of the cheese environment and can survive during cheese making causing mild to life-threatening symptoms after ingestion ( Coia et al . , 2001; Montet et al . , 2009; Frank et al . , 1977 ) . Using the E . coli transposon library , we ( i ) identified the set of genes important for growth alone in the cheese environment , ( ii ) identified the set of genes important for growth in pairwise conditions with each individual community member and ( iii ) identified the set of genes important for growth in the presence of the complete community . Characterization of the functions or pathways associated with growth in interactive versus non-interactive conditions were then used to infer the biological processes involved in interactions within the model microbiome . Additionally , we compared the set of genes important for growth in pairwise conditions with the ones important for growth in a community to investigate how microbial interactions change depending on the complexity of the interactive context . We also performed a similar RB-TnSeq analysis during non-interactive and interactive conditions using a transposon library we generated in the cheese-endogenous species Pseudomonas psychrophila JB418 . Finally , we measured changes in the transcriptional profile of E . coli during growth alone , growth in pairwise conditions , and within the community using RNAseq as a complementary approach to RB-TnSeq in defining microbial interactions . This analysis revealed a deep reorganization of gene expression whenever E . coli is in the presence of other species . This work revealed numerous interactions between species , such as metabolic competition for iron and nitrogen , as well as cross-feeding from fungal partners for certain amino acids . Our analysis showed that most of the metabolic interactions ( competition and cross-feeding ) observed in pairwise conditions are maintained and amplified by the addition of all partners in the community context . However , around half of the genetic requirements observed in pairwise conditions were no longer apparent in the community , suggesting that higher-order interactions emerge in a community .
We used the E . coli Keio_ML9 RB-TnSeq library from Wetmore et al . , 2015 , containing a pool of 152 , 018 different insertion mutants ( with a median of 16 insertion mutants per gene; covering 3728 of 4146 protein-coding genes ) , each associated with a unique 20 nucleotide barcode . This library was originally generated in and maintained on lysogeny broth medium ( LB ) and was used previously to identify genes required for growth across a variety of conditions ( Wetmore et al . , 2015; Price et al . , 2018 ) . To determine genes important for growth on our cheese-based medium , we grew the pooled library by itself on sterile cheese curd agar plates ( CCA: 10% freeze-dried fresh cheese , 3% NaCl , 0 . 5% xanthan gum , 1 . 7% agar ) , the same medium used in all further experiments and used previously to demonstrate that cheese communities could be successfully reconstructed in vitro ( Wolfe et al . , 2014 ) . As the library is composed of multiple insertion mutants for a gene , we expect the individual insertion mutants to be evenly distributed in the experimental environment , minimizing the effect on any individual insertion mutant due to stochastic processes such as genetic drift or localized effects related to spatial structure ( Hallatschek et al . , 2007 ) . During growth , we expect the library to modify the environment by taking up nutrients and excreting molecules ( waste products , enzymes , etc ) . Consequently , we expect that some genetic requirements will change during growth . Thus , to provide a comprehensive overview of the genetic requirements for growth , we grew the pooled library on CCA and collected samples after 1 , 2 and 3 days . For each time point , we harvested the library from the surface of the cheese plate , extracted genomic DNA , used PCR to amplify the barcoded regions of the transposons , and then sequenced these products to measure the abundance ( i . e . the number of sequencing reads associated with each barcode ) of each transposon mutant over time ( see Materials and Methods ) . The fitness of each insertion mutant was calculated as the log2 of the ratio of its abundance at a given timepoint compared to its abundance at T0 ( the inoculum ) . We calculated the raw fitness of a gene as the weighted average of the fitness of all insertion mutants of that gene . Gene fitness values were then normalized . First , fitness values are corrected to account for changes in copy number along the chromosome as insertions near the replication fork are expected to have higher copies in dividing cells . Then , fitness values were normalized based on the assumption that disruption of most of the genes leads to little or no fitness effect ( see Materials and Methods and ( Wetmore et al . , 2015 ) for details ) . Consequently , most of the fitness values are expected to be close to 0 , indicating that disruption of these genes leads to no particular growth modification compared to the rest of the library . Negative gene fitness values , however , identify mutants that are growing slower than the rest of the library and therefore , genes that are of particular importance for growth in the studied condition . A t-score , calculated as a moderated t-statistic , is determined for each gene fitness value to assess if the fitness value is reliably different from 0 ( see Materials and Methods and ( Wetmore et al . , 2015 ) for details ) . The RB-TnSeq pipeline from experimental set-up to gene fitness calculation is summarized in Figure 1—figure supplement 1 . At each timepoint , we were able to calculate the fitness and a corresponding t-score for 3298 protein-coding genes ( Figure 1A , Figure 1—source data 1 ) . Because we were interested in genes with a strong fitness defect ( significant negative fitness values ) , we first removed genes with an absolute t-score <= 3 . This t-score threshold was set to identify strong negative fitness values while minimizing potential false positives ( false discovery rate of 0 . 2% ) . The t-score assesses how reliably a fitness value is different from 0 . In each condition , most genes have no detectable fitness effect , and thus have a fitness value close to 0 . Thus , in our dataset , most of the genes below this t-score also have a fitness value close to 0 . On average , 97% of the fitness values were associated with a t-score that falls below our threshold . Within the fitness values that pass the t-score threshold , we then removed genes associated with positive fitness values . Thus , we only retained the genes whose deletion leads to a consistent growth defect for E . coli on CCA compared to the rest of the library . This filtering process revealed 160 genes that were important for E . coli growth alone on CCA ( Figure 1A ) . To identify the functions associated with these 160 genes , we mapped them to the KEGG BRITE database ( Figure 1B ) . 84 genes were assigned to KEGG modules and 64 of them were associated with E . coli metabolism . Within these metabolic genes , we found 28 genes associated with amino acid metabolism , specifically the biosynthesis of all amino acids except for proline , lysine and histidine . Quantification of free amino acids in our medium highlighted very low concentrations of all amino acids ( Figure 1—figure supplement 2 ) suggesting that a limited supply of free amino acids leads to a genetic requirement for amino acid biosynthesis . This is supported by the observation that both spoT and relA , regulators of the stringent response which can be triggered by amino acid starvation ( Cashel et al . , 1996 ) , are also associated with a negative fitness value . Additionally , we observed the importance of the regulator gcvR , that inhibits catabolism of glycine into C1 metabolism . In fact , GcvR inhibits the glycine cleavage complex , a multienzyme complex that oxidizes glycine ( Ghrist and Stauffer , 1995; Ghrist et al . , 2001 ) . Furthermore , mutants of the glycine cleavage complex displayed a significant positive fitness , suggesting that absence of glycine utilization through C1 metabolism is beneficial in our amino acid-deficient environment . Altogether , this observation also underlines that amino acids are limiting in the environment and that their biosynthesis and utilization control is important for growth . 19 of the 160 genes were associated with energy metabolism and , more specifically , with sulfur assimilation ( n = 7 genes ) and respiration ( n = 8 genes ) . Here , we deduce that importance of sulfur assimilation is directly caused by the lack of the amino acids cysteine and methionine , which are the major pools of sulfur-containing compounds in the cell . As a non-endogenous species , E . coli might not possess the adequate peptidases or proteases to degrade and use the highly available protein casein . Identification of two of the three genes of the Leloir pathway ( galE and galT ) , involved in the uptake and conversion of galactose into glucose , suggests that galactose might be a crucial nutrient for E . coli growth on CCA . Eight genes mapped to membrane transport and were associated with two specific pathways: ferric-enterobactin transport and glycine-betaine transport . Ferric-enterobactin transport allows the cells to scavenge iron in a low-iron environment ( Raymond et al . , 2003; Hider and Kong , 2010 ) . Iron is an essential micronutrient and cheese is known to be iron-limited ( Albar et al . , 2014 ) . Glycine betaine is used by cells as an osmoprotectant in high osmolarity environments . During cheese curd processing , high concentrations of NaCl are added ( Guinee , 2004 ) , and our CCA medium contains 3% NaCl to mimic these conditions . The importance for E . coli to maintain its cell osmolarity is also suggested by the requirement of genes responsible for the transport of the ions sodium , potassium and zinc . In our experiment , the fact that all of the mutants are pooled together limits our ability to identify genes whose phenotypes can be complemented by common goods ( molecules released in the environment ) produced by neighboring cells . For example , given that iron is limiting in cheese , we expect that enterobactin biosynthesis is an important pathway for growth in this environment . However , no genes from the enterobactin biosynthesis pathway ( entCEBAH , entD and entF ) had a significant negative fitness value ( average fitness of the enterobactin biosynthesis pathway: 0 . 1 ) , while individual growth of these enterobactin biosynthesis mutants from the KEIO collection was limited on CCA compared to a rich , non-iron-limited medium ( Figure 1—figure supplement 3 ) . In summary , functions of major importance for E . coli to grow alone in our experimental environment involved ( i ) response to low iron availability , ( ii ) response to osmotic stress and ( iii ) response to limited available nutrients ( specifically free amino acids ) . These required functions are consistent with recently published results on the requirements of the mammary pathogenic E . coli ( MPEC ) during growth in milk ( Olson et al . , 2018 ) except for resistance to osmotic stress which does not occur in milk . We also generated an RB-TnSeq library in the bloomy rind cheese endogenous species P . psychrophila JB418 and found comparable requirements for growth alone on cheese ( Figure 1—figure supplement 4 ) . To validate the results obtained with the RB-TnSeq library , we measured the fitness of individual knockout mutants from the E . coli Keio collection ( Baba et al . , 2006 ) . We tested 25 knockout mutants corresponding to genes with a strong growth defect observed after one day of growth . We carried out competitive assays between each knockout mutant and the wild-type strain on CCA . We calculated each knockout mutant fitness as the log2 of the fold change of its abundance after one day of growth . A z-score was also calculated to assess the confidence of that fitness . 21 of 25 knockout mutants displayed a fitness value lower than 0 with at least 95% confidence ( Figure 1—figure supplement 5 ) . The remaining four mutant strains ( brnQ , cysK , serA and trxA ) were associated with high fitness value variability across replicate experiments and had a lower z-score . Altogether , this supports the reliability and validity of RB-TnSeq results . The growth of the E . coli library alone allowed us to determine the baseline set of genes required for optimal growth in the model cheese environment . We next wanted to identify genes with negative fitness during growth when another species is present . First , we analyzed the growth of E . coli and the partner species . We grew E . coli for 3 days on CCA in the presence of either H . alvei , G . candidum or P . camemberti . In addition to belonging to distinct domains or phyla , these three partners are the typical members of a bloomy rind cheese community ( such as Brie or Camembert ) . The presence of E . coli did not influence the growth of any partner species ( Figure 2—figure supplement 1 ) . However , E . coli’s growth was reduced in the presence of each partner after three days of growth ( Figure 2A ) . We then determined the genes associated with negative fitness during E . coli growth in each pairwise condition using RB-TnSeq ( i . e . genes whose fitness value is negative and associated with an absolute t-score greater than three in the pairwise condition ) ( Figure 2B , Figure 2—source data 1 ) . As performed above , barcode frequencies were compared between T0 and after growth with each partner ( at days 1 , 2 and 3 ) . As our goal is to compare genetic requirements for growth in interactive and non-interactive conditions rather than to examine changes in requirements over time , we grouped genes with a significant negative fitness for at least one timepoint as a single set of genes for each pairwise condition . We identified 145 genes with negative fitness values in E . coli for growth with H . alvei , 142 genes for growth with G . candidum and 131 genes for growth with P . camemberti . Altogether they constitute a set of 153 genes that are required for optimal growth in at least one pairwise culture . Comparison of genes with negative fitness identified when E . coli is grown alone with the genes identified when E . coli is grown in pairwise conditions is expected to highlight differences brought about by the presence of another species ( Figure 2C ) . Consistent presence of multiple genes of the same pathway within only one of these sets of genes associated with negative fitness is likely to point out a pathway specifically important in one condition . Thus , we can infer possible interactions based on the different relevant pathways between interactive and non-interactive growth conditions . Altogether , the 153 genes with a negative fitness in pairwise conditions and the 160 genes for E . coli growth alone represent 235 unique genes ( Figure 2C ) . These can be divided into three groups of genes: ( i ) conserved negative fitness: genes with negative fitness in both growth alone and in all pairwise conditions ( n = 78 ) , ( ii ) pairwise-alleviated negative fitness: any gene found to have a negative fitness during E . coli growth alone that was not associated with a negative fitness in at least one of the three pairwise cultures ( n = 82 ) , and ( iii ) pairwise-induced negative fitness: any gene with negative fitness in the presence of at least one of the partners but not associated with a negative fitness during growth alone ( n = 75 ) ( Figure 2C and D and Figure 2—figure supplement 2 ) . We further focused on the pairwise-alleviated and pairwise-induced negative fitness as these groups contain genes potentially related to interactions . Genes whose negative fitness is alleviated by pairwise growth can highlight processes that are of importance for growth alone but no longer important because of the presence of a partner , thus suggesting interactions between E . coli and the partner . Just over half of the genes with negative fitness alone appeared to be relieved by the presence of a partner ( n = 82 genes , Figure 2C ) , suggesting major modifications of growth conditions following the introduction of a partner . We mapped these 82 alleviated genes to the KEGG BRITE database to identify functions and pathways that are no longer critical in the presence of a partner ( Figure 2E ) . 16 genes were associated with unknown or predicted proteins and did not map to any field of the database . Of the remaining genes , 45 mapped to modules of the KEGG orthology hierarchy . Most of the genes with alleviated negative fitness were associated with the KEGG metabolism module and are thus part of metabolic pathways . It is especially evident that pairwise growth leads to major changes in the need for amino acid biosynthesis . For example , 6 out of the 8 genes of valine and isoleucine biosynthetic pathways are no longer associated with a negative fitness during pairwise growth ( Figure 3C ) . In addition , 2 genes of arginine biosynthesis , 2 genes of methionine biosynthesis as well as final steps of homoserine , aspartate and glutamate biosynthesis are no longer required . Moreover , ilvY , the transcriptional activator of valine and isoleucine biosynthesis was also among the genes no longer required for pairwise growth . Here , the dominant presence of amino acid biosynthesis genes in the alleviated functions suggests cross-feeding of the pathway end-products or intermediates which are either provided directly by the partner species or made more available in the environment as a consequence of the partner’s metabolic activity . Thus , our data suggest that pairwise growth may allow cross-feeding of the amino acids valine , isoleucine , arginine , methionine , homoserine , aspartate and glutamate . Isoleucine and methionine are also intermediates of cofactor biosynthesis , and genes associated with their biosynthesis were also mapped to metabolism of cofactors and vitamins . To understand if the genes with pairwise-alleviated negative fitness were related to a specific partner , we investigated how each partner contributed to this gene set ( Figure 2—figure supplement 2 ) . Of the 82 total genes , 36 were alleviated in all pairwise conditions , suggesting that any partner leads to the compensation of these requirements . They included genes associated with amino acid metabolism specific to homoserine and methionine biosynthesis . Of the remaining genes , eight were specifically not required in the presence of H . alvei , nine were specifically not required in the presence of G . candidum and nine were specifically not required in the presence of P . camemberti . Alleviation of leucine and valine biosynthesis was observed with both fungal partners , while biosynthesis of arginine appeared to be no longer required specifically in the presence of G . candidum . Fungal species are known to secrete proteases that digest small peptides and proteins ( Kastman et al . , 2016; Boutrou et al . , 2006b; Boutrou et al . , 2006a ) and may lead to increased availability of amino acids in the environment . We then analyzed the 74 genes with pairwise-induced negative fitness in order to identify functions or pathways that become important in the presence of a partner ( Figure 2E ) . These genes represent almost half ( 75 out of 153 – Figure 2C ) of the genes with negative fitness in pairwise conditions , suggesting that presence of a partner introduces new selection pressures . 33 genes mapped to KEGG orthology terms . Among this gene set are pathways associated with signal transduction , biofilm formation and drug resistance . They were related to three major responses: metabolic switch ( creB: carbon source responsive response regulator ) , response to stress and toxic compounds ( cpxA: sensory histidine kinase , oxyR: oxidative stress regulator , acrAB: multidrug efflux ) and biofilm formation ( rcsC and rcsB: regulator of capsular synthesis , pgaC: poly-N-acetyl-D-glucosamine synthase subunit ) . Biofilms are microbial structures known to provide resistance to different stresses , including resistance to antibiotics , and biofilm-inducing genes can be activated in the presence of stress events ( Landini , 2009 ) . The transcriptional regulator OxyR and the transduction system CpxA and CpxB are known coordinators of stress response and biofilm formation ( Gambino and Cappitelli , 2016; Dorel et al . , 2006 ) . While these genes represent only a small subset of the pairwise-induced gene set , they could suggest that partner species are producing toxic compounds or oxidative stress-inducing compounds . We again investigated if these responses were partner-specific ( Figure 2—figure supplement 2 ) . Of the 74 pairwise-induced negative fitness , 11 were found to have a negative fitness in the presence of all partners , 13 were specific to the presence of H . alvei , 24 were specific to the presence of G . candidum and 11 were specific to the presence of P . camemberti . Despite involving different genes , necessity of biofilm formation and response to toxic stress were associated with the presence of all partners . Finally , functional analysis of the conserved genes with negative fitness highlighted that functions associated with membrane transport , including resistance to high osmolarity and iron transport as well as functions associated with energy metabolism and aromatic amino acid biosynthesis were still important to grow in the presence of a partner ( Figure 2E ) . We performed similar pairwise assays using the RB-TnSeq library of P . psychrophila JB418 with H . alvei , G . candidum or P . camemberti . We identified a similar number of genes associated with pairwise-alleviated and pairwise-induced requirements ( Figure 1—figure supplement 4 ) as we did when using the E . coli library . As with E . coli , we can infer production of toxic stress by the partners as genes associated to DNA repair were identified with a negative fitness in pairwise conditions in the functional analysis . However , cross-feeding by fungal partners was not as striking as for E . coli . We next aimed to investigate the differences between genes with a negative fitness during growth in a community ( complex interactive condition ) and genes with a negative fitness during growth in associated pairwise conditions ( simple interactive conditions ) ( Figure 2—figure supplement 2 ) . We grew the E . coli library with the complete community composed of H . alvei , G . candidum and P . camemberti and we identified 126 genes with a reliable negative fitness ( Figure 3A , Figure 3—source data 1 ) . E . coli’s final biomass was more reduced by the presence of the community than by a single partner . However , the growth of each community member remained unaffected ( Figure 2—figure supplement 1 ) . We first identified community-induced and community-alleviated genes by comparing the genes with a negative fitness in the community with the genes with a negative fitness during growth alone . We identified 89 genes that had negative fitness for both community and alone ( conserved negative fitness ) , 37 genes with negative fitness only with the community ( community-induced negative fitness ) and 71 genes with negative fitness only for growth alone ( community-alleviated negative fitness ) . As with a single partner , the presence of a complex community potentially relieves some fitness effects while introducing new ones . Comparing community-induced and pairwise-induced genes can reveal if and how community complexity modifies the genes that are important in different interactive contexts compared to growth alone ( Figure 3B – Interaction-induced negative fitness ) . We identified 29 genes with a negative fitness in both pairwise and community growth compared to growth alone ( conserved interaction-induced negative fitness ) . These include genes associated with oxidative stress and biofilm formation . These genes are likely to be associated with pairwise interactions which are maintained in a community context . Meanwhile , eight genes appeared to be specifically associated with negative fitness in the presence of the community ( Figure 3B , community-specific induced genes ) , highlighting higher-order interactions that emerge from a higher level of complexity in the community composition . Interestingly , these genes represent only a small fraction ( 22% ) of the community-induced requirements , suggesting that most of the negative fitness effects observed in the community are derived from pairwise interactions . Finally , we identified 46 genes that have a negative fitness in pairwise conditions , but not during growth alone or within the community ( Figure 3B , pairwise-specific induced genes ) . These genes could be related to interactions that are either alleviated or counteracted in a community , either by the presence of a specific species or by the community as a whole . For example , some of the identified genes were associated with antimicrobial resistance , and , in a diverse community , other species could degrade the putative antimicrobial molecules or prevent the producing species from secreting it . Consequently , E . coli would be exposed to a lower level of antimicrobials , suppressing the necessity of a resistance gene . Thus , the complex pattern of requirements for these genes may reflect higher-order interactions . We next investigated if the interactions related to pairwise-alleviated negative fitness and community-alleviated negative fitness were similar ( Figure 3B – Interaction-alleviated negative fitness ) . 68 genes were no longer associated with a negative fitness in both pairwise conditions and with the community compared to growth alone ( conserved interaction-alleviated negative fitness ) . These genes may represent pairwise interactions maintained in the community context . Amino acid biosynthesis was highly represented within these genes and more specifically biosynthesis of valine , isoleucine , methionine , homoserine , aspartate and glutamate ( Figure 3C ) . This suggests that , despite the presence of more species , these amino acids are still cross-fed . We also identified 14 genes that no longer had a negative fitness in pairwise conditions compared to growth alone yet remained with a negative fitness in growth with the community ( pairwise-specific alleviated negative fitness ) . These 14 genes represent a small fraction of the pairwise-alleviated , thus suggesting that most of interactions related to pairwise-alleviation are maintained in the community . Finally , only three genes were specifically alleviated in the community ( community-specific alleviated fitness ) . This points out that presence of the full community does not lead to emergence of specific alleviation of fitness effects but that most of the fitness effect alleviations observed in the community are conserved from pairwise interactions . In both cases , these 14 pairwise-specific and three community-specific alleviated genes could highlight existence of more higher-order interactions . However , too few genes are involved to determine the exact nature of these interactions . Finally , we identified 75 genes with negative fitness in all conditions ( core negative fitness ) . These genes encompass functions including iron uptake and response to high osmolarity . Overall , they are associated with response to environmental parameters that other species do not alleviate . To summarize , the community-induced genes were mostly maintained from pairwise-induced genes . Similarly , the genes that were community-alleviated were highly similar to the pairwise-alleviated genes . However , we also observed emergence of higher-order interactions in the community condition as numerous interactions observed in pairwise conditions ( n = 46 + 14 ) were not conserved in the community condition and specific interactions ( n = 8 + 3 ) were observed in the community condition . Altogether , 58% of the interactions observed in the community were from pairwise interactions while 42% emerged from higher community complexity . Again , we carried out similar experiments and analysis using the P . psychrophila JB418 RB-TnSeq library generated in our laboratory . The results were highly similar to the ones observed with E . coli in terms of number of genetic requirements alleviated in the presence of the community compared to growth alone as well as the number genes specifically important to grow with the community compared to growth alone ( Figure 1—figure supplement 4 ) . Finally , we consistently observed importance of higher-order interactions , 61% of the observed interactions in the community were conserved from pairwise interactions and 39% were higher-order interactions . So far , we used a genome-scale genetic approach to investigate potential microbial interactions . As a complementary strategy , we generated transcriptomic data for E . coli during growth in each previously described condition . Changes in transcriptional profiles can be a powerful indicator of an organism’s response to an environment and have been used to identify E . coli pathways involved in interactions ( Croucher and Thomson , 2010; McAdam et al . , 2014; Galia et al . , 2017 ) . To measure E . coli gene expression , we extracted and sequenced RNA from each timepoint and condition of the same samples used for RB-TnSeq above ( after 1 , 2 and 3 days of growth when grown alone , in pairwise conditions or with the community ) . Comparison of transcriptional profiles suggests a strong reorganization of E . coli gene expression in response to the presence of a partner ( Figure 4A , Figure 4—source data 1 and Figure 4—figure supplement 1 ) . We first focused on the genes differentially expressed between growth in pairwise conditions and growth alone . We calculated the fold change of gene expression between pairwise growth and growth alone and identified differentially expressed genes by screening for adjusted p-values lower than 1% ( Benjamini-Hochberg correction for multiple testing ) and an absolute log2 of fold change ( logFC ) greater than 1 . To remain consistent with the analysis performed for the genetic requirements , we pooled the data of all timepoints after identifying the upregulated or downregulated genes for each timepoint . We found a total of 966 upregulated and 977 downregulated genes across all partners ( 482 upregulated genes and 478 downregulated genes in presence of H . alvei , 633 upregulated genes and 719 downregulated genes in presence of G . candidum , 626 upregulated genes and 694 downregulated genes in presence of P . camemberti , Figure 4A ) . Almost half of E . coli’s genome is subjected to expression modification , suggesting a global response to the presence of a partner . We further investigated if differential expression in pairwise conditions is partner-specific ( Figure 4—figure supplement 1 ) . Around half of E . coli gene expression regulation in the presence of a partner appears to be independent of which partner is present . Also , a number of genes were differentially expressed depending on the partner: 66 genes were specifically upregulated and 60 genes downregulated with H . alvei , 213 upregulated and 182 downregulated with G . candidum , and 183 upregulated and 161 downregulated with P . camemberti . Due to the larger gene set compared to RB-TnSeq , we performed KEGG pathway enrichment analyses on the differentially expressed genes in pairwise conditions to determine upregulated functions and pathways ( Figure 4B ) . First , almost all of the aminoacyl-tRNA-synthetases and functions associated with energy production were upregulated . Interestingly upregulation of energy production through aerobic respiration and the TCA cycle happened after 3 days of growth . Oxygen availability ( Gunsalus , 1992 ) and growth phase ( Wackwitz et al . , 1999 ) are the two known regulators of aerobic respiration . At day 3 , E . coli was observed to be in log phase when alone , whereas in the presence of a partner , and especially with P . camemberti , E . coli was observed to enter the stationary phase between day 2 and day 3 ( Figure 2 – figure supplement 1 ) . Therefore , upregulation of aerobic respiration is most likely associated with the growth stage difference between E . coli alone and with a partner . While these functions were upregulated regardless of the partner , more genes were upregulated in the presence of G . candidum than the other partners and thus , several pathways associated with nucleotide biosynthesis ( C1-pool by folate , purine metabolism , and pyrimidine metabolism ) were specifically upregulated with this partner . This suggests that either E . coli and G . candidum compete for nucleotide compounds from the environment or that presence of G . candidum leads to an increased demand of nucleotide compounds for E . coli’s metabolism and growth . We performed a similar KEGG pathway enrichment analysis on the downregulated genes in pairwise conditions . Pathways involved in the biosynthesis of amino acids , specifically tyrosine , phenylalanine , tryptophan , methionine , lysine , arginine , homoserine , leucine , glutamate , threonine and glycine , appeared to be the principal downregulated functions in the presence of a partner and more particularly with a fungal partner . Interestingly , some amino acid biosynthetic pathways were upregulated later in the growth but not significantly enriched in the enrichment analysis ( phenylalanine , tyrosine and leucine ) . Downregulation of amino acid biosynthesis suggests that the partner species generates amino acids available for cross-feeding . The observation of this interaction in the transcriptome data is consistent with our interpretation of RB-TnSeq results and reinforces the likelihood of such an interaction . However , late upregulation of some amino acid biosynthesis suggests that as the partner grows along with E . coli they eventually end up competing for amino acids , leading to biosynthesis upregulation . This late competition was unlikely to be detected by RB-TnSeq using our current analysis . To summarize , presence of a partner triggers a significant and dynamic reorganization of E . coli gene expression . Most of these modifications restructure E . coli metabolic activity: mostly in response to modification of growth phase , but also in response to nutrient availability changes and for example to benefit from cross-feeding and common goods . Next , we aimed to determine whether E . coli gene expression reorganization significantly changes when grown with the full community as compared to growth in pairwise conditions . To do so , we first calculated E . coli gene logFC at each timepoint between growth with the community and growth alone . We further analyzed genes with adjusted p-values lower than 1% ( Benjamini-Hochberg correction for multiple testing ) and absolute logFC greater than 1 . After pooling across timepoints , we identified 465 upregulated and 476 downregulated genes in the presence of the community versus growth alone ( Figure 4A ) . We then compared these genes to the 966 upregulated genes and 977 downregulated genes in pairwise conditions versus growth alone ( Figure 4B and C ) . First , 416 genes were found to be upregulated in both pairwise and community growth versus growth alone ( conserved upregulated genes ) . Enrichment analysis highlighted functions that were previously described as upregulated in most of the pairwise conditions: aminoacyl-tRNA-synthetase and energy metabolism ( Figure 4B ) . This suggests that certain interactions that E . coli experienced in pairwise conditions are conserved in the community context . To investigate if the addition of similar interactions from different partners leads to an amplified response , we explored if the magnitude of expression changes in these pathways is higher in the community . We performed differential expression analysis on the genes comparably regulated in pairwise conditions and with the community ( Figure 4—figure supplement 1 ) . 50 of the 416 conserved upregulated genes were significantly more upregulated in community growth compared to pairwise growth . Among them , sulfate assimilation genes were overrepresented . This suggests that similar pairwise interactions may be additive in the community , leading to a stronger transcriptional response . Next , we identified 549 genes that were specifically upregulated in pairwise conditions versus growth alone and not upregulated in community versus growth alone ( pairwise-specific upregulated genes ) . KEGG pathway enrichment analysis highlighted that these genes were mostly associated with quorum sensing , fatty acid metabolism and oxidative phosphorylation ( Figure 4B ) . This observation suggests that the presence of additional species in the community counteracts or prevents certain pairwise interactions . It supports the presence of higher-order interactions as highlighted with the RB-TnSeq experiments . Indeed , more than half of the upregulated genes observed in pairwise conditions are not conserved with the community . Finally , 49 genes were specifically upregulated during community growth versus growth alone ( community-specific upregulated genes ) . Emergence of specific expression patterns with the community also suggests the existence of higher-order interactions . However , these community-specific upregulated genes represent only a small fraction ( 10% ) of upregulated genes within the community . Thus , most expression upregulation observed with the community is conserved from expression upregulation observed in pairwise conditions . Genes specifically upregulated with the community were associated with the biosynthesis of valine , leucine , and isoleucine , pyrimidine metabolism as well as arginine and proline metabolism ( Figure 4B ) . Upregulation of certain amino acid biosynthesis pathways suggests that despite potential cross-feeding from individual partners , addition of many partners eventually leads to competition . Upregulation of pyrimidine , arginine and proline metabolism however is part of a larger response; the response to nitrogen starvation ( Figure 4D ) . This response facilitates cell survival under nitrogen-limited conditions . Specifically , upregulated genes included all the genes involved in the regulatory loop of the transcriptional regulator NtrC ( glnL ) and nitrogen utilization as well as NtrC transcriptional targets: the transcriptional regulator Nac ( nac ) , the operon rutABCDEFG involved in ammonium production by uracil catabolism , the astABCDE operon constituting the arginine degradation pathway ( AST pathway ) and the two regulators of the stringent response , relA and spoT . Thus , the presence of additional species in the community specifically triggers the activation of the response to nitrogen starvation , which suggests a potential higher competition for nitrogen in the community context . We performed a similar analysis on downregulated genes in pairwise conditions and with the community versus growth alone to investigate if transcriptional downregulation in pairwise and community conditions are similar ( Figure 4C ) . We identified 448 genes that were downregulated during both pairwise and community growth conditions versus growth alone ( conserved downregulated genes ) . Enrichment analysis pointed to the downregulation of amino acid biosynthesis as well as cysteine and methionine metabolism . Therefore , consistent with our RB-TnSeq data , this suggests that cross-feeding from a single partner is maintained in a more complex context . 527 genes were specifically downregulated in pairwise conditions and not with the community ( pairwise-specific downregulated genes ) . Despite the large number of genes , no specific functions were enriched . However , the rutABCDEFG and astABCDE operons associated with the response to nitrogen starvation were downregulated in each pairwise condition ( Figure 4D ) . Altogether , pairwise-specific downregulated genes represent 54% of the genes downregulated in pairwise conditions , thus strongly suggesting higher-order interactions . Here , the presence of the community may trigger a highly specific response that would otherwise be downregulated in the presence of only one species . Finally , also highlighting potential higher-order interactions , 28 genes were specifically downregulated when E . coli is grown with the community ( community-specific downregulated genes ) . However , this represents only 6% of the observed downregulated genes in the community , highlighting again that most of the gene expression regulations in the presence of the community are conserved from gene expression regulations pairwise interactions . To conclude , most of the changes in E . coli gene expression during growth with the community were similar to a subset of expression changes observed in pairwise conditions . Moreover , some of these changes were amplified in the community compared to pairwise . This suggests that while a large part of transcriptional regulation in the community results from pairwise interactions , similar interactions from different partners may be additive in the community and exert a stronger impact on transcription . Also , the observed changes in nitrogen availability-related transcription suggest that community growth may induce new metabolic limitations .
In this work , we used the model organism E . coli as a readout for microbial interactions in a model cheese rind microbiome . We used genome-scale approaches to determine the changes in E . coli’s genetic requirements and gene expression profiles in conditions with increasing levels of community complexity . Our analysis highlighted both important changes in E . coli’s genetic requirements between interactive and non-interactive conditions as well as deep reorganization of E . coli’s gene expression patterns . We identified a variety of interactive mechanisms in the different interactive contexts . Our data revealed that interactions within the community include both competitive and beneficial interactions . By reconstructing a community from the bottom up , we were able to investigate how interactions in a community change as a consequence of being in a more complex , albeit still simple , community . RNASeq and RB-TnSeq consistently showed that around half of the interactions in a community can be attributed to pairwise interactions and the other half can be attributed to higher-order interactions . Although community structure is argued to be predictable from pairwise interactions in specific cases , higher-order interactions are believed to be responsible for the general lack of predictability ( Billick and Case , 1994; Friedman et al . , 2017; Momeni et al . , 2017 ) . Similarly , such higher-order interactions have been shown to be responsible for the unpredictability of community function from individual species traits ( Sanchez-Gorostiaga et al . , 2018 ) . Our work demonstrates the existence and prevalence of these higher-order interactions even within a simple community . Together , RB-TnSeq and RNASeq provided insight into mechanisms of mutualism between microbial species in this model system . One major interaction mechanism appears to be cross-feeding of amino acids from fungal partners . Although amino acid biosynthesis pathways were strongly required when E . coli grew alone , the presence of fungal species , but not bacterial species , led to fitness effect alleviation and downregulation of amino acid biosynthesis . This suggests that fungi increase the availability of free amino acids in the environment . Cheese-associated fungal species are known to secrete proteases that can degrade casein , the major protein found in cheese ( Kastman et al . , 2016; Boutrou et al . , 2006b; Boutrou et al . , 2006a ) , and therefore may increase the availability of an otherwise limiting resource . Although our model system is based on cheese , interactions based on cross-feeding are widely observed in other environments , such as soil , the ocean or the human gut ( Freilich et al . , 2011; Pacheco et al . , 2018; Goldford et al . , 2018 ) . For example , in the gut microbiome , Bifidobacteria can ferment starch and fructooligosaccharides and produce fermentation products including organic acids such as acetate which can in turn be consumed by butyrate-producing bacteria like Eubacterium hallii ( Belenguer et al . , 2006; De Vuyst and Leroy , 2011; Flint et al . , 2012 ) . Cross-feeding of other nutrients in the gut has also been uncovered using a related approach ( INSeq ) which found that vitamin B12 from Firmicutes or Actinobacteria was important for the establishment of Bacteroides thetaiotaomicron in mice ( Goodman et al . , 2009 ) . Our results also revealed mechanisms of competition within the community . RNASeq highlighted that both siderophore production and uptake are upregulated in interactive conditions , suggesting that there is competition for iron between species . Competition for iron is frequently observed across many environments , including cheese , as iron is an essential micronutrient for microbial growth and often a limited resource ( Monnet et al . , 2012; Albar et al . , 2014; Stubbendieck and Straight , 2016; Traxler et al . , 2012 ) . Interestingly , although we were able to detect fitness defects for siderophore uptake using RB-TnSeq , we did not see fitness defects for siderophore biosynthesis mutants . Because RB-TnSeq relies on a pooled library of mutants , one of the limitations to this approach is that it is difficult to detect fitness effects for genes associated with the production of common goods . For example , in the pooled library , most cells have wild-type siderophore biosynthesis genes , and thus produce and secrete siderophores into the environment under iron limitation . A consequence of this is that any cell that has lost the ability to produce siderophores can readily access the siderophores produced by neighboring cells . In contrast , the genes for uptake of common goods should remain crucial , and accordingly , we do observe fitness defects in the siderophore uptake genes . For this reason , using RNASeq can help overcome some of the limitations , such as pooling effects , associated with RB-TnSeq . Interactions between species also appeared to lead to stressful growth conditions , as RB-TnSeq showed the need for genes to deal with growth in the presence of toxic compounds . G . candidum is known to produce and excrete D-3-phenyllactic acid and D-3-indollactic acid , which inhibit the growth of Gram-negative and Gram-positive bacteria in the cheese environment ( Boutrou and Guéguen , 2005; Dieuleveux et al . , 1998 ) . Also , strains of H . alvei isolated from meat have been shown to produce compounds inhibiting biofilm formation in Salmonella enterica serovar Enteritidis ( Chorianopoulos et al . , 2010 ) . To begin to understand the extent to which the interactions we detected with E . coli were specific to this species , or more general , we performed similar RB-TnSeq experiments with the cheese isolate Pseudomonas psychrophila . This comparative approach showed that some responses to growth with other species are conserved , such as those needed to survive stress conditions , while others differ between the two species such as amino acid cross-feeding . This further highlights the ability to detect the dynamic nature of interactions , which not only change with community complexity , but also with the composition of the community . While our analysis highlighted global changes occurring as a consequence of interactions , and some of the key underlying interaction mechanisms , many more aspects of the biology occurring within communities are likely to be uncovered even within this simple model system . For example , much of our current analysis is limited to well-characterized pathways with strong negative fitness effects , yet many uncharacterized genes were also identified as potentially involved in interactions . Further investigation of these genes could uncover novel interaction pathways . Additionally , analysis of the exact ways in which community members modify the growth environment , such as through the production of extracellular metabolites , will be important to fully understand the molecular mechanisms of interactions . Altogether , this study revealed the intricacy , redundancy and specificity of the many interactions governing a simple microbial community . The ability of E . coli to act as a probe for molecular interactions , the robustness of RB-TnSeq , and its complementarity with RNASeq open new paths for investigating molecular interactions in more complex communities , independently of the genetic tractability of their members , and can contribute to a better understanding of the complexity and diversity of interactions within microbiomes . Finally , our work provides a starting point for better understanding the exact nature of higher-order interactions , and how they impact microbial communities .
The following growth assays are distinct from the growths carried out for RB-TnSeq and fitness analysis ( see below ) . Assays have been performed in at least triplicates . Growth assays have been carried out for the E . coli JW0024 strain ( Baba et al . , 2006 ) and P . psychrophila JB418 during growth alone , in pairwise conditions with either H . alvei JB232 , G . candidum or P . camemberti and with the full community . E . coli was pre-cultured overnight in liquid LB-kanamycin ( 50 µg/ml ) at 37°C and P . psychrophila JB418 was pre-cultured overnight in LB at room temperature ( RT ) . Then , for growth alone assays , 1000 cells of E . coli or P . psychrophila JB418 were inoculated on a 96 well plate containing 200 µL of CCA per well . For pairwise growth assays , either E . coli or P . psychrophila JB418 was co-inoculated with either H . alvei JB232 , G . candidum or P . camemberti at a ratio of 1:1 cell ( 1000 cells of E . coli and 1000 cells of the partner ) . Finally , for growth assay with the community , E . coli or P . psychrophila JB418 have been co-inoculated with H . alvei JB232 , G . candidum and P . camemberti at a ratio of 10:10:10:1 cells . Growth assays were then carried out for 3 days at RT . Agar plugs from 96 well plates were harvested at T = 0 hr , 6 hr , 12 hr , 24 hr , 36 hr , 48 hr , 72 hr and 120 hr for E . coli growth assays and T = 0 hr , 12 hr , 24 hr , 48 hr and 72 hr for P . psychrophila JB418 growth assays . Agar plugs were homogenized in 1 mL of PBS1X-Tween0 . 05% and three dilutions were plated on different media to measure growth of each species ( see Table 2 ) . Plates were incubated for 24 hr at 37°C for E . coli and 2 days at RT for P . psychrophila JB418 . After incubation , colony forming units ( CFUs ) were counted to estimate the number of bacterial cells on the cheese curd agar plates . Growth alone of H . alvei JB232 , G . candidum and P . camemberti have also been carried out similarly to E . coli and P . psychrophila JB418 growth alone . P . psychrophila JB418 gDNA was sequenced using Pacific Biosciences ( PacBio ) , Oxford Nanopore Minion ( Oxford Nanopore , Oxford , UK ) and Illumina sequencing . PacBio library preparation and sequencing were performed by the IGM Genomics Center at the University of California San Diego . Nanopore library preparation and sequencing were done at the University of California , Santa Barbara as part of the Eco-Evolutionary Dynamics in Nature and the Lab ( ECOEVO17 ) . Illumina library preparation and sequencing were done at the Harvard University Center for Systems Biology . Canu was used to assemble the PacBio and nanopore reads ( Koren et al . , 2017 ) . Illumina data was then used to correct sequencing error using the software Pilon ( Walker et al . , 2014 ) . The assembled genome was annotated using the Integrated Microbial Genomes and Microbiomes ( IMG/M ) system ( Markowitz et al . , 2012 ) . The P . psychrophila JB418 genome is 6 , 072 , 477 nucleotides long . It contains a single circular chromosome of 5 . 85 Mb and 4 plasmids of 172 . 2 Kb , 37 . 7 Kb , 5 . 8 Kb and 2 . 4 Kb . 6060 genes including 5788 open reading frames were identified . This genome is publicly available on the IMG/M website as IMG Genome ID 2751185442 . P . psychrophila JB418 was mutagenized by conjugation with E . coli strain APA766 ( donor WM3064 which carries the pKMW7 Tn5 vector library containing 20 bp barcodes ) ( Wetmore et al . , 2015 ) . This donor strain is auxotrophic for diaminopimelic acid ( DAP ) . The full collection of the APA766 donor strain ( 1 mL ) was grown up at 37°C overnight at 200 rpm . Four 25 mL cultures ( each started with 250 µL of APA766 stock ) were grown in LB-kanamycin:DAP ( 50 µg/mL kanamycin and 60 µg/mL DAP ) . A 20 mL culture was started from an individual P . psychrophila JB418 colony in LB broth and grown at RT overnight at 200 rpm . E . coli donor cells were washed twice with LB and resuspended in 25 mL LB . Donor and recipient cells were then mixed at a 1:1 cell ratio based on OD600 measurements , pelleted , and resuspended in 100 µL . This was done separately for each of the four E . coli cultures . 40 µL were plated on nitrocellulose filters on LB plates with 60 µg/mL DAP . Two filters were used for each of the four conjugation mixtures ( eight total conjugations ) . The conjugations took place for 6 hr at RT . After 6 hr , the filters were each resuspended in 2 mL of LB broth and then plated on LB:kanamycin ( 50 µg/mL ) for selection of transconjugants . 20 plates were plated of a 1:2 dilution for each conjugation ( 160 plates total ) . Transconjugants were pooled and harvested after three days of growth on selection plates . The pooled mixture was diluted back to 0 . 25 in 100 mL of LB:kanamycin ( 50 µg/mL ) . The culture was then grown at RT to an OD600 of 1 . 3 before glycerol was added to 10% final volume and 1 mL aliquots of the library ( named JB418_ECP1 ) were made and stored at −80°C for future use . Library preparation was performed as in Wetmore et al . , 2015 with slight modifications ( Wetmore et al . , 2015 ) . The E . coli barcoded transposon library Keio_ML9 and the P . psychrophila strain JB418 library were used for RB-TnSeq fitness assays on CCA during growth alone , growth in pairwise condition with each bloomy rind cheese community member and during growth with the full community . Figure 1—figure supplement 1 provides a description of the fitness assays as well as fitness calculation . We used mutants from the Keio collection to validate the genes identified by RB-TnSeq as having a significant fitness in E . coli growth alone on CCA ( see list in Figure 1—figure supplement 5 ) . Each mutant was grown in a competition assay with the non-kanamycin resistant wild-type ( Keio ME9062 – ( Baba et al . , 2006 ) ) . 1000 cells of a specific mutant were inoculated with 1000 cells of the wild type ( WT ) on the surface of the same cheese plug in a 96 well plate containing 10% CCA , pH7 . The number of the mutant cells and the WT cells were calculated at T0 and day one after harvesting and homogenizing the cheese plug , plating serial dilutions and counting CFUs . Experimental fitness of each mutant was calculated as the log2 of the ratio of the mutant abundance ( mutant CFUs divided by total CFUs ( WT +mutant ) ) after 24 hr and its abundance at T0 . | Microorganisms live almost everywhere on Earth . Whether it is rainforest soil or human skin , each environment hosts a unique community of microbes , referred to as its microbiome . There can be upwards of hundreds of species in a single microbiome , and these species can interact in a variety of ways; some cooperate , others compete , and some can kill other species . Deciphering the nature of these interactions is crucial to knowing how microbiomes work , and how they might be manipulated , for example , to improve human health . Yet studies into these interactions have proven difficult , not least because most of the species involved are difficult to grow in controlled experiments . One environment that is home to a rich community of microbes is the outer surface of cheese , known as the cheese rind . The cheese rind microbiome is a useful system for laboratory experiments , because it is relatively easy to replicate and its microbes can be grown on their own or in combinations with others . To explore the nature of interactions in microbiomes , Morin et al . have now grown a large collection of E . coli mutants as members of simplified microbiomes based on the cheese rind . The mutant bacteria were grown on cheese either alone , paired with one other species , or alongside a community of three species . The aim was to see which mutants grew poorly when other species were present , thus allowing Morin et al . to identify specific genes that are important for interactions within the experimental microbiomes . Even in these simplified microbiomes , the microbes interacted in a variety of ways . Some microbes competed with E . coli for elements like iron and nitrogen; others cooperated by sharing the building blocks needed to make larger molecules . Many of the interactions that happened when E . coli was paired with one species were not seen when more species were added to the community . Similarly , some interactions were only seen when E . coli was grown alongside a community of microbes , and not when it was paired with any of the three species on their own . These findings show that complex interactions are present even in a simplified microbiome . This experimental approach can now be applied to other microbiomes that can be grown in the laboratory to examine whether the patterns of interactions seen are generalizable or specific to the cheese rind system . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Changes in the genetic requirements for microbial interactions with increasing community complexity |
Dense connectomic mapping of neuronal circuits is limited by the time and effort required to analyze 3D electron microscopy ( EM ) datasets . Algorithms designed to automate image segmentation suffer from substantial error rates and require significant manual error correction . Any improvement in segmentation error rates would therefore directly reduce the time required to analyze 3D EM data . We explored preserving extracellular space ( ECS ) during chemical tissue fixation to improve the ability to segment neurites and to identify synaptic contacts . ECS preserved tissue is easier to segment using machine learning algorithms , leading to significantly reduced error rates . In addition , we observed that electrical synapses are readily identified in ECS preserved tissue . Finally , we determined that antibodies penetrate deep into ECS preserved tissue with only minimal permeabilization , thereby enabling correlated light microscopy ( LM ) and EM studies . We conclude that preservation of ECS benefits multiple aspects of the connectomic analysis of neural circuits .
A number of recent technological advances have automated the collection of serial section electron microscopy ( EM ) data from the nervous system ( Briggman and Bock 2012 ) . Complete ( connectomic ) mapping of synaptic connectivity in these datasets , however , is hampered by the lack of automated analysis methods . The two most important goals for neuronal circuit reconstruction are the reliable reconstruction of neuronal morphologies and the identification of synapses . The delineation of morphologies has proven to be the most difficult step to automate . Current machine learning-based analysis methods enable semi-automated reconstructions ( segmentations ) of neurons that still require significant human effort to correct ( Helmstaedter et al . , 2013; Takemura et al . , 2013; Kim et al . , 2014 ) . Many of the errors encountered during the automated segmentation of neuronal morphologies are related to the dense packing of neurites in neuropil , which frequently leads to the artifactual merging of neighboring neurons ( Jain et al . , 2010 ) . Merging errors are particularly difficult to detect and correct , so most current approaches attempt to reduce merging errors by biasing the output of algorithms to generate over-segmentations of neurons ( Helmstaedter et al . , 2013; Takemura et al . , 2013; Kim et al . , 2014 ) . Over-segmentation splits neurons into many small objects that must then be reassembled into complete morphologies , a labor-intensive process . Here , we illustrate that automated segmentation error rates are improved by reducing the packing density of neurites in neuropil . The preparation of tissue for EM requires finding a compromise among several potential artifacts . One of the most significant , but often underappreciated , artifacts encountered during aldehyde-based fixation is the loss of extracellular space ( ECS ) . Although this artifact was described decades ago , it has been omitted from major texts describing the ultrastructure of the nervous system ( Peters et al . , 1991 ) . The artifact was first recognized by Van Harreveld and Malhotra ( 1967 ) and was convincingly demonstrated by comparing the appearance of aldehyde-fixed tissue with that of frozen tissue ( Van Harreveld and Steiner 1970a; 1970b; Harreveld and Fifkova 1975 ) . The ultrastructure of rapidly frozen tissue revealed appreciable ECS volume fractions of 15–25% , depending on the brain region ( van Harreveld and Khattab 1968; Harreveld and Fifkova 1975; Korogod et al . , 2015 ) , compared to less than 5% ECS in aldehyde-fixed tissue . The presence of large in vivoECS volume fractions was corroborated by complementary methods including brain conductivity measurements , tracer diffusion studies , and more modern high-pressure freezing experiments ( Rostaing et al . , 2004; Sykova and Nicholson 2008; Korogod et al . , 2015 ) . The primary cause of ECS loss is a net inward flux of ions during tissue fixation , which increases intracellular osmolarity , leading to a redistribution of water into cells and resulting in a swelling of cellular compartments ( Van Harreveld and Malhotra 1967 ) . Cragg subsequently developed a method to preserve ECS during chemical fixation using a simple ion substitution protocol that replaces extracellular sodium and chloride with a membrane-impermeant molecule such as sucrose prior to fixation ( Cragg 1979; 1980 ) , preventing the net inward ion flux that leads to swelling ( Figure 1A ) . 10 . 7554/eLife . 08206 . 003Figure 1 . ECS preservation in acutely isolated tissues . ( A ) A model of the changes brain tissue undergoes during chemical tissue fixation and possible interventions to preserve ECS . ( B ) ECS fraction of mouse olfactory bulb ( x ) , retina ( o ) , and cortex ( □ ) correlates with increasing fixative vehicle concentration . Dashed line is the linear fit to olfactory bulb data , solid line is the linear fit to cortex data ( olfactory bulb: R = 0 . 80 , p = 0 . 0006; cortex: R = 0 . 79 , p = 0 . 0113 ) . ( C–F ) EM images from the external plexiform layer of the mouse olfactory bulb from a perfused brain ( C ) or from acute sections fixed with increasing buffer concentrations ( D–F ) . ECS fractions are 0 . 6% ( C ) , 5 . 8% ( D ) , 11 . 3% ( E ) , and 23 . 9% ( F ) . Scale bars: 1 µm ( C–F ) . ECS: Extracellular space; EM: Electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00310 . 7554/eLife . 08206 . 004Figure 1—figure supplement 1 . Change of tissue dimensions and uniformity of ECS preserved sections . ( A ) Vibratome sections were cut from the cerebral cortex of acutely isolated brains ( for ECS preservation ) or from perfused animals . Section thickness was nominally 500 µm as determined by the slice thickness setting of the vibratome . Sections were stained for EM and the cross-sectional thickness measured from EM images ( mean +/- SEM; n = 20 slices from 6 animals , one-way ANOVA: p = 0 . 46 ) . ( B ) Sample images from the edges ( Bi , Biii ) and center ( Bii ) of a 500-µm-thick section of the mouse olfactory bulb . ECS is evenly preserved across the section . ECS: Extracellular space; EM: Electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00410 . 7554/eLife . 08206 . 005Figure 1—figure supplement 2 . Effect of temperature on the preservation of ECS . ( A , B ) Vibratome sections were cut from the olfactory bulbs of acutely isolated brains at either 4°C or 25°C and then fixed at 25°C in 0 . 1 M CB + 2% GA . ( C , D ) Vibratome sections were cut from the olfactory bulbs of acutely isolated brains at either 4°C or 25°C and then fixed at 25°C in 0 . 2 M CB + 2% GA . ECS preservation was similar under both conditions indicating the temperature of solutions during tissue dissection is not a major determinant for ECS preservation . Scale bars: 1 μm . ECS: Extracellular space; EM: Electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 005 Because ECS preservation results in a sparser packing of neurites , we hypothesized that error rates from the automated segmentation of ECS-preserved tissue would be reduced . We tested this hypothesis in the study presented here . Our goal was not to approximate the actual in vivo distribution of ECS . Rather , we reasoned that any manipulation of the ultrastructure that improved image segmentation , including perhaps exaggerating ECS volume fractions beyond physiological ranges , would be beneficial for connectomics so long as the manipulation did not alter synaptic connectivity . By utilizing an ECS-preserving chemical fixation protocol , we were able to titrate the ECS volume fraction in a variety of brain regions . We measured a significant reduction in automated segmentation error rates as the ECS fraction increased , indicating that ECS preservation improves the ability to automatically reconstruct neuronal morphologies from EM data . We then explored the second important goal of connectomics , the reliable identification of synapses . Electrical synapses are particularly difficult to identify in conventionally prepared ( swollen ) tissue ( Rash et al . , 1998 ) . We focused on identifying electrical synapses in a known retinal circuit and found that gap junctions were readily identifiable in ECS-preserved tissue , again due to the increased separation between neurites . Finally , we demonstrate that ECS preservation leads to significant improvements in the diffusion of antibodies into tissue sections with minimal disruption of ultrastructure , an important prerequisite for correlating immunofluorescence with EM . Therefore , we conclude that there are multiple practical benefits to preserving ECS that improve the connectomic analysis of neuronal circuits .
Cragg described an ECS preservation protocol for chemical immersion fixation that involved briefly bathing acute slices in an unbuffered sucrose solution and then fixing sections in a sucrose/phosphate buffer with glutaraldehyde . We initially applied Cragg’s protocol to acute brain sections of the mouse cortex , hippocampus and olfactory bulb and , while ECS was preserved , membranes appeared excessively wrinkled . We therefore developed an alternative strategy to preserve ECS by simply varying the osmolarity of the fixative buffer . Because the loss of ECS during chemical fixation ultimately is due to an imbalance in osmolarity across cellular membranes , we reasoned that adjusting the osmolarity of the fixation buffer would enable us to control the ECS fraction in fixed tissue ( Figure 1A ) . It is known that buffer molecules rather than fixative molecules contribute most to the physiological osmotic pressure ( Young 1935; Bone and Ryan 1972 ) . Overall , we noticed smoother membranes using the osmolarity-based approach compared to the sucrose/phosphate protocol . We observed that the preserved ECS fraction was positively correlated with the osmolarity of the fixation buffer ( Figure 1B , C–F ) . For example , a volume fraction of 23 . 9% in the external plexiform layer ( EPL ) of the olfactory bulb was observed ( Figure 1F ) when fixed with 2% glutaraldehyde buffered with 175 mM sodium cacodylate buffer ( CB , 343 mOsm ) compared to 5 . 8% ECS when fixed with 100 mM CB ( 204 mOsm , Figure 1D ) . By comparison , fixing tissue by transcardial perfusion , in which the blood–brain barrier remains intact , consistently yielded <1% ECS in the EPL ( Figure 1B , C ) . We observed that we could achieve unrealistically large ECS fractions of >25% in the olfactory bulb by raising the buffer osmolarity above 400–450 mOsm , but we began to observe significant membrane ruptures and excessive cellular shrinkage under these conditions . To examine whether overall tissue dimensions changed ( either swelled or shrinked ) as a function of the preserved ECS fraction , we measured the cross-sectional thickness of 500 µm cortical sections by EM ( Figure 1—figure supplement 1 ) . We did not observe a correlation between the fixative osmolarity and section thickness ( p=0 . 46 , one-way ANOVA ) , indicating macroscopic tissue dimensions are not affected by osmolarity-based ECS preservation . We also noted that ECS was preserved uniformly across 500-µm-thick sections , the maximum thickness that we tested ( Figure 1—figure supplement 1 ) . Preservation of ECS was similar whether tissue was sectioned at 4°C or 20°C ( Figure 1—figure supplement 2 ) , indicating the described protocol is suitable if alterations of neuronal morphology due to cold shock are a concern ( Kirov et al . , 2004; Bourne et al . , 2007 ) . We repeated the above experiments for various brain regions including the cerebral cortex and obtained similar results ( Figure 1B ) . The only exception was the mouse retina , for which we were unable to preserve ECS by changing the osmolarity of the fixative buffer . Replacing the fixative buffer with unbuffered sucrose , although , restored our ability to preserve ECS in an osmolarity-dependent way ( Figure 1B ) . One possible explanation for this is that cellular membranes in the retina are permeable to the cacodylate anion ( 138 g/mol ) , but not to sucrose ( 342 g/mol ) . This observation indicates that the protocol requires fine-tuning for different brain regions , but we were able to control ECS preservation in every region we have studied . Given the ability to predictably control ECS volume fractions , we next investigated whether the increased separation between neighboring neurites in ECS-preserved data would lead to lower error rates in the automated analyses of EM data . Qualitatively , we observed that ECS-preserved data are far easier for human annotators to analyze ( e . g . to trace neurons over long distances to create skeleton representations ) . We therefore asked whether such data also were quantitatively easier to segment by automated machine learning algorithms . We used 2D EM images from the EPL of the mouse olfactory bulb as representative samples containing densely packed neuropil . Samples were prepared by perfusion fixation , yielding an ECS fraction of 0 . 6% , or by acute slice fixation as described above , yielding ECS fractions of 5 . 8% , 11 . 3% , and 23 . 9% ( Figure 2A ) . We then manually labeled a portion of these images ( Figure 2A , second column ) and used the labeled data to train convolutional neural networks ( CNNs , see 'Materials and methods' ) . The networks were trained to classify each pixel as belonging either to intracellular space ( including cytosol and organelles ) , to plasma membrane , or to ECS . 10 . 7554/eLife . 08206 . 006Figure 2 . Automated 2D segmentation of extracellular space preserved data . ( A ) Four raw EM images ( first column ) of varying ECS fractions ( 0 . 6 , 5 . 8 , 11 . 3 , and 23 . 9% ) were analyzed . The lowest ECS fraction ( 0 . 6% ) is the perfused preparation while the others are the acute slice preparations . Pixels in each image were annotated ( second column ) as either intracellular ( white ) , extracellular ( gray ) , or plasma membrane ( black ) . Representative test image pixel classifications ( third column ) yielded network classification errors of 9 . 3 , 8 . 1 , 6 . 7 , and 7 . 2% for the four examples shown . Intracellular pixel probabilities were thresholded and segmented ( fourth column ) to yield minimum Rand errors of 0 . 11 , 0 . 06 , 0 . 06 , and 0 . 07 for the four examples shown . ( B ) Pixel classification errors plotted versus ECS fraction from an ninefold cross-validation , median values in red , shows no significant difference between perfused and acute preparations ( K-W test: p=0 . 05 ) . ( C ) Rand error plotted versus ECS fraction for the ninefold cross-validation , median values in red , shows a significant difference between perfused and acute preparations ( K-W test: p<0 . 01; Wilcoxon rank-sum test: p=0 . 0017 [5 . 8%] , p=0 . 0091 [11 . 3%] , p=0 . 031 [23 . 9%] ) . ( D ) Total number of splits and merger per object in each test image plotted versus ECS fraction for the ninefold cross-validation , median values in red , shows a significant difference between perfused and acute preparations ( K-W test: p<0 . 001; Wilcoxon rank-sum test: p=0 . 00049 [5 . 8%] , p=0 . 00035 [11 . 3%] , p=0 . 00035 [23 . 9%] ) . Scale bars: 2 μm . p-Values: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . ECS: Extracellular space; EM: Electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00610 . 7554/eLife . 08206 . 007Figure 2—source data 1 . Raw image and expert labels of 0 . 6% ECS data . TIFF stack viewable using ImageJ . ECS: Extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00710 . 7554/eLife . 08206 . 008Figure 2—source data 2 . Raw image and expert labels of 5 . 8% ECS data . TIFF stack viewable using ImageJ . ECS: Extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00810 . 7554/eLife . 08206 . 009Figure 2—source data 3 . Raw image and expert labels of 11 . 3% ECS data . TIFF stack viewable using ImageJ . ECS: Extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 00910 . 7554/eLife . 08206 . 010Figure 2—source data 4 . Raw image and expert labels of 23 . 9% ECS data . TIFF stack viewable using ImageJ . ECS: Extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01010 . 7554/eLife . 08206 . 011Figure 2—figure supplement 1 . Warping error of automated segmentations . ( A ) Split/merger curves parameterized by threshold of the network output . Shaded area is the standard error of the mean , lines are the mean number of splits/mergers per object , and the circles are the mean of the minimum total splits + mergers per object . ( B ) Network segmentations at the minimum warping error threshold , same regions as in Figure 2A . Warping errors were 0 . 68 ( 0 . 6% ECS ) , 0 . 32 ( 5 . 8% ECS ) , 0 . 52 ( 11 . 3% ECS ) , and 0 . 56 ( 23 . 9% ECS ) for the four examples shown . ( C ) Cumulative distributions of the number of ground truth intracellular pixels per object for the four datasets . The difference in these distributions motivated the use of a Bernoulli sampling procedure for calculating the adapted Rand Error metrics in Figure 2C . Scale bars: 2 μm . ECS: Extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 011 We then quantified the number of misclassified pixels , the classification error ( Figure 2A , third column ) , on test images . Similar pixel classification errors were observed regardless of ECS fraction ( Figure 2B ) , indicating that networks were equally able to learn the statistics of the images . We , however , were interested primarily in the degree to which ECS preservation affected the ability of neural networks to segment actual cells ( Jain et al . , 2010 ) . We therefore joined ( segmented ) regions of neighboring intracellular pixels and measured segmentation performance using two common segmentation metrics ( see 'Materials and methods' ) . We color-coded segmentations using a four-color map method ( Figure 2A , fourth column ) such that neighboring objects do not share a color . This approach helps to highlight the locations of two predominant error-types: mergers between neurons that should have been segmented as independent objects as well as the splitting of neurons into multiple objects . We observed a significant reduction in segmentation errors that correlated with increasing ECS fraction , regardless of the error metric we used ( Figure 2C , D , Figure 2—figure supplement 1 ) . Therefore , despite sharing similar pixel classification errors , images that contained some degree of ECS were easier to automatically segment . This is because ECS provides more separation between neighboring neurons leading to , for example , fewer mergers between cells . It remained possible that while 2D segmentation is improved in ECS data , 3D segmentation would not show a similar improvement perhaps because many ECS-preserved neurite diameters were smaller than their conventionally-fixed counterparts . We therefore collected 3D serial block-face scanning electron microscopy ( SBEM , Denk and Horstmann , 2004 ) data from two of the olfactory bulb samples shown in Figure 2A ( top and bottom rows ) . We collected 10 x 10 x 12 µm3 volumes from the 0 . 6% ECS tissue ( LowECS ) and the 23 . 9% ECS tissue ( HighECS ) at a voxel resolution of 9 . 8 x 9 . 8 x 25 nm3 ( Figure 3A , B ) ( Pallotto et al . , 2015 ) . We hand-labeled sub-volumes dispersed throughout the volumes as training data and then trained network architectures identical to those used for the 2D automated segmentation in order to classify each voxel . The resulting probability maps were then segmented to create automated labels as a function of threshold ( see 'Materials and methods' ) . 10 . 7554/eLife . 08206 . 012Figure 3 . Automated 3D segmentation of extracellular space preserved data . ( A ) A 10 x 10 x 12 μm3 3D SBEM cube ( left ) of perfused tissue ( LowECS ) , a dense skeletonization of all neurites in the volume ( middle ) , and an automated segmentation of the volume ( right ) . ( B ) A 10 x 10 x 12 μm3 3D SBEM cube ( left ) of ECS-preserved tissue ( HighECS ) , a dense skeletonization of all neurites in the volume ( middle ) , and an automated segmentation of the volume ( right ) . ( C ) Adapted Rand error of automated segmentations compared to dense skeletonizations as a function of the segmentation threshold for the HighECS ( brown ) and LowECS ( blue ) data volumes . ( D ) The fraction of merged skeleton nodes versus split skeleton edges as a function of network threshold . Solid points indicate the minimum sum of the error fractions . ( E ) The total error-free skeleton path length recovered as a function of network threshold . Shaded regions in C–E are the 99% confidence intervals based on a Bernoulli sampling of the data . Scale bars in A , B = 1 μm . SBEM: Serial block-face scanning electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01210 . 7554/eLife . 08206 . 013Figure 3—source data 1 . M0027_11 dense skeletonization , Knossos NML file . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01310 . 7554/eLife . 08206 . 014Figure 3—source data 2 . M0007_33 dense skeletonization , Knossos NML file . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01410 . 7554/eLife . 08206 . 015Figure 3—source data 3 . M0007_33 training data cubes . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01510 . 7554/eLife . 08206 . 016Figure 3—source data 4 . M0027_11 training data cubes . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01610 . 7554/eLife . 08206 . 017Figure 3—figure supplement 1 . Cumulative distributions for 3D hand-labeled volumes and skeletonizations of the LowECS ( blue ) and HighECS ( brown ) data volumes . ( A ) Cumulative distribution frequency ( CDF ) of number of intracellular ground truth voxels per object . The distribution medians are significantly different ( Wilcoxon rank-sum test , p <0 . 0001 ) . ( B ) CDF of path length per skeleton , medians 10 . 2 ( LowECS ) and 10 . 1 ( HighECS ) μm . The distributions are not significantly different ( two-sample Kolmogorov–Smirnov ( KS ) test , p = 0 . 91 ) . ( C ) CDF of the number of nodes annotated per skeleton , medians 16 ( LowECS ) and 17 ( HighECS ) nodes . The distributions not significantly different ( two-sample KS test , p = 0 . 77 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 017 To assess segmentation performance across the entire volumes , we densely skeletonized the neurites within each volume; these skeletons served as a ground truth ( GT ) measure of neurite continuity ( Figure 3A , B ) . Notably , the total number of skeletons ( LowECS=220 , HighECS=217 ) , the total skeletonized path length ( LowECS = 3 . 29 mm , HighECS = 3 . 40 mm ) and the median neurite path length ( LowECS = 10 . 2 [7 . 6 – 13 . 3] µm , HighECS = 10 . 1 [6 . 6 – 12 . 7] µm , median [IQR] ) were similar between the two datasets . The distributions of neurite path lengths were not significantly different between the two datasets ( Kolmogorov–Smirnov test , p=0 . 91 ) . Together these measurements indicate that the basic statistics of 3D neurite continuity are not altered for the HighECS data compared to the LowECS perfused tissue ( Figure 3—figure supplement 1 ) . We then varied the probability threshold for each network ( see 'Materials and methods' ) and measured the Rand error of the resulting segmentations at the node locations of the dense skeletons . Similar to the 2D results , we observed a significantly lower Rand error for the HighECS dataset ( Figure 3C ) . The threshold level that led to the minimum error rate was also lower for the HighECS dataset , indicating that fewer mergers were encountered compared to the LowECS data . We also measured the number of merged skeleton nodes and split skeleton edges as a function of network threshold ( Figure 3D ) . The two datasets yielded similar minimum edge split error rates , but the LowECS data achieved a lower node merger error rate . Finally , we measured the total error-free skeleton path length that was recovered as a function of network threshold . We recovered 40% and 47% of the total GT path length for the LowECS and HighECS datasets , respectively ( Figure 3E ) . We note that the error metrics were calculated based on a watershed of the network output and that no additional steps to improve the segmentations , such as supervoxel agglomeration ( Jain et al . , 2011; Nunez-Iglesias et al . , 2014 ) , were applied . In summary , the improvements in automated segmentation that were observed in 2D images were also found in a 3D analysis of ECS preserved compared to perfused tissue . We next explored whether ECS preservation would improve the identification of synaptic contacts between neurons . Chemical synapses indicated by clusters of presynaptic vesicles are relatively easy to identify in most brain regions . The identification of electrical synapses ( gap junctions ) , however , remains difficult even in the highest resolution electron micrographs . The number of missed electrical synapses based on thin section images has been estimated to be as high as 75–95% ( Rash et al . , 1998 ) . Unless gap junctions are captured in cross-section in an electron micrograph , electron tomography is usually required to positively identify a gap junction ( Rash et al . , 1998 ) . But performing electron tomography on every putative gap junction in a large EM volume would prohibitively slow data acquisition of even modestly large volumes . We hypothesized that gap junctions would be easier to identify in ECS-preserved tissue given the reduction in incidental ( non-synaptic ) contacts between cells . We tested this idea in the well-studied AII amacrine cell circuit of the mammalian retina ( Demb and Singer 2012 ) . The AII amacrine cell receives ribbon-type chemical synapses from rod bipolar cells and conventional chemical synapses from other amacrine cells . Importantly , it is known that AIIs are coupled electrically to each other as well as to the terminals of ON cone bipolar cells by gap junctions ( Hartveit and Veruki 2012 ) . We therefore sought to identify gap junctions on a portion of an AII amacrine cell reconstructed from 3D SBEM data . We volumetrically reconstructed an AII amacrine cell and then annotated all contacts onto a distal portion of its dendritic tree . Contacts were identified as membranes closely apposed ( within 50 nm ) to the AII’s dendrite ( n = 171 total contacts ) . For the majority of these contacts , we observed a clear cleft between the apposed membranes ( n = 150 cleft contacts ) , but , as well , we noted many instances in which we could not observe any cleft ( n = 21 tight contacts ) ( Figure 4A–D , see also Supplemental Data stacks ) . For each contact , we then traced the cell forming the contact through the 3D SBEM volume and classified each cell into four categories: AII amacrine cell , ON cone bipolar cell , presynaptic amacrine cell , or other ( see 'Materials and methods' ) . The ‘other’ category included a variety of cell types including bipolar , amacrine , ganglion and glial cells that , although making cleft contact with the AII , formed no obvious synapse ( e . g . no presynaptic clouds of vesicles were observed ) . Of the 150 cleft contacts , 13 were attributed to chemical synapses from amacrine cells . The remaining 137 cleft contacts assigned to the ‘other’ category were deemed to be incidental contacts or ribbon-type synapses ( Figure 4E , F ) . 10 . 7554/eLife . 08206 . 018Figure 4 . Extracellular space aids the identification of gap junctions . ( A ) Reconstructed dendrites of an AII amacrine cell ( gray ) and surface renderings of tight ( yellow ) and cleft ( black ) contacts . ( B-D ) Example EM images and volumetric reconstructions of ( B ) an AII ( gray ) to AII ( blue ) tight contact , ( C ) an AII ( gray ) to cone bipolar cell ( green ) tight contact ( C ) , and ( D ) a cleft contact identified as a chemical synapse between a presynaptic amacrine cell ( cyan ) and the AII cell ( gray ) . ( E ) Tight contact surface renderings color-coded by the identity of the synaptic partner: AII amacrine cell ( blue ) , cone bipolar cell ( green ) , presynaptic amacrine cell ( cyan ) . ( F ) Summary of the number of contacts color-coded by the cell type of the contacting cell: AII ( blue ) , cone bipolar ( green ) , presynaptic amacrine cell ( cyan ) , other ( black ) . Scale bars: 2 µm ( A , E ) , 1 µm ( B–D ) . EM: Electron microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01810 . 7554/eLife . 08206 . 019Figure 4—source data 1 . Example image stack of an A2 to A2 tight contact . TIFF stack viewable using ImageJ . Slice #128 in the stack indicates the location of the tight contact . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 01910 . 7554/eLife . 08206 . 020Figure 4—source data 2 . Example image stack of a cone bipolar to A2 tight contact . TIFF stack viewable using ImageJ . Slice #128 in the stack indicates the location of the tight contact . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 02010 . 7554/eLife . 08206 . 021Figure 4—source data 3 . Example image stack of a chemical synapse to A2 cleft contact . TIFF stack viewable using ImageJ . Slice #128 in the stack indicates the location of the cleft contact . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 02110 . 7554/eLife . 08206 . 022Figure 4—figure supplement 1 . Simple measures of contact geometry do not predict gap junctions . ( A ) Surface area and the volume of the convex hull bounding each contact are plotted for cleft contacts ( gray ) , AII amacrine cell tight contacts ( blue ) and cone bipolar cell tight contacts ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 022 We then asked whether the tight contacts were actually indicative of gap junctions ( electrical synapses ) onto the AII amacrine cell . Of the 21 tight contacts we annotated , 4 were formed with a neighboring AII amacrine cell and 17 were formed with ON cone bipolar cell terminals ( Figure 4E , F ) . Thus , in every instance , the tight contact was formed with a cell type known to be electrically coupled to the AII amacrine cell indicating that tight contacts in ECS preserved data are consistent with gap junctions . Moreover , AII-AII gap junctions were found in sublamina 5 of the IPL , and AII-ON cone bipolar gap junctions were found in sublaminae 3 and 4 , as previously reported ( Strettoia et al . , 1992 ) . We noted that the gap junctions were not identifiable based on contact geometry measurements , such as surface area , alone ( Figure 4—figure supplement 1 ) . The preservation of ECS therefore enabled the identification of gap junctions due to the relative scarcity of contacts amongst cells and the simple segregation of contacts into tight versus cleft categories . Finally , we explored whether ECS preservation could improve the ability to correlate fluorescent immuno-labeling of proteins with EM ultrastructure . A long-standing problem with preparing brain tissue for immunohistochemical observation is the incomplete diffusion of antibodies into large tissue volumes . Antibodies are large proteins and therefore strong permeabilization of tissue is typically required for labeling thick tissue sections . Often , the degree of permeabilization required to label thick samples destroys the fine tissue ultrastructure . We hypothesized that the preservation of ECS would allow antibodies to penetrate deeper into tissue and would , therefore , require only minimal permeabilization . We tested this idea in the mouse retina using antibodies against the vesicular acetylcholine ( VAChT ) and GABA ( VGAT ) transporters , two antibodies that generate distinct labeling patterns in the retina . Following a fixation and minimal permeabilization procotol that we developed in order to preserve ECS and decent tissue ultrastructure ( see 'Materials and methods' ) , we incubated 200-µm-thick retinal sections with primary and secondary antibodies for relatively short durations ( 9 hr each ) and measured the labelling depth . We observed deeper penetration of both antibodies in ECS-preserved retina ( Figure 5A–C ) . VAChT penetrated 28 . 1 ± 2 . 4 µm in ECS-preserved samples versus 6 . 8 ± 0 . 5 µm in control samples; VGAT penetration was 16 . 9 ± 1 . 0 µm versus 6 . 1 ± 0 . 4 µm for ECS-preserved and control samples , respectively . The labeling patterns were consistent with the known expression patterns of the two transporters in the retina ( Koulen 1997; Johnson et al . , 2003 ) . We processed these same pieces of retina for EM and confirmed that the penetration depth correlated with the ECS fraction ( Figure 5D , 0 . 2% vs . 19 . 2% ECS fraction ) . Furthermore , the ultrastructure of the weakly permeabilized tissue was well preserved ( Figure 5D ) . 10 . 7554/eLife . 08206 . 023Figure 5 . Extracellular space preservation increases access to antibodies . ( A ) Flat-mounted retinas were immunolabeled for vACHT with and without ECS preservation . Retinas were then cross-sectioned and confocal images were taken to directly visualize penetration depth . ( B ) Same as A , but immunolabeled for vGAT . ( C ) Fullwidth half-maximal penetration distances ( mean +/- SEM ) for retinas with and without ECS preservation ( n = 20 retina slices from two animals , unpaired Student’s t-test , vACHT: ***p<0 . 0001 , vGAT: ***p<0 . 0001 ) . ( D ) EM images collected from the regions of interest in panel B indicated by dotted rectangles . Scale bars: 50 μm ( A , B ) , 1 μm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08206 . 023
The major bottleneck for automating connectivity mapping in large neuronal circuits is no longer data acquisition ( Briggman and Bock 2012 ) , but rather data analysis . Manual data analysis allows targeted reconstructions of small population of neurons ( Briggman et al . , 2011 ) but is unfeasible for dense reconstructions . Automating image segmentation is therefore necessary for any reasonable scalability; trained human annotators , however , will still be required to correct the results of automated image processing algorithms ( Helmstaedter et al . , 2013; Takemura et al . , 2013; Kim et al . , 2014 ) . Therefore , any reduction in the error rates of automated segmentations of large EM datasets will directly reduce the required human effort to correct , or ‘proofread’ , errors . Our findings indicate that preserving ECS aids the connectomic analysis of neuronal circuits in multiple ways . The loss of ECS during the conventional chemical fixation of brain tissue leads to densely packed neurites and tightly apposed plasma membranes that complicate automated analyses of EM data . We reasoned that preserving some degree of ECS would improve automated image segmentation methods due to the increased separation of non-synaptic membrane appositions . We emphasize that our goal was not to preserve in vivo-like ECS or neuronal volumes . Indeed , freezing tissue is likely the only way to approximate in vivo-like morphologies ( Korogod et al . , 2015 ) . Our goal was rather to measure image segmentation performance as a function of increasing ECS volume fractions . In other words , because changes in ECS are unavoidable during the chemical fixation of brain tissue , why not adjust the artifact to be more beneficial to us from a data analysis standpoint ? Indeed , it was possible that actually exceeding in vivo ECS volume fractions would lead to the best segmentation performance . On the contrary , we found that while preserving ECS improved segmentation performance , the performance did not continue to improve with increasing ECS fractions ( Figure 2C , D ) . One of the reasons for the diminishing returns on segmentation performance is the shrinkage of cells that accompanies increasing ECS fractions which makes segmenting thin processes , particularly glial sheets , more error prone . There , therefore , appears to be an optimal ECS fraction from the standpoint of image segmentation that balances the increased separation between neurites with the shrinkage of cellular compartments . Modest ECS fractions of 6% already significantly improve segmentation performance compared to fractions of <1% ( Figure 2C , D ) , typical of standard perfusion fixation protocols . Most current EM segmentation strategies rely on generating over-segmentations ( neurons are split into many objects ) to avoid accidental mergers between neighboring neurons . These over-segmented objects then are joined together algorithmically ( Jain et al . , 2011; Nunez-Iglesias et al . , 2013; 2014 ) and/or by human annotators ( Takemura et al . , 2013; Kim et al . , 2014 ) . The segmentation of ECS-preserved data results in fewer mergers , leading to better initial segmentations , and therefore should reduce the human effort required to generate accurate reconstructions . We modified a state-of-the-art CNN originally designed for natural image classification ( Krizhevsky et al . , 2012 ) to analyze EM data . The CNNs we trained achieved similar pixel classification error rates regardless of the ECS fraction of a sample ( Figure 2B ) , but segmentation errors were reduced substantially by increasing ECS fractions ( Figure 2C , D and Figure 3 ) . In other words , pixel classification errors have a more detrimental impact on segmentation when cells are tightly packed versus when they are separated by ECS . This basic observation should hold regardless of the particular analysis approach used ( Rigoll et al . , 2008; Jurrus et al . , 2010; Turaga et al . , 2010; Berning et al . , 2015 ) . Our analysis showed that the improved segmentation performance in 2D images also extends to 3D segmentation ( Figure 3 ) . In order to prepare tissue with the ECS volume fractions we desired , we modified an existing protocol for ECS preservation in chemically fixed tissues . The modification we made involved simply changing the osmolarity of the buffer used during fixation . This approach , however , was not successful in the mouse retina; we had to rather change to unbuffered sucrose fixation to preserve ECS in the retina . We have successfully preserved ECS uniformly across 500-µm-thick acute brain slices , the thickest slices we tested . High-pressure freezing is an alternative to chemical fixation that also preserves ECS ( Rostaing et al . , 2004; Sykova and Nicholson 2008 ) . The maximum slice thickness that is adequately preserved using this technique , however , is at most 200 µm: too thin to contain many neuronal circuits . The chemical preservation protocol we used is also far less expensive to implement than high-pressure freezing and is simple to evaluate in different brain regions and species . A 500-µm section is a sufficiently large volume to contain many of the local neuronal circuits in the brain , such as a cortical column . Because improvements in data acquisition speeds eventually will allow larger volumes to be acquired , extending uniform ECS preservation to , for example , a whole mouse brain is an important goal . A previous study described a transcardial perfusion-based approach to preserve ECS throughout the brain involving transient opening of the blood-brain barrier and ion replacement ( Cragg 1980 ) , but the uniformity of the ECS fraction across the brain was not evaluated . Given the benefits we observed , further exploration and validation of this whole-brain ECS preservation method is warranted . The unambiguous identification of gap junctions is a long-standing problem in the analysis of brain circuits by EM . When ECS is lost , conventional transmission EM with pixel resolutions of a few nm can resolve gap junctions only when they are oriented such that they are captured in cross-section; electron tomography is required to resolve gap junctions oriented obliquely to the image plane ( Rash et al . , 1998 ) . Performing electron tomography on each putative gap junction in a large tissue volume is impractical . We therefore determined whether ECS preservation aided gap junction detection at a moderate pixel resolution ( 10 nm ) using SBEM ( Denk and Horstmann , 2004 ) data . We found that subsets of all contacts onto an AII amacrine cell in the mouse retina are tight contacts with no visible cleft . We identified all these tight contacts as being formed with cells known to be coupled electrically to AIIs ( Figure 4 ) . Although we performed this analysis manually , we anticipate a relatively straightforward method for automating the detection of tight contacts between cells . Along the same lines , the automated identification of chemical synapses may also be aided simply because there are fewer incidental contacts between cells in ECS-preserved data . Importantly , the analysis of the AII amacrine cell circuitry demonstrates that the known synaptic connectivity is present in ECS-preserved tissue . Correlating fluorescence immunohistochemistry with EM data can add significant functional information to purely structural descriptions of circuits . However , there is usually a tradeoff between the degree of permeabilization required for antibody penetration and the quality of ultrastructural preservation . The preservation of ECS retains large diffusion pathways , allowing deeper penetration of large antibody proteins into tissue ( Figure 5 ) . The increased access to antibodies allowed for the use of a minimal permeabilization pre-embedding protocol that preserves satisfactory tissue ultrastructure for subsequent EM examination of antibody-labeled tissue . The ECS-preserved EM images of the brain we report here appear substantially different from the vast majority of published EM data over the last 50 years . Practitioners of rapid freezing-based tissue preservation , however , have appreciated the loss of ECS as a major artifact in conventional EM tissue processing for decades . We wish to , in particular , note the pioneering work of Anthonie Van Harreveld who first elucidated this artifact in EM and whose careful analyses directly led to the ECS-preserving interventions that we employed . The relatively minor changes to existing tissue fixation protocols that preserve ECS ultimately benefit several aspects of the connectomic analysis of neuronal circuits .
We fixed and examined tissue from a variety of brain regions of C57BL/6 mice , aged 9-12 weeks in accordance with NIH animal ethics guidelines . For ECS preservation in acute slices , mice were deeply anesthetized using a mixture of ketamine hydrochloride ( 100 mg/kg IP ) and xylazine hydrochloride ( 10 mg/kg IP ) or isoflurane ( Forane , Baxter , Deerfield , IL ) and then swiftly decapitated . We cut 300–500-µm slices on a vibratome ( Leica ) according to the procedure of ( Bischofberger et al . , 2006 ) . Slices were cut at 4°C or at 20°C in ACSF containing ( in mM ) : 124 NaCl , 3 KCl , 1 . 3 MgSO4 . 7H2O , 84 NaHCO3 , 1 . 25 NaH2PO4 . H20 , 20 glucose , 2 CaCl2 . 2H2O and equilibrated with a 95% O2 – 5% CO2 gas mixture . Slices were then transferred into the fixative solution at 20°C , containing 2% glutaraldehyde ( GA , Electron Microscopy Sciences , Hatfield , PA ) buffered ( pH 7 . 4 ) with one of the following sodium-cacodylate buffer ( CB , Sigma-Aldrich , St . Louis , MO ) concentrations ( in mM ) : 100 , 150 , 175 , or 200 . Slices were fixed for 8 hr , rinsed with CB buffer of the same concentration used for fixation for 4 hr and then stained for EM . For perfusion-fixed tissue , we followed the rapid perfusion approach of ( Tao-Cheng et al . , 2007 ) . Mice were transcardially perfused with 2% GA , 2% paraformaldehyde ( PFA , Electron Microscopy Sciences ) , in 150 mM CB buffer ( pH 7 . 4 ) . The brain was extracted and postfixed for 8 hr in the same fixative solution . Subsequently , the tissue was rinsed in 150 mM CB buffer for 4 hr . Tissue sections from the perfused brains were cut on a vibratome ( Leica , Germany ) at 300–500 µm and stained for EM . For isolated retinas , retinas were isolated from the eye cup in Ames solution ( Sigma-Aldrich ) , flat mounted on filter paper and then incubated for 5 min at 20°C in a modified Ames solution . The modified Ames solution was prepared by diluting Ames 1:1 with distilled water and then adding 5% sucrose . We then transferred retinas into a fixative at 20°C containing 2% GA in unbuffered sucrose ( for ECS preservation ) or 2% GA buffered with 150 mM CB ( pH 7 . 4 ) . Retinas were fixed for 1 hr , rinsed with 150 mM CB buffer and stained for EM . Retina tissue samples were taken from an eccentricity approximately halfway between the optic disk and the peripheral edge of the retina . We used the ROTO ( reduced osmium-thiocarbohydrazide-osmium ) en bloc staining protocol suitable for SBEM previously described ( Briggman et al . , 2011 ) . Briefly , samples were stained in a solution containing 1% osmium tetroxide , 1 . 5% potassium ferrocyanide , and 150 mM CB for 2 hr at room temperature . The osmium stain was amplified with 1% aqueous thiocarbohydrazide ( 1 hr at 50°C ) , and then 2% aqueous osmium tetroxide ( 1 hr at room temperature ) . The samples were then stained with 2% aqueous uranyl acetate for 12 hr at room temperature and lead aspartate for 2–12 hr at room temperature . Samples were embedded in Epon resin . All 2D EM images were acquired from ultrathin ( 50–100 nm ) sections mounted on copper TEM grids in a scanning electron microscope with a field-emission cathode ( Nova NanoSEM 450 , FEI Company , Netherlands ) using a solid-state back-scattered electron detector . Incident beam energies were 2 . 0–2 . 5 kV and pixel resolution was typically 9 . 8 nm . For quantification of ECS in 2D , we randomly selected 9 . 8 x 9 . 8 µm2 regions from EM images of dense neuropil and manually labeled ECS pixels . Labeling was performed blinded to the fixation conditions . We intentionally avoided annotating regions containing cells bodies or blood vessels that would distort ECS fraction estimates due to their large volumes . For olfactory bulb data , we collected images from the neuropil of the EPL . For retina data , we imaged the neuropil of the inner plexiform layer . For cerebral cortex , we imaged neuropil from layers 2/3 . The ECS percentage was measured as the fraction of labeled ECS pixels in the annotated region . Flat-mounted retinas were incubated in the modified ACSF solution for 5 min at 20°C and then fixed for 1 hr with 2% PFA + 0 . 01% GA in either 7 . 5% sucrose ( for ECS preservation ) or 150 mM CB ( pH 7 . 4 ) . Retinas were then rinsed 3 x 15 min in 150 mM CB at room temperature , embedded in 3% agarose prepared in 150 mM CB and vibratome sectioned into 200µm slices . The slices were cut at an eccentricity approximately halfway between the optic disk and the peripheral edge of the retina . Slices were rinsed in 50 mM glycine in 150 mM CB for 30 min , and in a 300 mOsm PB-BSA washing buffer containing 120 mM phosphate buffer ( pH 7 . 4 ) , 0 . 5% BSA ( Sigma-Aldrich ) , and 0 . 05% sodium azide ( Sigma-Aldrich ) for 2x 10 min . They were then transferred to a blocking solution containing 120 mM PB ( pH 7 . 4 ) , 1% BSA , 10% normal donkey serum ( NDS , Abcam , United Kingdom ) , 0 . 5% Tween 20 ( Sigma-Aldrich ) , and 0 . 05% sodium azide for 1 hr at 20°C . Primary antibody staining was performed on free floating agitated slices for 9 hr at 4°C with antibodies targeting either the vesicular GABA transporter ( Synaptic Systems , Germany , cat no . 131003 , rabbit anti-VGAT ) or the vesicular acetylcholine transporter ( Synaptic Systems , cat no . 139103 , rabbit anti-VAChT ) at dilutions of 1:250 and 1:300 , respectively , in the blocking solution except with 3% NDS ( instead of 10% NDS ) . Retinas were then rinsed in PB-BSA for 2 hr at 4°C and then stained with a donkey anti-rabbit fluorescent secondary antibody , DyLight 650 ( Abcam ) , at 1:300 dilution in the blocking solution ( with 3% NDS ) for 9 hr at 4°C . Retinas were then rinsed in PB-BSA 2x 10 min and transferred to 150 mM CB . The slices were then re-embedded in 3% agarose in 150 mM CB and cut in half to assay the degree of antibody penetration . Retinal cross-sections from the cut surface were imaged on a confocal microscope ( Carl Zeiss AG , Germany ) with a 633-nm laser and a LP650 emission filter . All images were acquired under identical confocal imaging parameters . The antibody penetration distance was measured by averaging the signal across the depth of the inner plexiform layer and quantifying the full-width half-maximum ( FWHM ) distance of the staining profiles from the two edges . The FWHM distances from both edges were averaged to yield one data point per tissue slice . Following confocal imaging , all retinas fixed again in 150 mM CB + 2% GA and were subsequently stained for EM using the method above . For the 3D segmentation analysis , we collected two volumes from samples obtained from the EPL of the mouse olfactory bulb containing 0 . 6% ( dataset M0027_11 , LowECS ) or 23 . 9% ( dataset M0007_33 , HighECS ) 2D ECS fractions . Data were collected using a custom serial block-face microtome designed by K . L . Briggman . The specimens were cut out of the flat-embedding blocks and re-embedded in Epon Hard on aluminium stubs . The samples were then trimmed to a block face of ~200 μm wide and ~300 μm long . The samples were imaged in a scanning electron microscope with a field-emission cathode ( NanoSEM 450 , FEI ) . Back-scattered electrons were detected using a concentric segmented back-scatter detector . The incident electron beam had an energy of 2 . 4 keV and a current of ~200 pA . Images were acquired with a pixel dwell time of 2 μs and size of 9 . 8 nm × 9 . 8 nm which corresponds to a dose of about 42 electrons/nm2 . Imaging was performed at high vacuum , with the sides of the blocks sputter coated with a 100-nm-thick layer of gold . The section thickness was set to 25 nm . Five hundred and twelve consecutive block faces were imaged from each sample , resulting in aligned data volumes of 4096 × 3536 × 512 voxels ( corresponding to an approximate spatial volume of 40 × 35 × 12 . 8 µm3 ) . Then , 10 × 10 × 12 µm3 regions were selected from the two volumes for the 3D segmentation analysis . For the gap junction analysis of the retina , a 73 . 2 x 27 . 6 x 57 . 6 µm3 ( 6100 x 2300 x 2300 voxels ) block of ECS-preserved mouse retina ( dataset k0731 , 17 . 4% ECS ) was acquired by SBEM at a resolution of 12 x 12 x 25 nm3 using a microtome designed by W . Denk . The dataset spanned the inner plexiform layer including the ganglion cell layer and a portion of the inner nuclear layer . The retina was cut out of the flat-embedding block and re-embedded in Epon Hard on aluminium stubs . The retina was then trimmed to a block face ~200 μm wide and ~200 μm long . The samples were imaged in a scanning electron microscope with a field-emission cathode ( UltraPlus , Carl Zeiss ) . Back-scattered electrons were detected using a custom-designed detector based on a special silicon diode ( AXUV , International Radiation Detectors , Torrance , CA ) combined with a custom-built current amplifier . The incident electron beam had an energy of 1 . 4 keV and a current of ~1 nA . Images were acquired with a pixel dwell time of 0 . 45 μs and size of 12 nm × 12 nm , which corresponds to a dose of ~31 electrons/nm2 . Imaging was performed at high vacuum , with the sides of the blocks sputter coated with a 100-nm-thick layer of gold . The section thickness was set to 25 nm . All data sets were split into cubes ( 128 × 128 × 128 voxels ) for viewing in KNOSSOS ( www . knossostool . org , Helmstaedter , et al . , 2011 ) . An AII amacrine cell was initially identified by tracing a cell that received a ribbon type synapse from a rod bipolar cell . A portion of this AII cell was volumetrically reconstructed manually using ITK-Snap ( www . itksnap . org , [Yushkevich et al . , 2006] ) . Contacts onto the surface of the AII from neighboring cells were manually annotated and each contact was classified as either a tight contact , based on the absence of any discernable gap between the two membranes , or a cleft contact for membranes that came within ~50 nm of each other . K . L . B and J . H . S . then independently skeletonized the cells forming these contacts using Knossos ( www . knossostool . org ) . Cells were skeletonized until each cell’s identity could be unambiguously determined as either: an AII amacrine cell ( based on visualizing connectivity with rod bipolar cell terminals ) , a cone bipolar cell ( based on presence of ribbon synapses and the absence of rod bipolar cell connectivity ) , a presynaptic amacrine cell ( based on presence of a conventional chemical synapse at the contact ) , or ‘other’ . The ‘other’ category included ganglion cell dendrites , amacrine cell dendrites , rod or cone bipolar cell axons not forming synaptic junctions , and Muller glial cells . None of the cells in the ‘other’ category formed a chemical synaptic contact onto the AII amacrine cell . For 2D segmentation comparisons , test and training images were obtained from EM images from the EPL of the mouse olfactory bulb containing 0 . 6% , 5 . 8% , 11 . 3% , or 23 . 9% ECS . A portion of each of these images ( 1920 x 1728 pixels , 18 . 8 µm x 16 . 9 µm ) was hand-labeled using ITK-Snap , and each pixel was classified as either: membrane ( MEM ) , ECS or intracellular space ( ICS ) . GT integer labels were obtained from these hand-labeled images by running connected components ( four connectivity ) on ICS pixels only . The labeled region was subdivided into nine sub-images of size 640 x 576 pixels to perform a ninefold cross-validation of neural network performance . For each EM image , nine independent CNNs were trained on eight of the sub-images and then tested on the remaining sub-image . For 3D segmentation comparisons , training cubes were obtained from the M0027_11 and M007_33 data volumes . We densely hand-labeled voxels from six sub-volumes ( each 128 x 128 x 128 voxels , 1 . 25 x 1 . 25 x 3 . 20 µm3 ) contained within the data volumes . Sub-volumes were labeled so that ICS voxels from distinct neurites had different integer labels and so that ECS voxels were all labeled with the same integer label . Manual labels were automatically corrected by , in order: ( 1 ) removing very small foreground ( ICS and ECS ) and background ( MEM ) components of less than 5 pixels in xy , xz , and yz ( orthogonal ) image slices ( eight connectivity ) ; ( 2 ) removing adjacency between ICS components with different labels in orthogonal image slices ( eight connectivity ) by setting pixels on each side of any adjacencies to MEM; ( 3 ) finding ICS components with the same label value that were not connected in 3D and assigning different label values to unconnected components ( six connectivity ) ; and ( 4 ) removing very small 3D ICS labels that contained less than 27 voxels . The M0027_11 and M0007_33 data volumes were densely skeletonized using Knossos for comparison with full 3D segmentations . Only neurons were skeletonized , glial processes ( identified based on cellular morphology ) and ECS were not skeletonized . We used a CNN with a parallelized GPU implementation ( Krizhevsky et al . , 2012; Zeiler and Fergus 2013 ) modified from ( https://github . com/akrizhevsky/cuda-convnet2 ) . Networks were trained on 3GB Nvidia GTX 780Ti and 6GB GTX TITAN Black graphics cards ( EVGA ) . The CNNs processed 2D images to learn labels for both 2D and 3D EM data . Training images were sampled from randomized pixels in the EM data such that the network was trained to learn , in parallel , the identity of all the pixels ( ICS , ECS , or MEM ) in a 16 x 16 pixel patch located in the center of 64 x 64 pixel input images ( Jain et al . , 2007; Ciresan et al . , 2012 ) . We trained seven-stage convolutional networks containing ( 1 ) an input convolutional layer with 96 kernels of size 7 , response normalization across 31 feature maps , and max pool subsampling of size 3 with stride of 2 ( overlapped pooling resulting in an overall downsampling factor of 2 ) ; ( 2 ) a convolutional layer with 256 kernels of size 5 and the same normalization and subsampling as the first layer; ( 3 , 4 ) two convolutional layers with 384 kernels of size 3; ( 5 ) a convolutional layer with 256 kernels of size 3 and the same subsampling as the first layer; ( 6 , 7 ) two fully connected hidden layers of size 4096 each with a dropout rate of 0 . 5 ( Hinton et al . , 2012 ) ; and ( 8 ) an output layer of size 768 ( 16x16 output pixels x 3 pixel identities ) . Each output represents the conditional probability that a particular output pixel belongs to one of ICS , ECS or MEM ( Bishop 1995 ) . Convolutional layer units used the ReLU nonlinearity ( Krizhevsky et al . , 2012 ) while fully connected hidden layer units used a linear activation function . The output layer used the logistic nonlinearity and gradients were calculated using a cross-entropy error function for multiple independent attributes ( Bishop 1995 ) . Average gradients were calculated and back-propagated for each mini-batch size of 128 image examples . Weights in the fully connected and output layers were updated with a weight decay of 0 . 02 and all weights and biases used a weight momentum term of 0 . 9 ( Krizhevsky et al . , 2012 ) . Randomized training images were augmented using simple transformations of the input image , which included all combinations of transposes and reflections for a total of eight possible augmentations ( in two dimensions ) ( Ciresan et al . , 2012 ) . For 2D image segmentations , nine networks were trained for each ECS dataset using a ninefold cross-validation described above . For 3D volume segmentations , five networks were trained for each volume using all six sub-volumes of training data . Three networks were trained using images taken from the xy planes of each volume ( based on the original orientation of the EM data ) . Two networks were trained by reslicing the original volumes to create xz and yz oriented images . The final voxel probabilities used to create EM segmentations were a weighted average of the five networks , with the probabilities from the xz and yz networks weighted at 0 . 5 relative to the xy networks . The randomized presentation of images to the network was balanced so that the center pixel ICS , ECS , or MEM identities were presented with equal probability . Because this procedure changes the prior probabilities and does not guarantee the same equal balancing for all 256 ( 16 x 16 ) output pixels , the output probabilities of test images were reweighted using Bayes’ rule ( Vucetic and Obradovic 2001 ) . The test probabilities ( the target prior ) represented the actual frequency of ICS , ECS and MEM voxels over all training data . Training frequencies used in the reweighting procedure were measured by tallying the count of ICS , ECS , and MEM pixels that were actually presented to each of the 256 outputs during training . Networks were trained using 0 . 8 million images per training epoch . Learning rates for all weights were initialized at 0 . 00005 and for all biases at 0 . 0001 . Different randomizations of the training data were presented during each epoch . The learning rate decayed exponentially but discretely 16 times per epoch with a time constant of 5 epochs . All networks were run for 10 training epochs . Network output probabilities for each output were exported for test voxels and the output was classified as either MEM , ECS , or ICS based on the maximum probability ( winner-take-all ) . To create segmented labels , we used a customized watershed procedure that relied on iteratively running connected components to find local peaks within the ICS and ECS CNN probability outputs for each pixel . ICS and ECS probabilities were set to zero if they were not the winning probability , and then probabilities were binarized at increasing thresholds . For 2D segmentations the thresholds were 0 . 3 , 0 . 4 , 0 . 5; then from 0 . 5 to 0 . 99 in steps of 0 . 01; and then 0 . 995 , 0 . 999 , 0 . 9995 , 0 . 9999 , 0 . 99995 , 0 . 99999 , 0 . 999995 , and 0 . 999999 . For 3D segmentations thresholds were 0 . 3 to 0 . 9 in steps of 0 . 1; then 0 . 95 , 0 . 975 , 0 . 99; and then 0 . 995 , 0 . 999 , 0 . 9995 , 0 . 9999 , 0 . 99995 , 0 . 99999 , 0 . 999995 , and 0 . 999999 . At each binarization step , connected components ( four connectivity for 2D , six connectivity for 3D ) were created for all pixels above the threshold . Any components that fell below a minimum size threshold parameter ( Tmin ) of 64 pixels for 2D or 256 voxels for 3D were accumulated in a binary mask . Tests for the sensitivity of this parameter revealed that it had less than 1% impact on all metrics measured across the ranges tested ( Tmin = 8 , 16 , 32 , 64 , 128 , 256 ) . This step essentially was a method for detecting local peaks in the probabilities . Moving from low to high thresholds , connected components was run on the logical OR of the accumulated binary mask and the remaining binarized pixels above the current threshold level . These components were then dilated toward filling in the surrounding region of voxels with the matching winner-take-all identity using a topological warping technique that does not allow components to become split or merged ( Kong and Rosenfeld 1989; Legland et al . , 2011 ) . Any remaining ICS or ECS winner-take-all voxels were merged to the nearest ( Euclidean distance ) ICS or ECS component , respectively . For display of 2D segmentations , we used four colors to label the different components so that no neighboring components were the same color ( Appel and Haken 1977 ) . Pixel categorization error was calculated as the fraction of correctly labeled pixel identities from the CNNs ( winner-take-all max probability out of ICS , ECS , MEM ) compared to GT test images . To compare components of automated segmentations and GT segmentations , we used the Adapted Rand Error ( ARE ) and warping error to compare ICS components only . The ARE was calculated as 1 minus the F-score of the precision and recall of the Rand Index ( Hubert and Arabie 1985; Jain et al . , 2010 ) . In order to account for significantly different GT component sizes ( pixels per component ) , which can bias the ARE when comparing different ECS datasets , we applied a Bernoulli sampling procedure that sampled ( without replacement ) GT ICS components and pixels within those components . We used the minimum number of components ( n = 71 ) across the 36 possible sub-images ( four ECS fractions x nine sub-images ) as the expected number of components selected on each sampling iteration . A portion of pixels within each of the sampled components was then selected . We used the minimum number of pixels per component ( n = 20 ) as the expected number of pixels selected from each component . The ARE was then calculated using only the selected pixels for a particular sampling iteration . We used the median ARE of 1000 samples as the final metric for this procedure . We also compared 2D segmentations using the warping error ( Jain et al . , 2010 ) . Difference pixels were classified by the type of topologic error that they represented , and only split and merger pixels were counted . Connected split and merger pixels ( i . e . , belonging to the same components being split or merged ) were counted as a single split or merger by running connected components ( eight connectivity ) on pixel errors . We divided the number of splits and mergers by the number of GT components in each test image to fairly compare between datasets ( yielding splits per GT component and mergers per GT component ) . For 3D segmentations , we compared automated segmentations with the dense skeletonization for each data volume . Using the skeletons as GT , we calculated: ( 1 ) a skeleton node-based ARE; ( 2 ) the fraction of split edges and merged nodes and ( 3 ) the total error-free path length ( TEFPL ) . Overlap between skeletons and segmented 3D supervoxels was calculated as a confusion matrix , where rows correspond to skeletons and columns to supervoxels . Each entry in the matrix is the number of nodes on a given skeleton that are contained within a particular supervoxel . The sum total of all entries is therefore equal to the total number of skeleton nodes ( n=5130 LowECS , n=5364 HighECS ) . The ARE was then calculated based on this confusion matrix using the same calculation as that used for pixel-wise 2D segmentation . Bernoulli sampling was not required in this case as the number of nodes per skeleton was not statistically different between the datasets ( Figure 3—figure supplement 1A ) . Any nodes that overlapped with membrane areas of the segmentations were represented in the confusion matrix as belonging to a single membrane supervoxel . Both ICS and ECS supervoxels were used in creating the confusion matrix . Overall , only a very few number of skeleton nodes fell into either membrane ( n=8 , LowECS; n=12 , HighECS ) or ECS supervoxel locations ( n=0 , LowECS; n=1 , HighECS ) . We also used the supervoxel/skeleton confusion matrix to calculate the fraction of merged nodes . Summing along the rows of the logical confusion matrix ( true at locations with at least one node ) creates a marginal that indicates which supervoxels overlap with more than one skeleton . All nodes that are within these supervoxels were then counted as merged nodes . The number of merged nodes can then range from zero to the total number of nodes ( if all nodes were part of a single supervoxel ) , which we divided by to get the fraction of merged nodes . To calculate the fraction of split edges , we iterated over the edges of skeletons . An edge was split if both nodes belonging to the edge were located in different supervoxels . If either node was located in a membrane area , the edge was also considered as being split . The number of split edges can then range from zero to the total number of edges ( if all supervoxels only contained a single node ) , which we divided by to get fraction of split edges . To calculate TEFPL , an edge was error free if it was not split and if neither of its nodes was a merged node . We summed the path lengths of all error free edges to get TEFPL , which was expressed as a percentage of the total path length in each volume . All statistical analyses were performed using GraphPad Prism ( GraphPad Software ) or Matlab ( The Mathworks ) . For comparison between two groups of equal size , the non-parametric Kruskal–Wallis test was applied to assess whether there was any significance between medians , then the Wilcoxon rank-sum test was used for pairwise comparisons in the cases where the Kruskal–Wallis test was significant ( Figure 2B–D ) . For comparison between two groups of equal size , unpaired t-tests were used ( Figure 5C ) . For correlation coefficients , the Student's t-test was used with n-2 degrees of freedom ( Figure 1B ) . Confidence intervals for 3D skeleton-based error metrics were calculated using a Bernoulli sampling with an expected value of 95% of the minimum number of skeletons ( n = 206 ) . | The brain consists of billions of neurons that are connected into many different circuits . Mapping the connections between these neurons could help researchers to understand how the nervous system works . A method commonly used to do so is to preserve samples of brain tissue in chemical fixatives , and then image thin slices of this tissue using powerful microscopes . As each tissue sample contains many neurons , computer algorithms have been developed to analyze the microscope images and automatically identify the neurons and the connections they make . However , these algorithms often make 'segmentation errors' that researchers need to manually correct: for example , overlapping neurons may be counted as a single neuron , or a neuron may be marked into several segments . Correcting these errors is a time-consuming and tedious task that limits how much of the brain can be currently mapped . Future algorithm improvements will hopefully reduce the number of errors; Pallotto , Watkins et al . explored an alternative approach by making the images themselves easier to analyze using existing algorithms . The chemicals used to preserve brain tissue often suck out the fluids that fill the spaces between the neurons , causing these 'extracellular spaces' to shrink . Pallotto , Watkins et al . have now developed a method of preserving tissue that maintains more space between the neurons , and used this method to preserve samples of mouse brain with different amounts of extracellular space . Pallotto , Watkins et al . found that the algorithm used to analyze the images of these samples made far fewer segmentation errors on samples that contained more extracellular space . It was also easier to identify the connections between different neurons in these samples . The next challenge will be to extend these methods to preserving extracellular space across whole brains . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Extracellular space preservation aids the connectomic analysis of neural circuits |
Maintenance of a quiescent and organotypically-differentiated layer of blood vessel-lining endothelial cells ( EC ) is vital for human health . Yet , the molecular mechanisms of vascular quiescence remain largely elusive . Here we identify the genome-wide transcriptomic program controlling the acquisition of quiescence by comparing lung EC of infant and adult mice , revealing a prominent regulation of TGFß family members . These transcriptomic changes are distinctly accompanied by epigenetic modifications , measured at single CpG resolution . Gain of DNA methylation affects developmental pathways , including NOTCH signaling . Conversely , loss of DNA methylation preferentially occurs in intragenic clusters affecting intronic enhancer regions of genes involved in TGFβ family signaling . Functional experiments prototypically validated the strongly epigenetically regulated inhibitors of TGFβ family signaling SMAD6 and SMAD7 as regulators of EC quiescence . These data establish the transcriptional and epigenetic landscape of vascular quiescence that will serve as a foundation for further mechanistic studies of vascular homeostasis and disease-associated activation .
Endothelial cells ( EC ) are long-lived cells of the mesodermal lineage that line the inside of all blood vessels forming a single layer of organotypically-differentiated cells ( Augustin and Koh , 2017; Deanfield et al . , 2007; Rafii et al . , 2016; Regan and Aird , 2012 ) . Unlike unidirectional differentiation programs of short-lived epithelial cells , EC acquire a quiescent state during the transition to adulthood from which they can switch back to the activated phenotype resulting in disease-associated angiogenesis ( Hobson and Denekamp , 1984; Marcelo et al . , 2013 ) . In fact , the maintenance of the quiescent vascular phenotype is of pivotal importance for human health . Loss of this EC phenotype is a hallmark of vascular dysfunction and a commonality of diverse life-threatening diseases ranging from sepsis , atherosclerosis to cancer ( Carmeliet and Jain , 2011; Gimbrone and García-Cardeña , 2016 ) . Likewise , major and socioeconomically important chronic diseases are characterized by a loss of the quiescent EC phenotype , like age-dependent macular degeneration and diabetic retinopathy . Vascular targeted therapies for such conditions are mostly aimed at interfering with EC activation programs ( Krzystolik et al . , 2002; Tanabe et al . , 2017 ) . Therefore , it is crucial to get a better understanding of the pathways involved in acquisition of vascular quiescence to improve vessel targeted therapies . While the mechanisms of sprouting angiogenesis , network formation and remodeling have been molecularly unraveled in increasing detail , the molecular mechanisms of acquisition of EC quiescence are still poorly understood ( Adams and Alitalo , 2007; Herbert and Stainier , 2011; Jain , 2003; Korn and Augustin , 2015 ) . EC quiescence is an active process , as it not only involves the absence of activators but also microenvironmental factors as well as cell intrinsic mechanisms . Matricellular factors , contact with pericytes and hemodynamic forces act as microenvironmental determinants of vascular maturation ( Armulik et al . , 2011; Baeyens et al . , 2016; Davis et al . , 2015; Stratman et al . , 2017; Yousif et al . , 2013 ) . To switch from an activated to a resting phenotype , numerous signaling pathways in EC including VEGF , NOTCH , FGF , TGFβ , angiopoietin and semaphorin signaling need to be properly regulated ( Marcelo et al . , 2013; Patel-Hett and D'Amore , 2011 ) . To add another layer of complexity , there is an intricate crosstalk between these pathways and their integration is not completely understood ( Ehling et al . , 2013; Holderfield and Hughes , 2008; Kim et al . , 2011 ) . Furthermore , while some pathways are described to have clear pro- or anti-angiogenic effects , the impact of other signaling pathways is rather contextual or even still controversially discussed . For instance , TGFß family signaling has been described to exert both , pro-and anti-angiogenic functions and to contextually cooperate with NOTCH signaling in the regulation of angiogenesis ( Cai et al . , 2012; Dyer et al . , 2014; Goumans et al . , 2003; Goumans et al . , 2002; Larrivée et al . , 2012; Mallet et al . , 2006; Mouillesseaux et al . , 2016; Suzuki et al . , 2008 ) . Thus , better molecular understanding of the way crucial EC-associated signaling pathways are regulated during the post-angiogenic acquisition of vessel quiescence will yield better mechanistic insights into vascular function during health and disease . The mechanisms of EC activation during health and disease have been studied in substantial detail on the expression level of individual genes or pathways as well as genome-wide , for example genome-wide screens of regulators of angiogenesis or inflammation ( Cannon et al . , 2013; del Toro et al . , 2010; Massaro et al . , 2015; Zhang et al . , 2012 ) . Yet , no unbiased genome-wide systems biological effort from in vivo isolated EC has been made so far to identify the transcriptomic signatures associated with the acquisition and maintenance of EC quiescence . DNA methylation acts in concert with other epigenetic marks in defining chromatin states which create gene active and silenced compartments in the genome ( Jones , 2012; Soshnev et al . , 2016 ) . Moreover , considering the increasingly recognized important roles of epigenetic alterations in the control of cellular differentiation processes ( Boland et al . , 2014; Cabezas-Wallscheid et al . , 2014; Lipka et al . , 2014 ) , crucial molecular mediators of vessel maturation and EC quiescence may also be regulated by epigenetic mechanisms , which in EC have similarly not been studied in an unbiased genome-wide approach . The present study consequently aimed at obtaining and comparing transcriptomic and epigenomic signatures of EC from infant and young adult mice in order to identify most prominently regulated signaling pathways of vascular quiescence .
To identify factors and pathways regulated during acquisition of endothelial cell ( EC ) quiescence in an unbiased genome-wide manner , RNA-seq from freshly FACS-sorted lung EC isolated from infant ( p8-p10 ) and adult mice ( 8–12 weeks ) was performed ( Figure 1A and Figure 1—figure supplement 1A ) . The purity of cell sorting was controlled by the validation of marker gene expression ( Figure 1—figure supplement 1B ) . The phenotypic shift in marker expression ( CD31/PECAM1 and CD34 ) between infant and young adult EC reflected the size of the corresponding EC ( Figure 1—figure supplement 1C ) . EdU incorporation assays verified the almost complete arrest of cell proliferation in adult EC ( Figure 1B and Figure 1—figure supplement 1D ) . Comparative transcriptome analysis identified 2 , 216 differentially expressed genes with pronounced regulation of genes mediating the induction of immune mechanisms , metabolism and phagosome pathways as well as the repression of genes controlling cell cycle , developmental angiogenesis , extracellular matrix interactions and cell contraction ( Reactome-based analysis ) ( Figure 1C ) . To gain further insight into the mechanisms of EC quiescence , the protein interaction network ( interactome ) of the differentially regulated genes was generated ( Figure 1—figure supplement 2A ) . The analysis of this interactome led to the identification of several pathways presumably relevant during the acquisition of EC quiescence ( Figure 1—source data 1 ) . Importantly , it resulted in further specification of immune system regulation , as JAK-STAT signaling was most significantly enriched ( p=5 . 7×10-11 ) in an EC quiescence network with most factors being induced , putatively mediating the repression of the pro-angiogenic growth factors Fgf7 and Fgf10 ( Figure 1—figure supplement 2B and Figure 1—source data 1 ) . Further network clustering identified densely connected protein communities within the interactome ( Figure 1—figure supplement 2C ) . Functional annotation of the genes within each cluster revealed unique ( SEMA , TGFB and ERBB signaling ) enriched biological functions compared to Reactome-based analysis . Altogether , interactome analysis confirmed and extended the functional annotation of the genes differentially regulated during acquisition of EC quiescence . In line with the EdU assays , prominent regulators of cell cycle arrest were validated to be down-regulated during acquisition of EC quiescence in lung EC ( Figure 1D and Figure 1—source data 2 ) as well as in brain and heart EC ( Figure 1—figure supplement 1E ) . Altogether , these data suggest that the comparison of the transcriptomic profile of infant and adult EC displayed valid quiescence-dependent changes . Corresponding to the arrest of EC proliferation , the transcriptomic screen identified distinct changes of angiogenic molecule expression . Sixteen pro-angiogenic molecules were down-regulated and eight anti-angiogenic molecules were up-regulated during acquisition of vascular quiescence ( Figure 2A ) . This involved the downregulation of angiogenic signal transducing cell surface receptors ( Tgfbr2 , Bmpr1 , Notch3 and Fgfr1 ) as well as the upregulation of quiescence signal transducing receptors , most notably Cd36 and the angiopoietin receptor Tek ( encoding for Tie2 protein ) , probably the functionally best characterized maturation receptor of EC ( Augustin et al . , 2009 ) . Several autocrine and paracrine acting secreted angiogenic molecules were identified as differentially expressed: up-regulation of the semaphorins Sema3C and Sema3G as well as thrombomodulin ( Thbd ) and downregulation of Fgf7 , Fgf10 , placental growth factor ( Pgf ) and Wnt5a . Comparative quantitative PCR ( qPCR ) of selected candidate molecules validated gene expression changes identified by RNA-seq confirming the high robustness of the sequencing data ( Figure 2B and Figure 2—figure supplement 1A ) . Among these angiogenesis-regulating molecules , the TGFβ pathway was identified as the most concordantly downregulated signaling pathway with repression of matrix-associated latent TGFβ-binding proteins ( Ltbp2 , Ltbp3 , Ltbp4 ) and receptors ( Tgfbr2 , Bmpr1 ) as well as upregulation of the inhibitory Smads , Smad6 and Smad7 ( Figure 2C ) . Correspondingly , upstream analysis of the complete set of differentially expressed genes including activation state prediction of the corresponding upstream regulator revealed TGFß1 as the most significantly inhibited growth factor ( Figure 2D ) . Collectively , the data uncovered complex transcriptional changes of angiogenic molecules governing EC quiescence . They highlight a series of pathways that have all been associated with EC activation , most notably the TGFβ pathway . To assess the impact of epigenetic remodeling in the vascular system during acquisition of vascular quiescence , we analyzed the expression of chromatin modifying enzymes ( CME ) ( Plass et al . , 2013 ) . About 10% of CMEs were differentially expressed with the majority being repressed in adult EC ( Figure 2—figure supplement 1B and C ) , including the de novo DNA methyltransferase Dnmt3a and four additional DNA methylation modifiers ( Gadd45a , Tet1 , Fos , Uhrf1 ) ( Figure 2—figure supplement 1D ) . This suggested that the patterns of DNA methylation were more dynamic in infant compared to adult EC . It is now accepted that the DNA methylome gives a precise picture of the epigenetic state , defining active and inactive regions within genomes ( Cabezas-Wallscheid et al . , 2014; Jones , 2012; Lipka et al . , 2014; Soshnev et al . , 2016 ) . Consequently , tagmentation-based whole genome bisulfite sequencing ( T-WGBS ) of biological triplicates of infant and young adult EC samples enabled the identification of quiescence-dependent DNA methylation changes at single CpG nucleotide resolution ( Figure 1A and Figure 3—figure supplement 1A ) . T-WGBS data were highly reproducible in MassARRAY using independent biological replicates ( Figure 3—figure supplement 1B and C ) . Bioinformatic analysis revealed a significant increase of CpGs with >80% methylation level in adult EC compared to infant EC ( p=6 . 87×10−6 ) demonstrating a global increase in CpG methylation during the transition to vascular quiescence ( Figure 3A ) . In total , DNA methylation changes resulted in more than 1 . 4M differentially methylated CpGs ( DMCs ) . Retaining only the consecutive DMCs ( 68 , 184 ) with a 10% methylation difference showed a significant intragenic enrichment ( Figure 3—figure supplement 1D ) . These DMCs gave rise to 18 , 333 differentially methylated regions ( DMRs ) ( Figure 3B ) . The distribution of methylation differences among the DMRs demonstrated a more prominent loss ( 20–60% methylation difference between infants and young adult ) than gain of methylation ( 10–40% methylation difference ) ( Figure 3C ) . Applying a threshold of 10% methylation difference , young adults acquired a gain of methylation in 54% of DMRs and a loss of methylation in 46% of DMRs during acquisition of quiescence ( 20% gain and 80% loss , respectively , when applying 30% as threshold ) ( Figure 3B ) . Both , loss and gain of methylation DMRs were depleted on transcription start sites ( TSS ) , but enriched up- and downstream ( 10–100 kb distant ) from the TSS ( Figure 3D ) . There was an overall significant enrichment of DMRs in introns ( OR 2 . 65 , p<0 . 0001 ) , most prominent in intron 1–3 of a given gene ( Figure 3E and Figure 3—figure supplement 1E ) . Notably , hypomethylated DMRs were overrepresented in introns and underrepresented in intergenic regions ( Figure 3—figure supplement 1F ) . Overlap of the DMRs with lung-identified genomic regulatory features ( Shen et al . , 2012; Yue et al . , 2014 ) further specified the DMRs to coincide with putative enhancer regions ( H3K4me1 positive regions , enhancer-promoter units [EPU] , DNase hypersensitive regions ) but not with transcription start sites ( CpG islands , TSS defined by FANTOM consortium , H3K4me3 positive regions ) ( Figure 3F ) . This could be visualized in an exemplary way in genome browser views depicting CpG islands and putative enhancer regions together with the identified DMRs ( Figure 3G and Figure 4E ) . Essentially , DNA methylation profiling suggested a prominent loss of methylation at intronic enhancers during acquisition of quiescence in lung EC . Methylome analysis identified strong DNA demethylation of putative intronic enhancer regions suggesting that DMRs may have regulatory potential on gene expression . As enhancer regions could influence the expression of distant genes , up to three nearby genes were assigned to the quiescence-dependent DMRs by applying the GREAT tool to identify genes likely affected by differential methylation ( McLean et al . , 2010 ) . The same tool was also used to identify significantly enriched biological functions among the genes next to quiescence-dependent regulated DMRs . The functional annotation of genes next to hypomethylated DMRs revealed signaling pathways controlling vascular patterning ( semaphorin interactions ) and development/homeostasis ( TGFβ family signaling ) ( Figure 4A ) . Notably , 94% of the hypomethylated DMRs affecting genes of these pathways were located in intragenic regions with eight DMRs per gene on average ( Figure 4—figure supplement 1A , left ) resulting in DMR clustering ( more than five DMRs per gene locus ) in more than 36% of the affected genes ( Figure 4B ) . Contrary , 56% of the hypermethylated DMRs were located in intragenic regions with on average two DMRs per gene ( only 5 . 3% of genes harbored five or more hypermethylated DMRs ) . Furthermore , genes in the vicinity of gain of methylation DMRs were annotated as endothelial fate control genes ( NOTCH signaling ) ( Figure 4A and B , Figure 4—figure supplement 1A right ) . In summary , in line with transcriptome analysis , genes of TGFß family and semaphorin signaling were similarly affected by differential DNA methylation . Apart from the functional annotation of genes located in the vicinity of DMRs , these analyses revealed an overrepresentation of DMRs within certain gene loci . Notably , this clustering of DMRs was mostly restricted to loss of methylation DMRs . Correspondingly , DMR clustering was most obvious in induced angiogenic molecules ( Figure 4C ) . The ranking of all genes according to the number of intragenic DMRs further demonstrated that ( i ) 10% ( 446 ) of the affected genes contained more than five intragenic DMRs corresponding to 44% ( 5 , 180 ) of all intragenic DMRs and ( ii ) 68% of the clustered DMRs showed loss of methylation ( Figure 4D , bottom ) . This analysis highlighted the existence of large intragenic DMR clusters that undergo concerted differential DNA methylation in about 450 genes during transition to endothelial cell quiescence . Furthermore , enrichment analysis among this ranked list of genes confirmed that DMR clusters were overrepresented in genes of TGFβ and semaphorin signaling , particularly in Smad6 , Smad7 and Sema6a with up to 61 DMRs in one gene ( Figures 3G 4D , top , and 4E ) . The correlation of quiescence-related expression and methylation profiles by gene set enrichment analysis ( GSEA ) revealed an enrichment of genes with hypomethylated DMRs ( young adult vs . infant EC ) among the genes induced in adult EC independent of the genomic loci of one or several DMRs . Conversely , genes with hypermethylated DMRs were overrepresented among the genes upregulated in infant EC ( Figure 4—figure supplement 1C ) . In line with this , there was an unexpected high overlap of differentially expressed genes and genes affected by DNA methylation changes . Notably , we identified an increasing overlap of transcriptome and epigenome , when scoring genes on the basis of intragenic DMRs and even more so when scoring based on clustered DMRs ( Figure 4F ) . This correlation implies that genes with intragenic clusters of loss of methylation DMRs are more likely to be induced during acquisition of vascular quiescence than repressed , while those genes with clustered gain of methylation DMRs are more likely to be repressed . Together , these data suggest that vascular quiescence is accompanied by a prominent loss of methylation in intragenic DMR clusters preferentially affecting and regulating the expression of genes involved in the regulation of TGFβ family , semaphorin and NOTCH signaling in this process . Both , the transcriptome and the methylome analysis had highlighted TGFβ family members as relevant molecules during vascular quiescence . Among these , Smad6 and Smad7 contained the largest intragenic cluster of hypomethylated DMRs and showed significant induction during transition to quiescence in lung EC ( Figure 2B , Figure 2—figure supplement 1A and Figure 4B ) . We therefore focused further functional experiments prototypically on these molecules . SMAD6 and SMAD7 inhibit TGFβ family signaling by different mechanisms ( ten Dijke and Arthur , 2007 ) . Corresponding to the observed lung endothelial upregulation of Smad6 and Smad7 during acquisition of quiescence , we detected a strong decrease in R-SMAD ( receptor-regulated SMADs , SMAD2/3 and SMAD1/5/8 ) phosphorylation in adult lung tissue confirming an inhibition of TGFβ family signaling in adult mice ( Figure 5A ) . The upregulation of Smad6 and Smad7 during acquisition of vascular quiescence was not restricted to the lung , but also detected in EC isolated from adult mouse brain and heart ( Figure 5B and Figure 5—figure supplement 1A ) . Correspondingly , R-SMAD phosphorylation was decreased in these organs suggesting that TGFβ signaling may vessel bed independently act as global regulator of vascular quiescence ( Figure 5—figure supplement 1B and C ) . As reduced signaling could also be mediated by decreased ligand levels , BMPs and TGFB1 were determined in different organs ( Figure 5—figure supplement 1D ) . TGFβ family ligands were consistently higher in adult organs indicating that regulation of vascular quiescence is achieved through cell intrinsic blocking of the downstream signaling machinery , that is by downregulation of receptors and upregulation of inhibitory SMADs . To further examine if the expression of SMAD6 and/or SMAD7 meditates resistance of EC towards ligand-dependent receptor activation , HUVEC were lentivirally transduced and analyzed in vitro . Forced expression of SMAD6 or SMAD7 in cultured EC , but not gene silencing , led to the reduction of SMAD2/3 and SMAD1/5/8 phosphorylation upon ligand stimulation ( Figure 5C and D and Figure 5—figure supplement 2A ) . Cells overexpressing both , SMAD6 and SMAD7 showed reduced viability upon prolonged cell culture , which was confirmed by MTT assay ( Figure 5E and Figure 5—figure supplement 2B ) . Proliferation and migration of EC overexpressing SMAD6 and SMAD7 was reduced confirming the limited angiogenic potential of SMAD6 and SMAD7 expressing cells ( Figure 5F and G and Figure 5—figure supplement 2C and D ) . Together , these data validate based on expression in vivo and functional experiments in vitro the negative regulation of the TGFβ family signaling as critical for the acquisition of endothelial quiescence during development .
Loss of the quiescent EC phenotype is at the heart of vascular dysfunction that is associated with some of the most common and most devastating diseases in humans . Vascular quiescence and the switch from quiescent to activated EC therefore need to be tightly regulated . How the resting phenotype is achieved on the genome-wide epigenetic and transcriptomic level has not been studied so far . This study presents the first unbiased systems biology based profiling of primary EC isolated from infant and adult mice , revealing surprising transcriptomic and epigenetic signatures . It thereby establishes a roadmap for further molecular dissection of mechanisms of vascular quiescence . Specifically , extensive bioinformatic data analysis and functional validation demonstrated that the acquisition of EC quiescence ( i ) is associated with distinct transcriptomic changes affecting pro- and anti-angiogenic molecules with prominent down-regulation of TGFß family signaling molecules and ( ii ) is accompanied by gain of DNA methylation and more prominent loss of DNA methylation at intragenic enhancer regions preferentially affecting genes of TGFß family and semaphorin signaling . In general , the identified gene expression changes ( iii ) correlate with changes in DNA methylation in corresponding loci . Notably , ( iv ) most hypomethylated DMRs were concentrated in large intragenic clusters , with the gene loci of the inhibitory SMADs ( Smad6 and Smad7 ) among the genes containing the largest clusters . Cellular experiments ( v ) confirmed SMAD6 and SMAD7 as regulators of endothelial quiescence . Together , this study highlights the central role of suppression of TGFβ family signaling during acquisition of vascular quiescence and establishes a database enabling the datamining of transcriptomic and epigenetic changes associated with the acquisition of the quiescent EC phenotype . Although the importance of epigenetic mechanisms regulating vascular function has long been recognized ( Dunn et al . , 2015; Matouk and Marsden , 2008; Yan and Marsden , 2015 ) , studies of DNA methylation in EC are still in their infancy . Only few genome wide systems biology studies have been performed to dissect epigenetic changes in the vasculature and even less studies have systematically correlated epigenetic changes to transcriptomic signatures . It has been described that proximal promoter DNA methylation regulates EC gene expression and thereby controls EC fate ( Shirodkar et al . , 2013 ) . Similarly , flow-dependent DNA methylation changes in vessel wall cells have been shown to be enriched in GC-rich , 5’ untranslated , and exonic regions ( Davies et al . , 2014; Jiang et al . , 2015a; Jiang et al . , 2015b ) . These previous studies in EC have determined an enrichment of promoter-associated changes in DNA methylation that coincide with gene regulation . Here , we generated genome-wide high resolution DNA methylation data of primary infant and young adult EC which enabled the identification of an unusual abundance of intragenic ( mainly intronic ) demethylation events with numerous genes harboring large intronic DMR clusters . Although little is known about the impact of intragenic DNA methylation changes on gene expression , some studies suggest a role in alternative promoter usage or splicing ( Kulis et al . , 2013; Maunakea et al . , 2013; Maunakea et al . , 2010; Shukla et al . , 2011 ) . Conversely , the quiescent-dependent DMRs identified in the present study overlapped with putative intragenic enhancer features . Moreover , the genome-wide association with gene expression data revealed a good correlation of intragenic hypomethylation and gene induction or hypermethylation and gene repression , respectively . The presence of clustered DMRs was even more likely associated with differential expression changes of the affected gene , suggesting that we have identified intronic regions with high regulatory potential during the acquisition of vascular quiescence . Notably , while we identified a more prominent loss than gain of methylation during acquisition of quiescence , atherosclerosis progression in human appears to be associated with a distinct gain of methylation signature ( Zaina et al . , 2014 ) . Altogether , these data highlight the important role of epigenetic modifications in the regulation of EC functions and have led to the identification of distinct genomic regions with crucial regulatory potential that will serve as a foundation for further functional studies . Among the transcriptomically and epigenetically regulated genes were molecules of pathways with well-established EC-regulatory function including NOTCH , semaphorin , FGF , WNT and TIE signaling ( Marcelo et al . , 2013; Patel-Hett and DAmoreD'Amore , 2011 ) . Yet , both unbiased genome-wide approaches of this study , RNA-seq and T-WGBS , converged on a crucial role of TGFß family signaling in the regulation of vascular quiescence . This finding is not novel per se , but TGFß family signaling has been described to result in both , pro- and anti-angiogenic functions in a very context dependent manner ( Cai et al . , 2012; Dyer et al . , 2014; Goumans et al . , 2003; Goumans et al . , 2002; Larrivée et al . , 2012; Mallet et al . , 2006; Suzuki et al . , 2008 ) . Here , we show in the physiologically most relevant vascular quiescence process ( i . e . the comparison of infant and young adult EC ) that many important signaling molecules of the TGFß family pathway are differentially regulated leading to reduced signal transduction via R-SMAD phosphorylation in adult tissue despite high ligand levels . As a decisive point , the inhibitory SMADs , Smad6 and Smad7 , were epigenetically inhibited by large intronic clusters of methylated DNA during early development and induced in adult EC . Cell culture based analysis confirmed that the expression of SMAD6 and/or SMAD7 in EC mediates resistance to ligand treatment thereby blocking TGFß family signaling . Interestingly , SMAD6 and SMAD7 have originally been identified as vascular SMADs ( Topper et al . , 1997 ) . Their induction upon laminar shear stress highlights the importance of blood flow in maintaining the quiescent vascular phenotype . Although ubiquitously expressed , they exert particularly prominent vascular function . Smad6 has recently been described to act anti-angiogenic in sprouting assays ( Mouillesseaux et al . , 2016 ) and Smad7 has been shown to inhibit peritoneal angiogenesis ( Peng et al . , 2013 ) . Correspondingly , genetic deletion of Smad6 or Smad7 leads to major cardiovascular defects with partial prenatal mortality ( Smad6 ) ( Galvin et al . , 2000 ) or to massive growth retardation with reduced viability ( Smad7 ) ( Tojo et al . , 2012 ) . In conclusion , the present study underlines the central role of TGFß family pathways during acquisition of the resting EC phenotype including prominent regulation of inhibitory SMADs on the epigenetic level . Remarkably , molecules of the VEGF pathway , the master regulator of angiogenesis induction , were both , transcriptionally and epigenetically not identified as being significantly differentially regulated during the transition to EC quiescence . Surprisingly , the ligand Vegfa was significantly higher expressed in adult EC compared to infant EC . High expression of VEGFA has also been described in the adult brain ( Licht and Keshet , 2013 ) . Autocrine VEGF has been proposed to be essential for EC survival thereby fulfilling homeostatic functions . It appears to serve as a rheostat of paracrine acting VEGF to possible even exert negative regulatory functions . In fact , analyses of genetically engineered mice specifically lacking EC-produced VEGF indicate that paracrine VEGF does not compensate for autocrine VEGF ( Lee et al . , 2007 ) . Conversely , expression of the VEGF decoy receptor Flt1/Vegfr1 ( Boucher et al . , 2017; Fong et al . , 1995; Kappas et al . , 2008; Kearney et al . , 2002 ) was in the present study significantly increased ( but below threshold of 1 . 75x ) during acquisition of quiescence . Since the expression of the primary VEGF signal transducing receptor Kdr/Vegfr2 was neither in RNA-seq nor in qPCR validation data significantly changed , this results in an increased VEGFR1/VEGFR2 ratio in adults . Altogether , these data imply that VEGF signaling is not dominantly regulated during the transition to vascular quiescence but instead undergoes specific and situation-specific dynamic regulation . In conclusion , the present study has for the first time established the transcriptomic and epigenetic landscape of vascular quiescence in primary lung EC that might also apply to EC of other tissues . These findings reveal that vascular quiescence is comprehensively regulated on both transcriptional and DNA methylation level converging on the prominent inhibition of TGFß family signaling in adult EC . Moreover , the acquisition of the resting EC phenotype is accompanied by a dynamic balance of activating and inhibiting pathways . These findings might in the future potentially pave the ways towards fundamentally novel strategies to therapeutically interfere with unwanted EC activation during disease . The study thereby lays a foundation for mechanistic and functional studies of individual genes and complex gene signatures involved in the transition from the activated to the quiescent phenotype and vice versa .
C57BL/6N female mice were obtained from Taconic ( Taconic Biosciences , Europe ) . After arrival , mice were kept in ‘Individually Ventilated Cages’ ( IVC ) to maintain the health status as declared by Taconic as ‘Murine Pathogen Free' . Lungs were processed for EC isolation 1–3 days after arrival . Animals had free access to food and water and were kept in a 12 hr light-dark cycle . All mice were handled according to the guidance of the Institute and as approved by the German Cancer research center ( No . DKFZ305 ) . No statistical methods were used to predetermine sample size . The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment . Endothelial cell ( EC ) isolation was performed as described previously ( Korn et al . , 2014 ) . In brief , mice were sacrificed and lung , heart and brain of infant or young adult mice were surgically removed and cut into small pieces . Tissue pieces were digested in Dulbecco´s Modified Eagle´s medium ( DMEM , ThermoFisher Scientific , Germany ) containing 1 . 25 mM CaCl2 , 200 U/ml Collagenase I ( Sigma , Munich , Germany ) and 10 µg/ml DNaseI ( Roche , Germany ) at 37°C for 45 min . Single-cell suspensions were prepared by passing the digestion mix through 18G and 19G cannula syringes and filtering through a 100 µm cell strainer . To remove the myelin from the brain cell suspension 22% BSA was added ( 1:1 v/v ) , and after centrifugation the upper myelin layer was discarded . Cells were stained for the negative markers PDPN , LYVE1 , PTPRC and LY76 for 30 min at 4°C in PBS/5% fetal calf serum ( FCS ) . Cells stained with negative markers were depleted by incubation with Dynabeads ( ThermoFisher Scientific , Germany ) for 30 min at 4°C on the rotator followed by magnetic removal . The remaining cells were positively stained with antibodies against CD31 and CD34 in PBS/5% FCS for 30 min at 4°C . Dead cells were excluded by phosphatidylinositol ( PI ) staining ( 1:2000 ) . PTPRC‐LYVE1‐LY76‐PDPN‐PI‐CD31+CD34+ cells were sorted with a BD FACSAria ll ( BD Biosciences , Heidelberg , Germany ) The following primary antibodies were used: rat anti-CD31 ( 551262 , BD Biosciences , diluted 1:100 ) , rat anti-CD34 ( 48–0341 , eBioscience , diluted 1:50 ) , rat anti-LY76 ( 561032 , BD Biosciences , diluted 1:200 ) , rat anti-LYVE1 ( 53–0443 , eBioscience , diluted 1:250 ) , hamster anti-PDPN ( 53–5381 , eBioscience , 1:100 ) , rat anti-PTPRC ( 553080 , BD Biosciences , 1:400 ) . Mice were injected s . c . with EdU ( 50 μg/g mouse ) 17 hr prior to sacrificing the animals . Lungs were removed and cells were isolated and surface-stained as described in the corresponding section . For additional EdU-labeling cells were fixed , permeabilized and stained using the Click-iT Plus Alexa Fluor 555 Picolyl Azide Toolkit ( ThermoFisher Scientific , Germany ) according to the manufacturer’s instructions . Organ cryosections were fixed in ice-cold methanol for 10 min at −20°C . Blocking was performed with 10% goat serum ( ThermoFisher Scientific , Germany ) for 1 hr at RT . The vasculature was stained for CD31 and proliferating cells for KI67 overnight at 4°C . The sections were subsequently incubated with the appropriate secondary antibody for 1 hr at RT . Pictures were taken using the Zeiss Cell Observer and image analysis was accomplished with Fiji . The following primary antibodies were used: rat anti-CD31 ( 553370 , BD Biosciences , 1:50 ) , rabbit anti-KI67 ( GTX 16667 , Gene Tex , 1:100 ) , goat IgG ( A11006 , ThermoFisher Scientific , 1:500 ) , goat IgG ( A11071 , ThermoFisher Scientific , 1:500 ) . HUVECs were purchased from PromoCell ( Heidelberg , Germany ) and cultured in Endopan 3 with 3% FCS and supplements ( PAN Biotech , Aidenbach , Germany ) at 37°C , 5% CO2 and high humidity . Passages < P6 were used for all experiments and cells were tested negative for mycoplasma contamination . For Lentiviral transduction ( pLenti-based overexpression or pGIPZ-based silencing of certain molecules ) , 1 × 105 cells were seeded and 24 hr later lentivirus was added for 16 hr . Media was changed to Endopan with supplements for 24 hr before starting the selection with puromycin ( P; 0 . 4 µg/ml ) or neomycin ( N; 200 µg/ml ) or both ( P 0 . 3 µg/ml and N 150 µg/ml ) for 3–4 days . The following lentiviral vectors were used: pLenti-CMV-EV-puromycinRes ( control ) , pLenti-CMV-EV-neomycinRes ( control ) , pLenti-CMV-SMAD6-puromycinRes ( overexpression ) , pLenti-CMV-SMAD7-neomycinRes ( overexpression ) , pGIPZ-ns-shRNA ( control ) , pGIPZ-shSMAD6-118 ( depletion ) , pGIPZ-shSMAD7-115 ( depletion ) . Cells were starved in Endopan without supplements for 2–4 hr before treating with 5 ng/ml BMP9 ( R and D Systems , Minneapolis , Minnesota , US ) or 10 ng/ml TGFB ( R and D Systems , Minneapolis , Minnesota , US ) for 30 min . For EdU incorporation analysis EdU was directly added to the subconfluent cells at a final concentration of 10 µM for 4 hr . Harvesting , fixation , permeabilization and staining were performed using the Click-iT EdU Flow Cytometry Assay Kit Alexa Fluor 647 ( ThermoFisher Scientific , Germany ) according to the manufacturer’s protocol . Cells were analyzed on a BD FACSCanto II ( BD Bioscience , Heidelberg , Germany ) . MTT assay was performed in 96well plates according to manufacturer’s instructions ( Roche , Germany ) . Migration analysis was performed by scratch assay in 24well plates . To prevent cell proliferation during migration analysis cells were treated with 10 µg/ml mitomycin C ( Sigma , Munich , Germany ) for 1 . 5 hr prior to scratching the monolayer . Pictures were taken either manually at certain time points ( as indicated in the figures ) at an Olympus IX71 Microscope ( 10x ) or automatically every hour at an Olympus Cell^R Microscope with cell incubation chamber ( 10x ) . Image analysis was accomplished with Fiji . Ligand level in tissue lysates were determined by ELISA according to the manufacture’s protocols . BMP6 was measured by the ELISA Kit for Bone Morphogenetic Protein 6 ( BMP6 ) ( SEA646Mu , Cloud Clone , Katy , Texas , US ) , BMP9 was measured by the Mouse BMP9 ELISA Kit ( ELM-BMP9 , RayBiotech Norcross , Georgia , US ) and TGFB1 was measured by the TGF-β1 Quantikine ELISA Kit ( MB100B , R and D systems , Minneapolis , Minnesota , US ) . For BMP6 and BMP9 determination in lung and brain 100 ng/µl total protein concentration , for BMP6 and BMP9 measurement in heart 10 ng/µl total protein concentration and for TGFB1 detection in lung and heart 500 ng/µl total protein concentration was utilized . Cells ( whole tissue fragments or HUVEC ) were lysed in RIPA buffer ( 1% NP-40 , 0 . 1% Sodium dodecyl sulphate , 0 . 5% Sodium deoxycholate , 10% glycerol , 5 mM EDTA ) supplemented with proteinase inhibitor mix ( Serva Electrophoresis , Heidelberg , Germany ) and sodium orthovanadate ( Sigma , Munich , Germany ) . Protein lysates were separated on 10% polyacrylamide-SDS gels and blotted on nitrocellulose membranes . Membranes were blocked with 3% BSA and incubated with the indicated primary antibodies at 4°C over night . Horseradish peroxidase-conjugated secondary antibodies were used for chemiluminescence detection . For detection , either an AGFA classic EOS developer ( AGFA , Mortsel , Belgium ) or an Amersham Imager 600 ( GE healthcare , Little Chalfont , UK ) was used . The following primary antibodies were used: rabbit anti-pSMAD1/5/8 ( D6656/Vli31 , Maine Medical Center , 1:2000 ) , rabbit anti-SMAD1 ( 9743S , Cell Signaling , 1:1000 ) , rabbit anti-pSMAD2/3 ( D6658 , Maine Medical Center , 1:2000 ) , rabbit anti-SMAD2/3 ( 5678 , Cell Signaling , 1:1000 ) , rabbit anti-ACTB ( sc-1616 , Santa Cruz Biotechnology , 1:5000 ) , goat IgG ( P0448 , Dako , 1:5000 ) . Total RNA of FACS-sorted mouse ECs was isolated with Arcturus PicoPure RNA Isolation Kit ( ThermoFisher Scientific , Germany ) and RNA of HUVECs was isolated with RNeasy Mini Kit ( Qiagen , Hilden , Germany ) according to the manufacturer’s protocols . cDNA was synthesized with QuantiTect Reverse Transcription Kit ( Qiagen , Hilden , Germany ) according to the manufacturer’s instructions . Subsequent qPCR was performed with TaqMan gene expression assay ( ThermoFisher Scientific , Germany ) , TaqMan Fast Advanced Mastermix ( ThermoFisher Scientific , Germany ) and the StepOnePlus Real-Time PCR System ( ThermoFisher Scientific , Germany ) . Actb ( mouse EC ) or HPRT ( HUVEC ) was used for data normalization . The following TaqMan gene expression assays ( ThermoFisher Scientific , Germany ) were used: Acta2 ( Mm00725412_s1 ) , Actb ( Mm00607939_s1 ) , Bmpr2 ( Mm00432134_m1 ) , Ccnb1 ( Mm03053893_gH ) , Ccnb2 ( Mm01171453_m1 ) , Cdk1 ( Mm00772472_m1 ) , Cyr61 ( Mm00487501_g1 ) , Fgfr1 ( Mm00438930_m1 ) , HPRT ( Hs02800695_m1 ) , Icam1 ( Mm00516023_m1 ) , Kdr ( Mm01222421_m1 ) , Notch3 ( Mm01345646_m1 ) , Nr2f2 ( Mm00772789_m1 ) , Ptprc ( Mm01293577_m1 ) , Sema3c ( Mm00443121_m1 ) , Smad6 ( Mm01171378_m1 ) , SMAD6 ( Hs00178579_m1 ) , Smad7 ( Mm00484742_m1 ) , SMAD7 ( Hs00998193_m1 ) , Stat1 ( Mm00439531_m1 ) , Tgfbr2 ( Mm00436977_m1 ) . Total RNA from mouse lung EC was isolated using Arcturus PicoPure RNA Isolation Kit ( ThermoFisher Scientific , Germany ) according to the manufacturer’s instructions . DNA was removed by treating with RNase-free DNase Set ( Qiagen , Hilden , Germany ) . Quality control was performed by Qubit ( ThermoFisher Scientific , Germany ) and Bioanalyzer ( Agilent , Waldbronn , Germany ) measurements . Sequencing library was generated with 10 ng of total RNA using the SMARTer Ultra Low RNA Kit for Illumina Sequencing ( Clontech , Mountain View , California , US ) according to manufacturer’s protocol . Sequencing reads ( 100 bp Paired-End ) were generated on the HiSeq2000 platform ( Illumina , San Diego , California , US ) with four samples per lane . The sequenced reads were aligned to the mouse reference genome mm10 using STAR ( Dobin et al . , 2013 ) allowing up to 10 mismatches . The average transcriptome coverage was 37 . 8 ± 4 . DEseq2 was used to test for differential gene expression ( mm10 , RefSeq gene annotation ) ( Love et al . , 2014 ) . RNA-seq data are available at the NCBI Gene Expression Omnibus ( GEO ) under accession number GSE86600 . Only transcripts with an RPKM ≥1 in at least one sample were considered for further analysis . Significantly differentially expressed genes were defined as more than 1 . 75-fold regulated ( p<0 . 05 ) . Hierarchical clustering was performed using GenePattern Software ( Reich et al . , 2006 ) . For functional annotation with Reactome gene sets , the Molecular Signature Database ( MSigDB ) ( Subramanian et al . , 2005 ) was used . Upstream analysis with activation state prediction was generated by using QIAGEN’s Ingenuity Pathway Analysis ( IPA , QIAGEN Redwood City ) . According to the manual , “the upstream regulator analysis is based on prior knowledge of expected effects between transcriptional regulators and their target genes stored in the Ingenuity Knowledge Base . The analysis examines how many known targets of each transcription regulator are present in the user’s dataset , and also compares their direction of change ( i . e . expression in the experimental sample ( s ) relative to control ) to what is expected from the literature in order to predict likely relevant transcriptional regulators . If the observed direction of change is mostly consistent with a particular activation state of the transcriptional regulator ( ‘activated’ or ‘inhibited’ ) , then a prediction is made about that activation state . ’ The list of chromatin modifying enzymes has been described previously ( Plass et al . , 2013 ) . Genomic DNA from sorted EC was isolated using the QIAamp DNA Micro Kit ( Qiagen , Hilden , Germany ) according to the manufacturer’s instructions . T-WGBS was essentially performed as described previously ( Wang et al . , 2013 ) using 30 ng genomic mouse DNA as input . Three sequencing libraries were generated per sample and each was sequenced 100 bp , paired-end on one lane of a HiSeq2000 ( Illumina , San Diego , California , US ) machine . Validation of selected T-WGBS data was performed by MassARRAY ( Agena Bioscience , Hamburg , Germany ) as described previously ( Sonnet et al . , 2014 ) using primers listed in the Supplementary file 1 . Forward primers for MassARRAY were 5' tagged with AGGAAGAGAG , reverse primers with CAGTAATACGACTCACTATAGGGAGAAGGCT . We used a mapping method as described earlier ( Johnson et al . , 2012 ) modified for reads from tagmentation-based whole-genome bisulfite sequencing . Briefly , the mm10 reference genome was transformed in silico for both the top strand ( C to T ) and bottom strand ( G to A ) . Before alignment , adaptor sequences were trimmed using SeqPrep ( St John , 2016; https://github . com/jstjohn/SeqPrep ) . Then the first read in each read pair was C-to-T converted and the 2nd read in the pair was G-to-A converted . The converted reads were aligned to a combined reference of the transformed top ( C to T ) and bottom ( G to A ) strands using BWA ( bwa-0 . 6 . 1-tpx ) ( Li and Durbin , 2009 ) with default parameters except the quality threshold for read trimming ( -q ) of 20 and the Smith-Waterman for the unmapped mate disabled ( -s ) . After alignment , reads were converted back to the original states and reads mapped to the antisense strand of the respective reference were removed . Duplicate reads were removed using Picard MarkDuplicates ( http://picard . sourceforge . net/ ) . Reads with alignment scores less than one were filtered before subsequent analysis . Total genome coverage was calculated using the total number of bases aligned from uniquely mapped reads over the total number of mappable bases in the genome . At each cytosine position , reads that maintain the cytosine status were considered methylated , and the reads that have cytosine converted to thymine were considered unmethylated . Only bases with Phred-scaled quality score of ≥20 were considered . For libraries prepared with the tagmentation protocol , first 9 bp of the second read and last 9 bp before the adapter in the first read were excluded from methylation calling . Bisulfite conversion rates were estimated using the methylation level at CH sites . DMR calling was performed as described previously ( Bauer et al . , 2016 ) with minor modifications . In brief , DMR calling was conducted on methylation data without smoothing ( due to sufficiently high coverage ) with the bsseq v1 . 60 package ( Hansen et al . , 2012 ) for R statistical software v3 . 2 . 2 . We applied a coverage filter that required a minimum coverage of 8x per CpG in at least two of the three samples per group . The DMR model comprised at least three consecutive significant CpGs ( each p<0 . 05 , t-test ) and a minimum difference of 10% in mean methylation between groups . Intersection with known regulatory genomic features was performed with Galaxy ( Afgan et al . , 2016 ) using published datasets ( H3K4me3 ChIPseq: ENCSR000CAR; H3K4me1 ChIPseq: ENCSR000CAQ; DNase Hypersensitivity: ENCSR000CNM ) and enhancer-promoter units ( EPU ) defined in mouse lung tissue ( Shen et al . , 2012; Yue et al . , 2014 ) . For data visualization tracks were loaded into the IGV browser ( Robinson et al . , 2011 ) . Motif searches for known transcription factor binding sites ( TFBS ) in DMRs and TFBS enrichment over background were analyzed with the ‘findMotifsGenome . pl’ script of the HOMER tools software package ( Heinz et al . , 2010 ) . Genomic annotation of DMRs to the nearest TSS was obtained with the ‘annotatePeaks . pl’-script of the HOMER tools software package ( Heinz et al . , 2010 ) to genome version mm10 . Functional annotation of up to three nearby genes of DMRs was accomplished using the GREAT tool ( McLean et al . , 2010 ) . For enrichment analysis among a ranked list of genes ( sorted according to the number of DMRs or rel . gene regulation ) the Gene Set Enrichment Analysis ( GSEA ) Software was applied ( Subramanian et al . , 2005 ) . Tagmentation-based WGBS data are available at the NCBI Gene Expression Omnibus ( GEO ) under accession number GSE87374 . To gain further functional insights into the mechanisms of EC quiescence , we generated a network from genes that were significantly differentially regulated ( p<0 . 05 ) . For generating such a network , we used the DIAMOnD algorithm ( Ghiassian et al . , 2015 ) , which creates a network starting from a set of proteins ( so-called seed set ) by integrating related proteins into the network based on their topological relevance to the initial seed set . DIAMOnD first ranks all proteins in the interactome according their connectivity significance with respect to the seed set . Next , the protein with the highest rank , that is , lowest p-value , is added to the initial network . The procedure is repeated with the extended seed set , until the incorporation of seed proteins saturates . Genes with a FC above 1 . 75 or below −1 . 75 , were mapped to their corresponding gene products and selected as seeds for generating the network , with a total of 972 genes being up and 2 , 251 genes being down regulated ( a total of 3 , 223 regulated genes ) . The mouse interactome comprised 7 , 267 proteins and 18 , 382 protein interactions , obtained by merging publicly available data from the major public protein-interaction databases . In the end , 227 ( 23 . 4% ) upregulated and 618 ( 27 . 5% ) downregulated proteins were covered in the assembled mouse interactome and used as seed proteins to generate the EC specific interaction network . Given the incompleteness of the mouse interactome , a large number of DIAMOnD iterations were necessary to connect all seed proteins through interactions , that is , 5 , 943 iterations were needed to link 99 . 3% of the proteins . In consequence , the resulting network was very large , covering about 93% of the mouse interactome . Since we were primarily interested in proteins functionally related to EC quiescence , we applied a strict threshold , terminating the network generation after the first steep increase in the proportion of incorporated seeds started to flatten ( iteration: 265 ) . The EC quiescence specific network with 745 proteins and 1 , 661 interactions is shown in the Figure 1—figure supplement 2 . A total of 484 seed proteins ( 57 . 3% ) were connected in the network through 261 linker proteins , which were of high functional interest with respect to EC quiescence . Next , we assessed whether proteins forming the EC quiescence network were functionally coherent with respect to the initial seed set . To this end , we performed a functional enrichment with g:Profiler ( Reimand et al . , 2016 ) of two protein sets , namely ( i ) proteins of the seed set covered by the interactome , and ( ii ) proteins of the network , and compared the respective outcomes focusing on KEGG pathways . As background set , we used the 7 , 267 proteins comprising the mouse interactome . As the reduced seed set might yield a distinct enrichment , we performed in addition a functional enrichment for proteins of the complete initial seed set . The complete list of enriched pathways for each protein is shown in Figure 1—source data 1 . In general , we observed a large overlap of enriched pathways between the seeds and the EC quiescence network . These findings emphasized that proteins captured in the network were of functional relevance with respect to proteins regulated in EC quiescence . To obtain more insights from the generated network , we performed network clustering identifying densely connected subgraphes , so called communities . The 23 detected cluster were ranging in size from two to 165 proteins . For each of the cluster , we performed functional enrichment analysis using g:Profiler ( Reimand et al . , 2016 ) . | The vascular system is made up of vessels including arteries , capillaries and veins that carry blood throughout the body . The inner surfaces of these blood vessels are lined with a thin layer of cells , called endothelial cells , which form a barrier and a communicating interface between the circulation and the surrounding tissue . Early in an organism’s life , when the vascular system is still growing , endothelial cells increase in number by dividing into more cells . In adulthood , as the vascular system reaches its full size , the endothelial cells maintain a stable number . As a result , an adult’s vascular system has a resting layer of endothelial cells that does not divide . This is known as vascular quiescence , and scientists know little about how the body achieves and maintains it . To unravel the mechanisms controlling vascular quiescence , Schlereth et al . studied endothelial cells taken from blood vessels in the lungs of newborn and adult mice . By comparing all the genes present at both developmental stages , the changes of gene activity in these cells could be measured . The results showed that the activity of genes strongly correlated with so called epigenetic changes in the genes involved in vascular quiescence . These are DNA modifications that can alter the function of a gene without affecting its underlying sequence . Two genes in particular ( Smad6 and Smad7 ) appeared to play an important role in vascular quiescence . Their corresponding proteins , SMAD6 and SMAD7 , inhibit another group of proteins ( TGFβ family ) important for cell growth . The results showed that the endothelial cells in adult mice produced more SMAD6 and SMAD7 than in young mice . Therefore , endothelial cells of adult mice stop to increase in number and to migrate . For the first time ever , Schlereth et al . have provided an extensive comparative analysis of gene activity and epigenetic changes to study vascular quiescence . The findings open a new chapter of vascular biology and will serve as a foundation for future research into the mechanisms of vascular quiescence . Problems in maintaining a resting layer of cells may lead to vascular dysfunction , which is associated with a wide range of diseases , such as stroke , heart disease and cancer making it a leading cause of death . In future , scientists may be able to develop new treatments that target specific molecules to help the body achieve a resting blood vessel system . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2018 | The transcriptomic and epigenetic map of vascular quiescence in the continuous lung endothelium |
Centrioles organise centrosomes and template cilia and flagella . Several centriole and centrosome proteins have been linked to microcephaly ( MCPH ) , a neuro-developmental disease associated with small brain size . CPAP ( MCPH6 ) and STIL ( MCPH7 ) are required for centriole assembly , but it is unclear how mutations in them lead to microcephaly . We show that the TCP domain of CPAP constitutes a novel proline recognition domain that forms a 1:1 complex with a short , highly conserved target motif in STIL . Crystal structures of this complex reveal an unusual , all-β structure adopted by the TCP domain and explain how a microcephaly mutation in CPAP compromises complex formation . Through point mutations , we demonstrate that complex formation is essential for centriole duplication in vivo . Our studies provide the first structural insight into how the malfunction of centriole proteins results in human disease and also reveal that the CPAP–STIL interaction constitutes a conserved key step in centriole biogenesis .
Centrioles are small cylindrical organelles whose outer walls contain a ninefold symmetric array of microtubule triplets . These structures form the basal bodies that template the assembly of cilia and flagella , and they also organise a proteinaceous matrix termed the pericentriolar material ( PCM ) to form centrosomes , the main microtubule organising centres in animal cells . These organelles play an important part in many aspects of cell organisation , and centriolar dysfunction is linked to a plethora of human diseases , including cancer , obesity , macular degeneration and polycystic kidney disease ( Nigg and Raff , 2009; Bettencourt-Dias et al . , 2011 ) . Recently , an unexpected genetic link has emerged between centriole/centrosome assembly and human brain size . Autosomal recessive primary microcephaly ( MCPH ) is a rare condition where patients are born with small brains ( Thornton and Woods , 2009 ) . All eight identified MCPH genes encode proteins that localise to centrioles and/or centrosomes/spindle poles ( Thornton and Woods , 2009; Hussain et al . , 2012 ) . It is unclear why mutations in these proteins are linked to such a specific neuro-developmental problem in humans , but it seems likely that some aspect of centriole/centrosome function must be particularly important for the proper proliferation of human neural progenitors ( Siller and Doe , 2009; Megraw et al . , 2011 ) . In support of this possibility , mutations in the centriolar components CPAP ( DSas-4 in Drosophila , here called dCPAP ) and STIL ( Ana2 in Drosophila , here called dSTIL ) in flies lead to defects in the asymmetric division of larval neural stem/progenitor cells ( Basto et al . , 2006 ) . Mutations in MCPH proteins in mice , however , lead to complex phenotypes that can include , but are not restricted to , microcephaly ( McIntyre et al . , 2012 ) . Moreover , compelling genetic links are now emerging between centrioles/centrosomes and DNA damage repair ( DDR ) pathways: mutations in certain MCPH genes and in genes encoding other centriole/centrosome proteins can lead to Seckel syndrome and MOPD , pathologies normally associated with defects in DDR ( Megraw et al . , 2011 ) . Thus , the cellular mechanisms that lead to pathology when centriole/centrosome proteins are mutated in humans remain unclear . Centrioles are complex structures , but work in several model systems revealed only a small number of conserved proteins to be important for centriole assembly . These include PLK4/SAK , SAS-6 , STIL/Ana2 , CPAP/CenpJ/SAS-4 , Cep152/Asl , and CEP135 ( Brito et al . , 2012; Gonczy , 2012 ) . Several studies have identified a complex web of putative interactions between these proteins ( Cizmecioglu et al . , 2010; Dzhindzhev et al . , 2010; Hatch et al . , 2010; Tang et al . , 2011; Vulprecht et al . , 2012; Lin et al . , 2013 ) . However , an understanding of centriole architecture and its assembly mechanisms will ultimately require high-resolution structures of the key centriolar components and their complexes . The power of combining structural studies with protein biochemistry and functional in vivo experiments has been demonstrated by work on SAS-6 . These studies revealed how SAS-6 homo-oligomerises to organise the central cartwheel ( Kitagawa et al . , 2011b; van Breugel et al . , 2011 ) , the earliest structurally defined intermediate in centriole assembly ( Brito et al . , 2012; Gonczy , 2012 ) , and suggested how SAS-6 might interact with SAS-5 , the proposed STIL homologue in worms ( Qiao et al . , 2012 ) . Additionally , high-resolution structures of Sak/Plk4 fragments have recently been solved ( Leung et al . , 2002; Slevin et al . , 2012 ) . However , equivalent studies with other core centriolar components or especially their complexes are currently missing , and how any of these proteins might be structurally and mechanistically compromised in MCPH is not known . Of particular interest in this regard is the putative centriolar CPAP–STIL complex , as mutations in both components result in MCPH ( Leal et al . , 2003; Bond et al . , 2005; Gul et al . , 2006; Darvish et al . , 2010 ) . CPAP and STIL are strictly required for centriole assembly: STIL at a very early stage ( Stevens et al . , 2010b; Tang et al . , 2011; Kitagawa et al . , 2011a; Arquint et al . , 2012; Vulprecht et al . , 2012 ) and CPAP slightly later ( Kirkham et al . , 2003; Leidel and Gonczy , 2003; Basto et al . , 2006; Kleylein-Sohn et al . , 2007; Dobbelaere et al . , 2008; Vulprecht et al . , 2012 ) , possibly by controlling the organisation ( Pelletier et al . , 2006; Dammermann et al . , 2008 ) and length of the centriolar microtubules ( Blachon et al . , 2009; Kohlmaier et al . , 2009; Schmidt et al . , 2009; Tang et al . , 2009; Kim et al . , 2012 ) . A direct interaction between STIL and CPAP has been observed in yeast-two-hybrid and pull-down experiments ( Tang et al . , 2011; Vulprecht et al . , 2012 ) . Intriguingly , a MCPH mutation ( E1235V ) in the conserved C-terminal domain of CPAP ( the so-called TCP-domain or G-Box ) appeared to weaken this yeast-two-hybrid interaction ( Tang et al . , 2011 ) . Tissue culture experiments suggested that this MCPH mutation might cause a partial loss-of-function of CPAP ( Kitagawa et al . , 2011a ) . However , the same study also found that the E1235V mutation results in an enhanced functionality of CPAP when overexpressed in vivo ( Kitagawa et al . , 2011a ) . To understand how CPAP and STIL interact and how the MCPH mutation affects CPAP functionality in vitro and in vivo , we undertook a detailed biochemical , structural and functional study of the putative CPAP–STIL complex .
Yeast-two-hybrid experiments suggested that a region of human CPAP comprising its conserved C-terminal TCP domain ( or G-box ) can interact with a ∼400 amino acid ( aa ) region ( residues 231–619 ) of human STIL ( Tang et al . , 2011 ) . To try to identify the region of STIL most likely to be involved in an interaction with CPAP , we carried out a sequence alignment with multiple metazoan STIL proteins ( Figure 1A , Figure 1—figure supplement 1 ) . This analysis revealed a short ( ∼40 aa ) highly conserved proline-rich region ( CR2 ) ( Figure 1A ) within this interval . To test whether this region of STIL could bind to the CPAP TCP domain , we recombinantly produced the TCP domain of Danio rerio CPAP and used isothermal titration calorimetry ( ITC ) to test its ability to bind to a fragment of D . rerio STIL that spanned CR2 ( residues 404–448 ) ( Figure 1D , Table 1 ) . The two proteins formed a 1:1 complex with a KD of ∼2 μM . Next , we further split the peptide to test the binding contribution from its N-terminal ( residues 411–428 ) and C-terminal region ( residues 429–448 ) . The N-terminal region exhibited an only slightly weaker binding ( KD ∼4 μM ) to the TCP domain , whereas the C-terminal region showed a very weak binding ( KD > 500 μM ) ( Figure 1D; Table 1 ) . We conclude that the CPAP TCP domain binds to a short conserved motif in STIL ( CR2 ) with a potentially biologically significant affinity , and that the majority of the binding affinity comes from interactions with residues within the first proline-rich region in CR2 . 10 . 7554/eLife . 01071 . 003Figure 1 . Biochemical and structural characterisation of the CPAP TCP domain and its interaction with STIL . ( A ) Schematic representation of D . rerio CPAP and STIL . CPAP is a 1124 amino acid ( aa ) protein with three predicted coiled coil ( cc ) domains and a C-terminal TCP domain . STIL is a 1263 aa protein with one predicted cc domain and several conserved regions ( CR ) . The proline-rich CR2 domain is enlarged and coloured according to Consurf conservation scores ( Glaser et al . , 2003 ) from cyan ( variable ) to burgundy ( conserved ) . The constructs used in this study are indicated by bars . ( B ) Two views of the TCP domain structure ( green ) in complex with the STIL peptide ( orange ) , rotated by 180° . Images on the left of each view show a ribbon representation and images on the right show the TCP domain as a molecular surface coloured according to Consurf conservation scores . Note the presence of a conserved patch ( dashed circle ) along the edge of the TCP domain where the STIL peptide is bound . This patch contains aromatic residues ( black sticks ) that would be well placed to interact with conserved prolines in the C-terminal part of the STIL CR2 region that we had to omit for crystallisation . ITC experiments ( Figure 1D ) suggest that these putative additional contacts would only contribute weakly to overall binding . ( C ) Detailed view of the D . rerio CPAP–STIL interaction interface coloured according to Consurf conservation scores . Interface residues are shown in sticks , and the TCP domain is shown as a semi-transparent molecular surface . Contact residues are labelled in green ( CPAP ) and orange ( STIL ) . Dotted yellow lines indicate hydrogen-bonds . The dark orange sphere represents a bound water molecule . ( D ) ITC analysis using the STIL constructs shown in Figure 1A . The excess heat measured on titrating STIL into CPAP at 25°C was fitted to a single set of binding sites model . Fitted KD values are indicated together with their standard deviations . ( E and F ) Ribbon models of the apo-structures of the D . rerio CPAP TCP domain: ( E ) WT apo-structure; ( F ) E1021V ( MCPH mutation ) apo-structure ( V1021 represented as red spheres ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00310 . 7554/eLife . 01071 . 004Figure 1—figure supplement 1 . Multiple sequence alignment of the conserved proline-rich region of STIL ( CR2 ) . The numbering refers to D . rerio STIL . The alignment is coloured by conservation according to the Consurf conservation score from cyan ( variable ) to burgundy ( conserved ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00410 . 7554/eLife . 01071 . 005Figure 1—figure supplement 2 . The TCP domain sequence repeats . ( A ) Alignment of the sequence repeats of D . rerio CPAP937–1124 . Residues are coloured according to the Clustalx colour scheme . R , Repeat . Repeat 1 was not visible in the electron density map of D . rerio apo-CPAP937–1124 but could be seen partially in the structure of the complex between D . rerio CPAP937–1124 and STIL408–428 . ( B ) Sequence logo of the CPAP937–1124 repeat with the relative residue frequencies at each position . Prominent features of this repeat are two PDG motifs and the high frequency of aromatic residues adjacent to the first PDG motif in position 6 of the repeat . ( C ) Left: ribbon presentation of the D . rerio apo CPAP937–1124 structure with its sequence repeats rainbow-coloured from N- to C-terminus . R , Repeat . An individual structural repeat consists of a β-hairpin . The aromatic residues found in position 6 of the repeat are shown in black sticks . These aromatic residues run along the edge of one side of the β-sheet , where the proline-rich STIL peptide binds . The PDG motifs frequently constitute the β-turns of the TCP repeats . Boxed are three of these turns that are presented on the right as a close-up . In this close-up , residues of the PDG motif are labelled and shown in sticks . The Asp residue in this motif hydrogen-bonds ( dotted black lines ) to the main-chain of the ( n ) + 1 neighbouring residue . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00510 . 7554/eLife . 01071 . 006Figure 1—figure supplement 3 . The TCP domain of CPAP is predominantly monomeric in solution . ( A ) Panel showing size exclusion chromatography coupled to multi-angle light scattering ( SEC-MALS ) chromatograms . CPAP937–1124 was injected at concentrations of approximately 70 μM ( light grey ) , 460 μM ( dark grey ) and 2 . 4 mM ( black ) . The corresponding chromatogram traces ( thin lines ) show the refractive index signal . The concentrations of the CPAP937–1124 monomer measured at the peaks are indicated , and heavy solid lines across the peaks show the calculated molar masses . When averaged across the central 50% of the peaks , these molar masses were 22 kDa , 23 kDa , and 29 kDa , respectively . The theoretical molecular weight of a CPAP937–1124 monomer is 22 kDa . The Rh of CPAP937–1124 determined at the intermediate concentration was 2 . 9 ± 0 . 15 nm , which is significantly larger than expected for a globular protein of this mass and thus is consistent with the extended crystallographic structure . SEC-MALS measurements were performed using a Wyatt Heleos II 18 angle light scattering instrument coupled to a Wyatt Optilab rEX online refractive index detector . Detector 12 in the Heleos instrument was replaced with Wyatt's QELS detector for dynamic light scattering measurement . Samples ( 100 μl ) were resolved on a Superdex S-200 10/300 analytical gel filtration column ( GE Healthcare , Little Chalfont , UK ) running at 0 . 5 ml/min in 25 mM bis Tris pH 7 . 2 , 100 mM NaCl buffer before passing through the light scattering and refractive index detectors in a standard SEC-MALS format . Protein concentration was determined from the excess differential refractive index based on 0 . 186 RI increment for 1 g/ml protein solution . The concentration and the observed scattered intensity at each point in the chromatograms were used to calculate the absolute molecular mass from the intercept of the Debye plot using Zimm's model as implemented in Wyatt's ASTRA software . Autocorrelation analysis of data from the dynamic light scattering detector was also performed using Wyatt's ASTRA software , and the translational diffusion coefficients determined were used to calculate the hydrodynamic radius using the Stokes-Einstein equation and the measured solvent viscosity of 9 . 3 e-3 Poise . ( B ) Small-angle X-ray scattering ( SAXS ) experiment with approximately 20 μM D . rerio CPAP937–1124 in 25 mM bis Tris pH 7 . 2 , 100 mM NaCl , 2 mM DTT . Shown in blue is the experimentally measured SAXS curve of CPAP937–1124 with the experimental error indicated by black bars . The orange line shows the fitted theoretical SAXS curve of CPAP937–1124 derived from its crystal structure . Fitting was done using CRYSOL ( Svergun et al . , 1995 ) and resulted in a χ-value of 1 . 416 . A . U . , arbitrary units . At higher CPAP937–1124 concentrations the fit became less good due to the tendency of CPAP937–1124 to self-associate at these concentrations as revealed by a gradual increase of the derived Rg values . SAXS data were collected at the European Synchrotron Radiation Facility ( ESRF ) , Grenoble , France , at beamline ID14–3 . Measurements were done at 10°C at a wavelength of 0 . 931 Å with the standard beamline settings using a PILATUS 1M detector ( Dectris , Baden , Switzerland ) . To minimise radiation damage , a flow cell was used for the measurements . Collected data was buffer subtracted using PRIMUS ( Konarev et al . , 2003 ) and the beamstop shadow removed by cutting the data at a q-value of 0 . 055 nm−1 . Above a q-value of 3 . 8 nm−1 the data became too noisy to be interpretable and the data were therefore cut at this value . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00610 . 7554/eLife . 01071 . 007Figure 1—figure supplement 4 . The TCP domain resembles engineered peptide-assembly mimics used to study β-rich self-assemblies . Side-by-side comparison of the apo-structure of the D . rerio TCP-domain of CPAP ( left ) with the structure of an engineered peptide-assembly mimic based on Borrelia OspA ( right ) that is used to study β-rich self-assemblies ( PDB code 2FKJ , chain A ) . Structures are shown as ribbon presentations and are rainbow-coloured from N- to C-terminus . Note that the conformation of the peptide-self-assembly mimic is maintained by two globular domains capping both ends of its β-sheet . In contrast , the TCP domain entirely lacks a hydrophobic core . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00710 . 7554/eLife . 01071 . 008Figure 1—figure supplement 5 . Multiple sequence alignment of the TCP domain of CPAP . The numbering refers to D . rerio CPAP . The alignment is coloured by conservation according to the Consurf conservation score from cyan ( variable ) to burgundy ( conserved ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 00810 . 7554/eLife . 01071 . 009Table 1 . Characterisation of the CPAP:STIL interaction in vitroDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 009Danio rerio STIL peptide in syringeDanio rerio CPAP937–1124 TCP domain in cellNumber of binding sites ( N ) SD NKD ( μM ) SD KD ( μM ) ΔH ( kcal/mol ) SD ( kcal/mol ) n ( number of measurements ) Factor change in KDSTIL404–448WT1 . 070 . 041 . 90 . 2−10 . 10 . 351STIL411–428WT0 . 980 . 0340 . 3−11 . 30 . 532STIL429–448WT0 . 970 . 08540130−6 . 30 . 62∼280Binding parameters between D . rerio CPAP and various D . rerio STIL constructs obtained from ITC experiments . Fitting was performed with N as a variable . Constraining N to a fixed value of 1 during fitting produced KD values that were within the experimental error of those tabulated here . To understand how CPAP and STIL interact at the molecular level , we obtained the crystal structures of the TCP-domain of D . rerio CPAP937–1124 , both on its own and in a complex with D . rerio STIL408–428 ( Figure 1B , C , E; Table 2 , Table 3 ) . In both structures , the TCP domain adopts a nearly identical conformation , suggesting that no significant conformational change occurs in CPAP upon binding to STIL ( RMSD = 1 . 5 Å ± 0 . 2 Å over 148 ± 4 Cα pairs ) . The TCP domain folds into a single layer β-sheet comprising ∼20 consecutive antiparallel strands connected by type I β-turns and is stabilised by an extensive hydrogen-bonding network . The resulting sheet shows a twist of approximately 13° ( i . e . , the angle between the consecutive , hydrogen-bonded strands ) , slightly lower than the average value of 20° observed for typical β-sheets ( Chothia , 1973; Murzin , 1992 ) . Individual β-hairpins correspond to previously noted ( Islam et al . , 1993; Hung et al . , 2000 ) repeats in the TCP domain sequence; the turns of these hairpins are often constituted by a PDG motif explaining the high frequency of proline and glycine residues in this domain ( Figure 1—figure supplement 2 ) . Crystal packing interactions involve only small protein interfaces , suggesting that the protein is biologically active as a monomer . Indeed both small-angle X-ray scattering ( SAXS ) and size-exclusion chromatography—multi angle light scattering ( SEC-MALS ) experiments demonstrate that the TCP domain is predominantly monomeric in solution ( Figure 1—figure supplement 3 ) . The structure of the TCP domain represents an unusual , novel architecture . It is reminiscent of the β-sheet conformation proposed to exist within amyloid fibrils and resembles engineered water-soluble peptide self-assembly mimics ( PSAMs ) used to study β-rich self-assemblies ( Makabe et al . , 2006 ) . In contrast to these PSAM structures whose conformation is maintained by two globular domains capping both ends of the β-sheet , the TCP domain stably exists on its own . ( Figure 1—figure supplement 4 ) . The TCP domain structure lacks a defined hydrophobic core typical for globular domains , and both sides of its β-sheet are exposed to the solvent and well hydrated . The structure of the CPAP–STIL complex revealed that the STIL peptide binds in a polyproline II helical conformation along one edge of the TCP domain β-sheet . The STIL peptide binds to CPAP by four main mechanisms ( Figure 1C ) . First , three STIL prolines ( P417 , P421 , and P423 ) pack against aromatic CPAP residues ( F978 , Y996 , and F1015 ) in a way that resembles target motif recognition by other described proline-rich motif ( PRM ) binding domains ( Kay et al . , 2000 ) . Second , R418 ( STIL ) makes a cation-π interaction with the phenyl ring of Y994 ( CPAP ) . Third , STIL R418 is further involved in a water-mediated hydrogen bonding network that includes CPAP residues H1003 and T1005 . Finally , sidechain–mainchain interactions are formed between CPAP residues Y994 , Q1019 , and E1021 and the bound STIL peptide . The CPAP and STIL residues involved in this interaction are highly conserved across metazoans ( Figure 1C ) . Sequence conservation of the TCP domain is not confined to this section of our structure but extends further along the same edge of the sheet ( Figure 1B ) . This additional conserved region contains aromatic residues that are arranged similar to those that pack against the proline residues of the bound STIL peptide in our crystal structure ( Figure 1B , C ) . Intriguingly , the C-terminal part of STIL’s CR2 region ( omitted to obtain diffraction grade crystals ) contains two highly conserved proline residues ( P435 and P438 in D . rerio ) that would be well positioned to bind to these aromatic residues in an analogous way ( Figure 1A , B ) . Thus , we speculate that the entire CR2 region of STIL spanning from residue 417 to residue 438 ( D . rerio ) may be bound all along the edge of the TCP domain . Although our ITC experiments suggest that these putative additional contacts are insufficient to establish strong binding between STIL and CPAP ( Figure 1D ) they may contribute cooperatively to the CPAP–STIL interaction once the N-terminal proline-rich region in CR2 established binding . We conclude that the TCP domain of CPAP adopts a unique extended open β-sheet conformation that recognises a series of conserved prolines in the CR2 region of STIL . The involvement of CPAP E1021 in the interaction with STIL in zebrafish is potentially significant , as the equivalent residue in human CPAP ( E1235 ) is mutated to valine in some MCPH patients . To test whether this mutation disrupts the organisation of the TCP domain , we obtained the crystal structure of D . rerio CPAP937–1124 carrying the E1021V mutation ( Figure 1F; Table 2 ) . The structure of the wild-type and the mutant TCP domain were virtually identical ( RMSD = 0 . 1 Å over 142 Cα pairs ) demonstrating that the TCP domain structure was not compromised . To test whether this mutation perturbed the interaction with STIL , we purified WT and various other mutant forms of D . rerio CPAP937–1124 in which we valine substituted residues that our crystal structure suggested to be important for binding ( Figure 2B ) . Circular dichroism ( CD ) spectra indicated that the mutant forms of the TCP domain were correctly folded with a predominantly β-type profile ( Figure 2—figure supplement 1 ) . ITC experiments with WT D . rerio STIL404–448 showed that the mutation of residues F978 , Y994 , and F1015 decreased the binding strength by ∼20 to 40-fold ( Figure 2A , left; Table 4 ) , while mutation of E1021 decreased the binding strength by approximately eightfold . In contrast , mutation of T986 , which is not predicted to be in the interaction interface , did not detectably perturb binding . 10 . 7554/eLife . 01071 . 010Figure 2 . Mutational analysis of the CPAP:STIL interaction in vitro and conservation of the interaction across species . ( A ) Graphs showing the binding constants ( KD ) determined by ITC for the interaction between WT and mutant constructs of CPAP937–1124 and STIL404–448 . Left panel , WT and various mutant forms of CPAP937–1124 binding to WT STIL404–448 ( T986 is a non-interacting residue included as a negative control ) . Error bars , standard deviation . Right panel , WT and various mutant STIL404–448 constructs binding to WT CPAP937–1124 ( N422 is a non-interacting residue included as a negative control ) . Error bars , standard deviation . The wild-type measurements are the same as shown in Figure 1D and are shown again for comparison to the mutants . ( B and C ) Close-up view of the CPAP ( green ) :STIL ( orange ) interaction interface from D . rerio ( B ) and Drosophila ( C ) . Interface residues are shown as sticks , in yellow is the Glutamate residue in Drosophila and D . rerio CPAP that is equivalent to E1235 in human CPAP ( mutated in MCPH ) . Residues of the D . rerio protein mutated for ITC experiments are ringed in green ( CPAP ) or red ( STIL ) . Dotted black lines indicate hydrogen-bonds . The conserved bound water molecule is shown as a red sphere . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 01010 . 7554/eLife . 01071 . 011Figure 2—figure supplement 1 . Characterisation of the D . rerio TCP domain mutants and STIL peptide mutants used for thermodynamic analysis . ( A ) Coomassie stained SDS-PAGE gel of purified , recombinant CPAP937–1124 and its mutants ( as labelled in colours above the gel ) . ( B ) Buffer-subtracted circular dichroism ( CD ) spectra of D . rerio CPAP937–1124 and its mutants in 10 mM Na-Phosphate pH 7 . 3 at approximately 200 μg/ml . Spectra were recorded on a JASCO J-810 from 260 to 190 nm in 0 . 2 nm steps at 20°C and are colour-coded as in ( A ) . The data were cut at 200 nm as the detector was saturated below this wavelength for some constructs as indicated by a high HT voltage . ( C ) Coomassie stained SDS-PAGE gel showing purified , recombinant D . rerio STIL404–448 and its mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 011 We also purified mutant forms of the D . rerio STIL404–448 peptide and tested their ability to interact with the WT D . rerio TCP domain in ITC experiments ( Figure 2A , right , Figure 2—figure supplement 1C; Table 4 ) . Alanine substitution of P417 , R418 or P421 decreased the binding strength by ∼10 to 20-fold and alanine substitution of P423 by approximately twofold to threefold . In contrast , the mutation of residue N422 , which is not predicted to be in the interaction interface , did not compromise binding . Taken together these results lend strong support to our structural model and indicate that the E1021V MCPH mutation leads to roughly an order of magnitude decrease in affinity of the CPAP–STIL interaction . The sequence conservation of the CPAP TCP domain ( Figure 1—figure supplement 5 ) and the CR2 region of STIL ( Figure 1—figure supplement 1 ) suggests that this interaction may be conserved . To confirm this , we solved the crystal structure of the TCP domain from Drosophila melanogaster DSas-4 ( dCPAP ) ( residues 700–901 ) in complex with the region of Ana2 ( dSTIL ) equivalent to CR2 ( residues 1–47 ) ( Table 2 , Table 5 , Table 6; Figure 2C ) . The dSTIL–dCPAP interaction interface in this structure was highly similar to the D . rerio complex ( inter-species alignments of the structures yielded an average pairwise RMSD of 1 . 2 ± 0 . 2 Å across an average of 118 ± 4 Cα pairs ) . Indeed , all copies of the complex obtained in the structures from both species superimposed well and exhibited the same four major groups of binding interactions as described for the D . rerio structure . This conservation includes the contact made by the E792 residue in dCPAP ( the equivalent of the E1235 residue in human CPAP that is mutated in MCPH ) . Together , these data allow us to determine a consensus CPAP binding motif in metazoan STIL proteins ( PRxxPxP , Figure 1—figure supplement 1 ) and suggest that the described CPAP–STIL interaction constitutes a highly conserved step in centriole biogenesis . Since the binding mechanism of CPAP and STIL is conserved between zebrafish and Drosophila , we turned to D . melanogaster as a model system to address the functional relevance of this interaction in vivo . In flies , the lack of dCPAP or dSTIL leads to centriole loss and a consequent severe uncoordinated ( unc ) phenotype due to the lack of basal bodies and so cilia in Type I sensory neurons . These flies lack all mechano- and chemo-sensation and , although viable , they usually die shortly after eclosion , as they cannot feed or move in a coordinated fashion ( Kernan et al . , 1994; Basto et al . , 2006; Wang et al . , 2011 ) . We examined the ability of various GFP-tagged versions of dCPAP and dSTIL to rescue the centriole loss observed in these mutants and assayed their ability to localise to centrosomes in the presence of endogenous dCPAP or dSTIL ( Figure 3 ) . 10 . 7554/eLife . 01071 . 012Figure 3 . The interaction between dCPAP and dSTIL is essential for centriole duplication in Drosophila . ( A ) Schematic view of the complex between dCPAP ( green ) and dSTIL ( magenta ) with the residues mutated in MC1 ( cyan ) , MC2 ( brown ) and MC3 ( dark purple ) indicated as coloured sticks . The MCPH residue E792 is circled in red . Note that MC1 and MC2 are mapped onto the Drosophila structure ( dark-green backbone ) , while MC3 had to be mapped onto the backbone of the D . rerio structure ( light green backbone ) . Although highly conserved between Drosophila and D . rerio ( Figure 1—figure supplement 5 ) this region was not visible in the electron density map of the Drosophila structure probably due to its partial unfolding to enable packing interactions within the crystal . ( B–M ) Panels show representative still images taken from movies of Drosophila embryos expressing the indicated dCPAP-GFP or dSTIL-GFP constructs . Note that all analyses were performed in the presence of endogenous WT dCPAP or dSTIL , and that all images were acquired with the same microscope settings at the same stage of the cell cycle . ( B–F ) dSTIL-GFP constructs localise to centrosomes at similar levels . ( G–M ) All mutant dCPAP-GFP constructs localise to centrosomes , but at strongly reduced levels compared to wild-type dCPAP-GFP . ( N ) Graphs show the percentage of cells with 0 , 1 , 2 , and 3 centrosomes in the genotypes analysed ( as indicated ) . All dSTIL-GFP and dCPAP-GFP constructs were analysed in their respective mutant backgrounds . Note that this experiment was performed blind . ( O–Q′′ ) Panels show third instar larval brain cells of various genotypes in metaphase . Cells were stained for the centriolar protein Asterless ( Asl—green ) and the PCM component Centrosomin ( Cnn—red ) and DNA ( blue ) . Wild-type metaphase cells have two centrosomes ( O ) , whereas centrosomes are mostly absent in third instar larval brain cells from dCPAP mutants ( P ) . As an example , representative images of dCPAP mutant cells expressing the dCPAP_E792V-GFP construct are shown that were scored with 2 ( Q ) , 1 ( Q′ ) or no ( Q′′ ) centrosomes . Scale bars = 3 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 01210 . 7554/eLife . 01071 . 013Figure 3—figure supplement 1 . Protein expression levels of GFP-tagged dCPAP and dSTIL constructs in dCPAP or dSTIL mutant Drosophila brain cells and quantification of their centriole/centrosome numbers . Panels show western blots of third instar larval brain samples probed with antibodies against GFP ( recognising the fusion proteins ) , actin ( as a loading control ) and dCPAP ( recognising both the fusion proteins and the endogenous protein ) ( highlighted by an arrow ) . All fusion proteins were expressed in their respective mutant backgrounds , as indicated by the black lines . Left panel , the dCPAP fusion proteins were expressed at approximately equal levels ( when compared to the actin control ) but were all moderately overexpressed compared to the endogenous dCPAP protein . Right panel , dSTIL_WT , dSTILΔN and dSTILP11AR12A were expressed at slightly higher levels than dSTIL_P11A and dSTIL_R12A , and all GFP fusion proteins were strongly overexpressed when compared to endogenous dSTIL ( data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 013 We first expressed a version of dSTIL that lacks the first 45 aa ( including the PRxxPxP motif required for the interaction with dCPAP ) . GFP-tagged wild-type dSTIL ( dSTIL_WT-GFP ) served as a control . Both proteins were expressed at similar levels and localised strongly to centrosomes in the presence of endogenous dSTIL ( Figure 3B , C; Figure 3—figure supplement 1 ) . Only the wild-type version , however , was able to rescue the unc phenotype and the centriole loss phenotype of the dSTIL mutant ( Figure 3N; Table 7 ) . To further characterise the dCPAP binding domain in vivo we mutated the first proline and arginine of the PRxxPxP motif of dSTIL to alanine , both separately and in combination ( P11A , R12A , and P11A:R12A , Figure 2C ) . All three constructs strongly localised to centrosomes in the presence of endogenous dSTIL ( Figure 3D–F , Figure 3—figure supplement 1 ) . Both single mutants rescued the unc phenotype of the dSTIL mutation while the double mutant failed to do so ( data not shown ) . The single mutants P11A and R12A were also able to partially rescue the centriole loss phenotype , whereas the double mutant P11A:R12A showed only a poor rescue ( Figure 3N; Table 7 ) . These data strongly suggest that the interaction with dCPAP is essential for dSTIL function in centriole assembly . We next deleted the entire TCP-domain of dCPAP ( dCPAP_ΔC ) , or expressed GFP fusion proteins carrying mutation clusters ( MCs ) altering 3–4 residues in different regions of the TCP domain ( Figure 3A ) . Mutation clusters were designed that targeted central ( dCPAP_MC1 ) or peripheral ( dCPAP_MC2 ) residues in the dSTIL binding domain , as well as residues that are predicted to not significantly be involved in complex formation ( dCPAP_MC3 ) , according to the crystal structure and the ITC data ( Figure 3A , Figure 2A , Figure 1D; Table 1 ) . We also analysed dCPAP_E792V-GFP lines , which carried the MCPH equivalent mutation E792V ( E1235V in humans and E1021V in zebrafish CPAP ) . All transgenic dCPAP-GFP proteins were expressed at approximately equivalent levels in vivo , but were moderately overexpressed compared to endogenous dCPAP ( Figure 3—figure supplement 1 ) . Wild-type dCPAP-GFP localised strongly to centrosomes and rescued both the unc phenotype and the centriole loss phenotype ( Figure 3G , N; Table 7 ) . Strikingly , the rescuing ability of the mutant constructs strongly correlated with the predicted strength of dSTIL binding . dCPAP_ΔC-GFP failed to rescue , dCPAP-MC1 and dCPAP-MC2 rescued poorly , the MCPH mutation E792V showed an intermediate phenotype , while dCPAP-MC3 exhibited a robust rescue ( Figure 3N; Table 7 ) . Interestingly , when compared to wild-type dCPAP-GFP , all mutant constructs ( including dCPAP_MC3 ) localised only weakly to centrosomes ( Figure 3H–M ) . Together , these data suggest that the interaction between dCPAP and dSTIL is a key step in centriole assembly and is essential for centriole duplication . Furthermore , they indicate that low total levels of dCPAP at centrosomes might be sufficient for centriole duplication , as long as some interaction with dSTIL is maintained . It has been proposed that SAS-5 is the C . elegans homolog of the STIL proteins in flies and vertebrates , but there is little sequence homology between these proteins ( Stevens et al . , 2010a ) . We failed to identify an unambiguous PRxxPxP motif in worm SAS-5 , so we tested whether the TCP domain of SAS-4 ( the C . elegans CPAP homologue ) is functionally important . We used the Mos single-copy insertion system ( MosSCI; Frøkjær-Jensen et al . , 2008 ) to generate transgenic lines with single-copy transgenes under the control of sas-6 regulatory sequences integrated at a specific site on chromosome II ( Figure 4A ) . Transgenes were generated expressing GFP fusions with either WT SAS-4 ( SAS-4WT::GFP ) or a form in which the C-terminal TCP domain ( aa 557–808 ) had been deleted ( SAS-4ΔTCP::GFP ) ; both transgenes contained a 497 bp resequenced region in their N-terminal coding region ( preserving codon usage ) that rendered them resistant to RNAi-mediated depletion ( Figure 4A ) . 10 . 7554/eLife . 01071 . 014Figure 4 . The TCP domain of C . elegans SAS-4 is required for its interaction with SAS-5 , its localisation to centrioles , and for centriole assembly . ( A ) Schematic illustration of the MosSCI system used for generating single-copy sas-4 transgene insertions . ( B ) A schematic illustration of the monopolar spindle assay for centriole duplication in C . elegans embryos . Panels show maximum intensity projections of representative fluorescence confocal z-series taken of sas-4 ( RNAi ) embryos expressing either WT or ΔTCP SAS-4::GFP . Transgenic SAS-4WT::GFP localises to sharp foci representing the centrioles , whereas SAS-4ΔTCP::GFP localises diffusely to the pericentriolar material . Bar , 10 μM . ( C ) Graphs show the quantification of second division monopolar spindles ( left ) and embryonic viability ( right ) after sas-4 ( RNAi ) and rescue with either a WT or ΔTCP sas-4::gfp transgene . ( D ) Panels show autoradiographs ( top panel ) and a Coomassie stained gel from a Ni-NTA pull-down experiment with 35S-labelled in vitro translated SAS-4 fragments ( prey ) and SAS-5-6xHis fragments ( baits ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 014 SAS-4 depletion by RNAi prevents centriole assembly , resulting in a signature phenotype characterised by a normal first mitotic division followed by monopolar spindles during the second division ( Figure 4B , C; O’Connell et al . , 2001 ) . This phenotype arises because the sperm that fertilise the SAS-4-depleted oocytes carry two normal centrioles , since sperm are produced prior to introduction of the dsRNA . The two sperm-derived centrioles organise two centrosomes , and the first mitotic division appears normal . If no new centrioles form , each daughter cell inherits a single sperm-derived centriole leading to monopolar spindles in the second division and 100% embryonic lethality ( Figure 4C ) . Both the monopolar spindle phenotype and embryonic viability were fully rescued by the WT sas-4::gfp transgene ( aa 1–808 ) , but not by the ΔTCP transgene ( aa 1–556 ) ( Figure 4B , C ) . While SAS-4WT::GFP targeted to centrioles in the absence of the endogenous protein , SAS-4ΔTCP::GFP did not , and instead exhibited a diffuse accumulation in the pericentriolar material ( Figure 4B ) . Thus , the SAS-4 TCP domain is required for SAS-4 to accumulate at centrioles and become incorporated into the microtubule-containing outer centriole wall . To determine if the failure of SAS-4ΔTCP::GFP to become incorporated in the centriole outer wall could be due to an inability to interact with SAS-5 , we performed a pull-down assay to determine whether 35S-labelled in vitro translated SAS-4 fragments could interact with the N-terminal or C-terminal regions of SAS-5 bound to beads ( Figure 4D ) . In vitro translated full-length SAS-4 interacted specifically with the N-terminal domain ( aa 1–202 ) of SAS-5 . Interestingly , we could not further narrow down the region of SAS-4 required for this interaction . Neither the SAS-4 N-terminal nor C-terminal region ( which includes the TCP domain ) alone could be pulled down by SAS-5 . This result suggests that although the TCP domain is required for SAS-4 to interact with SAS-5 , it is not sufficient . Together , these data suggest that a TCP domain-dependent interaction between SAS-4/CPAP and SAS-5/STIL is conserved and essential for centriole duplication in C . elegans , but that the precise interaction interface may have diverged .
Only a small set of conserved centriolar proteins is essential for centriole assembly ( Brito et al . , 2012; Gonczy , 2012 ) and some of these proteins , like CPAP and STIL , have been linked to microcephaly in humans ( Leal et al . , 2003; Bond et al . , 2005; Gul et al . , 2006; Thornton and Woods , 2009; Darvish et al . , 2010 ) . However , there is currently little structural understanding on how these proteins interact with one another , how mutations in them cause microcephaly in humans and how these interactions are regulated . Here we have solved the crystal structures of the CPAP–STIL complex from zebrafish and Drosophila . We showed that the CPAP TCP domain folds into an elongated open-sided β-meander that consists of ∼20 consecutive antiparallel β-strands connected by type I β-turns . β-meanders are frequently found in β-barrels , β-propellers and some α+β proteins . However , what , to our knowledge , makes the TCP domain structure unique amongst naturally occurring proteins is that it solely consists of a freestanding meander β-sheet that entirely lacks a defined hydrophobic core and is not flanked by other globular domains that pack against it . We show that the TCP domain is predominantly monomeric in solution and self-interacts in its crystallised form only through small interfaces that are not conserved . Thus , despite some reminiscence to β-sheets observed in amyloid fibrils it is unlikely that the TCP domain self-associates in a similar manner . Instead , we demonstrate that the TCP domain of CPAP constitutes a novel proline-rich-motif ( PRM ) recognition-domain that specifically binds to a short target motif in STIL . Although the overall sequence identity of the CPAP and STIL proteins between Drosophila and zebrafish is relatively low ( ∼22% and ∼13% , respectively ) , our structural analysis revealed that the interaction interface is conserved , confirming the previous proposal that fly Ana2 is the functional homologue of vertebrate STIL ( Stevens et al . , 2010a ) . Our characterisation of the binding interface also allowed us to define a consensus-binding site ( PRxxPxP ) for the CPAP TCP domain in STIL that is conserved across metazoa . Our mutational analysis of the interface demonstrates a remarkable correlation between the ability of mutant proteins to bind to one another in vitro and their ability to support centriole assembly in vivo , providing compelling support for our structural model of the metazoan CPAP–STIL complex . These data strongly suggest that the interaction between CPAP and STIL is a conserved , essential step in centriole biogenesis . A schematic model that places this interaction in the context of a possible centriole assembly pathway is shown in Figure 5 . 10 . 7554/eLife . 01071 . 015Figure 5 . A schematic representation of protein interactions within the inner region of the centriole . In this illustration , interactions whose crystal structure have been determined are highlighted by green boxes—all other interactions are inferred from biochemical and genetic studies and so are depicted in cartoon form . The cartwheel central hub comprises SAS-6 ( red ) ( Nakazawa et al . , 2007; Kitagawa et al . , 2011b; van Breugel et al . , 2011 ) . The spokes extending outward from the hub consist of a homodimeric SAS-6 coiled-coil , which extends ( van Breugel et al . , 2011 ) into a region known as the ‘pinhead’ ( cyan in low magnification view , left ) , where CEP135 ( grey ) may act as a linker between SAS-6 , CPAP and microtubules ( Hiraki et al . , 2007; Roque et al . , 2012; Lin et al . , 2013 ) . CPAP ( dark blue ) localises more towards the periphery of the centriole ( Mennella et al . , 2012; Sonnen et al . , 2012; Lukinavičius et al . , 2013 ) , where its N-terminal part may interact directly with both Asterless/CEP152 ( Cizmecioglu et al . , 2010; Dzhindzhev et al . , 2010 ) ( orange arrow ) and microtubules ( Hsu et al . , 2008 ) ( green arrow ) . In contrast STIL ( yellow ) localises more towards the interior of centrioles ( Arquint et al . , 2012 ) , and appears to function upstream of CPAP in centriole biogenesis ( Tang et al . , 2011; Vulprecht et al . , 2012 ) . Thus , we propose that the C-terminal TCP domain of CPAP interacts with the conserved region 2 ( CR2 ) of STIL towards the interior of the centriole and that this interaction is crucial for CPAP/STIL function at centrioles . The orientation of STIL in centrioles is unknown . DOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 015 The high degree of sequence divergence between vertebrate STIL , Drosophila Ana2 and C . elegans SAS-5 suggests that STIL homologs are under particularly strong lineage-specific selection . Despite the many sequence changes between Drosophila Ana2 and vertebrate STIL , our work suggests that the interaction interface between Ana2/STIL and dSAS-4/CPAP TCP domain has been retained , highlighting its importance . Even in C . elegans , which is the most divergent of the functionally characterised STIL homologs , our work indicates that the SAS-4 TCP domain is essential for centriole assembly , and that a TCP-domain dependent interaction between SAS-4 and SAS-5 has been conserved . Nevertheless , as the SAS-4 TCP domain is not sufficient for interaction with SAS-5 and we were unable to identify a PRxxPxP interaction motif in worm SAS-5 , more work will be needed to understand the SAS-4—SAS-5 interaction in C . elegans and its relationship to the CPAP–STIL interaction in other metazoans . A surprising aspect of our findings is that the E792V ( MCPH ) mutant and all three of the mutation clusters ( MCs ) that we analysed in dCPAP localise poorly to centrosomes . For the E792 , MC1 , and MC2 mutations this could be expected , as these are all predicted to perturb the interaction between dCPAP and dSTIL ( as is the case with similar mutations in zebrafish CPAP in our in vitro binding assays ) , and this would be predicted to perturb the recruitment of dCPAP to centrioles . The MC3 cluster , however , is not predicted to lie in a strong interaction interface and , unlike the MC1 and MC2 mutation clusters , it can rescue the centriole duplication defect in dCPAP mutant cells nearly as efficiently as the WT protein . Possibly , an interaction with another protein that plays some part in recruiting dCPAP to centrioles might be perturbed by these mutations . Alternatively , similar to the situation with C . elegans SAS-4 ( Dammermann et al . , 2008 ) , dCPAP may localise to both centrioles and the PCM . It might therefore be PCM and not centriole recruitment that is affected by the mutation clusters . If this were the case it would be hard to discern an additional partial loss of centriole recruitment , as this loss would be masked by the PCM pool of dCPAP , especially under conditions of moderate overexpression of dCPAP . Importantly , however , our findings demonstrate that even very reduced amounts of centrosomal dCPAP can support robust centriole duplication as long as this protein can interact efficiently with dSTIL . Our studies provide the first structural insight into the nature of the link between centrioles and human microcephaly . It is unclear why mutations in genes encoding key centriole or centrosome proteins can lead to such a specific neuro-developmental disorder in humans . It is widely assumed that some aspect of centriole/centrosome function must be particularly important in human neural progenitor cells , and that the failure of these cells to proliferate in an appropriate manner underlies the small brain size in affected individuals ( Megraw et al . , 2011 ) . One possibility , based on the fact that these neural progenitors seem to divide asymmetrically ( Siller and Doe , 2009; Megraw et al . , 2011 ) , is that centrioles/centrosomes may play a particularly important role in properly orienting the spindle during asymmetric divisions , and division orientation could in turn be required for the maintenance of neuronal progenitors . This appears to be the case in flies , where mutations in dCPAP/DSas-4 and dSTIL/Ana2 lead to defects in the asymmetric division of the neural stem/progenitor cells ( Basto et al . , 2006; Wang et al . , 2011 ) . However , there are other possible explanations . Human neural progenitor cells form primary cilia , for example , and signalling through the cilium could be perturbed if centriole assembly is perturbed ( Han and Alvarez-Buylla , 2010; Megraw et al . , 2011 ) . Moreover , several studies have linked centriole and centrosome malfunction to defects in DNA damage repair ( DDR ) pathways ( Megraw et al . , 2011 ) , and mutations in MCPH genes can also lead to more severe phenotypes in humans that may be related to DDR pathway malfunction ( Al-Dosari et al . , 2010; Kalay et al . , 2011; Megraw et al . , 2011 ) . A previous analysis of the behaviour of various CPAP mutant proteins ( modelled on MCPH mutations ) in human cells revealed some surprising findings ( Kitagawa et al . , 2011a ) . The deletion of the TCP domain or the mutation of E1235 to Valine did not effect the localisation of CPAP to the centriole , although centriole duplication was compromised by both mutations . Moreover , overexpression of the E1235V mutant protein was able to promote centriole overgrowth to a greater extent than the WT protein , suggesting that it may have acquired some enhanced functionality . The structures we report here reveal that E1235 is one of the several residues involved in the binding interface with STIL , making an important sidechain–mainchain contact . This structural model explains how the E1235V mutation can compromise complex formation , and we have confirmed that this is the case with zebrafish proteins in vitro . Moreover , the equivalent mutation in flies leads to inefficient centriole assembly , but this process is not abolished . Taken together , our data strongly suggest that it is a partial failure in centriole assembly that is the primary cause of microcephaly in these patients . The challenge now is to understand how inefficient centriole assembly leads to microcephaly in humans .
D . rerio CPAP937–1124 was cloned from D . rerio cDNA . Proteins were expressed in Escherichia coli BL21 ( DE3 ) Rosetta as N-terminally His-tagged constructs , and purified via immobilised metal ion affinity chromatography ( NiNTA; Qiagen , Hilden , Germany ) , proteolytic tag cleavage , followed ( optionally ) by size-exclusion chromatography and ion-exchange chromatography using standard methods . The selenomethionine derivative protein was expressed in selenomethionine supplemented M9 medium and purified in the same way . Purified constructs contained the sequence GPHM at the N-termini that stem from the cloning and protease cleavage sites . D . rerio STIL404–448 was cloned from IMAGE clone 7147918 and expressed in E . coli C41 BL21 ( DE3 ) , fused to two His-tagged lipoyl domains from Bacillus stearothermophilus dihydrolipoamide acetyltransferase at both the N- and C terminus . The peptide was purified via NiNTA chromatography , proteolytic cleavage of the His-lipoyl domains , and ion-exchange chromatography . The purified constructs contained a G ( GG for D . rerio STIL404–448 and its point-mutants ) at their N-terminus and the sequence EFGENLYFQ ( ENLYFQ for D . rerio STIL408–428 and D . rerio STIL411–428 ) at their C-terminus . These extra sequences stem from the cloning and protease cleavage sites . Mutations of the D . rerio constructs were introduced into the expression vectors by site-directed mutagenesis . Codon-optimised ( GeneArt , Carlsbad , CA ) Drosophila dSTIL1–47 was genetically fused to the N-terminus of Drosophila dCPAP700–901 via 3-way ligation . The fusion protein was expressed in E . coli B834 ( DE3 ) as an N-terminally His-tagged fusion , and purified via NiNTA chromatography , proteolytic tag cleavage and size exclusion chromatography . The selenomethionine derivative protein was expressed using SelenoMethionine Medium ( Molecular Dimensions , Newmarket , UK ) and purified in the same way . 10 . 7554/eLife . 01071 . 016Table 2 . Native dataset analysis and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 016D . rerio CPAP937–1124 WTD . rerio CPAP937–1124 E1021VD . rerio CPAP937–1124 + D . rerio STIL408–428 complexD . melanogaster dSTIL1–47 − dCPAP700–901 fusion complexBeamlineDiamond I02MRC-LMB Cambridge UKDiamond I04Diamond I04Space groupP21P21P21P1Wavelength ( Å ) 0 . 97861 . 54180 . 97950 . 9795Monomers in the asymmetric unit1123Unit cell dimensions ( Å ) a = 52 . 34; b = 36 . 44; c = 56 . 44; α = 90 . 00; β = 117 . 31; γ = 90 . 00a = 52 . 12; b = 36 . 48; c = 56 . 46; α = 90 . 00; β = 117 . 47; γ = 90 . 00a = 60 . 25; b = 67 . 47; c = 61 . 65; α = 90 . 00; β = 113 . 92; γ = 90a = 58 . 64; b = 69 . 91; c = 69 . 98; α = 86 . 96; β = 88 . 64; γ = 67 . 69Resolution ( Å ) 29 . 48–1 . 736 . 48–1 . 956 . 35–2 . 7/2 . 2 ( anisotropy ) 64 . 60–2 . 57Completeness ( overall/inner/outer shell ) 99 . 7/99 . 4/100100/99 . 6/10099 . 9/99 . 5/99 . 997 . 6/93 . 6/97 . 2Rmerge ( overall/inner/outer shell ) 0 . 074/0 . 030/0 . 9290 . 096/0 . 028/1 . 0930 . 101/0 . 053/1 . 0080 . 091/0 . 069/0 . 512Rpim ( overall/inner/outer shell ) 0 . 029/0 . 012/0 . 3690 . 039/0 . 012/0 . 4560 . 050/0 . 027/0 . 5050 . 061/0 . 035/0 . 449Mean I/σI ( overall/inner/outer shell ) 14 . 6/39 . 9/2 . 013 . 8/43 . 4/1 . 87 . 6/19 . 0/1 . 47 . 6/16 . 3/1 . 7Multiplicity ( overall/inner/outer shell ) 7 . 2/7 . 0/7 . 36 . 8/6 . 6/6 . 64 . 8/4 . 7/4 . 93 . 1/2 . 9/3 . 1Number of reflections19 , 94114 , 34921 , 89231 , 911Number of atoms1595151531763924Waters1901145465Rwork/Rfree ( % data used ) 19 . 9/24 . 4 ( 5 . 1% ) 20 . 9/26 . 7 ( 5 . 0% ) 23 . 4/27 . 7 ( 5 . 0% ) 24 . 5/26 . 3 ( 5 . 05% ) rmsd from ideal values: bond length/angles0 . 011/1 . 4780 . 009/1 . 3100 . 015/1 . 6190 . 007/0 . 900Mean B value26 . 5231 . 55359 . 6970 . 80Correlation coefficient Fo-Fc/Fo-Fc free0 . 961/0 . 9420 . 955/0 . 9260 . 954/0 . 9330 . 854/0 . 836Molprobity Score0 . 97 ( 100th percentile ) 1 . 2 ( 99th percentile ) 1 . 70 ( 96th percentile ) 1 . 40 ( 100th percentile ) 10 . 7554/eLife . 01071 . 017Table 3 . SeMet D . rerio CPAP937–1124 dataset analysis and phasing statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 017BeamlineESRF ID 23–1Space groupP21Wavelength ( Å ) 0 . 9791 ( Peak ) 0 . 9794 ( Inflection ) 0 . 9393 ( Remote ) Unit cell dimensions ( Å ) a = 52 . 39 b = 36 . 53 c = 56 . 34 α = 90 . 00 β = 117 . 28 γ = 90 . 00a = 52 . 59 b = 36 . 60 c = 56 . 48 α = 90 . 00 β = 117 . 24 γ = 90 . 00a = 52 . 49 b = 36 . 55 c = 56 . 38 α = 90 . 00 β = 117 . 26 γ = 90 . 00Resolution ( Å ) 36 . 56–1 . 736 . 56–1 . 736 . 56–1 . 7Completeness ( overall/inner/outer shell ) 100 . 0/99 . 7/100 . 0100/99 . 2/100100/99 . 7/100Rmerge ( overall/inner/outer shell ) 0 . 09/0 . 048/1 . 2960 . 127/0 . 047/2 . 8400 . 092/0 . 046/1 . 370Rpim ( overall/inner/outer shell ) 0 . 042/0 . 031/0 . 5520 . 056/0 . 028/1 . 2010 . 041/0 . 026/0 . 580Mean I/sd ( I ) ( overall/inner/outer shell ) 10 . 7/26 . 0/1 . 58 . 9/26 . 4/0 . 710 . 8/27 . 4/1 . 4Multiplicity ( overall/inner/outer shell ) 7 . 2/7 . 0/7 . 37 . 2/6 . 9/7 . 37 . 2/7 . 0/7 . 3Se sites found/expected5/7Overall FOM0 . 30610 . 7554/eLife . 01071 . 018Table 4 . Characterisation of the CPAP:STIL interaction in vitroDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 018Danio rerio STIL404–448 peptide in syringeDanio rerio CPAP937–1124 TCP domain in cellNumber of binding sites ( N ) SD NKD ( μM ) SD KD ( μM ) ΔH ( kcal/mol ) SD ( kcal/mol ) n ( number of measurements ) Factor change in KDWTWT1 . 070 . 041 . 90 . 2−10 . 10 . 351WTF978V0 . 700 . 093710−235420WTT986V1 . 010 . 071 . 90 . 2−10 . 60 . 431WTY994V1 . 000 . 336814−9 . 53 . 8536WTF1015V0 . 930 . 137018−10 . 42 . 7337WTE1021V0 . 910 . 13162−8 . 10 . 638WTWT1 . 070 . 041 . 90 . 2−10 . 10 . 351P417AWT1 . 060 . 02371 . 3−11 . 50 . 2320R418AWT1 . 120 . 02191−8 . 80 . 1410P421AWT1 . 160 . 03170 . 3−9 . 80 . 249N422AWT1 . 090 . 030 . 70 . 05−12 . 30 . 340 . 4P423AWT1 . 160 . 054 . 60 . 3−10 . 90 . 442 . 4Tables show the binding parameters between various D . rerio CPAP and STIL constructs obtained from ITC experiments . The measurements of the WT STIL404–448—WT CPAP937–1124 interaction are identical to each other and identical to those shown in Table 1 and are only presented again to allow easier comparison within each table . Fitting was performed with N as a variable . Constraining N to a fixed value of 1 during fitting produced KD values that were within the experimental error of those tabulated here . In control measurements on wild-type material and a selection of mutants of both CPAP and STIL , the experimental configuration was reversed with CPAP protein titrated into STIL peptide in the ITC cell . These experiments gave similar values for N , KD and ΔH to the standard configuration reported here . 10 . 7554/eLife . 01071 . 019Table 5 . D . melanogaster dSTIL1–47-dCPAP700–901 crystallisation conditionsDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 019CrystalProtein concentration ( mg/ml ) Mother liquorµl protein:µl Mother liquorµl seed stockNative6 . 18100 mM MES/imidazole mix pH 6 . 5 , 30 mM MgCl2 , 30 mM CaCl2 , 20% ethylene glycol , 10% PEG 80000 . 15:0 . 05–Semet15 . 00100 mM MES/imidazole mix pH 6 . 5 , 20% ethylene glycol , 10% PEG8000 , 0 . 2 M racemic glutamic acid , 0 . 2 M glycine , 0 . 2 M racemic serine , 0 . 2 M racemic alanine , 0 . 2 M racemic lysine HCl0 . 1:0 . 1–Semet25 . 29100 mM MES/imidazole mix pH 6 . 5 , 14% ethylene glycol , 7% PEG8000 , 30 mM NaNO3 , 30 mM NaPO4 , 30 mM NH4SO40 . 3:0 . 10 . 05Semet35 . 29100 mM MES/imidazole mix pH 6 . 5 , 14% ethylene glycol , 7% PEG8000 , 30 mM NaNO3 , 30 mM NaPO4 , 30 mM NH4SO40 . 3:0 . 10 . 05Semet45 . 29100 mM MES/imidazole mix pH 6 . 5 , 16% ethylene glycol , 8% PEG8000 , 30 mM NaNO3 , 30 mM NaPO4 , 30 mM NH4SO40 . 3:0 . 10 . 0510 . 7554/eLife . 01071 . 020Table 6 . D . melanogaster dSTIL1–47-dCPAP700–901 SeMet dataset analysisDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 020Semet1-PEAKSEMET1-LREMSemet2-PeakSemet3-PEAKSemet3-INFLSemet4-PeakBeamlineDiamond IO4Diamond IO4Diamond IO3Diamond IO3Diamond IO3Diamond IO3SpacegroupP1P1P1P1P1P1Wavelength0 . 97950 . 99990 . 97920 . 97910 . 97940 . 9791Unit cell dimensions ( Å ) a = 59 . 31 b = 70 . 02 c = 70 . 01 α = 87 . 65 β = 89 . 24 γ = 67 . 37a = 59 . 14 b = 70 . 24 c = 70 . 13 α = 87 . 62 β = 89 . 12 γ = 67 . 35a = 58 . 47 b = 70 . 15 c = 69 . 99 α = 87 . 08 β = 88 . 41 γ = 67 . 60a = 58 . 56 b = 70 . 03 c = 70 . 14 α = 86 . 93 β = 88 . 39 γ = 68 . 09a = 58 . 72 b = 70 . 06 c = 70 . 28 α = 86 . 84 β = 88 . 47 γ = 68 . 36a = 59 . 01 b = 70 . 17 c = 70 . 15 α = 87 . 16 β = 88 . 64 γ = 67 . 58Resolution ( Å ) 54 . 74–3 . 5064 . 77–3 . 5064 . 80–3 . 4470 . 04–3 . 5070 . 17–4 . 6064 . 80–3 . 36Completeness ( overall/inner/outer ) 98 . 1/93 . 8/98 . 398 . 4/98 . 0/98 . 297 . 8/91 . 6/93 . 798 . 3/95 . 3/97 . 697 . 8/79 . 1/89 . 997 . 6/91 . 0/97 . 4Rmerge ( overall/inner/outer ) 0 . 093/0 . 055/0 . 1180 . 086/0 . 041/0 . 2380 . 17/0 . 076/0 . 5180 . 152/0 . 086/0 . 3360 . 116/0 . 039/0 . 1890 . 125/0 . 037/0 . 433Rpim ( overall/inner/outer ) 0 . 071/0 . 047/0 . 1360 . 062/0 . 029/0 . 1720 . 078/0 . 040/0 . 2290 . 075/0 . 043/0 . 1840 . 087/0 . 034/0 . 1410 . 100/0 . 038/0 . 323I/σI ( overall/inner/outer ) 9 . 1/16 . 9/5 . 610 . 9/24 . 7/4 . 97 . 3/18 . 8/3 . 59 . 1/27 . 2/3 . 66 . 0/20 . 8/4 . 66 . 9/22 . 5/2 . 7Multiplicity ( overall/inner/outer ) 3 . 9/3 . 8/3 . 93 . 9/3 . 8/3 . 87 . 0/7 . 0/7 . 16 . 0/6 . 7/5 . 33 . 5/3 . 6/3 . 53 . 5/3 . 4/3 . 6No . unique reflections12 , 83212 , 88413 , 31512 , 797564114 , 46110 . 7554/eLife . 01071 . 021Table 7 . Quantification of centriole/centrosome numbers in dCPAP or dSTIL mutant larval brain cells expressing the indicated WT or mutant constructsDOI: http://dx . doi . org/10 . 7554/eLife . 01071 . 021GenotypeNumber of brainsTotal number of cellsCells with centrosome number ( % ) 0123WT129442 . 12 . 695 . 20 . 0dCPAP866195 . 24 . 20 . 60 . 0dCPAP_WT-GFP97152 . 86 . 390 . 50 . 4dCPAP_ΔC-GFP13114795 . 13 . 81 . 00 . 0dCPAP_MC1-GFP1196864 . 930 . 14 . 80 . 3dCPAP_MC2-GFP17105343 . 542 . 214 . 10 . 3dCPAP_MC3-GFP1618704 . 513 . 381 . 90 . 3dCPAP_MC1-3-GFP1188890 . 18 . 01 . 80 . 1dCPAP_E792V-GFP9101513 . 731 . 154 . 80 . 4dSTIL169984698 . 11 . 90 . 00 . 0dSTIL719998099 . 20 . 80 . 00 . 0dSTIL_WT-GFP64245 . 79 . 085 . 10 . 2dSTIL_ΔN-GFP1388488 . 19 . 81 . 90 . 1dSTIL_P11A-GFP9100811 . 041 . 048 . 00 . 0dSTIL_R12A-GFP97097 . 031 . 062 . 00 . 0dSTIL_P11AR12A-GFP972741 . 043 . 016 . 00 . 0 Native D . rerio CPAP937–1124 was crystallised in sitting drops in 80 mM Tris pH 8 . 5 , 160 mM MgCl2 , 20% PEG-4000 , 18% Glycerol , 1 mM DTT at 19 . 5°C . The drops were set up using 1 μl of the protein solution and 0 . 5 μl of the reservoir solution . Crystals were mounted after 3 days and flash-frozen in liquid nitrogen . D . rerio CPAP937–1124 E1021V was crystallised in sitting drops in 80 mM Tris pH 8 . 5 , 160 mM MgCl2 , 24% PEG-4000 , 20% glycerol at 19 . 5°C . Crystals were mounted after 3 days and flash-frozen in liquid nitrogen . SeMet D . rerio CPAP937–1124 crystals were obtained using the sitting drop method with a reservoir solution of 80 mM Tris pH 8 . 5 , 160 mM MgCl2 , 26% PEG-4000 , 18% glycerol , 1 mM DTT at 19 . 5°C . Drops were set up using 1 μl protein solution and 1 μl of reservoir solution . Native CPAP937–1124 crystals were used for streak-seeding into these drops and crystals allowed to grow for 7 days before mounting and flash-freezing them in liquid nitrogen . Crystals of the complex of D . rerio CPAP937–1124 and D . rerio STIL408–428 were initially obtained using the LMB screening set-up with the Clear Strategy 2 pH 8 . 5 screen ( MDL , Newmarket , UK ) . Crystals were used to streak seed into sitting drops consisting of 1 μl of a reservoir solution of 100 mM Tris pH 8 . 5 , 200 mM CaAcetate , 17% PEG-2000 MME and 1 μl of protein/peptide mixture ( 0 . 25 μl protein + 0 . 75 μl peptide ) at 19 . 5°C . Crystals were grown for 2 days before mounting them in 100 mM Tris pH 8 . 5 , 200 mM CaAcetate , 17% PEG-2000 MME , 25% glycerol and flash-freezing them in liquid nitrogen . The protein concentrations of D . rerio CPAP937–1124 used for crystallisations were measured by the Bradford assay with BSA as a standard and were 39 . 9 mg/ml ( apo-CPAP937–1124 ) , 46 . 2 mg/ml ( CPAP937–1124 E1021V ) , 30 . 6 mg/ml ( SeMet CPAP937–1124 ) and 81 mg/ml CPAP937–1124 ( 3 . 7 mM , CPAP/STIL complex ) . The concentration of STIL408–428 ( CPAP/STIL complex ) was determined by amino acid analysis and was 11 . 8 mg/ml ( 3 . 8 mM ) . Native D . melanogaster dSTIL1–47-dCPAP700–901 was crystallised using the sitting drop approach , using the Morpheus screen ( Molecular Dimensions ) . Crystals grew after approximately 3 weeks ( Table 5 , ‘Native’ ) . Crystals were mounted after approximately 4 weeks . SeMet D . melanogaster dSTIL1–47-dCPAP700–901 was initially crystallised using the Morpheus screen ( Molecular Dimensions ) . Crystals typically grew after 3–4 weeks . Some crystals were used for microseeding of further screens including an optimisation screen . Seed stock was generated using a Seed bead kit ( Hampton , Aliso Viejo , CA ) . Details of crystallisation conditions are shown in Table 5 . Native data were collected as described in Table 2 . All D . rerio datasets were integrated and scaled using MOSFLM ( Leslie and Powell , 2007 ) and Scala ( Evans , 2006 ) respectively . The D . rerio CPAP937–1124 structure was solved by MAD in CRANK ( Ness et al . , 2004; Cowtan , 2006 ) , resulting in clear electron density into which an initial model was built using ArpWarp ( Langer et al . , 2008 ) . Phenix . refine ( Afonine et al . , 2005 ) and REFMAC ( Murshudov et al . , 2011 ) were used to refine the model against the native dataset with manual building done in Coot ( Emsley and Cowtan , 2004 ) . D . rerio CPAP937–1124 E1021V was solved by molecular replacement in Phaser ( McCoy et al . , 2007 ) using a poly-alanine model derived from the WT model . The model was further built and refined as described for the WT structure . The complex of D . rerio CPAP937–1124 and D . rerio STIL408–428 was solved by molecular replacement using Phaser ( McCoy et al . , 2007 ) with a distorted model of the D . rerio CPAP937–1124 WT apo-structure . Refinement yielded clear density for the residues of STIL shown here . The model was further built and refined as described for the other D . rerio structures . D . melanogaster dSTIL1–47-dCPAP700–901 data was scaled using Xia2 ( Winter , 2010 ) . Phasing was carried out using all SeMet datasets ( Table 6 ) in autoSHARP ( Vonrhein et al . , 2007 ) , using SHELXC/D ( Sheldrick , 2008 ) for heavy atom finding , SHARP for site refinement/phasing and SOLOMON ( Abrahams and Leslie , 1996 ) for density modification . This resulted in an experimental density map within which a CHAINSAW ( Stein , 2008 ) model based on the D . rerio complex structure could be manually placed , using heavy atom sites as a guide . Experimental density corresponding to the dSTIL peptide could be easily seen . Further refinement cycles allowed the remaining copies of the monomer to be placed and trimmed . Refinement and model building were carried out in autoBUSTER ( Bricogne et al . , 2011 ) and Coot ( Emsley and Cowtan , 2004 ) respectively . All ITC measurements were performed using an auto-iTC 200 instrument ( GE Healthcare , Little Chalfont , UK ) in 50 mM HEPES pH 7 . 5 , 100 mM NaCl at 25°C . Samples were stored by the instrument in 96-well microtiter plates at 5°C prior to loading and performing the titrations . Standard experiments used 19 × 2 μl injections of STIL peptide into CPAP protein preceded by a single 0 . 5 μl pre-injection . Heat from the pre-injection was not used during fitting . Data were analysed manually in the Origin software package provided by the manufacturer and fit to a single set of binding sites model . All measurements were corrected using control ITC experiments in which the peptide studied was injected into buffer only . The small endothermic heats of injection in these experiments were fitted to a linear function that was subsequently subtracted from the equivalent integrated heats of the peptide–protein binding experiment before fitting . The concentration of CPAP in the cell was typically 40 μM but varied maximally between 20 and 100 μM . The concentration of STIL used in the syringe was typically 700 μM but varied maximally between 600 and 2600 μM depending on the affinity of the peptide interaction being studied . C . elegans strains carrying single-copy sas-4 transgenes were generated using MosSCI ( Frøkjær-Jensen et al . , 2008 ) . To render the transgenes RNAi-resistant , a 500 bp region at the 5′ end of the sas-4 genomic sequence was re-encoded . The engineered sas-4 sequence was cloned into pCFJ151 with the promoter and 3′ UTR from sas-6 , as well as a C-terminal GFP tag . pCFJ151 contains homology arms that direct transposase-mediated insertion of intervening sequence into the ttTi5606 Mos1 site on Chromosome II . Transgene integration was confirmed by PCR of regions spanning each side of the insertion . The genotypes of the strains used are: unc-119 ( ed9 ) III; ltSi85[pOD1550; Psas-6::SAS-4 reencoded::GFP; cb-unc-119 ( + ) ]II for WT SAS-4; and unc-119 ( ed9 ) III; ltSi177[pOD1551; Psas-6::SAS-4 ( 1-556 ) reencoded::GFP; cb-unc-119 ( + ) ]II for SAS-4ΔTCP . Double-stranded sas-4 RNA was generated as described ( Oegema et al . , 2001 ) using DNA templates prepared by PCR . For experiments to quantify monopolar spindle formation , L4 hermaphrodites were injected with dsRNA and incubated at 20°C for 40 hr prior to dissection for imaging . For lethality assays , worms were maintained at 20°C . L4 hermaphrodites were injected with dsRNA and singled 24 hr post-injection . Adult worms were removed from the plates 48 hr post-injection , and hatched larvae and unhatched embryos were counted 24 hr later . For light microscopy to identify monopolar or bipolar second division cells , images were acquired using an inverted Zeiss Axio Observer Z1 system with a Yokogawa spinning-disk confocal head ( CSU-X1 ) , a 63X 1 . 4 NA Plan Apochromat objective , and a QuantEM:512SC EMCCD camera ( Photometrics ) . Adult worms were dissected in M9 buffer , and embryos were mounted onto 2% agarose pads for imaging . 11 × 1 μm z-stacks were collected in the GFP channel ( 100 ms , 20% power , no binning ) , along with one central DIC section . | Organisms—and individual tissues—grow and develop by dividing their cells . However , the process of cell division does not have to be symmetric , and the fates of the cells can be very different if cellular contents , including RNAs or proteins , are exclusively retained in the ‘mother’ or passed to her ‘daughter’ . Organelles known as centrioles can play an important part in influencing whether cell division is symmetric or asymmetric . Centrioles contain ordered assemblies of various proteins , and mutations in some of these proteins can cause developmental defects in humans . For example , mutations in the centriolar proteins CPAP and STIL cause a syndrome known as microcephaly , in which the brain is smaller than normal . Although CPAP and STIL are known to bind each other , how they interact on a molecular level to form centrioles—and how this interaction is disrupted in microcephaly—is not well understood . Cottee et al . have now used structural and biochemical assays to explore how these two proteins bind to each other , and have identified specific amino acid residues that enable this interaction . These residues are highly conserved across many organisms , and a mutation in one of them has previously been associated with microcephaly in humans . Now , Cottee et al . demonstrate that this mutation weakens the interaction between CPAP and STIL in vitro . To explore these processes in vivo , Cottee et al . studied mutant fruit flies in which the interactions between CPAP and STIL were weaker than normal , and found that these mutations prevented the normal formation of centrioles . Furthermore , there was a striking correlation between the ability to form centrioles in fruit flies and the ability of CPAP and STIL to bind each other , based on the structural model and in vitro binding studies . Cumulatively , these findings reinforce the importance of CPAP and STIL in centriole formation , and suggest that one reason for the development of microcephaly may be defects in the proper formation of centrioles . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Crystal structures of the CPAP/STIL complex reveal its role in centriole assembly and human microcephaly |
Cancer cells demand excessive nutrients to support their proliferation but how cancer cells sense and promote growth in the nutrient favorable conditions remain incompletely understood . Epidemiological studies have indicated that obesity is a risk factor for various types of cancers . Feeding Drosophila a high dietary sugar was previously demonstrated to not only direct metabolic defects including obesity and organismal insulin resistance , but also transform Ras/Src-activated cells into aggressive tumors . Here we demonstrate that Ras/Src-activated cells are sensitive to perturbations in the Hippo signaling pathway . We provide evidence that nutritional cues activate Salt-inducible kinase , leading to Hippo pathway downregulation in Ras/Src-activated cells . The result is Yorkie-dependent increase in Wingless signaling , a key mediator that promotes diet-enhanced Ras/Src-tumorigenesis in an otherwise insulin-resistant environment . Through this mechanism , Ras/Src-activated cells are positioned to efficiently respond to nutritional signals and ensure tumor growth upon nutrient rich condition including obesity .
The prevalence of obesity is increasing globally . Obesity impacts whole-body homeostasis and is a risk factor for severe health complications including type 2 diabetes and cardiovascular disease . Accumulating epidemiological evidence indicates that obesity also leads to elevated risk of developing several types of cancers ( Calle et al . , 2003; Renehan et al . , 2008; Arnold et al . , 2014 ) . However , the mechanisms that link obesity and cancer remain incompletely understood . Using Drosophila , we recently developed a whole-animal model system to study the link between diet-induced obesity and cancer and provided a potential explanation for how obese and insulin resistant animals are at increased risk for tumor progression ( Hirabayashi et al . , 2013 ) . Drosophila fed a diet containing high levels of sucrose ( high dietary sucrose or ‘HDS’ ) developed sugar-dependent metabolic defects including accumulation of fat ( obesity ) , organismal insulin resistance , hyperglycemia , hyperinsulinemia , heart defects and liver ( fat body ) dysfunctions ( Musselman et al . , 2011 , 2013; Na et al . , 2013; Na et al . , 2015 ) . Inducing activation of oncogenic Ras and Src together in the Drosophila eye epithelia led to development of small benign tumors within the eye epithelia . Feeding animals HDS transformed Ras/Src-activated cells from benign tumor growths to aggressive tumor overgrowth with tumors spread into other regions of the body ( Hirabayashi et al . , 2013 ) . While most tissues of animals fed HDS displayed insulin resistance , Ras/Src-activated tumors retained insulin pathway sensitivity and exhibited an increased ability to import glucose . This is reflected by increased expression of the Insulin Receptor ( InR ) , which was activated through an increase in canonical Wingless ( Wg ) /dWnt signaling that resulted in evasion of diet-mediated insulin resistance in Ras/Src-activated cells . Conversely , expressing a constitutively active isoform of the Insulin Receptor in Ras/Src-activated cells ( InR/Ras/Src ) was sufficient to elevate Wg signaling , promoting tumor overgrowth in animals fed a control diet . These results revealed a circuit with a feed-forward mechanism that directs elevated Wg signaling and InR expression specifically in Ras/Src-activated cells . Through this circuit , mitogenic effects of insulin are not only preserved but are enhanced in Ras/Src-activated cells in the presence of organismal insulin resistance . These studies provide an outline for a new mechanism by which tumors evade insulin resistance , but several questions remain: ( i ) how Ras/Src-activated cells sense the organism's increased insulin levels , ( ii ) how nutrient availability is converted into growth signals , and ( iii ) the trigger for increased Wg protein levels , a key mediator that promotes evasion of insulin resistance and enhanced Ras/Src-tumorigenesis consequent to HDS . In this manuscript , we identify the Hippo pathway effector Yorkie ( Yki ) as a primary source of increased Wg expression in diet-enhanced Ras/Src-tumors . We demonstrate that Ras/Src-activated cells are sensitized to Hippo signaling , and even a mild perturbation in upstream Hippo pathway is sufficient to dominantly promote Ras/Src-tumor growth . We provide functional evidence that increased insulin signaling promotes Salt-inducible kinases ( SIKs ) activity in Ras/Src-activated cells , revealing a SIKs-Yki-Wg axis as a key mediator of diet-enhanced Ras/Src-tumorigenesis . Through this pathway , Hippo-sensitized Ras/Src-activated cells are positioned to efficiently respond to insulin signals and promote tumor overgrowth . These mechanisms act as a feed-forward cassette that promotes tumor progression in dietary rich conditions , evading an otherwise insulin resistant state .
Ras/Src tumors were generated in the developing Drosophila eye epithelium by pairing targeted expression of the activated dRas1 isoform ras1G12V with targeted knockout ( Lee and Luo , 1999 ) of the negative regulator of Src , C-terminal src kinase ( csk−/− ) . Feeding animals a diet containing 1 . 0 M sucrose ( high dietary sucrose or ‘HDS’ ) transformed these Ras/Src-activated cells from benign growths to aggressive tumors associated with emergent tumor spread to other parts of the body ( Figure 1A , B ) ( Hirabayashi et al . , 2013 ) . In HDS-fed animals , Ras/Src-activated cells promoted gene expression of InR through increased canonical Wg-dependent signaling , leading to increased insulin sensitivity in Ras/Src-activated cells in otherwise insulin resistant animals . Expression of a constitutive active isoform of Insulin Receptor ( inrCA ) in Ras/Src-activated cells ( inr CA , ras1G12V;csk−/− ) was sufficient to promote elevation of Wg levels and tumor growth even in a control diet , establishing an InR-Wg-InR amplification circuit that promotes aggressive tumorigenesis ( Figure 1E ) ( Hirabayashi et al . , 2013 ) . 10 . 7554/eLife . 08501 . 003Figure 1 . Yorkie Activity is Required for Increased Wg Expression in Diet-enhanced Ras/Src-tumors . ( A–G ) Developmental stage matched third instar larvae with the genotype , ( A , B ) ras1G12V;csk−/− , ( C ) ras1G12V;csk−/− , wgRNAi , ( D ) ras1G12V;csk−/− , wts , ( E ) inrCA , ras1G12V;csk−/− , ( F ) inrCA , ras1G12V;csk−/− , wgRNAi , and ( G ) inrCA , ras1G12V;csk−/− , wts , raised on indicated diets . Images were taken at the same magnification . Scale bar , 500 μm . ( A′–G′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . ( H ) Percent pupariation of animals from indicated genotypes and diets . Column bars represent the mean of three independent experiments . Error bars denote s . e . m . Total n was 166 , 431 , 309 , 291 , 204 , 200 , and 251 from left to right . Asterisks indicate statistically significant difference ( *p < 0 . 01 t-test ) . Numerical data are available in Figure 1—source data 1 . ( I–L ) Wg staining ( red ) of eye tissue from ( I ) ras1G12V;csk−/− , ( J ) ras1G12V;csk−/− , wts , ( K ) inrCA , ras1G12V;csk−/− , and ( L ) inrCA , ras1G12V;csk−/− , wts animals raised on indicated diets . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 00310 . 7554/eLife . 08501 . 004Figure 1—source data 1 . Percent pupariation of animals from indicated genotypes and diets . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 00410 . 7554/eLife . 08501 . 005Figure 1—figure supplement 1 . Effect of reducing Wg or over-expressing Wts in the eye tissue . ( A , B ) Reducing Wg by RNAi ( wgRNAi ) did not affect normal eye tissue growth of the late third instar larvae . Developmental stage matched wgRNAi third instar larvae raised on ( A ) control diet , and ( B ) HDS . ( C , D ) Over-expression of Warts kinase led to small clones . Developmental stage matched wts third instar larvae raised on ( C ) control diet , and ( D ) HDS . Images were taken at the same magnification . Scale bar , 500 μm . ( A′–D′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 00510 . 7554/eLife . 08501 . 006Figure 1—figure supplement 2 . Yorkie target genes are upregulated in diet-enhanced Ras/Src-tumors . ( A–C ) Myc staining ( red ) of ( A ) ras1G12V;csk−/− in control diet , ( B ) ras1G12V;csk−/− in HDS , and ( C ) inrCA , ras1G12V;csk−/− in control diet . ( D–F ) Cyclin E staining ( red ) of ( D ) ras1G12V;csk−/− in control diet , ( E ) ras1G12V;csk−/− in HDS , and ( F ) inrCA , ras1G12V;csk−/− in control diet . Images were taken at the same magnification . Scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 006 Reducing Wg by RNAi ( wgRNAi ) did not affect normal eye tissue growth of the late third instar larvae ( Figure 1—figure supplement 1 ) . However , reducing Wg in Ras/Src-activated cells ( ras1G12V;csk−/− , wgRNAi ) fed HDS or in InR/Ras/Src-activated cells ( inr CA , ras1G12V;csk−/− , wgRNAi ) fed a control diet significantly suppressed tumor growth ( Figure 1C , F ) . As a result , whereas only 35 . 9% of ras1G12V;csk−/− animals in HDS and 42 . 6% of inrCA , ras1G12V;csk−/− animals in control diet initiated pupariation , most ras1G12V;csk−/− , wgRNAi animals and inrCA , ras1G12V;csk−/− , wgRNAi animals successfully pupariated ( Figure 1H ) . These observations identify Wg as an essential mediator of diet-enhanced Ras/Src-tumors or InR/Ras/Src-tumors . However , the factors that elevate Wg expression in diet- or InR-activated Ras/Src-tumors has not been identified . The Hippo pathway is an evolutionarily conserved signaling pathway that regulates tissue growth and cell fate ( Harvey and Tapon , 2007; Halder and Johnson , 2011 ) . The Hippo pathway regulates growth through the transcriptional co-activator Yki , a Drosophila homolog of mammalian YAP/TAZ ( Huang et al . , 2005 ) . The core pathway kinase effector Warts ( Wts ) phosphorylates Yki and inhibits its activity by sequestering Yki in the cytoplasm ( Dong et al . , 2007; Zhao et al . , 2007; Oh and Irvine , 2008 ) . Conversely , loss of components in the core Hippo complex results in translocation of Yki into the nucleus where it regulates factors that promote proliferation and inhibit cell death ( Goulev et al . , 2008; Wu et al . , 2008; Zhang et al . , 2008 ) . Yki activation has been previously associated with increased expression of Wg ( Cho et al . , 2006 ) . Inhibition of Yki activity by over-expressing Wts led to small clones ( Figure 1—figure supplement 1 ) . Similarly , over-expression of Wts in Ras/Src-activated cells ( ras1G12V;csk−/− , wts ) fed HDS or in InR/Ras/Src-activated cells ( inrCA , ras1G12V;csk−/− , wts ) fed a control diet led to a strong suppression of tumor growth and animal lethality ( Figure 1D , G and H ) . Importantly , increased Wg expression was lost in these clones , indicating that Yki is required for the increased Wg expression observed in diet- or InR-activated Ras/Src-tumors ( Figure 1I–L ) . Myc and cyclin E are well-established transcriptional targets of Yki in Drosophila ( Tapon et al . , 2002; Udan et al . , 2003; Neto-Silva et al . , 2010 ) . Myc and cyclin E were strongly elevated in eye clones of ras1G12V;csk−/− animals fed HDS and in inrCA , ras1G12V;csk−/− animals fed a control diet ( Figure 1—figure supplement 2 ) compared to controls . As previously reported , diap1 gene expression—assessed by the Yki transcriptional reporter diap1-lacZ—was strongly increased in the ras1G12V;csk−/− clones of animals raised in HDS compared to animals fed a control diet ( ( Hirabayashi et al . , 2013 ) Figure 2A , B ) . Upon closer examination , diap1 gene expression was at most mildly increased in most ras1G12V;csk−/− clones in animals fed a control diet ( Figure 2A , arrowheads ) . This increase in Yki activity was not sufficient to promote tumor overgrowth: ras1G12V;csk−/− clones of animals raised in a control diet were progressively eliminated from the tissue by apoptotic cell death ( Hirabayashi et al . , 2013 ) . Activation of insulin signaling pathway alone ( inrCA ) failed to elevate diap1 expression in a control diet ( Figure 2C ) . However , the triple combination ( inr CA , ras1G12V;csk−/− ) led to strongly elevated diap1 gene expression , including in animals fed a control diet ( Figure 2D ) . These results indicate that activation of Yki is an emergent property of Ras and Src co-activation , and increased insulin signaling further promotes Yki activity in Ras/Src-activated cells . 10 . 7554/eLife . 08501 . 007Figure 2 . Ras/Src-activated Cells are Sensitive to Perturbations in the Hippo Signaling . ( A–D ) β-galactosidase ( β-gal ) staining ( red ) of eye tissue from ( A , B ) ras1G12V;csk−/− , diap1-lacZ , ( C ) inrCA;diap1-lacZ , ( D ) inrCA , ras1G12V;csk−/− , diap1-lacZ animals raised on indicated diets . Scale bars , 50 μm . ( E–H ) Developmental stage matched third instar larvae raised on control diet with the genotype , ( E ) ex+/− , ( F ) ras1G12V;csk−/− , ( G , H ) ex+/− , ras1G12V;csk−/− . Images were taken at the same magnification . Scale bar , 500 μm . ( E′–H′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . ( I ) Percent pupariation of animals from indicated genotypes . Column bars represent the mean of three independent experiments . Error bars denote s . e . m . Total n of 389 , 238 , and 206 from left to right . Asterisks indicate statistically significant difference ( *p < 0 . 01 t-test ) . Numerical data are available in Figure 2—source data 1 ( J ) β-galactosidase ( β-gal ) staining ( red ) of ex+/− , ras1G12V;csk−/− animals raised on control diet . Scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 00710 . 7554/eLife . 08501 . 008Figure 2—source data 1 . Percent pupariation of animals from indicated genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 008 The FERM domain protein Expanded ( Ex ) is both an upstream regulator of the Hippo pathway and a transcriptional target of Yki , forming a negative feedback loop ( Hamaratoglu et al . , 2006 ) . An enhancer trap fly line in which a lacZ gene is inserted in the ex locus ( ex674; [Boedigheimer and Laughon , 1993] ) can therefore be used to reduce ex activity as well as a readout of Yki transcriptional activity . Removing a functional genomic copy of ex ( ex+/− ) did not affect normal eye tissue growth ( Figure 2E ) . Surprisingly , reducing ex in ras1G12V;csk−/− animals ( ex+/− , ras1G12V;csk−/− ) dominantly promoted growth of Ras/Src-activated cells in a control diet ( Figure 2F-H ) . As a consequence , 40% of ex+/− , ras1G12V;csk−/− animals in control diet failed to initiate pupariation , dying as larvae with overgrown eye tissue ( Figure 2I ) . Immunostaining using anti-β-galactosidase antibody indicated that ex gene expression was strongly increased in ex+/− , ras1G12V;csk−/− clones of animals raised in a control diet , demonstrating Yki activation in these clones ( Figure 2J ) . Together these results provide compelling evidence that Ras/Src-activated cells are functionally linked to Hippo pathway activity , as even a subtle perturbation of upstream Hippo signaling is sufficient to dominantly promote tumor Ras/Src-tumor overgrowth . To determine whether diet-enhanced Ras/Src-tumors promote Yki activity through the core Hippo signaling pathway , we performed Western-blot analysis using an antibody to the phosphorylated Serine-168 residue of Yki , a standard indication of Wts kinase activity . As anticipated , phosphorylation of Yki was significantly reduced in genotypically wts−/− eye tissues . Dietary sucrose did not alter Yki phosphorylation in lacZ-expressing control clones ( Figure 3A ) . In ras1G12V;csk−/− animals , however , HDS led to a reduction in Yki phosphorylation to a level comparable to loss of wts ( Figure 3A ) . Phosphorylation was similarly reduced in eye tissues of inrCA , ras1G12V;csk−/− animals fed a control diet , indicating that increased insulin signaling is sufficient to suppress Wts kinase activity in Ras/Src-activated cells ( Figure 3A ) . We did not observe significant changes in total Yki levels ( Figure 3A ) . Our results indicate that ( i ) Ras/Src-tumors in the presence of HDS or ( ii ) InR/Ras/Src-tumors in a control diet promote Yki activity through inhibition of Wts kinase activity . 10 . 7554/eLife . 08501 . 009Figure 3 . Salt-inducible Kinases are Required for Diet-enhanced Ras/Src-tumorigenesis . ( A ) Extracts from dissected eye tissues of third instar larvae were examined by immunoblotting using antibodies against phospho-Sav ( p-Sav; * indicates p-Sav specific band; the upper band is a non-specific band showed as an internal loading control ) , phospho-Yki ( p-Yki ) , total Yki ( Yki ) , and Syntaxin ( Syt ) . ( B , C ) Developmental stage matched third instar larvae raised on HDS with the genotype , ( B ) ras1G12V;csk−/− , and ( C ) sik2/3RNAi , ras1G12V;csk−/− . ( D , E ) Developmental stage matched third instar larvae raised on control diet with the genotype , ( D ) ras1G12V;csk−/− , and ( E ) sik2/3RNAi , ras1G12V;csk−/− . Images were taken at the same magnification . Scale bar , 500 μm . ( F , G ) ras1G12V;csk−/− animals raised on HDS containing ( F ) 0 . 05% DMSO , or ( G ) 25 μΜ HG-9-91-01 . ( H , I ) inrCA , ras1G12V;csk−/− animals raised on control diet containing ( H ) 0 . 05% DMSO , or ( I ) 25 μΜ HG-9-91-01 . Images were taken at the same magnification . Scale bar , 500 μm . ( B′–I′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . ( J ) Percent pupariation of DMSO or HG-9-91-01 treated animals from indicated genotypes and diets . Column bars represent the mean of three independent experiments . Error bars denote s . e . m . Total n of 139 , 76 , 123 , and 72 from left to right . Asterisks indicate statistically significant difference ( *p < 0 . 01 t-test ) . Numerical data are available in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 00910 . 7554/eLife . 08501 . 010Figure 3—source data 1 . Percent pupariation of animals from indicated genotypes and diets . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 01010 . 7554/eLife . 08501 . 011Figure 3—figure supplement 1 . Akt mediates activation of SIKs in Ras/Src-tumors . ( A , B ) Developmental stage matched third instar larvae raised on HDS with the genotype , ( A ) ras1G12V;csk−/− , and ( B ) ras1G12V;csk−/− , akthypo/hypo . ( C , D ) Developmental stage matched third instar larvae raised on control diet with the genotype , ( C ) inrCA , ras1G12V;csk−/− , and ( D ) inrCA , ras1G12V;csk−/− , akthypo/hypo . Images were taken at the same magnification . Scale bar , 500 μm . ( A′–D′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . ( E ) Extracts from dissected eye tissues of third instar larvae were examined by immunoblotting using antibodies against phospho-Sav ( p-Sav; * indicates p-Sav specific band; the upper band is a non-specific band showed as an internal loading control ) and Syntaxin ( Syt ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 01110 . 7554/eLife . 08501 . 012Figure 3—figure supplement 2 . Reducing SIK2/3 by RNAi did not affect normal eye tissue growth . ( A , B ) Developmental stage matched third instar larvae raised on HDS with genotypes , ( A ) lacZ , and ( B ) sik2/3RNAi . Images were taken at the same magnification . Scale bar , 500 μm . ( A′ , B′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 012 Salt-inducible kinases ( SIKs ) were recently shown to regulate wing tissue growth through Hippo pathway activity in Drosophila ( Wehr et al . , 2013 ) . Phosphorylation of Salvador ( Sav ) at Serine-413 by SIK led to dissociation of the Hippo complex and activation of Yki ( Wehr et al . , 2013 ) . In the eye tissue of ras1G12V;csk−/− larvae fed HDS , Serine-413 phosphorylation of Sav was strongly upregulated ( Figure 3A ) . Similarly , phosphorylation of Sav was strongly elevated in inrCA , ras1G12V;csk−/− animals fed a control diet , demonstrating that increased insulin signaling in Ras/Src-activated cells promotes SIK activity ( Figure 3A ) . Previous studies in mammals and Drosophila have shown that SIKs are activated by Akt ( Dentin et al . , 2007; Wang et al . , 2011; Choi et al . , 2015 ) . Reducing Akt activity through a hypomorphic allele of akt ( ras1G12V;csk−/− , akthypo/hypo ) suppressed Ras/Src-tumor growth and reduced phosphorylation of Sav in animals raised on HDS ( Figure 3—figure supplement 1 ) . Similarly , tumor growth and phosphorylation of Sav was suppressed in inrCA , ras1G12V;csk−/− , akthypo/hypo animals fed a control diet ( Figure 3—figure supplement 1 ) . These results demonstrate that—in ras1G12V;csk−/− larval eye tissue fed HDS and inrCA , ras1G12V;csk−/− animals fed a control diet—activation of SIKs are mediated by Akt . To examine whether the SIKs are required for diet-enhanced Ras/Src-tumorigenesis , we used a transgenic RNA-interference line that targets both sik2 and sik3 transcripts for knockdown ( Wehr et al . , 2013 ) . Reducing SIK2/3 in ras1G12V;csk−/− animals ( sik2RNAi , ras1G12V;csk−/− ) suppressed diet-enhanced Ras/Src-tumor growth ( Figure 3B , C ) . Importantly , reducing SIK2/3 by itself did not significantly affect normal eye tissue growth of animals fed HDS ( Figure 3—figure supplement 2 ) , indicating that SIK2/3 is functionally required for Ras/Src-tumor growth in the presence of HDS . Feeding HG-9-91-01 , a potent inhibitor of SIKs ( Clark et al . , 2012 ) , led to suppression of tumor growth and animal lethality in both ras1G12V;csk−/− animals fed HDS and in inrCA , ras1G12V;csk−/− animals fed a control diet ( Figure 3F-J ) . We conclude that activation of SIKs is functionally required for diet-enhanced Ras/Src-tumorigenesis . Conversely , expression of a constitutive active isoform of SIK2 ( sik2CA ( Wehr et al . , 2013 ) ) in Ras/Src-activated cells ( ras1G12V;csk−/− , sik2CA ) was sufficient to promote Ras/Src-dependent tumor overgrowth even in a control diet ( Figure 4A , B ) . Western blot analysis confirmed increased phosphorylation of Sav and reduced phosphorylation of Yki in ras1G12V;csk−/− , sik2CA tumors ( Figure 4C ) . Wg expression was strongly upregulated in ras1G12V;csk−/− , sik2CA tumors in animals raised on a control diet ( Figure 4D , E ) , further linking SIKs to Hippo pathway activity . Taken together , these results demonstrate that SIKs provide the upstream Hippo signal that mediates Ras/Src-tumorigenesis in diet-induced obese animals . 10 . 7554/eLife . 08501 . 013Figure 4 . Activation of Salt-inducible Kinase Promotes Ras/Src-tumor Growth . ( A , B ) Developmental stage matched third instar larvae raised on control diet with the genotype , ( A ) ras1G12V;csk−/− , and ( B ) ras1G12V;csk−/− , sik2CA . Images were taken at the same magnification . Scale bar , 500 μm . ( A′ , B′ ) Matching dissected eye epithelial tissue stained with DAPI ( red ) . Images were taken at the same magnification . Scale bar , 500 μm . ( C ) Extracts from dissected eye tissues of ras1G12V;csk−/− and ras1G12V;csk−/− , sik2CA animals fed a control diet were examined by immunoblotting using antibodies against phospho-Sav ( p-Sav; * indicates p-Sav specific band; the upper band is a non-specific band showed as an internal loading control ) , phospho-Yki ( p-Yki ) , total Yki ( Yki ) , and Syntaxin ( Syt ) . ( D , E ) Wg staining ( red ) of eye tissue from ( D ) ras1G12V;csk−/− , and ( E ) ras1G12V;csk−/− , sik2CA animals raised on control diet . Scale bars , 50 μm . ( F ) Model of diet-enhanced tumorigenesis of Ras/Src-activated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 08501 . 013
We previously demonstrated that Ras/Src-activated cells preserve mitogenic effects of insulin under the systemic insulin resistance induced by HDS-feeding of Drosophila ( Hirabayashi et al . , 2013 ) . Evasion of insulin resistance in Ras/Src-activated cells is a consequence of a Wg-dependent increase in InR gene expression ( Hirabayashi et al . , 2013 ) . In this study , we identify the Hippo pathway effector Yki as a primary source of the Wnt ortholog Wg in diet-enhanced Ras/Src-tumors . Mechanistically , we provide functional evidence that activation of SIKs promotes Yki-dependent Wg-activation and reveal a SIK-Yki-Wg-InR axis as a key feed-forward signaling pathway that underlies evasion of insulin resistance and promotion of tumor growth in diet-enhanced Ras/Src-tumors ( Figure 4F ) . In animals fed a control diet , we observed at most a mild increase in Yki reporter activity within ras1G12V;csk−/− cells ( Figure 2A ) . A previous report indicates that activation of oncogenic Ras ( ras1G12V ) led to slight activation of Yki in eye tissue ( Ohsawa et al . , 2012; Enomoto and Igaki , 2013; Enomoto et al . , 2015 ) . Activation of Src through over-expression of the Drosophila Src ortholog Src64B has been shown to induce autonomous and non-autonomous activation of Yki ( Enomoto and Igaki , 2013 ) . In contrast , inducing activation of Src through loss of csk ( csk−/− ) failed to elevate diap1 expression ( data not shown ) . Our results indicate that activation of Yki is an emergent property of activating Ras plus Src ( ras1G12V;csk−/− ) . However , this level of Yki-activation was not sufficient to promote stable tumor growth of Ras/Src-activated cells in the context of a control diet: Ras/Src-activated cells were progressively eliminated from the eye tissue ( Hirabayashi et al . , 2013 ) . It was , however , sufficient to sensitize Ras/Src-activated cells to upstream Hippo pathway signals: loss of a genetic copy of ex—which was not sufficient to promote growth by itself—dominantly promoted tumor growth of Ras/Src-activated cells even in animals fed a control diet ( Figure 2G-I ) . These data provide compelling evidence that Ras/Src-transformed cells are sensitive to upstream Hippo signals . SIK was recently demonstrated to phosphorylate Sav at Serine-413 , resulting in dissociation of the Hippo complex and activation of Yki ( Wehr et al . , 2013 ) . SIKs are required for diet-enhanced Ras/Src-tumor growth in HDS ( Figure 3C ) . Conversely , expression of a constitutively activated isoform of SIK was sufficient to promote Ras/Src-tumor overgrowth even in a control diet ( Figure 4B ) . Mammalian SIKs are regulated by glucose and by insulin signaling ( Wang et al . , 2008 , 2011 ) . However , a more recent report indicated that glucagon but not insulin regulates SIK2 activity in the liver ( Patel et al . , 2014 ) . Our data demonstrate that increased insulin signaling is sufficient to promote SIK activity through Akt in Ras/Src-activated cells ( Figure 3A , Figure 3—figure supplement 1 ) . We conclude that SIKs couple nutrient ( insulin ) availability to Yki-mediated evasion of insulin resistance and tumor growth , ensuring Ras/Src-tumor growth under nutrient favorable conditions . Our results place SIKs as key sensors of nutrient and energy availability in Ras/Src-tumors through increased insulin signaling and , hence , increased glucose availability . SIK activity promotes Ras/Src-activated cells to efficiently respond to upstream Hippo signals , ensuring tumor overgrowth in organisms that are otherwise insulin resistant . One interesting question is whether this mechanism is relevant beyond the context of an obesity-cancer connection: both Ras and Src have pleiotropic effects on developmental processes including survival , proliferation , morphogenesis , differentiation , and invasion , and these mechanisms may facilitate these processes under nutrient favorable conditions . From a treatment perspective our data highlight SIKs as potential therapeutic targets . Limiting SIK activity through compounds such as HG-9-91-01 may break the connection between oncogenes and diet , targeting key aspects of tumor progression that are enhanced in obese individuals .
UAS-ras1G12V , UAS-inrA1325D ( inrCA ) , UAS-wts , ex697 ( ex-lacZ ) , Diap1j5C8 ( diap1-lacZ ) , akt04226 flies were obtained from the Bloomington Drosophila Stock Center . UAS-wgRNAi and UAS-sik2RNAi flies were obtained from Vienna Drosophila RNAi Center . The following stocks were kindly provided to us: FRT82B , cskQ156Stop by A . O'Reilly and M . Simon; ey ( 3 . 5 ) -FLP1 by G . Halder; UAS-sik2S1032A ( sik2CA ) by N . Tapon; wtsX1 by C . Pfleger . To create eyeless-driven green fluorescent protein ( GFP ) –labeled clones , flies with the genotype ey ( 3 . 5 ) -FLP1; act > y+>gal4 , UAS-GFP; FRT82B , tub-gal80 were crossed with flies with the following genotypes: ( a ) UAS-ras1G12V; FRT82B , cskQ156Stop/TM6b; ( b ) UAS-ras1G12V; FRT82B , cskQ156Stop , UAS-wgRNAi/TM6b; ( c ) UAS-ras1G12V; FRT82B , cskQ156Stop , UAS-wts/TM6b; ( d ) UAS-inrA1325D , UAS-ras1G12V; FRT82B , cskQ156Stop/TM6b; ( e ) UAS-inrA1325D , UAS-ras1G12V; FRT82B , cskQ156Stop , UAS-wgRNAi/TM6b; ( f ) UAS-inrA1325D , UAS-ras1G12V; FRT82B , cskQ156Stop , UAS-wts/TM6b; ( g ) FRT82B , UAS-wgRNAi; ( h ) FRT82B , UAS-wts; ( i ) UAS-ras1G12V; diap1-lacZ , FRT82B , cskQ156Stop/TM6b; ( j ) UAS-inrA1325D , UAS-ras1G12V; diap1-lacZ , FRT82B , cskQ156Stop/TM6b; ( k ) UAS-inrA1325D; diap1-lacZ , FRT82B/TM6b; ( l ) ex697; FRT82B/SM6-TM6b; ( m ) ex697 , UAS-ras1G12V; FRT82B , cskQ156Stop/SM6-TM6b; ( n ) FRT82B , wtsX1/TM6b; ( o ) UAS-lacZ; FRT82B; ( p ) UAS-sik2RNAi , UAS-ras1G12V; FRT82B , cskQ156Stop/TM6b; ( q ) UAS-ras1G12V; UAS-sik2CA , FRT82B , cskQ156Stop/TM6b; ( r ) UAS-sik2RNAi; FRT82B; ( s ) UAS-ras1G12V; FRT82B , cskQ156Stop , akt04226 /TM6b; ( t ) UAS-inrA1325D , UAS-ras1G12V; FRT82B , cskQ156Stop , akt04226 /TM6b . Cultures were carried out on Bloomington semi-defined medium ( described by the Bloomington Drosophila stock center ) with modifications . Detailed recipes for control diet and HDS is previously described ( Musselman et al . , 2011 ) . The following final concentrations of carbohydrates were included: 0 . 15 M sucrose ( control diet ) and 1 . 0 M sucrose ( HDS ) . Cultures were performed at 25°C . These procedures were performed as previously described ( Hirabayashi et al . , 2013 ) . Primary antibodies used for immunofluorescence were: mouse anti-Wingless ( DSHB: Developmental Studies Hybridoma Bank , Iowa City , IA , United States ) , mouse anti-Cyclin E ( DSHB ) , mouse anti-β-galactosidase ( DSHB ) , rabbit anti-Myc ( Santa Cruz Biotechnology , Dallas , TX , United States ) . Western blots were probed with antibodies against Yki ( gift from K Irvine ) ( Oh and Irvine , 2008 ) , phospho-Yki ( pS168 ) ( gift from D . Pan ) ( Dong et al . , 2007 ) , phospho-Sav ( pS413 ) ( gift from N Tapon ) ( Wehr et al . , 2013 ) , and Syntaxin ( DSHB ) . HG-9-91-01 ( MedChem Express , Princeton , NJ , United States ) was solubilized in DMSO and diluted directly into the fly medium and vortexed extensively to obtain a homogeneous culture . | Around the world , obesity has become a much more common condition . It is a serious health concern , which can increase a person's risk of developing type 2 diabetes , heart disease and certain types of cancer . People who develop type 2 diabetes become insensitive to a hormone called insulin . This hormone normally helps the body to process sugar , and so insensitivity to insulin causes excess sugar to build up in the blood . The excess sugar may provide the extra nutrients cancer cells need to grow . In 2013 , researchers fed a high sugar diet to fruit flies that had been genetically engineered to develop eye tumors to study how obesity caused by a high sugar diet affects tumor growth . The high sugar diet caused the tumors to grow more aggressively . This happened because normal cells became insensitive to insulin , but the tumor cells didn't . This allowed the tumor cells to use the extra sugar to fuel their growth . The experiments showed that the tumor cells had more insulin receptors than normal cells because a molecular switch that controls the receptors was turned on . But it wasn't exactly clear how the cancer genes and excess sugar flipped that switch . Now , Hirabayashi and Cagan—who were both involved in the 2013 work—show that together cancer genes and excess sugar turn on a protein in the flies that senses sugar . This protein , called Salt-inducible kinase , blocks a cellular mechanism that normally limits the growth of cells . With this check on cellular growth blocked , the molecular switch that boosts the number of insulin receptor turns on . This in turn allows the excess sugar to fuel rapid growth of the tumor . In this way , tumor cells know when the sugars are available and make sure they grow in a nutrient-rich condition such as obesity . In the future , scientists may use this new information to develop treatments that help stop the growth of obesity-linked tumors . But first it must be confirmed whether excess sugar and cancer genes behave the same way in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"cell",
"biology"
] | 2015 | Salt-inducible kinases mediate nutrient-sensing to link dietary sugar and tumorigenesis in Drosophila |
A loss of the checkpoint kinase ataxia telangiectasia mutated ( ATM ) leads to impairments in the DNA damage response , and in humans causes cerebellar neurodegeneration , and an increased risk of cancer . A loss of ATM is also associated with increased protein aggregation . The relevance and characteristics of this aggregation are still incompletely understood . Moreover , it is unclear to what extent other genotoxic conditions can trigger protein aggregation as well . Here , we show that targeting ATM , but also ATR or DNA topoisomerases , results in the widespread aggregation of a metastable , disease-associated subfraction of the proteome . Aggregation-prone model substrates , including Huntingtin exon 1 containing an expanded polyglutamine repeat , aggregate faster under these conditions . This increased aggregation results from an overload of chaperone systems , which lowers the cell-intrinsic threshold for proteins to aggregate . In line with this , we find that inhibition of the HSP70 chaperone system further exacerbates the increased protein aggregation . Moreover , we identify the molecular chaperone HSPB5 as a cell-specific suppressor of it . Our findings reveal that various genotoxic conditions trigger widespread protein aggregation in a manner that is highly reminiscent of the aggregation occurring in situations of proteotoxic stress and in proteinopathies .
The PI3K-like serine/threonine checkpoint kinase ataxia telangiectasia mutated ( ATM ) functions as a central regulator of the DNA damage response ( DDR ) and is recruited early to DNA double-strand breaks ( DSBs ) by the MRE11/RAD50/NBS1 ( MRN ) complex ( Shiloh and Ziv , 2013 ) . Defects in ATM give rise to ataxia-telangiectasia ( A-T ) , a multisystem disorder that is characterized by a predisposition to cancer and progressive neurodegeneration ( McKinnon , 2012 ) . Impaired function of ATM has also been linked to a disruption of protein homeostasis and increased protein aggregation ( Corcoles-Saez et al . , 2018; Lee et al . , 2018; Liu et al . , 2005 ) . Protein homeostasis is normally maintained by protein quality control systems , including chaperones and proteolytic pathways ( Hipp et al . , 2019; Labbadia and Morimoto , 2015 ) . Together , these systems guard the balance of the proteome by facilitating correct protein folding , providing conformational maintenance , and ensuring timely degradation . When the capacity of protein quality control systems becomes overwhelmed during ( chronic ) proteotoxic stress , the stability of the proteome can no longer be sufficiently guarded , causing proteins to succumb to aggregation more readily . Proteins that are expressed at a relatively high level compared to their intrinsic aggregation propensity , a state referred to as ‘supersaturation , ’ have been shown to be particularly vulnerable in this respect ( Ciryam et al . , 2015 ) . A loss of protein homeostasis and the accompanying widespread aggregation can have profound consequences , and is associated with a range of ( degenerative ) diseases , including neurodegeneration ( Kampinga and Bergink , 2016; Klaips et al . , 2018; Ross and Poirier , 2004 ) . The characteristics and relevance of the aggregation induced by a loss of ATM are still largely unclear . Loss of MRE11 has recently also been found to result in protein aggregation ( Lee et al . , 2021 ) , and since MRE11 and ATM function in the same DDR pathway , this raises the question whether other genotoxic conditions can challenge protein homeostasis as well ( Ainslie et al . , 2021Huiting and Bergink , 2020 ) . Here , we report that not just impaired function of ATM , but also inhibition of the related checkpoint kinase ataxia telangiectasia and Rad3-related ( ATR ) , as well as chemical trapping of topoisomerases ( TOPs ) using chemotherapeutic TOP poisons leads to widespread protein aggregation . Through proteomic profiling , we uncover that the increased protein aggregation induced by these genotoxic conditions overlaps strongly with the aggregation observed under conditions of ( chronic ) stress and in various neurodegenerative disorders , both in identity and in biochemical characteristics . In addition , we find that these conditions accelerate the aggregation of aggregation-prone model substrates , including the Huntington’s disease-related polyglutamine exon 1 fragment . We show that the widespread protein aggregation is the result of an overload of protein quality control systems , which cannot be explained by any quantitative changes in the aggregating proteins or by genetic alterations in their coding regions . This overload forces a shift in the equilibrium of protein homeostasis , causing proteins that are normally kept soluble by chaperones to now aggregate . Which proteins succumb to aggregation depends on the ground state of protein homeostasis , including the wiring of chaperone systems in that cell . Finally , we provide evidence that the protein aggregation induced by genotoxic stress conditions is amenable to modulation by chaperone systems: whereas inhibition of HSP70 exacerbates aggregation , we also provide a proof of concept that aggregation can be rescued in a cell line-specific manner by increasing the levels of the small heat shock protein HSPB5 ( αB-crystallin ) .
Aggregated proteins are often resistant to solubilization by SDS , and they can therefore be isolated using a step-wise detergent fractionation and centrifugation method . We isolated 1% SDS-resistant proteins ( from here on referred to as aggregated proteins ) and quantified these by SDS-PAGE followed by in-gel protein staining . In line with previous findings ( Lee et al . , 2018 ) , we find that knocking out ATM in both U2OS and HEK293 results in an increase in protein aggregation ( Figure 1A and B , Figure 1—figure supplement 1A–C ) . Transient chemical inhibition of ATM ( 48–72 hr prior to fractionation; Figure 1—figure supplement 1D ) resulted in an increase in aggregated proteins in HEK293T cells as well ( Figure 1C and D ) . Using the same experimental set-up , we examined the impact on aggregation of targeting other DDR components . This revealed that chemical inhibition of the checkpoint signaling kinase ATR also enhanced protein aggregation ( Figure 1C and D ) . Inhibition of tyrosyl-DNA-phosphodiesterase 1 ( TDP1 ) , which repairs various 3′-blocking lesions including topoisomerase 1 ( TOP1 ) cleavage complexes , had no clear effect on protein aggregation ( Figure 1C and D ) . This could be a result of functional redundancy or limited TOP1 trapping occurring under unstressed conditions in a timeframe of 72 hr . We therefore also directly targeted TOPs using the chemotherapeutic compounds camptothecin ( CPT ) and etoposide ( Etop ) . The genotoxic impact of CPT and Etop is a well-documented consequence of their ability to trap ( i . e . , ‘poison’ ) respectively TOP1 and TOP2 cleavage complexes on the DNA , resulting in DNA damage ( Pommier et al . , 2010 ) . Strikingly , we found that transient treatment with either compound caused a particularly strong increase in protein aggregation ( Figure 1C and D ) , which was dose-dependent ( Figure 1E and F ) . Treatment of U2OS cells with CPT led to a dose-dependent increase in aggregation as well , although at higher doses compared to HEK293T cells ( Figure 1—figure supplement 1E ) . Inhibition of poly ( ADP-ribose ) polymerases 1–3 ( PARP1-3 ) , involved in single-strand break repair , did not increase aggregation ( Figure 1C and D ) . Recently , it was reported that PARP inhibition reduces the enhanced aggregation triggered by a loss of ATM ( Lee et al . , 2021 ) , something we find as well for CPT-treated cells ( Figure 1—figure supplement 1F , G ) . Neither CPT nor Etop treatment in HEK293T cells had any effect on aggregation within the first 24 hr ( Figure 1G and H ) . This reveals that the increased aggregation occurs only late and argues that it does not stem from any immediate , unknown damaging effect of either CPT or Etop on mRNA or protein molecules . Together , these data indicate that the increased protein aggregation triggered by targeting ATM , ATR , and TOPs is a late consequence of genotoxic stress . To investigate the nature of the proteins that become aggregated after genotoxic stress , we subjected the SDS-insoluble protein aggregate fractions and whole-cell lysates ( WCL ) of control ( DMSO ) and CPT-treated HEK293T cells to label-free proteomics ( Figure 2—figure supplement 1A ) . Using a stringent cutoff ( Benjamini–Hochberg corrected p<0 . 05; –1>log2fold change >1; identified in >1 repeats of CPT-treated cells ) , we determined that 122 proteins aggregated significantly more after CPT treatment compared to only 29 proteins that aggregated less ( Figure 2A , Supplementary file 1 ) . These 122 proteins aggregate highly consistent ( Supplementary file 1 ) . Most of them were not identified as aggregating in untreated cells , implying that they are soluble under normal conditions . We next used the same MS/MS approach to investigate the aggregation triggered by inhibition of ATM in HEK293T cells . We detected 39 proteins that aggregated significantly more in ATM-inhibited cells compared to control cells and 5 proteins that aggregated less ( Figure 2B , Supplementary file 1 ) . Surprisingly , only one protein was found to aggregate more after both CPT treatment and inhibition of ATM ( Figure 2C ) , suggesting that these genotoxic conditions drive the aggregation of different proteins . However , we noted that several proteins that aggregated more after CPT treatment were absent in the dataset of ATM inhibition ( Supplementary file 1 ) , suggesting that we may have only identified the most abundantly aggregating proteins . We selected MCM7 , TUBA1A , and HDAC1 , three proteins that were identified in our MS/MS analysis , to consistently aggregate more in CPT-treated HEK293T cells but that were not picked up in the ATM inhibition MS/MS analysis and confirmed that all three aggregated more in CPT-treated HEK293 cells ( Figure 2D ) . MCM7 , TUBA1A , and HDAC1 also aggregated more than in unstressed conditions after treatment of HEK293 cells with ATM inhibitor or when ATM was knocked out completely ( Figure 2D ) . These findings indicate that these different genotoxic conditions drive aggregation similarly , although the most prominent aggregating proteins differ . Next , we investigated the aggregation caused by a loss of ATM or CPT treatment in U2OS cells , a cell line that has been used previously as well to study the effect of a loss of ATM on protein aggregation . We found 210 proteins that aggregated more in ATM KO cells , while 53 proteins aggregated less ( Figure 2—figure supplement 1B , Supplementary file 1 ) . Of these 210 proteins , 114 were also found to aggregate more in ATM-depleted U2OS cells in a recent study by Lee et al . , 2021 . Treatment of U2OS cells with CPT resulted in 106 proteins aggregating more and 61 proteins to aggregate less ( Figure 2—figure supplement 1C , Supplementary file 1 ) . Close to 20% of proteins that aggregated more after CPT treatment also aggregated in ATM KO U2OS cells ( 20/106 ) ( Figure 2C ) . Similar to the induced aggregation in HEK293T cells , proteins that aggregate more in U2OS ATM KO or CPT-treated cells appear to be largely soluble in wild-type cells , but now aggregate consistently ( Supplementary file 1 ) . At first glance , protein aggregation caused by a ( functional ) loss of ATM or CPT treatment in U2OS cells seemed to be quite different from the aggregation observed in HEK293T cells . Not a single protein was found to aggregate more in all four different conditions , only two proteins aggregated more in both HEK293T and U2OS cells after CPT treatment , and only one protein overlapped between ATM KO U2OS cells and ATM-inhibited HEK293T cells ( Figure 2C ) . Interestingly , a GO term analysis of the aggregating proteomes did reveal overlap across the different treatments and the two cell lines ( Figure 2E , Figure 2—figure supplement 1D–I ) . Cytoskeleton-related terms , including microtubule and microfibril , are enriched across the different aggregating proteomes ( Figure 2—figure supplement 1D–F , H ) . However , enrichment of most GO terms is restricted to a specific treatment and/or cell line . For example , proteins that aggregated more in CPT-treated HEK293T cells are enriched for nucleotide binding terms , most prominently RNA binding , which is highly enriched among proteins that aggregate after ATM inhibition as well ( Figure 2E ) . Specifically CPT treatment in HEK293T cells drives the aggregation of mitochondrial components ( Figure 2—figure supplement 1D ) . In U2OS cells , proteins that aggregate more after a loss of ATM or CPT treatment appear to be enriched for components involved in cell-cell contact , including cell adhesion and cellular membrane processes ( Figure 2—figure supplement 1F–H ) . . As protein aggregation can manifest vastly different in distinct cell types ( David et al . , 2010; Freer et al . , 2016 ) , we examined which proteins aggregated consistently in HEK293T and U2OS cells , regardless of the presence or absence of genotoxic stress . Importantly , within each cell line , these ‘consistently’ aggregating proteins show a very high overlap between experiments ( ~80% overlap , see also Supplementary file 1 ) . Based on this , we defined a ‘baseline aggregating fraction’ for each cell line . This consisted of aggregating proteins that were not changed upon the genotoxic treatments: these proteins were detected in at least two experimental replicates of both treated and untreated cells and exhibited p-adjusted values of >0 . 05 in t-test comparisons , consistent with no significant effect ( Supplementary file 1 ) . This revealed that 66% ( 118/179 ) of the HEK293T baseline fraction aggregates in the U2OS baseline as well ( Figure 2—figure supplement 1J ) . Importantly , 62% ( 99/160 ) of the proteins that aggregated more in CPT- or ATM inhibitor-treated HEK293T cells are also part of the U2OS baseline ( Figure 2F ) . This indicates that in U2OS cells afar bigger cluster of proteins ends up in aggregates , even under normal conditions . Indeed , silver staining revealed that in unstressed U2OS cells protein aggregation is substantially more prominent than in untreated HEK293T cells ( Figure 2G ) . This is also reflected in MCM7 , TUBA1A , and HDAC1 , all three of which aggregate strongly already in ( untreated ) wild-type U2OS cells ( Figure 2D ) . These findings indicate that the lack of overlap between proteins that aggregate after CPT- or ATM inhibitor treatment in HEK293T and proteins that aggregate in U2OS after CPT treatment or in ATM KO cells is primarily a reflection of a different proteome and a different background aggregation in these two cell lines . These data indicate that the genotoxic conditions of TOP1 poisoning and ATM loss have a cell line-dependent impact on protein aggregation . In both HEK293T and U2OS cells , protein aggregation does not appear to be limited to a specific location or function but affects proteins throughout the proteome . This suggests that the aggregation is primarily driven by the physicochemical characteristics of the proteins involved . A key determinant of aggregation is supersaturation . Protein supersaturation refers to proteins that are expressed at high levels relative to their intrinsic propensity to aggregate , which makes them vulnerable to aggregation . Supersaturation has been shown to underlie the widespread protein aggregation observed in age-related neurodegenerative diseases , and in general aging ( Ciryam et al . , 2015; Ciryam et al . , 2019; Freer et al . , 2019; Kundra et al . , 2017; Noji et al . , 2021 ) . The relevance of supersaturation is underlined by the notion that evolutionary pressures appear to have shaped proteomes along its lines , so that at a global level protein abundance is inversely correlated with aggregation propensity ( Tartaglia et al . , 2007 ) . To determine the role of protein supersaturation in the aggregation observed in our experiments , we first defined a control group of proteins that were not identified as aggregating ( NIA ) for HEK293T cells to serve as a benchmark . This group consisted of all proteins that were only identified in the HEK293T WCL , and not in the SDS-insoluble fractions ( see also Supplementary file 1 ) . We next examined the intrinsic aggregation propensities of proteins using the aggregation prediction tools TANGO ( Fernandez-Escamilla et al . , 2004 ) and CamSol ( Sormanni and Vendruscolo , 2019 ) . Surprisingly , we found that aggregated proteins have in general a slightly lower ( for the baseline aggregation ) or equal ( for CPT- and ATM inhibitor-induced aggregation ) intrinsic propensity to aggregate compared to NIA proteins ( Figure 3A , Figure 3—figure supplement 1A ) . However , even proteins with a low intrinsic propensity to aggregate can be supersaturated and be vulnerable to aggregation , when they are expressed at sufficiently high levels . Interestingly , our MS/MS analysis revealed that proteins that aggregate in CPT- or ATM-inhibited-treated HEK293T cells are in general highly abundant compared to NIA proteins in ( Figure 3B ) . Cross-referencing the aggregated proteins in our datasets against a cell-line-specific NSAF reference proteome ( Geiger et al . , 2012 ) confirmed this ( Figure 3—figure supplement 1B ) . After performing RNA sequencing on the same HEK293T cell samples that we used for our MS/MS analysis ( Figure 2—figure supplement 1A , Supplementary file 2 ) , we found that genes coding for the aggregating proteins are in general higher expressed than genes coding for NIA proteins ( Figure 3C ) . To evaluate whether these proteins are indeed supersaturated , we used the method validated by Ciryam et al . , which uses transcript abundance and aggregation propensity as predicted by TANGO to estimate supersaturation ( Ciryam et al . , 2013 ) . Using the RNA-sequencing data ( Supplementary file 2 ) , we confirmed that aggregating proteins are in general indeed more supersaturated than NIA proteins ( Figure 3D–F ) . Cross-referencing our data against the composite human supersaturation database generated by Ciryam et al . yielded a similar picture ( Figure 3—figure supplement 1C ) . Although the relative supersaturation of aggregating proteins in HEK293T cells is intriguing , our data also indicates that most supersaturated proteins did not become SDS-insoluble , even after treatment with CPT or ATM inhibition ( Figure 3F ) . Supersaturation only relates to overall protein concentration per cell , but within a cell , local protein concentrations can differ . A prime example of this is the partitioning of proteins in so-called biomolecular condensates through liquid-liquid phase separation ( LLPS ) . LLPS can increase the local concentration of proteins , which has been shown to be important for a wide range of cellular processes ( Lyon et al . , 2021 ) . However , it also comes with a risk of transitioning from a liquid to a solid , and even amyloid state . Indeed , a large amount of recent data have clearly demonstrated that proteins that engage in LLPS are overrepresented among proteins that aggregate in various proteinopathies ( reviewed in Alberti and Hyman , 2021 ) . Using catGRANULE ( Mitchell et al . , 2013; http://tartaglialab . com ) , we find that HEK293T baseline aggregation and both CPT- and ATM inhibitor-induced aggregation are indeed made up of proteins that have a higher average LLPS propensity than NIA proteins ( Figure 3G ) . HEK293T baseline and ATM inhibitor-induced aggregation are also enriched for proteins that have a high propensity to engage in LLPS-relevant pi-pi interactions , as indicated by both a higher average PScore and a larger percentage of proteins that have a PScore > 4 ( i . e . , above the threshold defined by Vernon et al . , 2018; Figure 3—figure supplement 1D and E ) . Inversely , dividing NIA proteins into supersaturated and non-supersaturated subgroups reveals that they have a similarly low average LLPS propensity ( Figure 3—figure supplement 1F and G ) . This points out that a high LLPS propensity can discriminate supersaturated proteins that are prone to aggregate from supersaturated proteins that are not . Upon examining the proteins that aggregate in U2OS cells , we found further support for this . Baseline aggregation in U2OS cells is also made up of supersaturated , LLPS-prone proteins ( Figure 3—figure supplement 1H–R ) . Despite the baseline aggregation being far more pronounced in U2OS cells than in HEK293T cells , many supersaturated proteins are not SDS-insoluble in U2OS cells , even in cells treated with CPT or in cells lacking ATM ( Figure 3—figure supplement 1O ) . In U2OS ATM KO cells , proteins that aggregate more are supersaturated compared to U2OS NIA proteins ( Figure 3—figure supplement 1M ) ; for CPT-treated cells , this is not the case ( Figure 3—figure supplement 1M–O ) . Proteins that aggregate more in U2OS ATM KO cells also have a higher general propensity to engage in LLPS as predicted by PScore and catGRANULE ( Figure 3—figure supplement 1P and Q ) , while proteins that aggregate more in CPT-treated cells are enriched for proteins with a PScore > 4 ( Figure 3—figure supplement 1R ) . From this , we conclude that both CPT treatment and a loss of ATM further exacerbate the aggregation of LLPS-prone and supersaturated proteins in a cell-type-dependent manner . A GO term analysis of our RNAseq data revealed a striking lack of overlap in transcriptional processes altered upon treatment with CPT or loss of ATM function in either HEK293T or U2OS cells ( Figure 3—figure supplement 2 , Supplementary file 2 ) . Intriguingly , only two transcripts ( one up and one down ) were significantly altered ( –1 > log2FC > 1 ) after ATM inhibition in HEK293T cells despite the enhanced aggregation occurring in these cells . This further underlines that the enhanced aggregation after CPT treatment or a loss of ATM function is mostly driven by the physicochemical characteristics of the proteins involved . Our data shows that a substantial number of inherently similarly vulnerable proteins aggregate under the genotoxic conditions of CPT treatment or ATM loss . Their consistent aggregation across independent repeats argues against the possibility that this is caused by any genotoxic stress-induced DNA sequence alterations in their own coding regions as these would occur more randomly throughout the genome . Moreover , we find that the increased aggregation can also not be explained by any changes in abundance of the proteins involved , resulting for example from DNA damage-induced transcriptional dysregulation , as very limited overlap exists between proteins that aggregate and proteins with an altered expression upon CPT treatment or ATM loss ( see Figure 4A for HEK293T and Figure 4—figure supplement 1A for U2OS ) . Instead , our data indicate that a long-term consequence of these genotoxic conditions is a global lowering of the aggregation threshold of proteins . As a result , more and more LLPS-prone , supersaturated proteins that are normally largely soluble now start to aggregate , with the most vulnerable proteins aggregating first . This aggregation threshold appears to be inherently lower in U2OS cells compared to HEK293T cells , causing a large population of metastable proteins to aggregate already under normal conditions . Genotoxic stress in U2OS cells lowers the aggregation threshold even further , causing a ‘second layer’ of LLPS-prone proteins that are not even always supersaturated to aggregate also ( Figure 4—figure supplement 1B ) . This lowering of the aggregation threshold is highly reminiscent of ‘classic’ protein aggregation resulting from ( chronic ) proteotoxic stresses ( Weids et al . , 2016 ) and has been referred to as a disturbed ( Hipp et al . , 2019 ) or shifted protein homeostasis ( Ciryam et al . , 2013 ) . In line with this , we find that proteins that aggregate more in HEK293T and U2OS cells after CPT treatment or after a ( functional ) loss of ATM are enriched for proteins that have been reported to aggregate upon heat treatment of cells ( Figure 4B , Figure 4—figure supplement 1C; Mymrikov et al . , 2017 ) . In addition , they are enriched for constituents of stress granules ( Figure 4C , Figure 4—figure supplement 1D; http://rnagranuledb . lunenfeld . ca ) , cellular condensates that have been found to function as nucleation sites for protein aggregation ( Dobra et al . , 2018; Mateju et al . , 2017 ) . Protein aggregation after heat shock and the formation of aberrant stress granules also includes the aggregation of newly synthesized proteins ( Ganassi et al . , 2016; Xu et al . , 2016 ) . To assess the extent to which newly synthesized proteins aggregate in cells exposed to genotoxic conditions , we pulsed HEK293T cells with 35S-labeled cysteine and methionine 48 hr after CPT treatment ( Figure 4—figure supplement 1E ) . Notably , protein synthesis is reduced approximately threefold in CPT-treated cells ( Figure 4D ) . Despite this strong reduction in protein synthesis , radioactively labeled proteins were still clearly present in the aggregating fraction of CPT-treated cells . They were however not enriched in the aggregating fraction compared to control cells ( Figure 4D ) , indicating that the enhanced aggregation triggered by CPT is not explained through an accelerated aggregation of specifically newly synthesized proteins . A shift in protein homeostasis has also been suggested to be key to the build-up of protein aggregates during aging ( Ciryam et al . , 2014 ) and to the initiation of protein aggregation in a range of chronic disorders ( David et al . , 2010; Hipp et al . , 2019; Morley et al . , 2002 ) . Intriguingly , we find that proteins that aggregate after transient CPT treatment are enriched for constituents of various disease-associated protein aggregates ( Figure 4E ) . 67% ( 82/122 ) of them – or their mouse homologs – have already previously been identified in TDP-43 aggregates ( Dammer et al . , 2012; Zuo et al . , 2021 ) , Lewy bodies ( McCormack et al . , 2019 ) , or α-synuclein-induced aggregates ( Mahul-Mellier et al . , 2020 ) , or found to aggregate in Huntington’s disease ( HD ) ( Hosp et al . , 2017 ) or Alzheimer’s disease ( AD ) brains ( Hales et al . , 2016; Kepchia et al . , 2020 Figure 4—figure supplement 2 ) . An enrichment for HD and AD brain aggregating proteins was also observed among proteins that aggregate after inhibition of ATM ( Figure 4E ) . If genotoxic conditions indeed over time lead to a lowering of the aggregating threshold , this would predict that they can also result in an accelerated aggregation of aggregation-prone model substrates . For example , disease-associated expanded polyQ proteins are inherently aggregation prone , and they have been shown to aggregate faster in systems in which protein homeostasis is impaired ( Gidalevitz et al . , 2013; Gidalevitz et al . , 2010 ) . We went back to HEK293 cells and employed a line carrying a stably integrated , tetracycline-inducible GFP-tagged Huntingtin exon 1 containing a 71 CAG-repeat ( encoding Q71 ) . Transient targeting of ATM , ATR , and in particular TOPs , but not TDP1 , 24–48 hr prior to the expression of polyQ ( Figure 4—figure supplement 1F ) indeed accelerated polyQ aggregation in these cells ( Figure 4F , Figure 4—figure supplement 1G ) , closely mirroring the increased aggregation that we observed before ( Figure 1C and D ) . The accelerated polyQ aggregation under these conditions is also dose-dependent ( Figure 4G , Figure 4—figure supplement 1H and I ) , and it is not explained by changes in total polyQ levels ( Figure 4—figure supplement 1G–I ) . PolyQ aggregation is normally proportional to the length of the CAG repeat , which is intrinsically unstable . Importantly , we find no evidence that the accelerated polyQ aggregation induced by these genotoxic conditions can be explained by an exacerbated repeat instability ( Figure 4—figure supplement 1J ) . Next , we also used the same tetracycline-inducible system and experimental set-up to investigate the aggregation of the protein folding model substrate luciferase-GFP ( Figure 4—figure supplement 1K ) . We find that transient targeting of either ATM or TOP1 results in an enrichment of luciferase-GFP in the aggregated fraction ( Figure 4H ) . We noted that ATM inhibition and in particular CPT treatment resulted in an increased aggregation of multiple ( co ) chaperones in HEK293T cells ( Figure 5—figure supplement 1A and B ) . In U2OS cells , many ( co ) chaperones are already aggregating regardless of exposure to genotoxic stress , but still several chaperones aggregated significantly more in ATM KO cells or in cells treated with CPT ( Figure 5—figure supplement 1C and D ) . The overlap that exists between aggregating chaperones in each cell line suggests that this occurs mostly cell line specific ( Figure 5A , Figure 5—figure supplement 1E ) . These findings are interesting as chaperone systems have the ability to modulate aggregation ( Hartl et al . , 2011; Mogk et al . , 2018; Sinnige et al . , 2020; Tam et al . , 2006 ) . HSP70s ( HSPAs ) are among the most ubiquitous chaperones , and they have been shown to play a key role in maintaining protein homeostasis in virtually all domains of life ( Gupta and Singh , 1994; Hunt and Morimoto , 1985; Lindquist and Craig , 1988 ) . Upon cross-referencing the NIA and aggregating fractions against a recently generated client database of HSPA8 ( HSC70; constitutively active form of HSP70 ) and HSPA1A ( constitutively active and stress-inducible HSP70 ) ( Ryu et al . , 2020 ) , we find that HSPA8 and HSPA1A clients are enriched among aggregating proteins ( Figure 5B ) . We also mined the BioGRID human protein-protein interaction database using the complete KEGG dataset of ( co ) chaperones ( 168 entries ) . Although the transient and energetically weak nature of the interactions between many ( co ) chaperones and their clients ( Clouser et al . , 2019; Kampinga and Craig , 2010; Mayer , 2018 ) makes it likely that these interactions are underrepresented in the BioGRID database , it can provide additional insight into the presence of ( putative ) chaperone clients in the aggregating fractions ( Victor et al . , 2020 ) . We find that all aggregating fractions are enriched for ( co ) chaperone interactors compared to nonaggregating proteins ( NIA ) ( Figure 5C , Figure 5—figure supplement 2 ) . Aggregating proteins have reported interactions with a broad range of chaperone families , most notably HSP70s and HSP90s ( and known co-factors of these ) , and chaperonins ( TRiC/CCT subunits ) ( Figure 5D , Figure 5—figure supplement 2 ) . Intriguingly , several ( co ) chaperones that we found to aggregate themselves are among the most frequent interactors ( Figure 5D ) . This suggests that they were sequestered by protein aggregates as they engaged their client proteins , in line with what has been reported for disease-associated aggregation ( Hipp et al . , 2019; Jana et al . , 2000; Kim et al . , 2013; Mogk et al . , 2018; Yu et al . , 2019; Yue et al . , 2021 ) . Overall , we find that the relative levels of chaperone engagement of the different aggregating fractions largely reflect their respective supersaturation and LLPS propensities . When the capacity of chaperone systems is overloaded , this can eventually trigger a rewiring of chaperone systems . This plasticity allows cells to adapt to varying circumstances and proteotoxic stress conditions ( Klaips et al . , 2018 ) . In HEK293T cells , we find that treatment with CPT results in an overall upward shift of ( co ) chaperone expression levels , as measured in both our WCL MS/MS analysis ( 16 up , 8 down ) ( Figure 5E ) and in our RNAseq dataset ( 11 up , 5 down ) ( Figure 5F ) . Upregulated chaperones include HSPB1 , DNAJA1 , HSPA5 , HSPA8 , and HSP90AA1 ( Figure 5E ) , all of which are among the most frequent interactors of aggregating proteins in CPT-treated cells . The expression of most of these chaperones is regulated by the heat shock factor 1 ( HSF1 ) transcription factor ( Metchat et al . , 2009; Neueder et al . , 2017; Ostling et al . , 2007; Trinklein et al . , 2004 ) . HSF1 is indeed partially activated by CPT treatment in HEK293T cells ( Figure 5G ) . ATM inhibition in HEK293T cells resulted in a marginal HSF1 activation . This is in line with an overall less pronounced aggregation response after ATM inhibition in HEK293T cells , which appears to be insufficient to initiate a clear rewiring of chaperone systems ( Figure 5—figure supplement 1F , Supplementary file 2 ) . In U2OS cells , a loss of ATM or treatment with CPT appears to result in a more balanced rewiring of chaperone systems ( Figure 5—figure supplement 1G–J ) . In CPT-treated U2OS cells , protein levels of multiple chaperones are even lowered . Nevertheless , similar to HEK293T cells , many of the most frequent ( co ) chaperone interactors of the aggregating proteins in U2OS are found to aggregate themselves as well . These findings indicate that ( sufficient ) genotoxic stress induces a rewiring of chaperone systems in a cell and stress-specific manner . This rewiring is however insufficient to prevent the increased aggregation of metastable client proteins . We reasoned that the difference in aggregation between HEK293T and U2OS cells might also be reflected in different chaperone expression levels already under normal conditions . Indeed , a differential expression analysis between untreated HEK293T and untreated U2OS cells revealed a strong overall upward shift of ( co ) chaperone gene expression levels in the latter ( Figure 5—figure supplement 1K ) . For example , we found that gene expression levels of the small heat shock-like protein Clusterin ( CLU ) are >100-fold higher in wild-type U2OS compared to HEK293T cells , gene expression levels of the stress-inducible HSPA1A are >150-fold higher , and that of HSPB5 ( or CRYAB , i . e . , αB-crystallin ) were >400-fold higher . Interestingly , the differences in expression of chaperone systems in U2OS compared to HEK293T overlap with the changes occurring after CPT treatment in the latter . Out of the 13 ( co ) chaperones identified to be expressed differently in both ( RNAseq; –1 > log2FC > 1 ) , 12 are altered in the same direction ( Figure 5H ) . Our data suggest that the lowering of the aggregation threshold upon various genotoxic conditions is caused by an overload of chaperone systems , leading to a shift in protein homeostasis . We reasoned that targeting chaperone systems may then exacerbate aggregation . Indeed , mild HSP70 inhibition using the HSP70/HSC70 inhibitor VER-155008 after CPT treatment increased CPT-induced protein aggregation even further , while having no clear impact on aggregation in control cells ( Figure 6—figure supplement 1A ) . Similar results were obtained when we blotted the aggregated fractions for MCM7 and TUBA1A ( Figure 6A ) . We next reasoned that increasing chaperone capacity may also raise the aggregation threshold again . We screened an overexpression library of several major chaperone families , including HSPAs , J-domain proteins ( JDPs ) , and small heat shock proteins ( HSPBs ) for their ability to reduce the increased protein aggregation triggered by genotoxic conditions using U2OS ATM KO cells as a model ( Figure 6—figure supplement 1B ) . While most of these had no overt effect , overexpression of several JDPs reduced protein aggregation , including the generic anti-amyloidogenic protein DNAJB6b ( Aprile et al . , 2017; Hageman et al . , 2010 ) . However , we found that the small heat shock protein HSPB5 was especially effective . HSPB5 is a potent suppressor of aggregation and amyloid formation ( Delbecq and Klevit , 2019; Golenhofen and Bartelt-Kirbach , 2016; Hatters et al . , 2001; Webster et al . , 2019 ) . Its higher expression in U2OS cells compared to HEK293T cells , as well as its further upregulation in U2OS cells lacking ATM , suggests that it plays an important role in counteracting widespread protein aggregation in these cells . We generated U2OS cells that stably overexpress HSPB5 in both wild-type and ATM-deficient backgrounds ( Figure 6—figure supplement 1C ) , and confirmed that this drastically reduced the enhanced protein aggregation in the latter ( Figure 6B and C ) . HSPB5 overexpression also reduced ProteoStat aggresome staining and the occurrence of cytoplasmic FUS puncta ( without affecting overall FUS levels; see Supplementary file 1 ) , two other markers of a disrupted protein homeostasis ( Figure 6D–G; Neumann et al . , 2006; Shen et al . , 2011 ) . Although HSPB5 itself has never been linked to genome maintenance , we evaluated whether HSPB5 can mitigate the increased aggregation following a loss of ATM in U2OS cells by altering DNA repair capacity . However , we found no indication for this as the gamma irradiation-induced DNA lesion accumulation and subsequent resolution as measured by 53BP1 foci formation was not affected by HSPB5 expression in neither U2OS wild-type nor ATM KO cells ( Figure 6—figure supplement 1D and E ) . Moreover , neither HSPB5 , nor HSP70 nor HSP90 accumulated at either CPT- or gamma irradiation-induced DNA damage sites ( Figure 6—figure supplement 2A–C ) . This points out that the overload of chaperones is not due to a sequestering to DNA damage sites . HSPB5 is one of two chaperones that are transcriptionally upregulated in U2OS cells after either a loss of ATM or CPT treatment , and the only one that is not transcriptionally upregulated in CPT-treated HEK293T cells ( Figure 6—figure supplement 1F ) . Crucially , we found that overexpression of HSPB5 can reduce the enhanced protein aggregation in CPT-treated U2OS cells as well ( Figure 6H , I ) , but that stable overexpression of HSPB5 has no effect on the CPT-induced aggregation in HEK293 cells ( Figure 6—figure supplement 1G ) . These data emphasize that the rewiring of chaperone systems in response to genotoxic stress is tailored to each cell , depending largely on the ground state of protein homeostasis and concomitant aggregation that occurs .
Here , we report that TOP poisoning and functional impairment of ATM or ATR trigger a widespread aggregation of LLPS-prone and supersaturated proteins . Our data show that the aggregation of these metastable proteins is a consequence of an overload of chaperone systems under these genotoxic conditions . This is illustrated by the aggregation of certain chaperones , and the observation that specifically the ( putative ) clients of these chaperones aggregate as well . It is further supported by the notion that CPT treatment leads to a strong reduction in protein synthesis over time , something that has been reported for other forms of DNA damage as well ( Halim et al . , 2018 ) . Reduced protein synthesis is a well-known response to proteotoxic stress and is believed to lower the strong demand for chaperone capacity of nascent chains that are innately vulnerable to misfolding and aggregation ( Balchin et al . , 2016 ) . This indicates that despite a reduction in protein synthesis , genotoxic stress still leads to an overload of chaperone systems , causing protein homeostasis to shift . This effectively lowers the cell-intrinsic threshold of protein aggregation , and as a result , vulnerable proteins that are largely kept soluble under normal conditions now succumb more readily to aggregation . The accelerated aggregation of the model substrates polyQ- and luciferase that occurs in cells exposed to these conditions underlines this threshold change as well . The observed shift in protein homeostasis after genotoxic stress is strikingly reminiscent of what is believed to occur under conditions of ( chronic ) stress ( Weids et al . , 2016 ) and during many age-related neurodegenerative disorders ( David et al . , 2010; Hipp et al . , 2019; Morley et al . , 2002 ) . Supersaturated proteins have been found to be overrepresented in cellular pathways associated with these disorders ( Ciryam et al . , 2015 ) , and disease-associated aggregating proteins , including FUS , tau , and α-synuclein , are known to exhibit LLPS behavior ( reviewed in Zbinden et al . , 2020 ) . Indeed , we find that the proteins that aggregate in our experiments show a strong overlap in identity and function with stress-induced aggregation , and with the aggregation observed in various proteinopathies . The shift in protein homeostasis under genotoxic stress conditions can theoretically be caused by either an altered capacity of protein quality control systems or by an increased demand emanating from an altered proteome . These are , however , difficult to disentangle fully , in particular because they may form a vicious cycle of events , where ( co ) chaperones are increasingly sequestered as a growing number of proteins succumbs to aggregation ( Klaips et al . , 2018 ) . Either way , both result in a net lack of protein quality control capacity , which can be rescued by upregulating specific chaperones , and exacerbated by further decreasing chaperone capacity . The aggregation that occurs under the genotoxic conditions used in our study follows this pattern . Nevertheless , our data lead us to hypothesize that the overload of chaperone systems is largely caused by an increased demand for chaperone activity in the proteome . Multiple ( co ) chaperones that have been reported to interact frequently with the aggregating proteins are upregulated under the genotoxic conditions used in our study . Many of these ( co ) chaperones aggregate themselves as well . Crucially , further overexpression of one of the most upregulated chaperones in U2OS , HSPB5 , is able to largely bring aggregation in CPT-treated and ATM KO cells back down to the control level . The strong upregulation of several small heat shock ( -like ) proteins in U2OS cells , including HSPB5 , seems to point at a rewiring of the chaperone network in this cell line towards a more prominent reliance on this class of chaperones . Small heat shock proteins have been reported to act together in heterodimers and hetero-oligomers ( Aquilina et al . , 2013; Mymrikov et al . , 2020 ) . In addition , in vivo they rely on other chaperone systems such as the HSP70 machinery to efficiently counteract aggregation ( Mogk et al . , 2003; Reinle et al . , 2022; Zwirowski et al . , 2017 ) . The low expression of small heat shock proteins in general and of HSPB5 specifically in HEK293 ( T ) cells suggests that these cells might not be equipped to wield elevated levels of HSPB5 to prevent widespread aggregation of proteins . This may explain why elevated levels of HSPB5 have no effect on CPT-induced aggregation in HEK293 cells . The strong overlap between CPT-induced aggregation in HEK293T cells and baseline aggregation in U2OS cells also argues for an increased demand . U2OS is a cancer cell line ( osteosarcoma ) , whereas HEK293 ( T ) cells have a vastly different origin ( embryonic kidney ) . Cancer cells inherently exhibit elevated levels of protein stress , which has been attributed to an increased protein folding and degradation demand ( Dai et al . , 2012; Deshaies , 2014 ) . The notion that the rewiring of chaperone systems in response to CPT treatment in HEK293T cells mimics the difference in chaperone wiring between HEK293T and U2OS cells underlines this further . Our data indicate that any increased demand caused by these genotoxic conditions is , however independent of quantitative changes of the aggregating proteins themselves and likely also of any genetic alterations in their coding regions ( in cis genetic alterations ) . The accelerated polyQ aggregation – not accompanied by any enhanced CAG repeat instability – provides support for this . This is not necessarily surprising as proteins that aggregate as a consequence of an overload of the protein quality control do not have to be altered themselves . Previous studies have shown that during proteomic stress a destabilization of the background proteome can result in a competition for the limited chaperone capacity available , causing proteins that are highly dependent on chaperones for their stability and solubility to aggregate readily ( Gidalevitz et al . , 2010; Gidalevitz et al . , 2011 ) . In this light , it is interesting that the proteins that aggregate in our experiments are in general prone to engage in LLPS . LLPS is known to be regulated by RNA and often involves RNA-binding proteins . Strikingly , we find that the aggregation that occurs after genotoxic stress is enriched for RNA-binding proteins . This enhanced aggregation can also be mitigated by inhibiting PARylation , which plays a key role in the regulation of LLPS processes ( Duan et al . , 2019; McGurk et al . , 2018 ) . LLPS is a different biochemical process than protein aggregation ( with different underlying mechanisms and principles ) , but aberrant LLPS can drive the nucleation of insoluble ( fibrillar ) protein aggregates , for example , for polyQ ( Peskett et al . , 2018 ) . It is therefore believed that LLPS events need to be closely regulated and monitored to prevent aberrant progression into a solid-like state ( Alberti and Dormann , 2019 ) . Although data is so far limited , chaperones , and in particular small heat shock proteins , have been reported to play a pivotal role in the surveillance of biomolecular condensates . For example , the HSPB8-BAG3-HSPA1A complex has been found to be important for maintaining stress granule dynamics ( Ganassi et al . , 2016 ) , and recent work uncovered that HSPB1 is important to prevent aberrant phase transitions of FUS ( Liu et al . , 2020 ) . We find that HSPB8 , BAG3 , and HSPA1A are upregulated in HEK293T cells treated with CPT . Interestingly , although HSPB5 itself has so far not been shown to undergo LLPS , like HSPB1 , it has been found to associate with nuclear speckles ( van den IJssel et al . , 1998 ) , which are membraneless as well . HSPB5 has also been shown to be important to maintain the stability of the cytoskeleton ( Ghosh et al . , 2007; Golenhofen et al . , 1999; Yin et al . , 2019 ) , and we find that many proteins that aggregate upon genotoxic stress conditions are cytoskeleton ( -related ) components . A growing body of evidence indicates that cytoskeleton organization is regulated through LLPS processes ( reviewed in Wiegand and Hyman , 2020 ) . Our data thus indicate that genotoxic stress conditions can exacerbate the risk of aberrant progression of LLPS processes , which in turn may trigger an overload of chaperone systems . The increased protein aggregation that occurs after a loss of ATM – including in A-T patient brains – has been recently attributed to an accumulation of DNA damage ( Lee et al . , 2021 ) . As an impaired response to DNA damage is believed to be the primary driving force of A-T phenotypes ( Shiloh , 2020 ) , these findings have fueled the idea that a disruption of protein homeostasis may be an important disease mechanism in A-T . Our data provide further support for this as they show that the widespread aggregation caused by a loss of ATM follows a predictable pattern that overlaps strikingly with the aggregation that is believed to underlie many neurodegenerative disorders . Importantly , our findings also provide a proof of principle that other genotoxic conditions – including chemotherapeutic TOP poisons – can have a very similar impact . This points at the existence of a broader link between DNA damage and a loss of protein homeostasis . Although further research is needed to determine the full breadth and relevance of this link , our work may thus offer clues as to why besides impairments in ATM many other genome maintenance defects are characterized by often overlapping ( neuro ) degenerative phenotypes as well ( Petr et al . , 2020 ) .
Statistical testing was performed using GraphPad Prism software , except for label-free quantification ( LFQ ) proteomics and RNA sequencing , which were analyzed in R ( see their respective sections for more information ) . The statistical tests that were used are indicated in each figure legend . For experiments with pairwise comparisons , two-tailed Student’s unpaired t-test was used unless otherwise indicated . For experiments with multiple comparisons , a Kruskal–Wallis with Dunn’s post-hoc test ( when datasets did not pass normality testing ) or two-tailed Student’s unpaired t-tests with Bonferroni correction ( when indicated ) was performed . p-Values are shown for all experiments . All repetitions ( n ) originate from independent replicates; any representation of technical repeats in figures is explicitly mentioned in the accompanying legend . Gels and stains were processed and analyzed using ImageJ software ( Fiji ) . All parental cell lines were obtained from ATCC ( see Key resources table ) and are mycoplasma negative ( GATC Biotech GA , Konstanz , Germany ) . Cell lines were cultured in DMEM ( Gibco ) supplemented with 10% FBS ( Sigma-Aldrich ) , 100 units/ml penicillin , and 100 µg/ml streptomycin ( Invitrogen ) . HEK293 cells expressing inducible GFP-Httexon1-Q71 ( GFP-Q71 ) have been described previously ( Hageman et al . , 2010 ) , and HEK293 cells expressing inducible luciferase-GFP as well ( Hageman et al . , 2011 ) . U2OS and HEK293 ATM KO cells were generated using the ATM CRISPR/Cas9 KO and ATM HDR plasmids ( sc-400192 , sc-400192-HDR from Santa Cruz ) according to the manufacturer’s guidelines . Individual clones were picked and verified by PCR and Western blotting . For the generation of U2OS and HEK293 cells overexpressing HSBP5 , see later section . For Western blotting , proteins were transferred to either nitrocellulose or PVDF membranes , probed with the indicated antibodies , and imaged in a Bio-Rad ChemiDoc imaging system . For an overview of all antibodies used in this study , see Key resources table . For ( immuno ) staining , cells were grown on coverslips , fixed in 2% formaldehyde , permeabilized with 0 . 1–0 . 2% Triton-X100 , and incubated for 15 min with 0 . 5% BSA and 0 . 1% glycine solution in PBS . ProteoStat staining ( ENZO , ENZ-51023-KP050 ) was performed according to the manufacturer’s instructions . Primary antibody incubation ( see Key resources table ) was performed overnight at 4°C . After secondary antibody incubation , cells were stained with Hoechst ( Invitrogen , H1399 ) or DAPI as indicated , and mounted on microscopy slides in Citifluor ( Agar Scientific ) . Cells were observed using a confocal scanning microscope ( Leica ) , and images were processed using ImageJ software ( Fiji ) . The aggresome signature was defined as cells exhibiting both a curved nucleus and perinuclear presence of ProteoStat dye . For DNA repair kinetics experiments , cells were irradiated with 2 Gy ( IBL-637 irradiator , CIS Biointernational ) , fixed at the indicated timepoints post-irradiation , and stained for 53BP1 . For the localization experiments of HSPB5 , HSP70 , and HSP90 , U2OS cells were either irradiated with 2 Gy or treated for 24 hr with CPT ( 400 nM ) , and left to recover for 48 hr before immunostaining . See Table 1 ( and Key resources table ) for an overview of the drugs and concentrations used in this study . For genotoxic drug treatments , cells were treated with drugs in the indicated doses . The culture medium was replaced 24 hr after drug treatment , and after another 48 hr ( unless explicitly mentioned otherwise ) cells were harvested by scraping in PBS , centrifugation , and snap-freezing in liquid nitrogen . Cells were resuspended in ice-cold lysis buffer containing 25 mM HEPES pH 7 . 4 , 100 mM NaCl , 1 mM MgCL2 , 1% v/v Igepal CA-630 ( #N3500 , US Biological ) , cOmplete EDTA-free protease inhibitor cocktail ( Roche Diagnostics ) , and 0 . 1 unit/µl benzonase endonuclease ( Merck Millipore ) and left for 1 hr on ice with intermittent vortexing . Protein content was measured and equalized , and Igepal CA-630 insoluble proteins were pelleted by high-speed centrifugation ( 21 , 000 rcf , 45 min , 4°C ) . Protein pellets were washed with lysis buffer without Igepal CA-630 and redissolved in lysis buffer supplemented with 1% v/v SDS at room temperature ( RT ) in a Thermomixer R ( Eppendorf ) at 1200 rpm for 1–2 hr . SDS-insoluble proteins were then pelleted by high-speed centrifugation ( 21 , 000 rcf , 45 min ) . SDS-insoluble protein pellets were washed with lysis buffer without any detergent . For subsequent silver staining , pellets were solubilized in urea buffer ( 8 M urea , 2% v/v SDS , 50 mM DTT , 50 mM Tris/HCl pH 7 . 4 ) overnight at RT in a Thermomixer R ( Eppendorf ) at 1200 rpm . For subsequent Western blotting , pellets were solubilized in concentrated sample buffer ( 4% SDS , 125 mM Tris pH 6 . 8 , 100 mM DTT , 10% glycerol , bromophenol blue ) , boiled for 10 min , and left overnight RT in a Thermomixer R ( Eppendorf ) at 1200 rpm . Fractions were separated using SDS-PAGE , imaged using a Bio-Rad ChemiDoc imaging system , and analyzed using ImageJ software ( Fiji ) . Samples were reduced ( dithiothreitol 25 mM , 37°C , 30 min ) , alkylated ( iodoacetamide 100 mM , RT , 30 min in darkness ) and trypsin digested on S-trap columns ( Protifi ) using the S-Trap micro protocol ( https://files . protifi . com/protocols/s-trap-micro-long-4-7 . pdf ) . After elution , samples where dried up on speed-vac and resuspended in 25 µl of 0 . 1% ( v/v ) formic acid in water ( MS quality , Thermo ) . Mass spectral analysis was conducted on a Thermo Scientific Orbitrap Exploris . The mobile phase consisted of 0 . 1% ( v/v ) formic acid in water ( A ) and 0 . 1% ( v/v ) formic acid in acetonitrile ( B ) . Samples were loaded using a Dionex Ultimate 3000 HPLC system onto a 75 µm × 50 cm Acclaim PepMap RSLC nanoViper column filled with 2 µm C18 particles ( Thermo Scientific ) using a 120 min LC-MS method at a flow rate of 0 . 3 µl/min as follows: 3% B over 3 min; 3–45% B over 87 min; 45–80% B over 1 min; then wash at 80% B over 14 min , 80 to 3% B over 1 min and then the column was equilibrated with 3% B for 14 min . For precursor peptides and fragmentation detection on the mass spectrometer , MS1 survey scans ( m/z 200–2000 ) were performed at a resolution of 120 , 000 with a 300% normalized AGC target . Peptide precursors from charge states 2–6 were sampled for MS2 using Data Dependent Acquisition ( DDA ) . For MS2 scan properties , Higher Energy Collision Dissociation ( HCD ) was used and the fragments were analyzed in the orbitrap with a collisional energy of 30 % , resolution of 15000 , Standard AGC target , and a maximum injection time of 50 ms . MaxQuant version 1 . 6 . 7 . 0 was used for peptides and protein identification ( Tyanova et al . , 2016 ) and quantification with a proteomic database of reviewed proteins sequences downloaded from Uniprot ( 08/17/2020 , proteome:up000005640; reviewed:yes ) . Abbreviated MaxQuant settings: LFQ with minimum peptide counts ( razor + unique ) ≥ 2 and at least one unique peptide; variable modifications were oxidation ( M ) , acetyl ( protein N-term ) , and phospho ( STY ) ; carbamidomethyl ( C ) was set as a fixed modification with trypsin/P as the enzyme . ProteinGroup . txt from MaxQuant output was used for protein significance analysis via postprocessing in R ( R Core Team 2021 ) : potential contaminant and reversed protein sequences were filtered out , partial or complete missing values in either case or control replicates were imputed ( Dou et al . , 2020 ) in parallel 100 times , and subsequently averaged log2-transformed LFQ intensities were used for t-tests , including Benjamini–Hochberg-corrected , p-adjusted values . Log2 fold change for each protein record was calculated by subtracting the average log2 LFQ intensity across all replicates in control samples from the average log2 LFQ intensity across all replicates in case samples . To mitigate imputation-induced artifacts among significant proteins , only significant proteins detected and quantified in at least two replicates were considered: p-adjusted value ≤ 0 . 05 and , for cases ( log2 fold change ≥ 1 , replicates with nonimputed data ≥ 2 ) , or for controls ( log2 fold change ≤ –1 , replicates with nonimputed data ≥ 2 ) . RNA was isolated from cells with the AllPrep DNA/RNA Mini Kit from QIAGEN . RNA concentrations were measured on a NanoDrop . 150 ng of RNA was used for library preparation with the Lexogen QuantSeq 3′ mRNA-Seq Library Prep Kit ( FWD ) from Illumina . Quality control of the sequencing libraries was performed with both Qubit ( DNA HS Assay kit ) and Agilent 2200 TapeStation systems ( D5000 ScreenTape ) . All libraries were pooled equimolar and sequenced on a NextSeq 500 at the sequencing facility in the University Medical Center Groningen , Groningen , the Netherlands . Data preprocessing was performed with the Lexogen QuantSeq 2 . 3 . 1 FWD UMI pipeline on the BlueBee Genomics Platform ( 1 . 10 . 18 ) . Count files were loaded into R and analyzed with edgeR Robinson et al . , 2010 . Only genes with >1 counts in at least two samples were included in the analysis . Count data was normalized using logCPM for principal component analysis ( PCA ) . Differential gene expression analysis was performed using the likelihood ratio test implemented in edgeR . Cutoffs of an absolute log fold change > 1 and an FDR-adjusted p-value<0 . 05 were used to identify significantly differentially expressed genes . For MS/MS , GO term analyses were performed through Cytoscape within Python , with a redundancy cutoff of 0 . 2 . For RNA sequencing , GO term analyses were performed through Metascape ( webserver: https://metascape . org ) using default settings . Radioactive pulse labeling experiments were executed with cells subjected to the same CPT treatment regime as depicted in Figure 1—figure supplement 1D ( see also Figure 4D ) . After 48 hr of recovery , cells were starved of methionine and cysteine for 30 min DMEM without methionine and cysteine ( Gibco ) , see Key resources table , supplemented with 10% dialyzed FBS ( Sigma-Aldrich ) , 100 units/ml penicillin , and 100 µg/ml streptomycin ( Invitrogen ) . Then , 35S-met/cys ( Hartmann Analytics ) pulse labeling was performed for 10–40 min , and immediately after cells were harvested by scraping in PBS , centrifugation , and snap-freezing in liquid nitrogen . Protein fractionation was then performed as described . Autoradiography was performed by running samples on SDS-PAGE gels , gel drying , and placing gels on blank phosphor screens , shielded from light . After 1 week , phosphor screens were imaged using a Cyclone Plus Phosphor Image ( Perkin Elmer ) . 24 hr after seeding , stable tetracycline-inducible HTT Q71-GFP-expressing HEK293 cells were treated with the indicated genotoxic drugs listed in Table 1 , as described . Cell lysis , polyQ filter trap , and immunodetection were performed as described previously ( Kakkar et al . , 2016 ) , and results were analyzed using ImageJ software ( Fiji ) . DNA was isolated from HTT Q71-GFP-expressing HEK293 cells through MasterPure Complete DNA and RNA Purification Kit ( Epicentre ) according to the manufacturer’s instructions . The CAG repeat length analysis was performed by PCR with 100 ng of DNA in a 10 µl reaction volume containing AmpliTaq Gold Fast PCR Master Mix ( Applied Biosystems ) , and 0 . 2 µM of both forward ( HEK293TQ71F [FAM]: 5'-GAGTCCCTCAAGTCCTTCC-3' ) and reverse ( HEK293TQ71R: 5'-AAACGGGCCCTCTAGACTC-3' ) primers , flanking the CAG repeat tract . The samples were subjected to an initial denaturation step ( 95° C , 10 min ) , 35 amplification cycles ( 96°C , 15 s; 59 . 2°C , 15 s; 68°C , 30 s ) and a final extension of 72°C , 5 min . PCR was followed by capillary electrophoresis in a ABI3730XL Genetic Analyzer , and results were analyzed through GeneMapper Software V5 . 0 ( both Applied Biosystems ) . Retrovirus was produced in the Phoenix-AMPHO retroviral packaging cell line using a pQCXIN–HSPB5 vector as described before ( Schepers et al . , 2005 ) . Briefly , HEK293 , U2OS wild-type , and ATM KO cells were infected in the presence of 5 µg/ml polybrene ( Santa Cruz ) . Cells in which the HSPB5 vector integrated successfully were selected using G418 , and HSPB5 overexpression was confirmed via Western blotting . For an overview of online databases used in this study , see Table 2 . The MS/MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository ( Perez-Riverol , 2018 ) with the dataset identifier PXD030166 . The RNAseq data generated in this study are available through Gene Expression Omnibus with accession number GSE173940 . The R code for the MS/MS analysis can be found here ( copy archived at swh:1:rev:bda88adfdacefd6841d80c0c92e92b33b42c9b9c; LaCavaLab , 2022a ) and here ( copy archived at swh:1:rev:1d1711c210a0ac34f09499aa37c46989439ffcbe; LaCavaLab , 2022b ) . For the RNAseq differential expression analysis the R code can be found on github ( copy archived at swh:1:rev:e9e5879e270d8788d6f385159e2efcfd49e9c5e0; Huiting , 2022 ) . | Cells are constantly perceiving and responding to changes in their surroundings , and challenging conditions such as extreme heat or toxic chemicals can put cells under stress . When this happens , protein production can be affected . Proteins are long chains of chemical building blocks called amino acids , and they can only perform their roles if they fold into the right shape . Some proteins fold easily and remain folded , but others can be unstable and often become misfolded . Unfolded proteins can become a problem because they stick to each other , forming large clumps called aggregates that can interfere with the normal activity of cells , causing damage . The causes of stress that have a direct effect on protein folding are called proteotoxic stresses , and include , for example , high temperatures , which make proteins more flexible and unstable , increasing their chances of becoming unfolded . To prevent proteins becoming misfolded , cells can make ‘protein chaperones’ , a type of proteins that help other proteins fold correctly and stay folded . The production of protein chaperones often increases in response to proteotoxic stress . However , there are other types of stress too , such as genotoxic stress , which damages DNA . It is unclear what effect genotoxic stress has on protein folding . Huiting et al . studied protein folding during genotoxic stress in human cells grown in the lab . Stress was induced by either blocking the proteins that repair DNA or by ‘trapping’ the proteins that release DNA tension , both of which result in DNA damage . The analysis showed that , similar to the effects of proteotoxic stress , genotoxic stress increased the number of proteins that aggregate , although certain proteins formed aggregates even without stress , particularly if they were common and relatively unstable proteins . Huiting et al . ’s results suggest that aggregation increases in cells under genotoxic stress because the cells fail to produce enough chaperones to effectively fold all the proteins that need it . Indeed , Huiting et al . showed that aggregates contain many proteins that rely on chaperones , and that increasing the number of chaperones in stressed cells reduced protein aggregation . This work shows that genotoxic stress can affect protein folding by limiting the availability of chaperones , which increases protein aggregation . Remarkably , there is a substantial overlap between proteins that aggregate in diseases that affect the brain – such as Alzheimer’s disease – and proteins that aggregate after genotoxic stress . Therefore , further research could focus on determining whether genotoxic stress is involved in the progression of these neurological diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"genetics",
"and",
"genomics"
] | 2022 | Targeting DNA topoisomerases or checkpoint kinases results in an overload of chaperone systems, triggering aggregation of a metastable subproteome |
Stress is known to exert its detrimental effects not only by enhancing fear , but also by impairing its extinction . However , in earlier studies stress exposure preceded both processes . Thus , compared to unstressed animals , stressed animals had to extinguish fear memories that were strengthened by prior exposure to stress . Here , we dissociate the two processes to examine if stress specifically impairs the acquisition and recall of fear extinction . Strikingly , when fear memories were formed before stress exposure , thereby allowing animals to initiate extinction from comparable levels of fear , recall of fear extinction was unaffected . Despite this , we observed a persistent increase in theta activity in the BLA . Theta activity in the mPFC , by contrast , was normal . Stress also disrupted mPFC-BLA theta-frequency synchrony and directional coupling . Thus , in the absence of the fear-enhancing effects of stress , the expression of fear during and after extinction reflects normal regulation of theta activity in the mPFC , not theta hyperactivity in the amygdala .
Accumulating evidence from animal models shows that stress elicits divergent patterns of plasticity across brain regions ( Chattarji et al . , 2015 ) . For instance , repeated stress causes loss of dendrites and spines in the medial prefrontal cortex ( mPFC ) ( Shansky and Morrison , 2009 ) . In the basolateral amygdala ( BLA ) , by contrast , chronic stress strengthens the structural basis of synaptic connectivity through dendritic growth and spinogenesis ( Chattarji et al . , 2015 ) . Physiological and molecular measures of synaptic plasticity also exhibit these contrasting features . As useful as these studies have been in examining the effects of stress across biological scales , much of this evidence was gathered from postmortem analyses ( Chattarji et al . , 2015 ) . Less is known about how stress affects neural activity in the intact brain of behaving animals . Further , in many of these studies , stress-induced plasticity was viewed as stand-alone effects intrinsic to individual brain areas , despite extensive interconnections between them . Indeed , interactions between these brain areas together give rise to behaviors that are affected by stress ( Quirk and Mueller , 2008 ) . One such behavior involves the expression of fear memories , various facets of which depend on both the BLA and mPFC ( Sierra-Mercado et al . , 2011 ) . Repeated stress has been shown to enhance fear memories , as well as impair their extinction ( Miracle et al . , 2006; Suvrathan et al . , 2014 ) . These studies , however , first exposed animals to stress , and then subjected them to fear conditioning followed by extinction ( Izquierdo et al . , 2006; Miracle et al . , 2006 ) . In other words , these studies analyzed the impact of stress on both the formation and extinction of fear memories , not the latter alone . This experimental design has led to two broad classes of findings . In a majority of these studies , stress did not necessarily lead to higher freezing levels during fear recall . Even though expression of fear was not elevated by prior exposure to stress , it was resistant to subsequent extinction in stressed animals ( Noble et al . , 2017; Zhang and Rosenkranz , 2013 ) . In a few studies , however , the behavioral data suggest that prior exposure to stress may have also led to stronger fear memories manifested as higher levels of freezing in stressed animals at the beginning of extinction training ( Chauveau et al . , 2012; Hoffman et al . , 2014; Miracle et al . , 2006 ) . This too could have contributed to the subsequent deficit in fear extinction recall . An alternative strategy to test if stress impairs the extinction of fear memories would be for animals to form fear memories before stress exposure , thereby dissociating the effects of stress on acquisition versus extinction of conditioned fear . This would offer an opportunity to examine how stress specifically affects expression of fear during and after extinction , without the confounds of stronger fear memories caused by prior exposure to stress . Hence , the present study combines simultaneous behavioral and in vivo electrophysiological analyses to address this question .
Rats were first subjected to auditory fear conditioning ( Day 0 , Figure 1 ) at the end of which they exhibited significantly higher freezing behavior in response to the tone conditioned stimulus ( CS ) ( Figure 1b ) . These animals were then divided into two groups – one was subjected to 10 days of chronic immobilization stress ( Days 1–10 , Figure 1a ) while the other served as unstressed control . 24 hr after the end of chronic stress there was no difference in CS-induced freezing behavior between the two groups ( Day 11 , Block1 , Figure 1b ) . Thus , the recall of fear memory formed earlier was not affected by subsequent stress . This ensured that both stressed and unstressed animals were at the same levels of freezing when the extinction protocol was initiated after stress . Next , repeated tone presentations reduced freezing levels significantly such that both groups eventually underwent comparable extinction of fear , though the stressed rats were slower in achieving the same reduction in freezing ( Day 11 , Figure 1b ) . Notably , a day later stressed animals showed no difference in recall of extinction memory compared to unstressed animals ( Day 12 , Figure 1b ) . Thus , freezing levels during both fear and extinction recall were unaffected in stressed animals . This result differs from past reports of stress-induced deficits in fear extinction . However , as mentioned earlier , those earlier studies subjected animals to fear conditioning and extinction after exposure to stress ( Izquierdo et al . , 2006; Maren and Holmes , 2016; Miracle et al . , 2006 ) . Taken together , this suggests that the timing of stress may be a critical determinant of whether extinction recall is impaired . Thus , to examine this possibility we repeated the experimental design adopted in earlier studies by administering the same 10 day chronic immobilization stress protocol before ( Figure 1c , e ) the same fear conditioning paradigm depicted in Figure 1a . Consistent with earlier findings , prior exposure to chronic stress caused a significant impairment in the recall of fear extinction ( Figure 1e ) . However , these animals , unlike those used in Figure 1b ( i . e . conditioning before stress ) were not implanted with electrodes for simultaneous in vivo recordings . Hence , we repeated the behavioral experiments described in Figure 1b without surgical interventions related to in vivo recordings . This too yielded the same results as seen in the implanted animals – extinction recall was intact in stressed animals ( Figure 1d , f ) when they were subjected to conditioning before chronic stress . Next , we examined the neural basis of this result by recording local field potentials ( LFPs ) in these freely behaving rats ( Karalis et al . , 2016; Likhtik et al . , 2014 ) . While a role for potentiation of amygdalar neuronal responses to the tone CS in conditioned fear behavior is well established , in vivo recordings have also shown correlations of tone responses in the dorsal mPFC ( dmPFC ) with freezing behavior in fear conditioning and extinction . Taken together with earlier pharmacological inactivation studies , these findings identified an important role for the dmPFC in underlying conditioned fear responses and the expression of fear extinction ( Burgos-Robles et al . , 2009; Sierra-Mercado et al . , 2011 ) . Therefore , in addition to the BLA , we also monitored responses in the dmPFC ( Figure 2a , Figure 2—figure supplement 1 ) . We first analyzed CS-evoked LFPs in the BLA at three key behavioral time points described in Figure 1 – tone habituation before conditioning , fear recall and extinction recall ( Figure 2b–d ) . We measured auditory evoked potential ( AEP ) amplitudes as the difference between first peak and first trough in the AEP ( Figure 2—figure supplement 1 ) . During fear recall , AEP amplitudes ( Rogan et al . , 1997 ) were enhanced in both stressed and unstressed animals ( Figure 2c , e ) . However , while this increase was reversed in unstressed animals , it persisted in stressed rats even during extinction recall ( Figure 2c , e ) . Previous work also identified increase in CS-evoked theta power as a neural correlate of conditioned fear ( Likhtik et al . , 2014 ) . We found BLA theta power ( measured as the power of auditory evoked responses in the 2–12 Hz frequency band ) in unstressed animals also paralleled the increase , followed by decrease , in freezing during fear and extinction recall respectively ( Figure 2d–e ) . By contrast , BLA theta power remained high in stressed animals ( Figure 2d–e ) . Thus , despite stress-induced theta hyperactivity in the BLA , fear expression was not enhanced during extinction recall . To probe this further , we also analyzed the same LFP parameters in the dmPFC , which according to recent studies plays an important role in fear expression ( Karalis et al . , 2016; Likhtik et al . , 2014 ) . In the dmPFC of unstressed rats , AEP amplitude increased during fear recall , and this was reversed during extinction recall ( Figure 2f ) . Stressed animals , however , did not exhibit any change in dmPFC AEP amplitudes during either fear or extinction recall . Further , fear conditioning enhanced dmPFC theta power in both stress and unstressed animals ( Figure 2f ) . Interestingly , this was reversed in both groups during extinction recall . In other words , unlike the BLA , changes in theta power in the dmPFC , during fear and extinction recall , were not affected by stress . Moreover , in stressed animals , bidirectional modulation of theta power in the dmPFC , but not the BLA , accurately mirrored the changes in freezing , a behavioral expression of fear . Interestingly , the gradual decrease in dmPFC theta power paralleled the significant within-session reduction in freezing during the acquisition of extinction ( Day 11 , Figure 1b ) in both control and stressed animals ( Figure 2—figure supplement 2b , d ) . However , unlike the slower time course of extinction in the stressed rats , there was no difference in the time course of reduction in dmPFC theta power between stressed versus control animals . Similar analysis of BLA theta power during extinction learning ( Day 12 , Figure 1b ) did not reveal any significant differences that paralleled the gradual within-session decrease in freezing levels ( Figure 2—figure supplement 2b , c ) . Finally , there is growing appreciation of the importance of interactions between the mPFC and BLA , not just activity within these areas , in regulating fear behavior ( Karalis et al . , 2016; Lesting et al . , 2011; Likhtik et al . , 2014; Popa et al . , 2010 ) . This issue comes into sharp focus here because of the distinct effects of stress on the mPFC versus BLA . Thus , in light of recent reports that theta frequency oscillations synchronize dmPFC–BLA circuits during expression of fear behavior ( Karalis et al . , 2016; Likhtik et al . , 2014 ) , we investigated whether the tone-evoked increases in theta power ( Figure 2 ) were accompanied by enhanced theta-frequency synchrony between the two areas , and if this was in anyway affected by stress . Hence , we quantified CS-evoked coherence to assess moment-by-moment synchrony across LFPs recorded from the dmPFC and BLA for all three time points ( Figure 3a ) ( Likhtik et al . , 2014 ) . In unstressed rats , consistent with earlier reports , the CS elicited significantly higher theta coherence during fear recall ( Likhtik et al . , 2014 ) , and this increase persisted during extinction recall as well ( Figure 3b ) . Notably , in stressed animals , there was no change in BLA-dmPFC theta-frequency coherence ( Figure 3b ) . Thus , stress appears to completely suppress the dynamic , behaviorally relevant enhancement in BLA-dmPFC coherence that is seen during both fear and extinction recall in unstressed animals . In light of strong reciprocal connections between the mPFC and BLA , increases in theta synchrony have led earlier studies to analyze the direction of information flow between the two areas ( Karalis et al . , 2016; Likhtik et al . , 2014 ) . Hence , we estimated the directionality of functional connectivity and leads between the dmPFC and BLA using a previously validated method of calculating cross-correlations of instantaneous amplitude of filtered LFPs ( Adhikari et al . , 2010 ) . This reveals that theta activity in the dmPFC leads that in the BLA during recall of both fear and extinction memories in unstressed rats ( Figure 3c ) . However , this dmPFC-to-BLA directional influence is absent in stressed animals . Together , these data suggest that chronic stress causes a decoupling of activity between the two brain areas , as evidenced by a complete disruption of the increase in dmPFC-BLA theta synchrony and dmPFC-to-BLA directional influence normally seen during the recall of fear and extinction memories . This decoupling is also evident across all trial blocks during the acquisition of extinction the previous day ( Figure 3—figure supplement 1 ) . This breakdown in mPFC-BLA theta-frequency synchrony and directional coupling could be one reason why enhanced BLA theta activity was not manifested as higher freezing in stressed animals . The present study , specifically designed to administer chronic stress after the formation of fear memory , reveals that when stressed animals started extinguishing fear memories from the same level of freezing as their unstressed counterparts , their ability to recall extinction memory remained intact . This is in contrast to past findings wherein stressed animals exhibited a deficit in extinction recall when faced with the challenge of extinguishing fear memories strengthened by prior exposure to stress . This is consistent with earlier findings that repeated immobilization stress ( 4 h/day for 14 days ) administered after auditory fear conditioning did not affect the expression of previously acquired fear memories ( Meyer et al . , 2014 ) . However , the same immobilization stress , when administered before fear conditioning , elicited a robust increase in fear recall . Together , these results suggest that the timing of chronic stress exposure , with respect to fear conditioning , is a key determinant of how subsequent extinction of fear is affected by stress . This dissociation of the effects of stress on acquisition versus extinction of fear conditioning suggests that stress acts primarily on acquisition , and the earlier findings from pre-conditioning stress experiments were not the result of a deficit in fear extinction per se . Age may also influence how stress affects fear extinction in rats . For instance , deficient fear extinction was only observed in previously stressed adolescent , but not adult , rats ( Zhang and Rosenkranz , 2013 ) . Interestingly , despite no visible behavioral effect of stress on fear expression , our in vivo recordings reveal a robust impact of stress on amygdalar activity , as evidenced by enhanced theta activity in the BLA that failed to reverse even after the animals exhibited normal extinction recall ( Figure 2e ) . This is consistent with earlier findings on physiological and structural strengthening of excitatory synaptic connectivity , as well as reduced inhibitory tone , in the BLA after stress ( Suvrathan et al . , 2014 ) . Together , these changes point to possible synaptic mechanisms in the BLA for the stress-induced strengthening of subsequent encoding of fear memories . The same chronic stress , on the other hand , did not disrupt normal bidirectional modulation of mPFC theta activity , which in turn was reflected in normal freezing behavior during recall of fear and extinction ( Figure 2f ) . This is consistent with growing evidence for a pivotal role played by the mPFC in fear expression ( Dejean et al . , 2016; Sierra-Mercado et al . , 2011 ) . For instance , recent work has demonstrated strong correlations between mPFC theta-frequency oscillations and conditioning-induced freezing behavior ( Likhtik et al . , 2014 ) . Furthermore , we find normal mPFC theta activity to be decoupled from theta hyperactivity in the BLA , possibly reducing the latter’s influence on fear expression . This is similar to a report that even a single episode of stress can weaken functional connectivity between the two areas measured by resting state fMRI ( Liang et al . , 2014 ) . Indeed , such stress-induced disruptions in prefrontal-to-amygdala connectivity is also known to affect social interaction and anxiety-related behaviors in rodents ( Adhikari et al . , 2015; Hultman et al . , 2016 ) . Our findings , taken together with earlier behavioral studies ( Izquierdo et al . , 2006; Meyer et al . , 2014; Miracle et al . , 2006; Rau and Fanselow , 2009; Sierra-Mercado et al . , 2011; Suvrathan et al . , 2014 ) suggest a model for why the effects of stress on the recall and extinction of fear memories depend on the timing of stress exposure with respect to when the fear memory is formed and extinguished ( Figure 4 ) . Specifically , these findings suggest that the regulation of remote fear memories by stress ( i . e . pre-stress conditioning , Figure 4b ) is different from that of more recent fear memories ( i . e . post-stress conditioning , Figure 4c ) . As reported here , when fear memories were formed before exposure to the 10 day chronic stress , recall of fear extinction was not affected . Although stress caused a persistent increase in theta activity in the BLA , it did not affect bidirectional regulation of dmPFC theta activity ( Figure 4b ) . While stress-induced disruption in mPFC-BLA directional coupling and theta-frequency synchrony ( Figure 3 ) may explain why enhanced BLA theta activity did not result in stress-induced enhancement in freezing , another intriguing possibility arises from a recent report that neural circuits mediating the recall of fear memories shifts over time ( Do-Monte et al . , 2015 ) . Specifically , this study showed that optogenetic silencing of BLA reduced freezing when it was performed 6 hr , but not 7 days , after auditory fear conditioning . Since our post-conditioning stress protocol was carried out for 10 days , it may have outlasted BLA involvement in expression of fear and could also explain why enhanced BLA excitability was no longer reflected in freezing at that remote time point ( Day 11; Figure 4b ) . As a first step towards testing this possibility , we carried out in vivo infusions of muscimol directly into the BLA just before fear recall and extinction learning on Day 11 ( Figure 4—figure supplement 1 ) . Strikingly , this targeted inactivation of the BLA had no impact on fear expression at this time point in either stressed or control rats ( Figure 4—figure supplement 1b ) . Further , freezing levels in the muscimol-infused control and stressed rats did not differ from the freezing levels seen in stressed/control animals ( Figure 4—figure supplement 1c ) that did not receive any infusions in earlier experiments ( Figure 1f ) . Thus , we found that in vivo inactivation of the BLA no longer affects the recall of fear memories that were formed before the 10 day stress exposure . While more detailed analyses will be required to test all aspects of this model , these preliminary results suggest that fear memories formed 10 days ago no longer depend on BLA activity in either control or stressed animals . Consequently , the expression of fear after extinction reflects normal regulation of theta activity in the mPFC , not theta hyperactivity in the BLA ( Figure 4b ) . On the other hand , when fear memories were formed after exposure to the same chronic stress , the behavioural effects on freezing are very different , as has also been reported in earlier studies that employed a post-stress conditioning strategy ( Maren and Holmes , 2016; Miracle et al . , 2006; Rau and Fanselow , 2009 ) . Animals that were fear conditioned after stress show enhanced freezing during both acquisition and recall of extinction memory . What could be the neural basis for this difference ? Electrophysiological data from our post-conditioning stress experiments point to possible mechanisms that can be examined in future studies ( Figure 4c ) . For instance , in contrast to the pre-stress conditioning situation , stressed animals will recall and extinguish fear memories that were formed only 1 day ago . As a result , stress-induced BLA hyperactivity will be manifested as stronger fear memories that are resistant to subsequent extinction ( as evidenced by higher freezing during extinction recall in CIS rats , Figure 1e ) . Thus , now it is the sustained enhancement in BLA theta activity , not normal regulation of dmPFC activity , that is likely to drive higher freezing in stressed rats during extinction learning and extinction recall ( Figure 4c ) . Further , the detrimental effects of stress will be compounded by the disruption of mPFC-BLA theta-frequency coherence and directional coupling , which is likely to impair the mPFC’s ability to regulate BLA hyperactivity . Interestingly , pre-conditioning stress has been shown to cause subsequent enhancement in fear learning that can last up to 90 days later ( Rau and Fanselow , 2009 ) . This naturally raises questions about the involvement of the BLA in the enhanced expression of conditioned fear at remote time points . Our preliminary findings ( Figure 4—figure supplement 1 ) , taken together with the study by Do-Monte et al . ( 2015 ) , suggest that despite the effects of pre-conditioning stress on the encoding of fear memories , subsequent expression at later time points may no longer depend on BLA activity . This , in turn , raises the possibility that other brain areas that interact with the BLA in the regulation of fear , such as the dmPFC , may play a role in maintaining this expression at remote time points . Thus , a better understanding of the neural basis of earlier behavioral findings ( Izquierdo et al . , 2006; Meyer et al . , 2014; Miracle et al . , 2006; Noble et al . , 2017; Rau and Fanselow , 2009 ) will require detailed in vivo electrophysiological analyses , coupled with targeted inactivation , to examine the role of other brain areas that are engaged in stress-induced changes of fear expression . Finally , it is interesting to note that the chronic immobilization stress paradigm used in the present study was previously shown to strengthen functional connectivity from the amygdala to the hippocampus ( Ghosh et al . , 2013 ) , which undergoes stress-induced deficits similar to the mPFC ( Arnsten , 2015; Chattarji et al . , 2015; McEwen and Morrison , 2013 ) . In other words , although both the hippocampus and mPFC undergo similar forms of stress-induced deficits , the impact of stress on their individual interactions with the amygdala are strikingly different . A better understanding of these divergent features of aberrant interactions distributed across the amygdala-mPFC-hippocampal network , not just those confined within each area , may offer new insights into therapeutic interventions against the cognitive and emotional symptoms of stress-related psychiatric disorder .
Naïve 8–9 weeks old male Sprague-Dawley rats ( RRID: RGD_734476 ) weighing 300–350 grams at the start of the experiment ( National Centre for Biological Sciences , Bangalore , India ) and housed in groups of two were used in the study . They were maintained on a 14 hr/10 hr light/dark cycle and had access to water and a standard diet ad libitum . All experiments were conducted in accordance with the guidelines of the CPCSEA , Government of India and approved by the Institutional Animal Ethics Committee of National Centre for Biological Sciences . The experimental design comprised of experimental procedures conducted over a 4 week period . The animals were handled for 2–3 days to familiarize with the experimenter . This was followed by a surgery to implant bundle of electrodes in the basolateral amygdala and the dorsal medial prefrontal cortex of the animals . The animals were allowed to recover for 6–7 days after surgery . Next , the animals were subjected to a 15 day behavioral paradigm with simultaneous recording of local field potentials ( LFPs ) during behavior ( Figure 1a ) . The animals were initially habituated to the conditioning context on Day −1 and Day 0 . Next on Day 1 the animals were subjected to the tone habituation and fear conditioning protocol . Then on Day 2 the animals were randomly allotted to the chronic immobilization stress ( CIS ) or control groups . The animals in the CIS group were subjected to a 10 day chronic immobilization stress ( CIS; 2 h/day for 10 consecutive days ) from Day 2 to day 11 , whereas the animals in the control group were just handled once a day during the same period . Subsequently , on Day 12 , that is 24 hr after the end of CIS , the animals were subjected to fear recall and extinction training . On Day 13 , the animals were subjected to fear extinction recall session . After the end of the behavioral paradigm the animals were sacrificed and brains were collected for histological examination . For the behavior experiments without LFP recordings , the animals were handled for 2–3 days and then subjected to the same behavior protocol described above . For assessing fear memory formed prior to stress , the animals were subjected to tone habituation and conditioning on day 1 followed by the 10 day CIS paradigm . Subsequently , the animals were subjected to fear recall and extinction on Day 12 and extinction recall on Day 13 . For assessing fear memory formed after stress , the animals were first subjected to the CIS paradigm followed by tone habituation and conditioning on Day 11 . Subsequently , the animals were subjected to fear recall and extinction training on Day 12 and extinction recall on Day 13 . The animals were randomly allocated to either CIS or control groups . The animals were pair-housed and stressed and unstressed animals were separately housed . For the behavior experiments that did not involve BLA inactivation prior to fear recall ( see details below ) , the animals were handled for 2–3 days followed by surgery to implant stainless steel cannulae ( 24 gauge , Plastic One , Roanoke , Virginia , USA ) targeted at the basolateral amygdala ( BLA ) . Following the surgery , animals were single housed and allowed to recover for 6–7 days following the surgery . Next , the animals were subjected to tone habituation followed by fear conditioning on Day 1 . The animals were then randomly split into CIS and control groups . CIS rats were subjected to the same 10 day chronic immobilization stress paradigm . Subsequently , on Day 12 all the animals received intracranial infusion of muscimol into the BLA . 30 min after the muscimol infusion the animals were subjected to fear recall and extinction session . Finally , the animals underwent extinction recall on Day 13 . For recording LFPs from the dmPFC and BLA , rats were surgically implanted with formavar insulated nichrome wire ( 25 microns diameter; AM Systems , Carlsborg , WA , USA ) bundles unilaterally in the BLA and dmPFC . And for the BLA inactivation experiments the animals were implanted with stainless steel cannulae ( 24 gauge , Plastic One , Roanoke , Virginia , USA ) targeting the BLA bilaterally . Rats were induced into anesthesia with 5% isoflurane ( Forane , Asecia Queensborough , UK ) and then maintained in anesthesia with 1 . 5–2% isoflurane . The level of anesthesia was regularly monitored throughout the procedure using the pedal withdrawal reflex to toe pinch . The animal was placed and head fixed on a stereotaxic frame . Body temperature of rats was maintained with a heating pad . For implanting electrode bundles , burr holes were drilled at the stereotactic coordinates of the BLA ( stereotaxic coordinates were: 3 . 0 mm posterior to bregma and ±5 . 3 mm lateral to midline ( Paxinos and Watson , 2009 ) and the dorsomedial prefrontal cortex ( dmPFC; stereotaxic coordinates were: 3 . 0 mm anterior to bregma and ±0 . 5 mm lateral to midline , Paxinos and Watson , 2009 ) . A bundle of 8 formavar coated nichrome electrodes were then implanted using the stereotactic frame ( 8 . 3 mm and 3 . 4 mm ventral from the brain surface for BLA and dmPFC respectively ) . Head screws were implanted to anchor the implant . One head screw placed just behind the lambda on the skull was used as the ground electrode . For implanting cannulae , burr holes were drilled at the stereotactic coordinated of the BLA bilaterally ( stereotaxic coordinates were: 3 . 0 mm posterior to bregma and ±5 . 2 mm lateral to midline ( Paxinos and Watson , 2009 ) . Stainless steel guide cannulae ( 24 gauge , Plastic One , Roanoke , Virginia , USA ) were then implanted using the stereotactic frame ( 7 mm ventral from the brain surface ) . Dummy cannulae ( 28 gauge , Plastic One , Roanoke , Virginia , USA ) with 0 . 5 mm projection were inserted into the cannulae to prevent clogging . The implant was secured using anchor screws and dental acrylic cement . Rats were allowed to recover for 6–7 days following surgery . In the post-surgery period the animals were singly housed in separate cages . A total of 18 animals were implanted with electrodes . Three animals were excluded from the study because the positioning of the electrode bundles was incorrect . The location of the cannulae placement for the 15 animals used in the LFP experiments is shown in Figure 2a . A total of 16 animals were implanted with cannulae and one was excluded from the study due to an error with muscimol infusion . Rats in the CIS group were subjected to a chronic immobilization stress ( CIS ) paradigm ( Ghosh et al . , 2013 ) , consisting of complete immobilization for 2 hr per day ( before noon ) in rodent immobilization bags without access to either food or water , for 10 consecutive days . Fear conditioning and extinction took place in different contexts placed inside sound-isolation boxes ( Coulbourn Instruments , Whitehall , Pennsylvania , USA ) . Conditioning was performed in a box with metal grids on the floor ( context A: 12 inches wide ×10 inches deep ×12 inches high , no odour ) . Fear extinction training and extinction recall was performed in another context , a modified homecage ( context B: 14 inches wide ×8 inches deep ×16 inches high , mint odour ) . Lighting conditions and walls were different between the two contexts . All chambers were cleaned with 70% alcohol before and after each experiment . The behavior of the animals was recorded using a video camera mounted on the wall of the sound isolation box and a frame grabber ( sampled at 30 Hz ) . The videos were analyzed offline for further quantification of freezing behavior . Infrared LED cues were placed on the walls of the experimental chambers . These cues were activated in coincidence with auditory stimuli to monitor the tone-evoked freezing response offline . A programmable tone generator and shocker ( Habitest system , Coulbourn Instruments , Whitehall , Pennsylvania , USA ) were used to deliver tones and foot-shocks during the experiment . Foot-shocks were delivered through the metal grids on the floor of the conditioning chamber . The tone was played using a speaker ( 4 Ω , Coulbourn Instruments , Whitehall , Pennsylvania , USA ) placed inside the experimental chamber . During context habituation , the animals were allowed to explore context A for 25 min in each session . Next , in the tone habituation session ( Figure 1a ) the animals received five presentations of an auditory tone ( total duration of 30 s , 5 kHz auditory tone consisting of 30 pips of 100 millisecond duration at a frequency of 1 Hz; 5 millisecond rise and fall , 70 ± 5 dB sound pressure level ) in context A . This was immediately followed by fear conditioning protocol , where the tone ( CS ) was paired ( pairings , average inter-trial interval <ITI > =120 s , with a range of 80–160 s ) with a co-terminating 0 . 5 s scrambled foot shock ( US; 0 . 7 mA ) . In the fear recall and extinction session the animals were presented with the same tone ( CS ) for 15 times ( average inter-trial interval <ITI > =120 s , with a range of 80–160 s ) in the context B . Again , in the extinction recall session , the animals were subjected to the same CS 15 times again . Behavioral response was scored offline using video recordings of all the behavior sessions . Response to the auditory stimuli was evaluated in the form of freezing response . Freezing was defined as the absence of movement except due to respiration ( Blanchard and Blanchard , 1988 ) . The time spent freezing during the presentation of the tone was converted into a percentage score ( Figure 2c ) . The percentage freezing level was measured in every context/session for 30 s immediately before the presentation of the first tone trial to assess freezing in absence of an auditory stimulus . This was defined as the freezing in the pretone period ( Figure 2d ) . The pretone block represents freezing during the first pretone only . The tone habituation block represents freezing during the last two trials of tone habituation . The first and last trial blocks during conditioning represent the freezing during the first two and last two trials of the fear conditioning session . The trial blocks in the fear recall and extinction session and the extinction recall session represent freezing over blocks of two trials each ( 1 to 14 ) . All the animals were subjected to the recording of the local field potentials ( LFPs ) during tone habituation session , fear recall and extinction session and extinction recall session . Auditory-evoked potentials ( AEPs ) were recorded by connecting the microelectrodes to a unit gain buffer head stage ( HS-36-Flex; Neuralynx , Bozeman , Montana , USA ) and a data acquisition system Digilynx ( Neuralynx , Bozeman , Montana , USA ) . Neural data were amplified ( 1000 times ) and acquired at a sampling rate of 1 kHz followed by a band-pass filter ( 1–500 Hz ) using Cheetah data acquisition software ( Neuralynx , Bozeman , Montana , USA ) . Intra-amygdalar infusion of muscimol was performed 30 min prior to fear recall and extinction session on Day 12 . The infusion was performed using standard pressure injection methods . Infusion procedure was performed in the homecage . Injection cannulae with 1 mm projection ( 28 gauge , Plastic One , Roanoke , Virginia , USA ) were inserted through the guide-cannulae . The injection cannula was connected to a Hamilton syringe ( 10 μl ) using a polyethylene tubing ( Plastic One , Roanoke , Virginia , USA ) , which was mounted on an infusion pump ( Harvard Apparatus , Holliston , Massachusetts , USA ) . Rats were infused bilaterally with one hemisphere at a time . Muscimol ( 0 . 5 μl per side , 1 µg µl−1 in saline; Tocris Biosciences , Bristol , UK ) was infused at a rate of 0 . 1 μl min−1 . The injection cannula was taken out 5 min after the end of infusion , to allow the drug to diffuse into the tissue . After the completion of experiments , cannulae placement was confirmed using standard histological methods . | Patients with stress-related psychiatric disorders experience debilitating emotional symptoms , including excessive fear that they are unable to control . Decades of research have shown that such disorders have opposite effects on two key structures in the brain . Normally , a region called the amygdala helps to form fear-related memories , while the prefrontal cortex helps control these memories and eliminate them through a process called extinction . But brain imaging on patients with stress-related psychiatric disorders reveals a smaller prefrontal cortex , and an overactive amygdala . Animal studies confirm that chronic stress shrinks brain cells in the prefrontal cortex , but grows them bigger in the amygdala . These and other studies have led scientists to believe that stress causes people to both form stronger fear memories and then have difficulties getting rid of such memories . These include studies in which stressed animals were trained to fear a sound , and then followed while they learned to overcome that fear . One problem with many of these studies is that the animals were repeatedly stressed before the fear memory was formed . This made it hard to tease apart whether the strength of the fear memory itself or a problem extinguishing the fear memory were to blame for the animals’ difficulties overcoming their fear . Now , Rahman et al . address this problem by exploring how the timing of chronic stress affects how well an animal can overcome fear memories . In the experiments , some rats learned to fear a sound before exposure to chronic stress , while a second group was exposed to chronic stress first then learned to fear the sound . When the animals were stressed after they learned to fear the sound , they could still eliminate their fear . But the animals stressed before exposure to the fear-inducing sound struggled to extinguish the fear . Recordings of the brain activity in the rats that were exposed to stress after learning to fear the sound showed the amygdala remained overly active even after these animals overcame their fear . However , the stress did not seem to disrupt the normal activity of the prefrontal cortex in these rats . This shows that memories formed before stress reflect normal activity in the prefrontal cortex and not the abnormally high activity in the amygdala . Exposure therapy helps people overcome stress-disorder related fears . For the therapy to work , the prefrontal cortex must be able to extinguish fear . Rahman et al . show that this is possible when the fear-enhancing effects of prior stress are not in play . More studies exploring why prior stress makes fears stronger and harder to overcome may help scientists develop ways to make therapies for stress disorders more effective . | [
"Abstract",
"Introduction",
"Results",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2018 | Extinction recall of fear memories formed before stress is not affected despite higher theta activity in the amygdala |
Insects find food and mates by navigating odorant plumes that can be highly intermittent , with intensities and durations that vary rapidly over orders of magnitude . Much is known about olfactory responses to pulses and steps , but it remains unclear how olfactory receptor neurons ( ORNs ) detect the intensity and timing of natural stimuli , where the absence of scale in the signal makes detection a formidable olfactory task . By stimulating Drosophila ORNs in vivo with naturalistic and Gaussian stimuli , we show that ORNs adapt to stimulus mean and variance , and that adaptation and saturation contribute to naturalistic sensing . Mean-dependent gain control followed the Weber-Fechner relation and occurred primarily at odor transduction , while variance-dependent gain control occurred at both transduction and spiking . Transduction and spike generation possessed complementary kinetic properties , that together preserved the timing of odorant encounters in ORN spiking , regardless of intensity . Such scale-invariance could be critical during odor plume navigation .
Insects navigate odor landscapes that are often not smooth gradients ( Cardé and Willis , 2008; Riffell et al . , 2008 ) . Instead , turbulent airflows shape odor plumes into intermittent whiffs separated by stochastic durations of background air ( blanks ) . In the absence of reliable spatial gradients , navigating insects may combine the timing of whiff encounters ( Vergassola et al . , 2007 ) with sensation of wind direction ( Budick and Dickinson , 2006; Cardé and Willis , 2008; Duistermars et al . , 2009 ) to navigate odor plumes towards mates and food . Insect olfactory systems face dual challenges in detecting natural odor plumes . First , the intensity of whiffs is typically distributed according to a power law ( Murlis et al . , 1992 ) , with intense whiffs interleaved unpredictably with weak ones ( Riffell et al . , 2008 ) . Second , whiff durations and blank durations are also distributed as a power law over a wide range of time scales ( Celani et al . , 2014 ) . The encoding problem is aggravated by shifting local statistics of odor encounters , which change with wind speed ( Nagel and Wilson , 2016 , 2011 ) , position ( Justus et al . , 2002 ) , or environment ( Murlis et al . , 2000 ) . How does the olfactory system manage to encode whiffs of odors whose intensities and timing can vary over such wide ranges ? Several features of the olfactory system contribute to encoding odor stimuli of different intensities . A single odorant can be detected by multiple receptor types , with different sensitivities ( Hallem and Carlson , 2006 ) . Static compressive nonlinearities at both olfactory receptor neurons ( ORNs ) and their post-synaptic targets , the projection neurons ( PNs ) , selectively amplify weak signals and suppress responses to large signals ( Bhandawat et al . , 2007; de Bruyne et al . , 2001 ) . Glomerular mechanisms implement a type of divisive gain control that maintains PN sensitivity within the range of changing ORN responses ( Bhandawat et al . , 2007; Luo et al . , 2010; Olsen et al . , 2010; Olsen and Wilson , 2008 ) . Finally , transduction currents in response to odor pulses scale inversely with the intensity of the background signal , consistent with the Weber-Fechner Law ( Cafaro , 2016; Cao et al . , 2016 ) . However , whether ORN firing follows a similar scaling is unclear ( Cafaro , 2016; Martelli et al . , 2013 ) . Thus , although it is known that the input-output curve of ORNs to odor stimuli changes with odor background , how ORN gain ( from stimulus to firing rate ) scales with background signal intensity has not been characterized . Olfactory responses in insects can be fast . Transduction can be initiated within milliseconds of odor reaching the antenna ( Szyszka et al . , 2014 ) . The speed of the response is enhanced by ORN spike generation , which emphasizes changes in transduction currents ( Nagel and Wilson , 2011 ) , and by PNs ( Kim et al . , 2015 ) , which maintain fast information transmission from ORNs to PNs ( Jeanne and Wilson , 2015; Nagel et al . , 2015; Raccuglia et al . , 2016 ) . In contrast , adaptation to high intensity stimuli slows down transduction ( Cao et al . , 2016; Kaissling et al . , 1987; Nagel and Wilson , 2011 ) , a property that might make it difficult to reliably encode the timing of odor encounters . We investigated in vivo how Drosophila ORNs encode encounters with naturalistic odor plumes . To address this question , we first developed an odorant delivery system that reproducibly delivered odorants with naturalistic or Gaussian statistics with controlled means and variances . We simultaneously recorded the odorant stimulus using a fast photo-ionization detector ( PID ) , and recorded extracellularly from identified ORNs . We found that ORNs encoded broadly distributed naturalistic signals by using two mechansism: front-end nonlinearities that are inherent in receptor binding to ligand , as well as two adaptation mechanisms that are sensitive to the mean and variance of the stimulus . These adaptive mechanisms caused ORNs to rapidly desensitize following encounters with odorant whiffs , dynamically adjusting gain while responding to intermittent odorant stimuli . ORNs adapted to changes in the mean stimulus at the level of transduction by scaling gain inversely with the stimulus intensity , consistent with the Weber-Fechner Law . Variance-dependent gain control took place at both signal transduction and spiking machinery . While the transduction response time slowed down with increasing stimulus intensity , the spiking machinery sped up to compensate . These complementary kinetic changes caused the firing rate response time to remain invariant with stimulus intensity . This reveals a mechanism that could allow ORNs to preserve information about the precise timing of odor encounters over a wide range of rapidly changing signal intensities . A minimal two-state model of the activity of the Or-Orco complex ( olfactory receptor and co-receptor Orco ) with an adaptation mechanism that feeds back onto the free energy difference between active and inactive conformations reproduced Weber-Fechner scaling , slowdown of signal transduction kinetics , and responses to intermittent and Gaussian stimuli .
Odorant signals used to study ORN adaptation typically consist of long pulses or constant backgrounds of various intensities ( Cafaro , 2016; Cao et al . , 2016; Martelli et al . , 2013; Nagel and Wilson , 2011 ) . However , airborne stimuli encountered by flying insects can be intermittent with both the intensities of encounters and durations between encounters broadly distributed as power laws ( Celani et al . , 2014 ) . Since ORN transduction can be adapted by odorant pulses as brief as 35 ms on timescales as fast as 500 ms ( Cao et al . , 2016 ) , we asked to what extent the gain of ORNs could change dynamically during responses to naturalistic stimuli , amplifying responses to isolated whiffs of odorant , and suppressing responses to whiffs following dense clumps of whiffs . We measured the responses of ab3A and ab2A ORNs to naturalistic stimuli of ethyl acetate and 2-butanone . These odorants elicit spikes in these neurons ( Hallem and Carlson , 2006 ) , and are easy to control and measure ( Martelli et al . , 2013 ) ( Figure 1—figure supplement 1 ) . We used in vivo extracellular recording to record both the local field potential ( LFP ) and spikes from a single sensillum , with simultaneous measurement of the stimulus ( Figure 1—figure supplement 2 ) . Previous results have shown that: LFP responses are unaffected by the addition of TTX , which eliminates neural spiking , suggesting that LFP signals were generated upstream of the spiking machinery; and that LFP signals are unaffected when the neuron’s partner cell in the sensillum is genetically ablated , when that partner does not sense the odorant , suggesting that LFP signals are generated by the neuron of interest ( Nagel and Wilson , 2011 ) . Though the LFP could reflect activity of nearby sensilla , it serves as an imperfect but useful proxy for transduction activity in ORNs ( Johnston et al . , 1995; Kaissling , 1986; Nagel and Wilson , 2011; Su et al . , 2012 ) . The naturalistic stimulus we used was intermittent and consisted of brief odor whiffs of varied amplitude ( Figure 1a–b ) . Durations of whiffs and blanks were broadly distributed , with a power law of exponent −3/2 to match natural intermittent statistics of odor plumes ( Celani et al . , 2014 ) ( Figure 1—figure supplement 3 ) . ab2A and ab3A ORNs responded to whiffs with transient decreases in the local field potential ( LFP ) and corresponding increases in the firing rate ( Figure 1a–b ) . 10 . 7554/eLife . 27670 . 003Figure 1 . Adaptation and saturation modulate ORN responses to broadly distributed naturalistic stimuli . ( a ) Ethyl acetate odorant ( top ) elicits LFP ( middle ) and firing rate ( bottom ) responses from a ab3A ORN . ( b ) 2-butanone odorant ( top ) elicits LFP ( middle ) and firing rate ( bottom ) responses from a ab2A ORN . Insets in ( a–b ) show pairs of whiffs and the LFP and firing rate responses they elicit on an expanded timescale . All pairs of insets are shown at the same scale , for 400 ms around a whiff . ( c ) ab2 LFP responses vs . projected stimulus . ( d ) ab2A firing rate vs . projected stimulus . ( c ) and ( d ) show that ORN responses differ significantly from linearity . ( e ) ab2 LFP responses vs . whiff amplitude . ( f ) ab2A firing rate vs . whiff amplitude . n = 15 trials from 2 ORNs . 101 whiffs shown in ( e–f ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 00310 . 7554/eLife . 27670 . 004Figure 1—figure supplement 1 . Diagram of odor delivery device and calibration of Photo-Ionization Detector ( PID ) . 1 mL of pure odorant was placed in a 20 mL scintillation vial with a screw top . A computer-controlled Mass Flow Controller ( MFC ) forced air through this vial , which created an odorized airstream . This airstream was either directed into the main air flow or to waste ( vacuum ) using a solenoid valve . A PID ( inlet needle at the outlet of the main air tube ) recorded the gas phase concentration of the odorant stimulus as it was presented to the fly . We calibrated the PID by depleting a fixed , known volume of pure odorant at various flow rates , and integrating the resultant PID signal . Using the known densities and molar masses of these monomolecular odorants , we built maps from PID response in Volts onto the absolute odorant flux . This relationship was found to be linear for the two odorants tested . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 00410 . 7554/eLife . 27670 . 005Figure 1—figure supplement 2 . Example of simultaneously acquired primary data ( ab3A responses to ethyl acetate stimulus ) . ( a ) Seven repetitions of fluctuating ethyl acetate stimulus ( each gray trace is from a presentation to a different ab3A neuron in a different sensillum; mean shown in black ) . ( b ) Raw voltage recording from 7 different ab3 sensilla . ( gray traces , one from each neuron; mean shown in black ) . ( c ) One of the traces in ( b ) is filtered to visualize spikes . Note that spikes from the ab3A neuron are typically larger than spikes from the ab3B neuron , enabling us to sort them and measure the response of a single neuron in vivo . ( d ) Raster of ab3A spikes for 7 different ORNs . Arrow indicates trace shown in ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 00510 . 7554/eLife . 27670 . 006Figure 1—figure supplement 3 . Statistics of the ethyl acetate stimulus with naturalistic temporal structure . ( a ) Distribution of whiff intensities . ( b ) Distribution of whiff durations . ( c ) Distribution of blank durations . Predicted distributions from Celani et al . ( 2014 ) are shown in red lines ( a–c ) . c is the odor concentration ( whiff intensity ) . tw and tb are whiff and blank durations . ( d ) Mean vs . standard deviation of stimulus , computed in 400 ms non-overlapping blocks . ( e ) Correlation between mean and standard deviation of stimulus as a function of window length . Peak correlation observed for timescales ~400 ms . ( f ) Autocorrelation function of the stimulus . Shading indicates standard deviation across trials . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 00610 . 7554/eLife . 27670 . 007Figure 1—figure supplement 4 . Deviations from linearity persist even when filters extracted from Gaussian stimuli are used to project naturalistic stimulus . ( a ) LFP filters for ab2A ORNs responding to 2-butanone , extracted either form naturalistic stimuli ( black ) or from Gaussian stimuli ( red ) . ( b ) LFP responses to naturalistic stimulus vs . stimulus projected through filter computed from naturalistic stimulus ( Black filter in a ) . ( c ) LFP responses to naturalistic stimulus vs . stimulus projected through filter computed from Gaussian stimulus ( red filter in a ) . ( d ) Firing rate filters for ab2A ORNs responding to 2-butanone , extracted either from naturalistic stimuli ( black ) or from Gaussian stimuli ( red ) . ( e ) Firing rate responses vs . stimulus projected through filter computed from naturalistic stimulus ( black filter in d ) . ( f ) Firing rate responses vs . stimulus projected through filter computed from Gaussian stimulus ( red filter in d ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 007 Even though individual whiff intensities were broadly distributed ( First line in Figure 1a ) ORN responses to these whiffs were more even , so that responses to faint whiffs were amplified more than those to intense whiffs . To quantify these differences , we defined the gain of the neuron to be the change in the response for a unit change in the stimulus ( gain:=ΔR/ΔS ) . Since ORNs do not respond instantaneously to odorant stimuli ( Cao et al . , 2016; de Bruyne et al . , 2001; Martelli et al . , 2013; Nagel and Wilson , 2011 ) we fit linear filters to best predict the LFP and firing rate from the stimulus . We used these filters to make linear predictions of the responses from the stimuli ( Figure 1—figure supplement 4 ) . Changes in gain were therefore defined as deviations from the linear prediction of response from the stimulus , similar to ( Baccus and Meister , 2002; Kim and Rieke , 2001 ) . We visualized these gain changes by plotting the LFP responses against linear prediction of the LFP ( Figure 1c ) and the firing rate against the linear prediction of the firing rate ( Figure 1d ) . Each excursion in these plots corresponds to the ORN’s response to a single whiff . Excursions occurred with different slopes , suggesting that ORN gain changed frequently in time . Deviations from linearity persisted even when filters computed from Gaussian inputs were used to project the stimulus , suggesting that the existence of these deviations do not depend on the exact shape of the filter , but rather reflect a property of the ORN response not captured by the linear model ( Figure 1—figure supplement 4 ) . Variations in the gain ( ΔR/ΔS ) clearly do not arise solely from a static output nonlinearity , such as one associated with a linear-nonlinear transformation , since plotting neuron response against the projected stimulus ( Figure 1c–d ) does not yield a single transformative function . ( Dayan and Abbott , 2001 ) . We reasoned that changes in the gain could arise from input nonlinearities due to odor-receptor binding and channel opening . To visualize the nonlinearity between the stimulus and response , we plotted LFP and firing rate responses to each whiff in the naturalistic stimulus as a function of the amplitude of that whiff ( Figure 1e–f ) . A clear sigmoidal dependency is visible in the plot of LFP responses against whiff intensity , consistent with a front-end nonlinearity arising from receptor-odorant binding . However , in both the LFP and firing rates , responses to whiffs with similar intensities varied significantly , deviating from a single sigmoidal dose-response curve ( Figure 1e–f ) . What causes these deviations from the dose-response curve ? One possibility is that these deviations are due to random variability in the responses of the neuron . Another possibility is that these deviations are due to adaptation of the neuron to the stimulus history preceding each whiff , which may vary with every whiff . To distinguish between these possibilities , we collected whiffs that had similar amplitudes , and examined the LFP and firing rate responses they evoked ( Figure 2a–b ) . The amplitude of LFP and firing rate responses elicited by these whiffs varied inversely with the amplitude of the preceding stimulus: whiffs that occurred in isolation ( purple ) elicited the largest responses , while whiffs that followed earlier , large whiffs ( blue , red ) elicited the smallest responses , suggesting that ORN responses can be modulated by stimulus history . 10 . 7554/eLife . 27670 . 008Figure 2 . Adaptation and saturation modulate ORN responses to broadly distributed naturalistic stimuli . ( a ) . Ethyl acetate whiffs of similar size ( top ) elicit ab3 LFP responses ( middle ) and ab3A firing rate responses ( bottom ) with different amplitudes . ( b ) 2-butanone whiffs of similar size ( top ) elicit ab2 LFP responses ( middle ) and ab2A firing rate responses ( bottom ) with different amplitudes . Bar graphs in ( a ) and ( b ) show that ordering in LFP and firing rate response does not correlate with whiff amplitude , but correlates with the intensity of the preceding whiff . Colors on bar graph correspond to colors in time series on the left . Deviations in LFP ( c ) and firing rate responses ( d ) from the median response vs . mean stimulus in the preceding 300 ms . Deviations in LFP ( c , inset ) and firing rate responses ( d , inset ) from the median response vs . whiff amplitude . ( e ) Deviations from the median of LFP responses ( positive deviations: red , negative deviations: blue ) as a function of the amplitude of the previous whiff and the time since previous whiff . Positive and negative deviations are significantly different ( p=0 . 01 , 2-dimensional K-S test ) . ( f ) Deviations from the median of firing rate responses ( positive deviations: red , negative deviations: blue ) as a function of the amplitude of the previous whiff and the time since previous whiff . Positive and negative deviations are significantly different , ( p=0 . 001 , 2-dimensional K-S test on firing rate deviations ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 00810 . 7554/eLife . 27670 . 009Figure 2—figure supplement 1 . An NL model ( static input nonlinearity followed by a linear filter ) cannot reproduce context-dependence of LFP responses to similar-sized whiffs . Input nonlinearity ( a ) and filter ( b ) fit to ab2 LFP responses to 2-butanone naturalistic stimulus . The input nonlinearity is a Hill function S/ ( S+KD ) where S represents the input , and KD the half maximum value ) . The nonparametric filter and parametric nonlinearity are fit simultaneously in an iterative manner ( see Materials and methods ) . ( c ) Comparison of ab2 LFP responses and NL model predictions . ( d ) Linear filter extracted from the stimulus and the NL model prediction . Note that the filter is not the same as in ( b ) ; a filter extracted from an NL model is not guaranteed to be an unbiased estimate of the true one . ( e ) NL model responses vs . naturalistic stimulus projected through filter in ( d ) , showing that the NL model shows deviations from linearity similar to what is observed in the data ( cf . Figure 1c ) . ( f–g ) Context dependence of response in the ab2 data and model . ( f ) ab2 LFP responses to whiffs of similar size ( same data as in Figure 1h ) . Note that the responses to isolated whiffs ( purple , yellow ) are larger than the responses to repeated whiffs ( red , blue ) . ( g ) NL model responses to these whiffs . Note that the responses to isolated whiffs ( purple , yellow ) are smaller than the responses to repeated whiffs ( red , blue ) , the opposite of the trend visible in the data . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 009 To quantify this context-dependent modulation , we estimated deviations of the LFP and firing rate response to each whiff from the median response ( see Materials and methods ) . Deviations in LFP response to each whiff decreased with mean stimulus in the preceding 300 ms ( Figure 2c , ρ=−0 . 39 , p=0 . 01 , Spearman test ) , and were uncorrelated with the amplitude of the whiff that elicited them ( Figure 2c , inset , p=0 . 9 , Spearman test ) . Similarly , deviations in the firing rate responses to each whiff decreased with mean stimulus in the preceding 300 ms ( Figure 2d . ρ=−0 . 68 , p<10−5 , Spearman test ) , and were uncorrelated with the amplitude of the whiff that elicited them ( Figure 2d , inset , p=0 . 37 , Spearman test ) . To generalize beyond a particular timescale of the stimulus history , we parametrized the stimulus history of each whiff by the amplitude and time since the preceding whiff , and grouped estimated deviations from the median response into positive or negative ( Figure 2e–f ) . When response deviations were negative ( smaller than median responses , blue dots ) , the amplitude of the preceding whiffs tended to be larger , and the time since the previous whiff tended to be shorter . When response deviations were positive ( red dots ) , the amplitude of preceding whiffs tended to be smaller ( <Sprevious> is 0 . 52 V when deviations are positive vs . 1 . 09 when deviations are negative for firing rate responses and 1 . 04 V vs . 1 . 06 V for LFP responses ) , and the time since the previous whiff tended to be longer ( <tprevious> is 3025 ms vs . 361 ms for firing rate responses , and 2556 ms vs . 566 ms for LFP responses ) . What causes this context-dependent suppression of responses following preceding whiffs ? One possibility is a bi-lobed filter , with one positive and one negative lobe , followed by a rectifying nonlinearity . Such a filter is partly differentiating , and has been measured in linear models of the firing rate ( Kim et al . , 2011; Martelli et al . , 2013; Nagel and Wilson , 2011 ) and would lead to attenuated responses to the second of two closely spaced whiffs due to linear superposition . Such a mechanism may partly account for context dependent variation in firing rates . However , stimulus-to-LFP filters , computed for this stimulus and others , are mono-lobed ( Figure 2—figure supplement 1 ) , and appear purely integrating ( Nagel and Wilson , 2011 ) , ruling out contributions to dynamic modulation of LFP responses by this mechanism . A model with a static front-end nonlinearity and a mono-lobed filter fit to the LFP also cannot reproduce context-dependent adaptation observed in the LFP ( Figure 2—figure supplement 1 ) , suggesting that this context-dependent variation in response arises at least in part from ORNs dynamically varying gain in response to naturalistic stimuli . Since the mean and variance of naturalistic stimuli are correlated over many timescales , ( Figure 1—figure supplement 3 ) , it is unclear whether adaptation in this context is sensitive to the mean or the variance ( or to some other statistic ) of preceding whiffs . To determine how changing one moment of the stimulus distribution changed ORN gain , and to disambiguate the effect of receptor saturation from adaptation , we proceeded to other experiments using Gaussian stimuli with changing means ( Figure 3 ) and variances ( Figure 4 ) . 10 . 7554/eLife . 27670 . 010Figure 3 . ORNs decrease gain with stimulus mean , consistent with the Weber-Fechner Law . ( a ) Ethyl acetate stimuli with different mean intensities but similar variances . Stimulus intensity measured using a Photo-Ionization Detector ( PID ) , units in Volts ( V ) . Colors indicate mean stimulus intensity . ( b ) Corresponding stimulus distributions . ( c ) ab3A firing rate responses to these stimuli . ( d ) Corresponding response distributions . ( e ) ORN responses vs . stimulus projected through linear filters . Colored numbers indicate r2 between linear projections and ORN response . ( f ) ORN gain vs . mean stimulus for each trial . Red line is the Weber-Fechner prediction ( ΔR/ΔS∼1/S ) ( g ) After rescaling the projected stimulus by the gain predicted by the red curve in ( f ) , and correcting for an offset , ORN responses collapse onto one line . n = 55 trials from 7 ORNs in 3 flies . All plots except ( f ) show means across all trials . ( f ) shows individual trials . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01010 . 7554/eLife . 27670 . 011Figure 3—figure supplement 1 . Weber-Fechner Law broadly observed across odor-receptor combinations . ( a ) Standard deviation vs . mean of ethyl acetate stimulus in Figure 1 . ( b ) ORN gain estimated by the ratio of standard deviation of firing rate to standard deviation of stimulus , vs . mean stimulus in each trial . This model-free estimate of ORN gain ignores kinetics of response , but returns similar estimates of the gain ( cf . Figure 3f ) . Note that the units of gain estimated this way are the same . ( c–f ) ORN gain as a function of mean stimulus for various odor-receptor combinations . In all plots , the red line is a power law with slope −1 ( the Weber-Fechner Law ) . Data in panel a and b is the same as in Figure 3 . n = 121 trials from 16 ORNs in 6 flies . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01110 . 7554/eLife . 27670 . 012Figure 3—figure supplement 2 . Ability of NL models to reproduce observed change in input-output curves . ( a–c ) Static NL model responses . ( a ) The input nonlinearity of NL model is chosen to be a Hill function with n = 1 . ( b ) Filter of NL model , measured directly from the data . ( c ) NL model responses vs . projected stimulus . While these curves appear to change slope with increasing mean stimulus , mean responses also tend to increase ( purple … yellow ) . ( d–f ) Varying NL model responses , where the KD of the input nonlinearity is allowed to vary with the mean stimulus . ( d ) Input nonlinearities for stimuli with different mean ( colors ) . The KD of each curve is set to the mean stimulus of that trial . ( e ) Filter of NL model , same as in ( b ) . ( f ) Model responses vs . projected stimulus . Note that , like in the data ( cf . Figure 2e ) , the mean response remains relatively invariant with mean stimulus , and that curves get shallower with increasing mean stimulus . ( g ) Comparison of steady state gain ( slope of functions shown in ( a ) and ( d ) ) when KD is fixed ( black ) and when KD is allowed to vary with the mean stimulus ( red ) . When KD is fixed , the the relationship between gain and mean stimulus approaches a power law with exponent 2 ( gain ~ KD ( S+KD ) 2 ) . However , when KD varies with the mean stimulus , the steady state gain ~ 1S , which is the Weber-Fechner Law . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01210 . 7554/eLife . 27670 . 013Figure 3—figure supplement 3 . Projected stimulus rescaled by Weber-Fechner relation correlate with firing rates . ( a–d ) Firing rate vs . projected stimulus rescaled by Weber-Fechner relation ( as in Figure 3g ) for four additional odorant-receptor combinations . Red line is the line of unity . Same data asin Figure 3—figure supplement 1c–f . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01310 . 7554/eLife . 27670 . 014Figure 4 . ORNs decrease gain with stimulus variance . ( a ) . Stimulus intensity of a fluctuating ethyl acetate stimulus with nearly constant mean but a variance that switches between high and low every 5 s . Five independent trials ( out of 248 ) are plotted . ( b ) Distributions of stimulus intensity for the epochs of low ( blue ) and high ( red ) variance . ( c ) ab3A firing rate responses corresponding to the trials shown in ( a ) following the switch from low to high variance , which takes place at t = 0 s and from high to low , which takes place at t = 5 s . ( d ) Probability distributions of the response . ( e ) Solid lines are ORN input-output curves computed from a single filter from both low ( blue ) and high ( red ) variance epochs . Dashed lines are the cumulative distribution functions ( c . d . fs ) of the projected stimulus . ( f ) ORN gain as a function of the standard deviation of the stimulus , measured per trial for each epoch . ( g ) Measured gain plotted against the slope of the cumulative distribution function for each trial . ( h ) Instantaneous gain ( blue ) and stimulus contrast ( orange ) as a function of time since switch . Dashed lines indicate crossover times of stimulus contrast and instantaneous gain . The delay is ~130 ms . n = 248 trials from 5 ab3A ORNs in 2 flies . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01410 . 7554/eLife . 27670 . 015Figure 4—figure supplement 1 . Variance gain control in Gaussian stimuli . ( a ) While the dominant change between the two epochs is the change in variance ( by construction ) , the low variance trials also tend to have slightly higher means . ( b ) ORN gain estimated by dividing the standard deviation of the response by the standard deviation of the stimulus , for each trial , vs . the standard deviation of the stimulus ( cf . Figure 4f ) . ( c ) Input-output curves for the ab3A ORN uncorrected for the change in the mean stimulus . The blue curve intersects the red curve , and is steeper than the red curve , suggesting that gain during the low variance epoch is higher than the gain during the low variance epoch . ( d ) ORN gain during high and low variance epochs , without correcting for the change in the mean stimulus . Each trial appears in the plot as one blue point ( for the low variance epoch ) and one red point ( for the high variance epoch ) . ( e ) Filters used in this analysis . Filters backed out of low variance ( blue ) or high variance ( red ) epochs alone are very similar . Therefore , we averaged all filters ( black ) and used that averaged filter to project all the stimulus in this dataset . ( f ) Coefficient of determination ( r2 ) vs . the standard deviation of the stimulus . ~80% of trials had r2>0 . 8 . ( g ) Coefficient of determination ( r2 ) vs . trial-wise ORN gain in the high and low variance epoch . Dashed lines in ( f–g ) indicate the median r2 during the high and low variance epoch . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 015 A common strategy used by sensory systems to encode signals over a broad range of background intensities is to scale the response according to the Weber-Fechner law ( Fechner , 1860; Stevens , 1957; Weber , 1834 ) , that is , to control gain inversely with stimulus mean . In ORNs , transduction currents elicited by odorant pulses are reduced by preceding pulses ( Nagel and Wilson , 2011 ) and scale inversely with background intensity , consistent with the Weber-Fechner law ( Cafaro , 2016; Cao et al . , 2016 ) . It remains unclear whether the ORNs’ ultimate output — the firing rate — follows the same Weber-Fechner scaling ( Cafaro , 2016; Martelli et al . , 2013 ) . We therefore stimulated ab3A ORNs with a set of fluctuating ethyl acetate stimuli with increasing means ( Figure 3a ) but roughly constant variances ( Figure 3b , Figure 2—figure supplement 1a ) . ORNs responded to the stimulus with the smallest mean by modulating firing rates between 0–60 Hz ( Figure 3c ) . This response range progressively decreased with increasing mean stimulus intensity ( Figure 3d ) , though the mean response remained at ~30 Hz . To estimate ORN input-output curves , we plotted ORN responses against the stimulus projected through the normalized best-fit linear filter for each stimulus , estimated by least-squares fitting ( Baccus and Meister , 2002; Chichilnisky , 2001; de Boer and Kuyper , 1968; Rieke et al . , 1997 ) ( Figure 3e ) Input-output curves grew shallower with increasing mean stimulus . We defined the ORN gain to be the slope of the input-output curve at that mean stimulus , similar to ( Baccus and Meister , 2002 ) . ORN gain in each trial varied with the mean stimulus in that trial as an approximate power law with exponent –1 ( Figure 3f ) . ORN gains could also be estimated by the ratio of standard deviation of the response to the standard deviation of the stimulus . This measure yielded similar values of ORN gain , and also decreased as a power law with exponent −1 ( Figure 3—figure supplement 1b ) . This exponent is consistent with the Weber-Fechner Law , which postulates that the just noticeable difference between two stimuli is inversely proportional to the stimulus magnitude ( Stevens , 1957 ) . Rescaling the projected stimulus by the gain predicted by Weber’s Law collapsed all input-output curves onto a single curve ( Figure 3g ) . Can front-end or back-end nonlinearities reproduce the observed change of input ( stimulus ) -output ( firing rate ) curves ( Figure 3e ) ? Clearly , no single output nonlinearity can fit the data shown in ( Figure 3e ) , since a single function cannot fit all the input-output curves . Since a front-end nonlinearity is present ( Figure 1 ) , we asked whether a static nonlinear-linear ( NL ) model could reproduce this data , with the input nonlinearity parameterized by a Hill function S/ ( S+K ) where S represents the input , and K the half maximum value . ( Figure 3—figure supplement 2a–c ) . NL model responses increased with mean stimulus ( Figure 3—figure supplement 2c ) , unlike in the data ( Figure 3d–e ) . However , if the half maximum value of the Hill function was allowed to vary with the mean stimulus , the model could qualitatively reproduce the data , suggesting adaptation at the front-end nonlinearity ( Figure 3—figure supplement 2d–f ) . To determine if similar gain-control relative to mean signal intensity was broadly observed , we tested additional ORNs from the two major olfactory organs of the fly , the antenna and the maxillary palp ( ab2A , pb1A ) , and used ecologically relevant odorants from three different functional groups ( ketones: 2-butanone , alcohols: 1-pentanol , esters: isoamyl acetate ) in various combinations . In all five cases , the neurons decreased gain with increasing odorant concentration , and obeyed a roughly inverse scaling ( Figure 3—figure supplement 1c–f ) . Rescaling the projected stimulus by the Weber-Fechner relation collapsed all ORN responses onto a single curve , similar to Figure 3g ( Figure 3—figure supplement 3 ) . Thus in vivo , for various neurons and odorants , ORN firing rate followed the Weber-Fechner Law . In other sensory modalities , such as vision , some peripheral neurons adapt not only to the mean but also to the variance of the signal ( Baccus and Meister , 2002; Rieke , 2001 ) . We therefore asked whether ORNs adjust their gain in response to changes in the variance of the signal . We stimulated ab3A ORNs with fluctuating ethyl acetate stimuli in which the variance of the signal changed every 5 s ( Figure 4a ) , switching back and forth between high to low values , around a nearly constant mean ( Figure 4b; Figure 4—figure supplement 1a ) , a protocol used to study gain control in visual neurons ( Baccus and Meister , 2002; Fairhall et al . , 2006; Shapley and Victor , 1978; Rieke , 2001; Smirnakis et al . , 1997 ) . As expected , ORNs responded to input fluctuations by modulating their firing rate . Interestingly , ORN firing rate variance did not vary as much as the stimulus variance between epochs of high and low stimulus variances , suggesting that ORNs actively changed their gain to compensate for such input differences ( Figure 4c–d ) . ORN input-output curves during high variance epochs ( red ) were shallower than during low variance ( blue ) epochs ( Figure 4e ) . Trial-wise ORN gain decreased with the variance of the stimulus ( Figure 4f ) . ORN gains estimated by dividing the standard deviation of the response by the standard deviation of the stimulus showed a similar decrease in ORN gain with stimulus variance ( Figure 4—figure supplement 1b ) . A simple coding strategy maximizes a neuron’s information capacity by matching its input-output curve to the cumulative distribution function ( c . d . f ) of the stimulus ( Laughlin , 1981 ) . Like the c . d . f . s ( dashed lines ) , the input-output curves ( solid ) are steeper during the low variance epoch . On a trial-by-trial basis , ORN gain was correlated with the c . d . f slope ( r2 = 0 . 7 ) ( Figure 4g ) . However , as the input variance changed by a factor of 2 . 5 , the gain in the neuron only changed by a factor of 1 . 7 , not as much as would be required for optimal information encoding . In these experiments , the gain changed within ~130 ms following the change in stimulus variance ( Figure 4h ) . ORN responses arise through two sequential steps: odor transduction followed by spike generation ( Nagel and Wilson , 2011 ) . Does each step possess separate gain control mechanisms , or is gain control achieved solely at one step ? Previous studies place the mechanism of adaptation to mean stimulus at the level of signal transduction ( Cafaro , 2016; Cao et al . , 2016; Nagel and Wilson , 2011 ) . How the spiking machinery might influence gain control , and where adaptation to signal variance takes place , remain unknown . We reanalyzed the responses of ab3A to ethyl acetate signals ( Figure 2 and 3 ) and measured ‘transduction gain’ ( stimulus to LFP ) and ‘firing gain’ ( LFP to firing rate ) . Changing the stimulus mean alone changed gain in LFP ( Figure 5a–b , Figure 5—figure supplement 1 ) . However , gain at the spiking machinery was largely invariant to the ten-fold change in the mean stimulus ( p=0 . 41 , Spearman rank correlation ) , with a 1 mV change in LFP leading to a ~ 10 Hz change in the firing rate , consistent with earlier studies ( Nagel and Wilson , 2011 ) ( Figure 5c–d ) . Transduction gain , like ORN gain , scaled with the Weber-Fechner Law , for a variety of odor-receptor combinations ( Figure 5—figure supplement 1a-d ) consistent with previous studies ( Cafaro , 2016; Cao et al . , 2016 ) . In contrast , adaptation to the stimulus variance changed gain both at transduction and at spiking ( Figure 5e–h ) . Both gains changed by a factor of ~1 . 3 from the high to the low variance epoch ( p<0 . 001 , Wilcoxon signed rank test ) , contributing roughly equally to variance gain control ( Figure 5—figure supplement 1f ) . 10 . 7554/eLife . 27670 . 016Figure 5 . Mean gain control occurs primarily at transduction , and variance gain control occurs both at transduction and at the firing machinery . ( a ) Transduction input-output curves from stimulus to LFP . Colors indicate increasing mean stimulus . Filters and projections are computed trial by trial . ( b ) Transduction gain , measured from the slopes of these input-output curves , decreases with the mean stimulus . The red line is a power law with exponent −1 , ( Weber’s Law ) . ( c ) Input-output curves for the firing machine module . ( d ) Firing gain does not change significantly with mean stimulus . ( e ) Transduction input-output curves for low ( blue ) and high ( red ) variance stimuli . ( f ) Transduction gains in the low variance epoch are significantly higher than transduction gains in the high variance epoch ( p<0 . 001 , Wilcoxon signed rank test ) ( g ) Input-output curves of firing machinery during low variance stimuli . ( g ) Firing gain during low variance epochs are significantly higher than firing gains during high variance epochs ( p<0 . 001 , Wilcoxon signed rank test ) . Projections of stimulus are divided by the mean stimulus in each trial to remove the small effect Weber-Fechner gain scaling . Data in this figure is same as in Figures 3 and 4 . ( a , c , e , g ) Mean across all trials . ( b , d , f , h ) Individual trials . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01610 . 7554/eLife . 27670 . 017Figure 5—figure supplement 1 . LFP responses to fluctuating Gaussian ethyl acetate signals with increasing mean . Colors correspond to increasing mean odorant stimuli ( purple…yellow ) . The data in this figure corresponds to the data shown in Figure 3 . Increasing stimulus mean decreases the variance of the LFP responses , similar to the decrease in LFP responses seen in Figure 3c . Traces are mean subtracted . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 017 Stimulating ORNs with odorants evokes responses from both transduction and spiking machinery , making it hard to determine how independent gain control at the two modules are . To further pin-point the contributions of signal transduction and firing machineries to gain control , we expressed Chrimson channels ( Klapoetke et al . , 2014 ) in ab3A ORNs and activated them using red light , either in isolation or in combination with odorants . First , we used a fluctuating ethyl acetate stimulus to probe transduction and ORN gain while increasing the neuron’s firing rate using increasing backgrounds of red light ( Figure 6a–b ) . While increasing light levels elicited increasing firing rates ( Figure 6b inset ) , ORN and transduction gain did not vary with the intensity of supplemental light ( Figure 6a–b ) . This suggests that constitutive spiking activity does not feed back onto LFP adaptation , or overall ORN gain . 10 . 7554/eLife . 27670 . 018Figure 6 . Modularity of gain control revealed by optogenetic stimulation . ab3A ORNs in w; 22a-GAL4/+; UAS-Chrimson/+ flies can be activated by ethyl acetate odorant or by red light . ( a–b ) Fluctuating odor foreground and constant light background . ( a ) Transduction gain to fluctuating odor vs . background light stimulation intensity . ( b ) Overall ORN gain to fluctuating odor stimulus vs . background light stimulation intensity . ( b , inset ) ORN firing rate vs . background light intensity . ( c–d ) Fluctuating light foreground and constant odor background stimulus . ( c ) Input-output curves to fluctuating light stimulus for increasing background odor ( lighter colors indicate larger odor background ) . ( d ) ORN gain is invariant with background odor concentration . ( d , inset ) Odor-induced firing gain vs . background odor concentration . ( e–f ) Fluctuating light stimulus with different variances . ( e ) Input-output curves for high ( red ) and low ( blue ) variance light stimuli . ( f ) ORN gain as a function of the standard deviation of the light stimulus . ( a–b ) n = 75 trials from 13 ORNs . ( c–d ) n = 64 trials from 5 ORNs . ( e–f ) n = 21 trials from 3 ORNs . Lines link trials from a single ORN . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 018 Second , we used a fluctuating light stimulus to probe the spiking gain while stimulating the ORN and its receptors with increasing backgrounds of ethyl acetate odorant ( Figure 6c–d ) . While ethyl acetate backgrounds of increasing intensity increased ORN firing rate ( Figure 6d inset ) , they failed to change gain in the spiking machinery . Increasing odor backgrounds moved input-output curves along the y-axis ( Figure 6c ) , consistent with increasing firing due to background odor , but failed to change the slope of these curves , suggesting that ORN gain to the fluctuating light probe was not changed . This suggests that adaptation at transduction does not affect gain of the spiking machinery , consistent with our result that increasing odor backgrounds decreased gain at transduction , but not spiking ( Figure 5a–d ) . Thus , Weber-Fechner scaling in ORN gain control to stimulus mean is insulated from activity of the spiking machinery . Variance gain control exists in a wide range of neurons ( Baccus and Meister , 2002; Díaz-Quesada and Maravall , 2008; Nagel and Doupe , 2006; Rieke , 2001; Wark et al . , 2009; Zaghloul et al . , 2005 ) and in models of spiking neurons ( Gaudry and Reinagel , 2007; Hong et al . , 2007; Yu and Lee , 2003; Yu et al . , 2005 ) , suggesting that variance gain control could be an intrinsic property of spiking neurons . To determine if the spiking machinery alone could give rise to variance gain control , we stimulated ab3A ORNs that express Chrimson with fluctuating light stimuli of different variances at fixed mean . ORN input-output curves were steeper when the variance of the light stimulation was smaller ( Figure 6e–f ) , similar to the curves observed with odor stimulation ( cf . Figure 4e ) . We observed that gain changed by a factor of ~1 . 5 when the standard deviation of the light stimulus changed by a factor of ~3 , consistent with variance gain control occurring partly in the spiking machinery ( Figure 5e–h ) , though Chrimson channels might exhibit their own nonlinear activation properties . When navigating odor plumes , the precise timing of the encounter with the plume carries important information , which may be lost if adaptation changes the lag between signal and response in a concentration-dependent manner . The kinetics of ORN spiking in response to pulses of odorant are invariant to the pulse intensity and to the background intensity over a range of odorant concentrations ( Martelli et al . , 2013 ) . Paradoxically , adaptation to background odorants slows transduction current responses to odor pulses ( Cao et al . , 2016; Kaissling et al . , 1987; Nagel and Wilson , 2011 ) . We hypothesized that these seemingly contradictory results might be resolved if the ORN spiking machinery speeds up to compensate for the intensity-dependent slowdown in the LFP . We characterized responses to odorant stimuli on increasing backgrounds by measuring both ORN spike rates and LFPs . We computed cross correlation functions between the stimulus and the LFP for various stimulus backgrounds . For stimuli on low backgrounds , LFP cross-correlation functions peaked earlier , while for stimuli on larger backgrounds , LFP cross-correlation functions peaked later ( Figure 7a ) , consistent with previous results ( Cao et al . , 2016; Kaissling et al . , 1987; Nagel and Wilson , 2011 ) . Surprisingly , cross-correlation functions from the stimulus to the firing rate were similar between stimuli on low and high backgrounds ( Figure 7b ) , consistent with ( Martelli et al . , 2013 ) . This selective change in the kinetics of the LFP , but not the firing rate , occurred even though there was no change in the stimulus autocorrelation function from low to high stimulus ( Figure 7c ) . We defined the LFP and firing rate lags relative to the stimulus by the location of the peak of the cross-correlation function , and found that while LFP response lags increased with increasing odorant concentration ( p<10−2 , Spearman test ) , firing rate lags remained relatively invariant with odorant concentration ( p>0 . 1 , Spearman test ) ( Figure 7d–g ) . 10 . 7554/eLife . 27670 . 019Figure 7 . Adaptation to the mean slows down LFP , but not firing rate . ( a–d ) . Response of ab3A ORNs to Gaussian ethyl acetate stimuli on increasing backgrounds . ( a ) Cross correlation functions between ethyl acetate stimulus and ab3 LFP responses for low ( purple ) and high ( yellow ) background stimuli . ( b ) Cross correlation functions between ethyl acetate stimulus and ab3A firing rate responses for low ( purple ) and high ( yellow ) background stimuli . ( c ) Stimulus autocorrelation functions for low ( purple ) and high ( yellow ) background stimuli . ( d–g ) LFP and firing rate lags with respect to the stimulus vs . the mean stimulus for various odor-receptor combinations . LFP lags increase with mean stimulus , while firing rate lags do not . ( h ) Firing lags of ab3A ORNs expressing Chrimson channels vs . applied light power . In ( c–g ) , ρ is the Spearman correlation coefficient , and p is the corresponding p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 01910 . 7554/eLife . 27670 . 020Figure 7—figure supplement 1 . Variance gain control does not change response kinetics . ( a ) Stimulus autocorrelation functions , computed during high variance epochs ( red ) and during low variance epochs ( blue ) . ( b ) Autocorrelation time ( defined as the time the autocorrelation function first drops to 1/e ) vs . the standard deviation of the stimulus , for each trial . ( c ) Cross correlation functions from stimulus to LFP . The cross correlation functions are very similar between high ( red ) and low ( blue ) variance epochs . ( d ) LFP lag with respect to the stimulus , estimated from the location of the peak cross-correlation , vs . standard deviation of the stimulus . No significant change in lag was observed ( p=0 . 4 , t-test ) . ( e ) Cross correlation functions from stimulus to firing rate . The cross correlation functions are very similar between high ( red ) and low ( blue ) variance epochs . ( f ) Firing rate lag with respect to the stimulus , estimated from the location of the peak cross-correlation , vs . standard deviation of the stimulus . No significant change in lag was observed ( p=0 . 133 , t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 020 For firing rate lags to remain invariant with odorant concentration despite a slowdown in transduction , the kinetics of the spiking machinery need to speed up with increasing input to the cell . To test if this is the case , we stimulated ab3A ORNs expressing Chrimson with Gaussian red light stimuli with increasing means , and measured lags between the applied light stimulus and firing rate . Firing lags decreased with increasing light power concentration ( p<10−4 , Spearman test ) , suggesting that the ORN spiking machinery can speed up with increasing mean input currents ( Figure 7h ) , as would accompany increasing odor backgrounds ( Cao et al . , 2016 ) . If adaptation to the mean slows down transduction , which is compensated for at spiking , does adaptation to the stimulus variance also lead to similar compensatory kinetics ? We found that a three-fold change in the stimulus variance , despite leading to changes in LFP and firing gains ( Figures 4–5 ) , did not significantly change kinetics either at LFP or firing rate ( Figure 7—figure supplement 1 ) , consistent with our earlier results suggesting that mean and variance gain control have distinct mechanisms . How do adaptive mechanisms at transduction preserve both the Weber-Fechner Law and lead to response slowdowns ? In the following we show that a minimal two-state model of the olfactory receptor-olfactory co-receptor ( Or-Orco ) complex with an adaptation architecture similar to that of the bacterial chemotaxis system ( Asahina et al . , 2009; Barkai and Leibler , 1997; Emonet and Cluzel , 2008; Shimizu et al . , 2010 ) can reproduce the LFP responses to naturalistic and Gaussian stimuli , as well as Weber-Fechner Law and its accompanying response slow down . In our model , Or-Orco complexes can be active or inactive ( C and C* in Figure 8a ) and the active complex binds odorant S with higher affinity than the inactive complex . We assume that ligand ( un ) binding is fast compared to ( in ) activation rates ( w+ and w− in Figure 8b ) . The fraction a of active complexes therefore obeys the equation ( 1 ) dadt= ( 1−a ) w+ ( S , ε ) −a w− ( S , ε ) 10 . 7554/eLife . 27670 . 021Figure 8 . A modified two state receptor model reproduces Weber’s Law and adaptive slowdown in LFP responses . ( a ) . Or-Orco complexes ( C ) can be bound or unbound and active or inactive . ( b ) We assume ( un ) binding rates are much faster than ( in ) activation rates . Activity of the complex feeds back onto the free energy difference between active and inactive conformations , which also decreases the activation and inactivation rates of the complex ( Equations 1–4 ) . A mono-lobed filter converts receptor activity into LFP signals ( Equation 5 ) . We fit the model to Gaussian ( Figures 3 and 5 ) and naturalistic data ( Figures 1–2 ) . In these fits , α=12 . 5 s−1 , β=1 . 26 s−1 , εL=0 . 86 , Kon=0 . 1 V and Koff=400 V . ( c ) Model gain vs . mean stimulus . Red line is the Weber-Fechner prediction ( ΔR/ΔS∼1/S ) . ( c ) LFP gain vs . model gain . ( e ) Model response lag with respect to stimulus vs . mean stimulus . ( f ) LFP and model responses to naturalistic stimulus . ( g ) The model reproduces LFP responses to similar-sized whiffs that vary inversely with the size of preceding whiffs . ( cf . Figure 2 ) . ( h ) LFP responses vs . model responses for every whiff in the naturalistic stimulus . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 02110 . 7554/eLife . 27670 . 022Figure 8—figure supplement 1 . Steady state activity as a function of the stimulus background . At high stimulus background , the steady state activity of the receptor complex is a0 ( here , 1/2 ) . The model is unable to adapt perfectly to lower stimulus backgrounds , since ε is bounded by εL . This causes the steady state activity to decrease . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 02210 . 7554/eLife . 27670 . 023Figure 8—figure supplement 2 . Front-end adaptation followed by a LN model reproduces firing rate responses to Gaussian and naturalistic stimuli . ( a–f ) Model from stimulus to firing rate ( see Materials and methods ) fit to Gaussian and naturalistic stimuli . ( a ) Model responses vs . projected stimulus with increasing mean stimulus ( cf . Figure 3 ) . ( b ) Model gain vs . mean stimulus . Red line is the Weber-Fechner prediction ( ΔR/ΔS∼1/S ) . ( c ) Firing rate gain vs . model gain . ( d ) Firing rate and model responses to naturalistic stimulus . ( e ) The model reproduces variation in the firing rate responses to similar-sized whiffs ( cf . Figure 2 ) . ( f ) Firing rate responses vs . model responses for every whiff in the naturalistic stimulus . DOI: http://dx . doi . org/10 . 7554/eLife . 27670 . 023 where the rates of activation w+ ( S , ε ) and inactivation w− ( S , ε ) are nonlinear functions of the odor concentration S and of the free energy difference ε between the unbound active and inactive states ( Equations 3-4 in Materials and methods ) . The LFP is modeled as a linear filter acting on the activity ( Figure 8b , Equation 5 in Materials and methods ) . At steady state , Equation ( 1 ) reduces to a¯ ( S , ε ) =1/ ( 1+w− ( S , ε ) /w+ ( S , ε ) ) , where the bar indicates steady state . a¯ ( S , ε ) is a monotonically increasing function of the odor concentration S . Increasing the free energy difference ε shifts this function towards higher values of S , therefore reducing the sensitivity of the system . We model adaptation by assuming that activity of the Or-Orco complex controls the activity of factors that act on the complex to modify the free energy difference ε: ( 2 ) dεdt=β ( a−a0 ) where β is the rate of adaptation . Importantly , the rate of change of ε only depends on the activity a but not on the free energy difference ε . The architecture of this feedback is similar to that of the bacterial chemotaxis system and ensures that for increasing values of S , the changes in ε compensate for changes in free energy due to ligand binding ( Barkai and Leibler , 1997 ) . Thus , adaptation eventually returns a to the adapted value a0 providing Weber-Fechner scaling ( Shimizu et al . , 2010 ) ( as in Figure 3 ) . We assume that the free energy of the complex can only be changed within a finite range , and that the lower bound εL is reached for small values of S . Thus , in the absence of ligand , the steady state activity can be smaller than a0 . For non-zero values of S , the steady state activity first increases with background signal intensity ( Martelli et al . , 2013 ) , before it becomes independent of background intensity once it reaches a0 ( Figure 8—figure supplement 1 ) , as seen in Figure 3c . An important intrinsic property of this model is that adaptation to increasing background of odorant decreases the rates of activation and inactivation , w+ and w− , of the Or-Orco complex , providing a self-consistent explanation for the slowdown of the response kinetics of the LFP upon adaptation . It is interesting to note that this kinetic property emerges because ( 1 ) the switching rates are decreasing functions of the free energy difference ε and ( 2 ) the requirement of Weber-Fechner scaling , which causes the adapted value of ε to scale with the logarithm of the mean signal intensity ( see Materials and methods ) . The resulting model ( Equations 1–2 , and Equations 3–5 in Materials and methods ) contains six parameters plus another three for converting the signal from activity to LFP . We fit this model to LFP responses to the Gaussian and naturalistic stimuli . The model decreased gain with the mean stimulus background , consistent with Weber-Fechner Law ( Figure 8c ) , and predicted the observed decrease in the LFP gains well ( Figure 8d , r2=0 . 84 ) . In addition , response lags of this model with respect to the stimulus increased with the mean stimulus ( Figure 8e ) , similar to the slowdown observed in the LFP responses ( cf . Figure 7 ) . Finally , this model can also reproduce LFP responses to naturalistic , intermittent signals , approximating well the time trace ( Figure 8f , h ) and the dependence on previous whiffs ( Figure 8g , compare to Figure 2 ) . Since the spiking machinery compensates for the slowdown in LFP responses to preserve the timing of odorant encounters , we wondered if a simplification of this model that ignores the slowdown of the LFP kinetics upon adaptation could be used to predict firing rate: RF=N ( KF⊗ a¯ ( S , ε ) ) where N is a static nonlinearity , KF is a partially derivative-taking linear filter ( ⊗ indicates convolution ) , and a¯ ( S , ε ) the steady state solution of Equation ( 1 ) with ε obeying Equation 2 . This simplification reduces this model to a type of adaptive nonlinear-linear-nonlinear ( NLN ) model , which preserves Weber-Fechner Law and reproduces the firing rates of ORN in response to both naturalistic and Gaussian Stimuli ( Figure 8—figure supplement 2 ) . Thus it could be a useful tool in modeling ORN responses received by PNs , or in constructing computational models of the antennal lobe ( Assisi et al . , 2011; Bazhenov et al . , 2001; Berck et al . , 2016; Capurro et al . , 2012; Chong et al . , 2012; Hopfield , 1991; Kee et al . , 2015; Koulakov et al . , 2007; Luo et al . , 2010; Sanda et al . , 2016; Satoh et al . , 2010; Stevens , 2015 ) . The model reproduced the change in the input-output curves on increasing the mean stimulus ( Figure 8—figure supplement 2a , cf . Figure 3e ) and decreased gain inversely with the mean stimulus , consistent with the Weber-Fechner Law ( Figure 8—figure supplement 2b ) . The model reproduced the observed decrease in the ORN gains ( Figure 8—figure supplement 2c , r2=0 . 85 ) , and responses to naturalistic stimuli ( Figure 8—figure supplement 2d–f ) .
The Weber-Fechner Law has been observed in several sensory systems , including vision ( Burkhardt , 1994; Laughlin and Hardie , 1978; Nikonov et al . , 2006 ) , audition ( Riesz , 1928 ) , and somatosensation ( Holway and Pratt , 1936 ) . In olfaction , the Weber-Fechner Law was demonstrated at the LFP level ( Cafaro , 2016; Cao et al . , 2016 ) . Here we directly measured ORN firing rate and stimulus intensity and found that the ORN firing rate exhibited Weber-Fechner gain scaling relative to the mean stimulus intensity for five different odor-receptor combinations ( Figure 3 , Figure 3—figure supplement 1 ) . These data suggest that olfaction shares the Weber-Fechner Law with other sensory systems . What is the purpose of front-end Weber-Fechner gain scaling ? ORNs are capable of spiking up to ~300 Hz ( Hallem and Carlson , 2006 ) ; however , we found that with their compressive gain control , ORNs maintained firing rates between 0–50 Hz to fluctuating odor stimuli , even with a ten-fold increase in the mean stimulus . The ORNs’ postsynaptic partners , the projection neurons ( PNs ) ( Olsen et al . , 2010 ) , are most sensitive to ORN firing rates of below ~50 Hz ( Jeanne and Wilson , 2015 ) . Thus , gain scaling at ORNs could act to maintain ORN firing rates in the range that PNs are most sensitive to for a wide range of concentrations , avoiding saturation of the ORN-PN synapse . In principle , gain control in sensory systems could be affected by several moments of the stimulus distribution , measured over many timescales . In the visual system , gain control depends on stimulus mean and variance , and some studies have shown little dependence on higher moments like the skew and the kurtosis ( Bonin et al . , 2006; Tkačik et al . , 2014 ) . Cell-intrinsic variance gain control exists in a variety of systems , including the retina ( Beaudoin and Manookin , 2008; Zaghloul et al . , 2005 ) , lateral geniculate nucleus ( Lesica et al . , 2007 ) , auditory neurons ( Nagel and Doupe , 2006 ) , and cortex ( Díaz-Quesada and Maravall , 2008; Ringach and Malone , 2007 ) . Photoreceptors do not exhibit variance gain control , and variance adaptation arises only in the subsequent processing in bipolar cells and ganglion cells ( Baccus and Meister , 2002; Kim and Rieke , 2001; Rieke , 2001 ) . What could be the functional role of variance gain control in olfaction ? One possibility is to help ensure that ORN responses occupy a large fraction of their dynamic range ( Laughlin , 1981 ) . While we quantified our stimulus in terms of the first and second moments of the stimulus statistics in this study , these moments may not map simply onto the salient features that are most relevant to the fly’s encoding scheme . Variance gain control could therefore be a consequence of an adaptive representation that is important to the coding properties of the ORN , but remains unknown to us . Nonetheless , because variance gain control is distributed between transduction and spiking machinery ( Figure 5 ) , and mean gain control slows down transduction ( Figure 7 ) ( Cao et al . , 2016; Nagel and Wilson , 2011 ) but variance gain control does not ( Figure 7—figure supplement 1 ) , adaptation to the stimulus variance is mechanistically distinct from adaptation to stimulus mean . The results presented here , and the models that reproduce them , focus on the phenomenology of gain control in ORNs . Do these phenomenological results constrain possible mechanisms that could implement gain control in ORNs ? Weber-Fechner gain scaling ( Figure 3 ) can be reproduced by models using feed-forward loops ( Clark et al . , 2013; Goentoro et al . , 2009 ) , integral feedback ( Yi et al . , 2000 ) , or both ( Schulze et al . , 2015 ) . A detailed biophysical model of odor-receptor binding and channel opening has been proposed to account for transduction responses to odors ( Nagel and Wilson , 2011 ) . While this model can change gain via a negative feedback mechanism , it does not reproduce the Weber-Fechner law , or the slowdown of LFP kinetics upon adaptation . Here we showed that these features emerge if we assume that: ( i ) the activity a of the Or-Orco complex feeds back onto the free energy difference ε between the active and inactive state of the unbound Or-Orco complex , which in turns affects both the rates of activation and inactivation of the complex; and ( ii ) the rate at which ε is modified only depends on the activity a . In summary , the steady state activity in our model depends nonlinearly on the stimulus , reproducing the effects of saturation in responses to naturalistic stimuli . Adaptation shifts the effective half-maximum of the input nonlinearity to the right , recapitulating Weber-Fechner gain control; and decreases transition rates from active to inactive receptor complexes , reproducing slowing LFP responses with adaptation . Such an architecture reproduces the Weber-Fechner law and is similar to that of the bacterial chemotaxis system ( Barkai and Leibler , 1997 ) . There , adaptation is mediated by two antagonistic factors , one that acts on inactive complexes only , and another one that acts on active complexes ( Barkai and Leibler , 1997 ) . While the molecular architecture of the signaling pathway in ORNs has not been fully characterized , several studies have implicated calcium as a slow diffusible factor that could mediate adaptation to the mean stimulus ( Deshpande et al . , 2000; Störtkuhl et al . , 1999 ) . Decreasing extracellular calcium levels , or internal free calcium , breaks Weber-Fechner gain scaling at transduction ( Cao et al . , 2016 ) . Other mechanisms have also been implicated in adaptation of ORNs , like autoregulation of Orco via cAMP signaling ( Getahun et al . , 2013 ) . While slower adaptive processes also exist , our data on responses to naturalistic stimuli ( Figures 1 and 2 ) , and data from paired-pulse experiments ( Cao et al . , 2016 ) , suggest that some adaptation mechanisms act on fast timescales of several hundred milliseconds . Many models that decrease gain with increasing mean stimulus also speed up response kinetics ( Clark et al . , 2013; De Palo et al . , 2012; Nagel and Wilson , 2011; Schulze et al . , 2015; Seung , 2003 ) , describing well the phenomenology of other sensory systems where gain and response speed trade off ( Baylor and Hodgkin , 1974; Dunn et al . , 2007; Nagel and Doupe , 2006; Payne and Howard , 1981 ) . However , in olfactory systems , transduction kinetics slow down with increasing stimulus background , both in insect ORNs ( Figure 7 , ( Cao et al . , 2016; Nagel and Wilson , 2011 ) ) and in vertebrate ORNs ( Reisert and Matthews , 1999 ) . It is not trivial to devise systems in which kinetics slow down with increasing stimulus background . To exhibit this property , a system must increase its effective timescale of response with stimulus intensity , for example , by decreasing all reaction rates uniformly . Interestingly , our model also exhibits a slowdown in the LFP kinetics upon adaptation . This feature emerges intrinsically from the model architecture because: ( i ) the feedback of the activity onto the free energy difference ε affects both the activation and deactivation rates of the complex ( w+ and w− ) ; and ( ii ) the Weber-Fechner gain control causes ε to scale logarithmically with the stimulus , which in turn causes w+ and w− to decrease . Earlier work modelled the transformation from LFP to firing rates using a derivative-taking kernel ( Nagel and Wilson , 2011 ) . Here , we show that the temporal structure of these kernels depends on the adaptation state of the ORN , and must take derivatives on shorter timescales at higher stimuli to compensate for slowing transduction kinetics . Consistent with this , we see that the latency of spiking decreases increasing optogenetic drive ( Figure 7h ) . Under these conditions , the change in spiking latency is smaller than the increase in transduction lags we observe ( Figure 7d–g ) . While the mechanism of the speed up in spiking with increasing odor stimulus is not known , the neuron’s ability to spike with shorter latencies relative to transduction could depend on the adapted state of its receptors , the level of intracellular calcium , or the distance of its membrane potential from firing thresholds . What cellular mechanisms could give rise to gain control that is variance dependent ? ( Figure 4 ) . We found that both the transduction machinery and the spiking machinery of the ORNs exhibit variance-sensitive gain-control ( Figures 5 and 6 ) . Variance gain control after transduction could arise from the spike generating machinery . Hodgkin-Huxley ( HH ) model neurons exhibit variance gain control ( Hong et al . , 2008; Lundstrom et al . , 2008; Yu and Lee , 2003 ) . Simpler neuron models , like the FitzHugh-Nagamo model ( Hong et al . , 2007 ) , and the linear integrate-and-fire ( LIF ) model ( Yu and Lee , 2003 ) also exhibit variance-dependent gain control . In the visual system , non-spiking bipolar neurons show variance gain control ( Baccus and Meister , 2002; Rieke , 2001 ) so mechanisms for variance gain control in the absence of spike generation might be similar between these systems . Previous studies of olfactory adaptation employed conditioning and probe stimuli ( Cafaro , 2016; Martelli et al . , 2013; Nagel and Wilson , 2011 ) , which typically adapt neurons over many seconds or minutes before testing response properties with a short probe . Other studies using paired pulse protocols ( Cao et al . , 2016 ) found that responses to brief pulses of odorant reduced gain on timescales as brief as 500 ms , which is close to the timescale of the neural response to odors ( Kim et al . , 2011; Martelli et al . , 2013; Nagel and Wilson , 2011 ) . Similar fast timescales of gain control have been observed in the visual system ( Burns et al . , 2002; Baylor and Hodgkin , 1974 ) . We found that this fast gain control was employed by ORNs to dynamically control gain during responses to naturalistic odorant stimuli ( Figures 1–2 ) . Dynamic gain control allows ORNs to respond to the rapidly changing statistics of natural odor plumes , letting gain decrease quickly in response to a large whiff and then ramp up again to a subsequent small whiff . Dynamic inhibition in the antennal lobe ( Nagel et al . , 2015; Raccuglia et al . , 2016 ) would permit PNs to remain sensitive to these rapid changes in ORN firing rate , ensuring propagation of information about odor encounters to the brain . Insects follow odor plumes to their source to find food or reproductive mates ( Murlis et al . , 1992 ) . For flies , this task is challenging since they fly fast ( ~30 cm/s ) ( Tammero and Dickinson , 2002 ) and odor filaments are narrow ( Murlis et al . , 1992 ) . Even for relatively broad and static odor plumes , flies are within odor plumes so briefly that they experience plume contact and plume loss in quick succession ( 10–250 ms ) ( van Breugel and Dickinson , 2014 ) . Olfactory search behavior in this setting consists of rapid flight surges on encountering odor plumes , and stereotyped crosswind casts on losing odor plumes ( van Breugel and Dickinson , 2014 ) . Navigation based on odor intensities alone may not be possible , as odor intensities are not informative about the direction to the odor source at length scales longer than 10 cm ( Murlis et al . , 1992 ) . Indeed , there is a growing body of evidence underlining the importance of timing in olfaction ( Martelli et al . , 2013; Rebello et al . , 2014; Shusterman et al . , 2011; Smear et al . , 2013 , 2011 ) . In this context , it may be important for the fly to know precisely when it encountered an odor filament . Previous studies have shown that kinetics of transduction slowed during adaptation ( Cao et al . , 2016; Nagel and Wilson , 2011 ) , but kinetics of firing rate did not ( Martelli et al . , 2013 ) . Here we reproduced both findings and resolved this apparent contradiction . We discovered that the spiking machinery speeds the kinetics back up . These complementary kinetic mechanisms mean that the timing of short odorant encounters is preserved in neural encoding , regardless of intensity . Such an encoding scheme could aid insects in navigating odor plumes to their source . When a system responds identically , in amplitude and in kinetics , to stimuli that are different only in scale , the system is said to show fold change detection ( FCD ) ( Goentoro et al . , 2009 ) . FCD thus implies the Weber-Fechner law , but systems can obey the Weber-Fechner Law without showing FCD . Another requirement for FCD is that response kinetics remain invariant with respect to the mean stimulus intensity . Thus , ORN responses are intriguingly similar to the response phenomenology of FCD networks ( Goentoro and Kirschner , 2009 ) . Interestingly , olfactory adaptation is linked to flight in insects . Olfactory receptors ( ORs ) adapt and have co-evolved with flight ( Edwards , 1997; Getahun et al . , 2013; Jones et al . , 2005 ) , and occur only in flying insects ( Missbach et al . , 2014 ) . In contrast , the more ancient ionotropic receptors ( Missbach et al . , 2014 ) , found in all insects , do not appear to adapt to prolonged odor stimuli ( Cao et al . , 2016 ) . While ORs play an important role in larval olfactory navigation ( Hernandez-Nunez et al . , 2015; Mathew et al . , 2013; Schulze et al . , 2015; Gepner et al . , 2015 ) , the statistics of odor signals close to surfaces , and in the air , where flying insects encounter them , may be very different ( Murlis et al . , 1992 ) ( Martelli et al . , 2013 ) . Receptors capable of fast adaptation may allow flying insects to detect brief whiffs of airborne odors .
Flies were reared at 25°C on conventional fly medium ( Helfand and Carlson , 1989 ) . All experiments were performed on adult female flies 3–5 days post-eclosion . Unless otherwise mentioned , recordings were from ab3A neurons in Canton-S flies . In Figure 6 and Figure 7h , we recorded from ab3A ORNs in w; Or22a-GAL4/+; UAS-Chrimson/+ flies . In these flies , only ab3A ORNs were sensitive to light , while ab3B neurons and nearby ab2 sensilla were not . We used a Photo-Ionization Detector ( PID ) ( 200B , Aurora Scientific ) to measure the odor stimulus during every experiment . Stimulus measurements occurred simultaneously with all electrophysiology , and the tip of the PID probe was <1 cm of the odor delivery tube and the fly ( Figure 1—figure supplement 1 ) . The PID was calibrated by depleting known volumes of pure odorants , and the response of the PID was found to be approximately linear with odorant flux ( Figure 1—figure supplement 1 ) . However , due to gradual changes in the sensitivity of the PID detector , odor intensity measurements are not comparable across experiments . We measured the intensity of red light that we used to activate Chrimson at the location of the fly using a PM160 light power meter ( Thorlabs ) . We used this to construct a function mapping control signals to our LED to light power in µW , and transformed control signals into light power using this function . | Insects follow odor trails carried by the wind to find mates and sources of food . The turbulent motion of the air means that these odors tend to arrive in whiffs with varying intensities and durations , which makes it difficult to distinguish them . Insects use sensory cells called olfactory receptor neurons on their antenna to process odors . Specialized receptor proteins on the surface of these olfactory receptor neurons detect odor molecules and set off a cascade of events in these cells that ends with a signal being sent to the brain . Much is known about how insects detect and process different kinds of smells , but it remains less clear how their olfactory neurons process the timing and intensity of odor whiffs . Now , Gorur-Shandilya et al . report what happens in the olfactory receptor neurons of fruit flies when they have to compensate for variations in the duration and intensity of odor whiffs . In the experiments , fruit flies were exposed to two sweet-smelling odors . To do so , Gorur-Shandilya et al . built an apparatus that enabled them to control the airflow with enough precision that they could simulate the variability in the timing and intensity of natural odors in the air . The response of the flies’ olfactory receptor neurons to these smells was recorded . The experiments showed that the neurons could adapt to both the average intensity and the variance in intensity of odor signals . The ability of these neurons to adapt to the average intensity of the odors followed a specific pattern , which is also seen in sensory cells responsible for vision and touch . Adapting to the average strength of an odor slows down the first of two steps in its processing . However , the second step has a complementary mechanism to speed up signals to the brain , so the timing of an odor whiff is accurately captured regardless of how strong it is . Based on these results , Gorur-Shandilya et al . created a biophysical model that could reproduce the experimental data , including the slowdown in the first step . The experiments and the model may now help other scientists to investigate how different animals detect and process smells . For example , some insects are pests of agricultural crops , while other insects , such as mosquitos , spread diseases between people . A better understanding of how insects detect odors may help scientists to find ways to interfere with these processes to protect food crops and reduce the spread of tropical diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2017 | Olfactory receptor neurons use gain control and complementary kinetics to encode intermittent odorant stimuli |
Neurons face unique challenges of transporting nascent autophagic vacuoles ( AVs ) from distal axons toward the soma , where mature lysosomes are mainly located . Autophagy defects have been linked to Alzheimer’s disease ( AD ) . However , the mechanisms underlying altered autophagy remain unknown . Here , we demonstrate that defective retrograde transport contributes to autophagic stress in AD axons . Amphisomes predominantly accumulate at axonal terminals of mutant hAPP mice and AD patient brains . Amyloid-β ( Aβ ) oligomers associate with AVs in AD axons and interact with dynein motors . This interaction impairs dynein recruitment to amphisomes through competitive interruption of dynein-Snapin motor-adaptor coupling , thus immobilizing them in distal axons . Consistently , deletion of Snapin in mice causes AD-like axonal autophagic stress , whereas overexpressing Snapin in hAPP neurons reduces autophagic accumulation at presynaptic terminals by enhancing AV retrograde transport . Altogether , our study provides new mechanistic insight into AD-associated autophagic stress , thus establishing a foundation for ameliorating axonal pathology in AD .
Autophagy is the major cellular degradation pathway for long-lived proteins and organelles ( Nixon , 2013; Rubinsztein et al . , 2011; Schneider and Cuervo , 2014; Yue et al . , 2009 ) . Altered autophagy has been linked to several major age-related neurodegenerative diseases , including Alzheimer’s disease ( AD ) , that are associated with accumulation of misfolded protein aggregates ( Nixon , 2013; Rubinsztein et al . , 2011; Schneider and Cuervo , 2014; Yue et al . , 2009 ) . Ultrastructural analysis revealed that AD brains display a unique autophagic stress phenotype: autophagic vacuoles ( AVs ) massively accumulate and cluster within large swellings along dystrophic neurites ( Nixon et al . , 2005 ) , a typical amyloid β ( Aβ ) -associated phenotype not found in other neurodegenerative diseases ( Benzing et al . , 1993 ) . The mechanisms underlying such autophagic stress in AD neurons are largely unknown . As highly polarized cells with long axons , neurons face the special challenge of transporting AVs containing engulfed aggregated proteins and damaged organelles generated from distal processes toward the soma where mature acidic lysosomes are mainly located ( Nixon , 2013; Sheng , 2014 ) . In neurons , autophagosomes are continuously formed in distal axons ( Maday and Holzbaur , 2014; Maday et al . , 2012 ) . Recent studies established that nascent autophagosomes in distal axons move exclusively in the retrograde direction toward the soma for lysosomal proteolysis ( Cheng et al . , 2015a , 2015b; Lee et al . , 2011a; Maday and Holzbaur , 2016; Maday et al . , 2012 ) . Such retrograde transport is initiated by fusion of nascent autophagosomes with late endosomes ( LEs ) into amphisomes and is driven by LE-loaded dynein-Snapin ( motor-adaptor ) complexes ( Cheng et al . , 2015a ) . The unique accumulation of immature AVs in the dystrophic neurites of AD brains therefore raises the fundamental question of whether the AV transport and maturation events are affected in AD . Axonal transport defects have been implicated in AD ( Pigino et al . , 2009 , 2003; Stokin et al . , 2005; Tang et al . , 2012 ) . Aβ has been shown to interfere with axonal transport ( Decker et al . , 2010; Hiruma et al . , 2003; Pigino et al . , 2009; Rui et al . , 2006; Tang et al . , 2012; Vossel et al . , 2010 ) . Many studies have been focusing on the mechanisms underlying interruption of kinesin-mediated anterograde transport by Aβ , raising the fundamental questions as to whether dynein-mediated retrograde transport is also impaired in AD neurons and , if so , whether such transport defects contribute to AD-associated autophagic stress . Intracellular Aβ accumulation is closely correlated with the progression of AD in the early stages of AD ( LaFerla et al . , 2007; Li et al . , 2007 ) . Aβ was shown to be generated in the ER and Golgi and also trafficked into the cytosol via the endocytic pathway or passive transport , leading to the accumulation of intracellular Aβ ( Gouras et al . , 2005; LaFerla et al . , 2007 ) . Intracellular Aβ is enriched in both AD human brains and AD mouse models in association with dystrophic neurites and abnormal synaptic morphology ( LaFerla et al . , 2007; Spires-Jones and Hyman , 2014 ) . Intracellular Aβ was proposed to induce presynaptic dysfunction , which might be one of the pathophysiological origins of early AD ( Parodi et al . , 2010; Yang et al . , 2015 ) . Several lines of evidence indicate that Aβ1-42 is associated with LEs or multivesicular bodies ( MVBs ) and AVs in AD brains ( Takahashi et al . , 2004 , 2013 , 2002; Yu et al . , 2005 ) . Given that AD-linked autophagic stress is uniquely associated with Aβ generation and amyloid pathology ( Benzing et al . , 1993; Nixon , 2007 ) , this raises an important question as to whether intracellular Aβ accumulation augments AV retention in distal AD axons and impairs autophagic clearance . In the current study , we provide new evidence that AVs aberrantly accumulate in distal axons and at the presynaptic terminals of mutant hAPP Tg mice and AD patient brains . Amphisomes , rather than autophagosomes are predominantly retained within axons . We demonstrate impaired retrograde transport of amphisomes in live AD axons . Soluble Aβ42 oligomers are enriched in the distal axons of AD mouse brains accompanied by the accumulation of amphisomes . Moreover , we reveal for the first time the molecular interruption leading to such transport defects in AD neurons: direct interaction of oligomeric Aβ1-42 with dynein intermediate chain ( DIC ) disrupts the coupling of dynein-Snapin , a motor-adaptor complex essential for recruiting dynein transport machinery to LEs and amphisomes . This mechanism is further confirmed in Snapin KO mice . Snapin deficiency impedes the removal of AVs from distal axons and synapses and recapitulates AD-associated autophagic stress . More importantly , overexpression Snapin in mutant hAPP Tg neurons reduces autophagic retention in distal axons and presynaptic terminals by enhancing their retrograde transport . Snapin mutant defective in DIC-binding fails to rescue autophagic stress in AD axons , thus supporting our conclusion that defective retrograde transport is one of main mechanisms underlying the AD-linked autophagic stress . Thus , our study provides new mechanistic insights into how Aβ impairs dynein-mediated retrograde transport of LEs and amphisomes , thus leading to autophagic pathology in AD axons . Our study also establishes a foundation for future investigation into regulation of dynein-Snapin coupling to attenuate autophagic defects in AD brains .
To determine whether autophagy is altered in AD neurons , we first examined the hippocampi of both wild-type ( WT ) and hAPP transgenic ( Tg ) mice harboring the human AD Swedish and Indiana mutations ( Camk2α-tTA X tet-APPswe/ind ) ( Jankowsky et al . , 2005 ) . In WT mouse brains , the autophagic marker LC3 appeared as a diffused pattern in the hippocampal mossy fiber processes . However , in mutant hAPP Tg mouse brains , a majority of LC3 associated with vesicular structures reflecting clustered autophagic vacuoles ( AVs ) . The average number of LC3-labeled AV clusters per slice section was substantially increased relative to that of WT mouse brains ( WT: 7 . 0 ± 1 . 028; mutant hAPP Tg: 53 . 35 ± 3 . 78; p<1×10−10 ) ( Figure 1A , B ) . Given that the hippocampal mossy fibers are composed of axons and presynaptic terminals from granule cells in the dentate gyrus , this observation suggests aberrant accumulation of AVs in the axons and axonal terminals of mutant hAPP Tg mouse brains . To further test this possibility , we performed additional line of experiments and showed that AVs accumulated in the distal axons surrounding amyloid plaques ( Figure 1—figure supplement 1A , B ) . While 21 . 05% of LC3-marked AVs localized to MAP2-labeled dendrites , 83 . 11% and 91 . 64% of AVs co-localized with the presynaptic marker synaptophysin and along Neurofilament ( NF ) -labeled axons , suggesting that autophagic stress occurs predominantly in the axons and presynaptic terminals of AD mouse brains ( Figure 1—figure supplement 1A , B ) . 10 . 7554/eLife . 21776 . 003Figure 1 . Autophagic accumulation in the distal axons of mutant hAPP Tg mice . ( A and B ) Representative images ( A ) and quantitative analysis ( B ) showing accumulation of LC3-labeled autophagic vacuoles ( AVs ) in the hippocampal mossy fibers of eight-month mutant hAPP Tg mice . ( C and D ) Quantitative analysis ( C ) and representative images ( D ) showing amphisome retention in the hippocampal mossy fibers of hAPP mice . Note that LC3-labeled AV clusters were co-localized with cation-independent mannose 6-phosphate receptor ( CI-MPR ) , a late endosome ( LE ) marker , suggesting that those AVs are amphisomes in nature following fusion with LEs . ( E and F ) Representative TEM images ( E ) and quantitative analysis ( F ) showing abnormal retention of AVd-like organelles within enlarged neurites in the hippocampal regions of mutant hAPP Tg mouse brains . Note that dystrophic/swollen neurites contained predominantly AVd-like structures marked by arrows , which was not readily observed in wild-type ( WT ) mice . The average number of AVd per EM field was quantified . ( G and H ) Quantitative analysis ( G ) and representative TEM images ( H ) showing aberrant accumulation of AVd-like structures ( black arrows ) at presynaptic terminals in hAPP mice . AVd-like structures , indicated by arrows , were not readily observed in WT mouse brains . Percentage of presynaptic terminals containing AVd was quantified . ( I and J ) Abnormal synaptic retention of LC3-II and p62 ( autophagy markers ) , APP , and Aβ in mutant hAPP Tg mouse brains . Equal amounts ( 15 μg ) of synapse-enriched synaptosomal preparations ( Syn ) and post-nuclear supernatants ( PNS ) from WT and hAPP mice were sequentially immunoblotted on the same membrane after stripping between each antibody application . The purity of synaptosomal fractions was confirmed by the absence of EEA1 and GAPDH . The synaptosome/PNS ratio in AD mice was compared to those in WT littermates . Data were quantified from three independent repeats . Stx1: syntaxin 1; SYP: synaptophysin Scale bars: 25 μm ( A and D ) , 100 nm ( E ) , and 200 nm ( H ) . Data were quantified from a total number of imaging slice sections indicated on the top of bars ( B and C ) from three pairs of mice . The average numbers of AV clusters per section ( 320 μm × 320 μm ) and per EM field ( 10 μm × 10 μm ) were quantified ( B and C ) . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 00310 . 7554/eLife . 21776 . 004Figure 1—figure supplement 1 . Axonal autophagic stress in mutant hAPP Tg mouse brains . ( A and B ) Representative images ( A ) and quantitative analysis ( B ) showing that AVs predominantly accumulate in Neurofilament ( NF ) -labeled axons and synaptophagysin ( SYP ) -marked presynaptic terminals surrounding amyloid plaques in the hippocampal regions of mutant hAPP Tg mice . The percentage of LC3-labeled AV clusters co-localization with MAP2 , NF , and SYP was quantified , respectively ( B ) . ( C–E ) Representative images ( C and D ) and quantitative analysis ( E ) showing aberrant accumulation of LC3-marked amphisomes co-labeled by antibodies against Ubiquitin , p62 , or CI-MPR in the hippocampal mossy fibers and within dystrophic axons around amyloid plaques of mutant hAPP Tg mice . The percentage of LC3-labeled amphisomes co-localization with Ubiquitin , p62 , and CI-MPR was quantified , respectively ( E ) . ( F ) The frequency of AVs per EM field in the hippocampal regions of WT and hAPP mice . Data were quantified from a total number of imaging slice sections ( 320 μm × 320 μm ) indicated on the top of bars ( B and E ) , or a total number of EM fields ( 10 μm × 10 μm ) in parentheses ( F ) . Scale bars: 10 μm ( A ) and 25 μm ( C and D ) . Error bars: SEM . Student's t test: ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 004 To determine those AVs as autophagosomes or amphisomes following fusion with late endocytic organelles , we next performed co-immunstaining with an antibody against cation-independent mannose 6-phosphate receptor ( CI-MPR ) , a membrane protein preferentially located in late endosomes ( LEs ) ( Griffiths et al . , 1988 ) . Consistent with previous studies ( Takahashi et al . , 2004 , 2002; Ye and Cai , 2014 ) , CI-MPR-labeled late endocytic organelles abnormally accumulated along the neuronal processes of AD mouse brains ( Figure 1D ) . Surprisingly , the majority of LC3-labeled AVs co-localized with LEs in the hippocampal mossy fibers of AD mouse brains ( Figure 1D ) , suggesting those AVs as amphisomes in nature following fusion with LEs . The average number of amphisomes labeled by both LC3 and CI-MPR per slice section was significantly increased compared to WT mouse brains ( WT: 7 . 28 ± 0 . 62; mutant hAPP Tg: 45 . 41 ± 2 . 75; p<1×10−12 ) ( Figure 1C ) . Moreover , a significant number of LC3 clusters co-labeled with Ubiquitin , p62 , or CI-MPR were retained in the hippocampal mossy fibers and within swollen/dystrophic axons surrounding amyloid plaques ( Ubiquitin: 97 . 89% ± 0 . 23%; p62: 96 . 29% ± 0 . 43%; CI-MPR: 93 . 68% ± 0 . 51% ) ( Figure 1—figure supplement 1C–E ) . Our data suggests predominant accumulation of amphisomes in the distal axons of AD mouse brains . Using Transmission Electron Microscopy ( TEM ) , we examined AVs at the ultrastructual level based on the established AV morphological features: initial AVs ( AVi ) contain intact cytosol and/or organelles with a sealed double-membrane bilayer separated by an electron-lucent cleft , whereas late-stage degradative AVs ( AVd ) after fusion with late endocytic organelles are those containing small internal vesicles and/or organelles at various stages of degradation , electron-dense amorphous material ( Cheng et al . , 2015a; Klionsky et al . , 2012 ) . The AV-like structures were observed within dystrophic/swollen neurites in mutant hAPP mouse brains: most of those were AVd-like structures that were not readily found in WT mouse brains ( Figure 1E ) . We examined the frequency and average number of AVd-like organelles per EM field within the hippocampal regions of AD mice . EM images containing cell bodies were excluded from analysis . Over 70% of images from WT mice have zero or one AV present , whereas more than 90% of EM images from hAPP Tg mice have at least one AV per field ( Figure 1—figure supplement 1F ) . Moreover , compared to WT ( 1 . 09 ± 0 . 18; n = 35 ) , AD mice also displayed an increased incidence of AVs per EM field ( 5 . 45 ± 1 . 05; n = 40; p=0 . 000632 ) ( Figure 1F ) . This observation is consistent with our immunostaining results ( Figure 1A–D ) , suggesting autophagic accumulation in AD mouse brains . We next assessed AV distribution in the axonal terminals of AD mice . We found a striking number of AVd-like structures were retained within the presynaptic terminals of AD mice ( 53 . 33% ± 8 . 43%; n = 60; p=0 . 0069 ) relative to that of WT controls ( 18 . 33% ± 4 . 01%; n = 58 ) ( Figure 1G , H ) . To confirm these ultrastructural observations , we purified synapse-enriched synaptosomes using Percoll gradient centrifugation as previously described ( DiGiovanni et al . , 2012 ) . AD mouse brains displayed significantly increases of the synaptosomal preparations ( Syn ) /post-nuclear supernatants ( PNS ) ratio in APP ( 1 . 36 ± 0 . 03; p=0 . 007319 ) , p62 ( 1 . 17 ± 0 . 02; p=0 . 021517 ) , and LC3-II ( 4 . 2 ± 0 . 57; p=0 . 011328 ) , but not SYP ( 1 . 02 ± 0 . 07; p=0 . 73211 ) relative to those of WT littermates . The levels of human Aβ detected by 6E10 antibody showed a four-fold increase in synaptosomal fractions compared to that of PNS fractions in AD mice ( Figure 1I , J ) . This result suggested that APP , p62 , LC3-II , and Aβ are relatively enriched in the synaptic terminals of AD mice . Altogether , our TEM and light imaging data combined with biochemical analysis consistently indicate that amphisomes predominantly accumulate in the distal axons of AD mouse brains . Many recent studies demonstrated that autophagosomes are predominantly generated in distal axons , and undergo exclusively retrograde transport following fusion with LEs to form amphisomes for lysosomal proteolysis in the soma ( Cheng et al . , 2015a; Wong and Holzbaur , 2015 ) . We next assessed the distribution of AVs in cultured live cortical neurons from WT and mutant hAPP Tg mice harboring the human AD Swedish and Indiana mutations ( J20 ) ( Mucke et al . , 2000 ) . Neurons were co-transfected with autophagy marker GFP-LC3 and LE marker mRFP-Rab7 , followed by imaging at DIV17-19 . In WT neurons , GFP-LC3 was diffused in the cytoplasm in the form of cytosolic LC3-I , whereas Rab7-labeled LEs appeared as vesicular structures along axonal processes ( Figure 2A ) . To our surprise , GFP-LC3 associated with vesicles as lipidated LC3-II in mutant hAPP Tg neurons at basal condition ( Figure 2B ) . Similar to WT neurons , the majority of autophagosomes in AD neurons co-localized with Rab7-labeled LEs along the axons ( WT: 80 . 64% ± 4 . 36%; hAPP: 86 . 12% ± 2 . 48%; p=0 . 27849 ) ( Figure 2—figure supplement 1A ) , suggesting effective formation of amphisomes by fusion of these two organelles . However , compared to WT neurons , the density of axonal AVs , particularly amphisomes , was robustly increased in AD neurons ( autophagosomes per 100 μm length: WT 0 . 47 ± 0 . 11; hAPP 1 . 52 ± 0 . 39; p=0 . 01306; amphisomes per 100 μm length: WT 2 . 17 ± 0 . 23; hAPP 8 . 97 ± 0 . 68; p<1×10−12 ) ( Figure 2—figure supplement 1A ) . Axonal AV accumulation in hAPP neurons was further confirmed by its negative staining for MAP2 ( Figure 2—figure supplement 1B ) . 10 . 7554/eLife . 21776 . 005Figure 2 . Impaired retrograde transport of axonal amphisomes in mutant hAPP Tg neurons . ( A and B ) Axonal amphisomes predominantly accumulated in cultured cortical neurons derived from mutant hAPP Tg mice . Cortical neurons were co-transfected with GFP-LC3 and mRFP-Rab7 , followed by imaging at DIV17-19 . Images were taken from the distal axons of WT ( A ) and mutant hAPP Tg neurons ( B ) . Late endosomes ( LEs ) are positive for Rab7 alone , whereas amphisomes are positive for both Rab7 and LC3 . Arrow indicates amphisome co-labeled with LC3 and Rab7 . Arrowhead points to AV or LE alone . ( C–E ) Representative TEM images ( C ) and quantitative analysis ( D and E ) showing aberrant accumulation of AVs in neuronal processes and presynaptic terminals of mutant hAPP Tg neurons . TEM showing retention of AVd-like organelles at the axonal terminals of hAPP neurons at DIV18-19 . Arrows indicate AVd-like structures , which were not readily observed in WT neurons . Images were representative from 50–150 electron micrographs of neurons cultured from three pairs of WT and mutant hAPP mice . ( F–H ) Dual-channel kymographs showing impaired retrograde transport of amphisomes in hAPP neurons . Vertical lines represent stationary organelles . Slanted lines or curves to the right ( negative slope ) represent anterograde movement; to the left ( positive slope ) indicate retrograde movement . An organelle was considered stationary if it remained immotile ( displacement ≤5 μm ) . GFP-LC3 was diffused and LEs predominantly moved toward the soma in WT neurons , whereas the majority of amphisomes ( labeled by both LC3 and Rab7 ) remained stationary in the axons of hAPP neurons ( F ) . Note that amphisomes and LEs share similarly reduced retrograde motility in the same axons of hAPP neurons . Relative motility of LC3-labeled AVs in hAPP neurons and LEs in WT neurons and hAPP neurons were examined ( G ) . The average velocity and run length of LE retrograde transport in WT and hAPP neurons were quantified ( H ) . Data were quantified from the total number of vesicles ( v ) in the total number of neurons ( n ) indicated in parentheses from more than four experiments . Scale bars: 100 nm ( C ) , 5 μm ( A and B ) , and 10 μm ( F ) , and . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 00510 . 7554/eLife . 21776 . 006Figure 2—figure supplement 1 . Axonal accumulation of amphisomes containing engulfed ubiquitinated cargoes in mutant hAPP Tg neurons . ( A ) Increased density of axonal amphisomes following fusion with LEs was detected in hAPP neurons . hAPP neurons were comparable to WT neurons in the fusion of autophagosomes with LEs to form axonal amphisomes . ( B ) Aberrant accumulation of AVs in the MAP2-negative axon of mutant hAPP Tg neurons . ( C and D ) Representative blots ( C ) and quantitative analysis ( D ) showing increased levels of APP and C99 in mutant hAPP neurons . A total of 20 μg of cell lysates was sequentially detected on the same membrane . Relative protein levels were normalized by p115 and to that of neurons from WT littermates . ( E–G ) Accumulated AVs contained ubiquitinated cargos along the axons of hAPP neurons . Cortical neurons were co-transfected with GFP-LC3 and mRFP-Ubiquitin ( Ub ) , followed by imaging at DIV17-19 . Images were taken of the distal axons . Note that most of LC3-marked AVs were co-labeled with mRFP-Ub in WT and hAPP axons . Representative kymographs showing lack of movement of GFP-LC3 and mRFP-Ub along the axon of hAPP neurons ( E ) . The percentage of AVs co-labeled by ubiquitin was quantified in WT and hAPP axons ( G ) . ( H ) Rab5-marked early endosomes partially co-localized with AVs within the axon of mutant hAPP neurons . ( I and J ) Representative images and kymographs showing no significant change of early endosome trafficking in mutant hAPP axons . Note that early endosomes moved either a short distance or in an oscillatory pattern along axons of WT and hAPP neurons . Data were quantified from four independent repeats of four pairs of mice , and from the total number of vesicles ( v ) in the total number of axons ( n ) indicated in parentheses ( A , D , G , and J ) . Scale bars: 10 μm . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 006 We next examined cortical neuron ultrastructure at DIV18-19 from WT and mutant hAPP Tg mice . A striking number of AV-like organelles were found along AD neurites ( 1 . 68 ± 0 . 10 per 10 μm length; p<1×10−16 ) , a phenotype rarely detected in WT neurons ( Figure 2C , D ) . Consistently , we demonstrated an increased percentage of presynaptic terminals in AD neurons containing AVs ( 35 . 20% ± 5 . 0%; n = 58 EM fields; p<1×10−6 ) relative to that of WT controls ( 8 . 98% ± 2 . 41%; n = 54 ) ( Figure 2C , E ) . Thus , these axonal imaging data from cultured neurons are consistent with our in vivo evidence from AD mouse brains ( Figure 1 and Figure 1—figure supplement 1 ) , suggesting axonal accumulation of amphisomes . We also showed that the majority of those axonal AVs were co-labeled with mRFP-Ubiquitin ( Ub ) ( hAPP: 91 . 67% ± 1 . 26% from 49 axons; WT: 87 . 5% ± 7 . 14% from 20 axons; p=0 . 6331 ) ( Figure 2—figure supplement 1E–G ) . Thus , consistent with our observations in AD mouse brains ( Figure 1—figure supplement 1C–E ) , this result indicates that these AVs contained engulfed ubiquitinated cargoes . The aberrant AV retention in the distal AD axons may reflect defects in their retrograde transport toward the soma , thus reducing autophagic clearance . We next assessed the retrograde motility of axonal AVs in live mutant hAPP Tg neurons . In WT neurons , while GFP-LC3 was diffuse , a significant portion of Rab7-labled LEs ( 30 . 16% ± 1 . 33% ) moved in the retrograde direction toward the soma along the same axon ( Figure 2F , G ) . However , LEs in AD neurons displayed reduced retrograde motility in distal axons ( 16 . 86% ± 1 . 51%; p<1×10−8 ) . Such reduction was not found in anterograde transport of LEs ( p=0 . 35506 ) ( Figure 2G ) . Strikingly , amphisomes displayed the similar motility pattern: reduced retrograde ( 18 . 04% ± 1 . 74% ) , but not anterograde transport in the same axons of AD neurons ( Figure 2F , G ) . We also quantified the average retrograde velocity and run length of Rab7-marked organelles in WT and hAPP mutant neurons ( Figure 2H ) . Consistent with previous studies ( Castle et al . , 2014; Deinhardt et al . , 2006 ) , the average retrograde velocity and run length of Rab7-associated LEs in WT neurons were 0 . 40 ± 0 . 01 μm/sec and 56 . 97 ± 2 . 02 μm , respectively , which were significantly reduced in AD neurons ( velocity: 0 . 10 ± 0 . 007 μm/sec , p<1×10−14; run length: 17 . 95 ± 1 . 18 μm , p<1×10−12 ) . Altogether , these observations indicate that aberrant accumulation of amphisomes in distal AD axons may result from impaired retrograde transport . Moreover , we showed increased levels of APP ( 2 . 05 ± 0 . 29; p=0 . 015952 ) and C99 ( 6 . 13 ± 1 . 07; p=0 . 008705 ) , but not C83 ( 1 . 18 ± 0 . 10; p=0 . 16055 ) in mutant hAPP Tg neurons relative to those of neurons from WT littermates ( Figure 2—figure supplement 1C , D ) . We also examined the co-localization of LC3 with Rab5-labeled early endosomes in cultured neurons from mutant hAPP Tg mice . We found that about 46% of LC3-labeled AVs co-localized with early endosomes within the axon of mutant hAPP neurons ( 45 . 58% ± 2 . 24%; n = 47 , v = 875 ) ( Figure 2—figure supplement 1H ) . However , Rab5-marked early endosomes moved either a short distance , or in an oscillatory pattern along axons ( Figure 2—figure supplement 1I ) . While our observation is consistent with the results from previous studies ( Cai et al . , 2010; Chen and Sheng , 2013 ) , the motility of axonal early endosomes showed no significant change in mutant hAPP neurons relative to that of WT neurons ( WT: 67 . 53% ± 1 . 97; hAPP: 70 . 93% ± 2 . 31%; p=0 . 268 ) ( Figure 2—figure supplement 1J ) . A recent study reported that nascent AVs gain retrograde transport motility by recruiting LE-loaded dynein-Snapin motor-adaptor complexes after fusion with Rab7-associated LEs to form amphisomes ( Cheng et al . , 2015a , 2015b ) . Thus , our data supports the notion that fusion of AVs with Rab5-endosomes could be a transitional process before they further mature into Rab7-positive amphisomes to gain long-distance retrograde transport motility . Given that AD-associated autophagic stress is uniquely linked to Aβ generation and amyloid pathology ( Benzing et al . , 1993; Nixon , 2007 ) , we next addressed whether Aβ associates with these AVs in the distal axons of AD neurons . We showed that the majority of LC3-labeled AVs co-localized with anti-β amyloid antibody ( 6E10 ) -labeled APP and Aβ within dystrophic neurites , but not with the core of 6E10 antibody-marked fibrillar amyloid plaques ( Figure 3A ) . The co-localization of LC3-labeled AVs with 6E10 within neurites is 74 . 9% ± 2 . 37% in the hippocampal regions of AD mice . It is predictable that not all AVs are associated with APP and Aβ and that some autophagosomes may function in engulfing other autophagic cargos such as dysfunctional organelles and cytosolic components . 10 . 7554/eLife . 21776 . 007Figure 3 . Association of soluble Aβ oligomers with amphisomes in the dystrophic axons of AD mice . ( A ) AVs clustering within swollen/dystrophic neurites around an amyloid plaque enriched with 6E10 antibody-labeled APP , C99 , or Aβ deposits . ( B and C ) Representative images ( B ) and quantitative analysis ( C ) showing that soluble Aβ oligomers labeled by anti-A11 antibody was concentrated within Neurofilament ( NF ) -labeled axons surrounding amyloid plaques in mutant hAPP Tg mice . The percentage of soluble Aβ co-localization with 6E10 antibody-labeled Aβ or NF was quantified , respectively . ( D and E ) Quantitative analysis ( D ) and representative images ( E ) showing the association of soluble Aβ oligomers with amphisomes within dystrophic axons around amyloid plaques in the hippocampal regions of mutant hAPP mice . The percentage of oligomeric Aβ co-localization with CI-MPR , p62 , and Ubiquitin ( Ub ) was quantified , respectively . ( F–I ) Immuno-EM analysis ( I ) and quantification ( F , G , and H ) showing that soluble Aβ oligomers in the cytoplasm , marked by anti-A11 immuno-gold particles ( white arrows ) , associated with or surround AVd-like structures ( black arrows ) within the neurites of cultured mutant hAPP Tg neurons . Note that anti-A11 immuno-gold particles were also present within the AVd-like structures containing organelles and small vesicles along hAPP neurites . Anti-A11 immuno-gold particles were detected in the cytoplasm of neurites in WT neurons . The average numbers of the A11 gold grains per EM field ( 10 μm × 10 μm ) and per 10 μm neurite were quantified in WT and hAPP neurons , respectively . The percentage of AVd-like compartments surrounded by the gold particles was quantified in hAPP neurons or in the absence of the primary antibody . The co-localized pixels of individual markers with Aβ oligomers were indicated in white ( B and E ) . Data were quantified from a total number of imaging slice sections ( 320 μm × 320 μm ) indicated on the top of bars ( C and D ) and from a total number of EM fields ( F ) , neurites ( G ) , or AVs ( H ) indicated in parentheses from more than three experiments . Scale bars: 25 μm ( A ) , 10 μm ( B and E ) , and 200 nm ( I ) . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 00710 . 7554/eLife . 21776 . 008Figure 3—figure supplement 1 . Association of soluble Aβ1-42 oligomers with amphisomes in the distal axons of mutant hAPP Tg mice . ( A and B ) Representative images ( A ) and quantitative analysis ( B ) showing co-localization of amphisomes with soluble Aβ1-42 oligomers in NF-marked distal axons surrounding amyloid plaques in AD mouse brains . The percentage of Aβ1-42 co-localization with 6E10 , neurofilament ( NF ) , CI-MPR , p62 , and Ubiquitin was quantified , respectively . The co-localized pixels of individual markers with Aβ1-42 were indicated in white ( A ) . ( C and D ) Representative images ( C ) and quantitative analysis ( D ) showing accumulation of amphisomes in the NF-marked distal axons of mutant hAPP Tg neurons . The percentage of CI-MPR and Ub co-localization with NF was quantified , respectively . Scale bars: 10 μm . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 008 Next , to examine the association of Aβ oligomers with AVs in AD brains , we utilized the well-characterized A11 antibody that only recognizes soluble Aβ oligomers , but not APP and its cleaved products C99 , soluble monomer , or insoluble fibrils ( Jimenez et al . , 2008 , 2011; Kayed et al . , 2003; Zempel et al . , 2010 ) . We found that the A11 antibody preferentially labeled soluble Aβ oligomers within neurites at the periphery of amyloid plaques indicated by 6E10 ( Figure 3B ) . This observation suggests that A11-marked soluble Aβ oligomers surrounds fibrillar plaque cores ( Spires-Jones and Hyman , 2014 ) . Oligomeric Aβ is more enriched in axons and associated with amphisomes within the dystrophic axons of AD mice ( Figure 3B–E ) . The percentage of A11-labeled oligomeric Aβ co-localization is 91 . 0% ± 1 . 21% with 6E10 , 89 . 5% ± 3 . 84% with NF , 88 . 0% ± 1 . 33% with CI-MPR , 90 . 0% ± 1 . 18% with p62 , and 89 . 0% ± 1 . 16% with Ubiquitin , respectively . Amphisomes are also concentrated in NF-labeled distal axons ( the percentage of co-localization with NF: 82 . 61% ± 0 . 86% ( CI-MPR ) ; 80 . 74% ± 0 . 72% ( Ubiquitin ) ) ( Figure 3—figure supplement 1C , D ) . Using another well-characterized antibody ( AB5078P ) recognizing soluble Aβ1-42 oligomers , but not Aβ1-40 or high molecular weight insoluble forms of Aβ1-42 ( Agholme et al . , 2012; Kamal et al . , 2001; Muresan et al . , 2009; Takahashi et al . , 2013 ) , we observed similar results: the co-localization of oligomeric Aβ1-42 is 92 . 5% ± 1 . 66% with 6E10 , 90 . 4% ± 1 . 21% with CI-MPR , 88 . 9% ± 1 . 79% with p62 , and 93 . 2% ± 1 . 26% with Ubiquitin , respectively ( Figure 3—figure supplement 1A , B ) . 91 . 86% ± 0 . 38% of Aβ1-42 co-localized with NF-labeled distal axons surrounding amyloid plaques ( Figure 3—figure supplement 1A , B ) . To further confirm these imaging results at the ultrastructural level , we performed immuno-EM analysis in cultured WT and mutant hAPP Tg neurons using the A11 antibody that detect soluble Aβ oligomers . Consistent with a previous study ( Diomede et al . , 2014 ) , we found that the immuno-gold anti-A11 antibody-labeled oligomeric Aβ was mostly present in the cytoplasm of WT , a pattern similarly found in hAPP neurons ( Figure 3I ) . However , the average numbers of anti-A11 immuno-gold grains per μm2 EM field and per 10 μm neurites were markedly increased in AD neurons relative to those of WT littermate controls ( EM field: WT: 0 . 53 ± 0 . 06; hAPP: 6 . 71 ± 0 . 86; p<1×10−8; Neurite with gold grains: WT: 9 . 19 ± 0 . 91; hAPP: 94 . 54 ± 10 . 07; p<1×10−10 ) ( Figure 3F , G ) , suggesting that Aβ oligomers is enriched in the cytoplasm of AD axons . Moreover , ~57 . 3% of AVd-like structures were associated with or surrounded by oligomeric Aβ gold particles in the cytoplasm of hAPP axons ( Figure 3H ) . We also detected the presence of anti-A11 immuno-gold within AVd-like structures . Altogether , our data consistently indicate amphisome-associated Aβ1-42 oligomers in distal axons of AD neurons . Dynein is the primary motor that drives retrograde transport of both LEs and AVs from distal axons to the soma ( Cai et al . , 2010; Lee et al . , 2011a; Maday et al . , 2012 ) . Our previous study revealed that Snapin serves as an adaptor for the recruitment of dynein motors to LEs by binding to DIC ( Cai et al . , 2010 ) . Disrupting DIC-Snapin coupling impairs LE retrograde transport . We recently established that autophagosomes acquire their retrograde motility by recruiting LE-loaded dynein-Snapin ( motor-adaptor complex ) upon fusion of these two organelles ( Cheng et al . , 2015a ) . We next asked whether reduced AV retrograde transport is caused by impaired dynein motor recruitment . By immunoisolation of LEs/amphisomes using anti-Rab7-coated magnetic beads , we detected reduced attachment of dynein DIC onto the purified LEs/amphisomes in mutant hAPP mouse brains relative to WT littermates ( 0 . 27 ± 0 . 02; p=0 . 000686 ) ( Figure 4A , B ) . Conversely , there was a significant increase in LC3-II levels ( p=0 . 01061 ) along with enhanced hAPP in the purified Rab7-associated organelles of mutant hAPP mouse neurons ( Figure 4A , B ) , suggesting aberrant amphisomal retention . As controls , similar amounts of Rab7 and Snapin were associated with LEs from WT and mutant hAPP Tg mouse brains . Our study suggests that reduced attachment of dynein motors impairs AV retrograde transport of amphisomes toward the soma , thus leading to AV retention in distal AD axons . 10 . 7554/eLife . 21776 . 009Figure 4 . Oligomeric Aβ42-mediated interruption of dynein-Snapin coupling and recruitment of dynein motors to amphisomes . ( A and B ) Immunoisolation assays ( A ) and quantitative analysis ( B ) from three repeats showing reduced dynein attachment to amphisomes in sixteen-month mutant hAPP Tg mouse brains . Rab7-associated organelles were immunoisolated from light membrane fractions , followed by sequential immunoblotting on the same membranes with antibodies against the dynein intermediate chain ( DIC ) , LC3 , hAPP , Snapin , Rab7 , and EEA1 . Note that AD mouse brains exhibited reduced DIC and increased LC3-II levels in the purified amphisomal organelles . Data were quantified from three independent repeats of three pairs of mice . ( C ) Immunoprecipitation showing reduced Snapin-DIC coupling in COS7 cells expressing mutant hAPPswe , but not WT hAPP . ( D ) Direct interaction of Aβ1-42 with GST-DIC , but not GST-Snapin or GST . 6E10 antibody was used to detect Aβ . ( E and F ) GST-DIC specifically interacts with Aβ1-42 , but not Aβ1-40 or scrambled Aβ . ( G ) Aβ1-42 interaction with DIC was increased as Aβ concentration was elevated from 0 . 5 μM to 4 μM in the presence of the same amount of GST-DIC . ( H ) Aβ1-42 interferes with DIC-Snapin coupling in a dose-dependent manner . Note that the DIC-Snapin interaction was competitively interrupted in the presence of as low as 0 . 2 μM Aβ1-42 when the same amount of Snapin and DIC was used . ( I ) The dot blot using anti-β amyloid antibody showed that GST-DIC bound to oligomeric Aβ . ( J ) Oligomeric Aβ1-42 interrupts dynein-Snapin coupling and the recruitment of dynein motors to LEs and amphisomes in mouse brains . Membranous organelles were pulled down from light membrane fractions ( LMF ) of mouse brains by GST-DIC in the presence or absence of 2 μM oligomeric Aβ1-42 . Bead-bound membrane organelles were resolved by PAGE and sequentially detected with antibodies on the same membranes after stripping between applications of each antibody . Note that reduced tethering of dynein motors to LAMP-1 or Rab7-associated LEs and amphisomes in the presence of Aβ was specifically caused by impaired DIC-Snapin interaction because the dynein-dynactin ( p150Glued ) complex was not affected . The attachment of dynein motors to mitochondria showed no detectable change . TOM20: a mitochondrial outer membrane protein . The purity of the preparation pulled down by GST-DIC beads was confirmed by the absence of GAPDH . Results were representative from three independent repeats . ( K ) The DIC-Aβ complex was immunoprecipitated by anti-DIC antibody from mutant hAPP Tg mouse brains . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 009 We next examined whether Aβ overproduction interrupts dynein-Snapin coupling , and thus impedes AV transport in AD neurons . By co-immunoprecipitation assays we observed that less dynein DIC-Snapin complexes were detected in cells expressing mutant hAPPswe relative to WT hAPP ( Figure 4C ) , which suggests that Aβ overproduction interferes with the assembly of the DIC-Snapin complex . Next , we sought to ask whether Aβ interacts with dynein DIC or Snapin , thus competitively interrupting the dynein-Snapin coupling . To address this question , we prepared soluble Aβ1-42 oligomers and performed five lines of experiments . First , direct in vitro binding of Aβ1-42 to GST-DIC , but not GST-Snapin or GST , was detected ( Figure 4D ) . GST-DIC specifically interacts with Aβ1-42 , but not Aβ1-40 or scrambled Aβ ( Figure 4E , F ) . Second , Aβ1-42 interaction with DIC was increased as Aβ concentration was elevated from 0 . 5 μM to 4 μM in the presence of the same amount of GST-DIC ( Figure 4G ) . Third , as low as 0 . 2 μM Aβ1-42 is effective in interference with the assembly of DIC-Snapin complex; this interruption is in a dose-dependent manner when the same amount of Snapin and DIC was used ( Figure 4H ) . Fourth , DIC bound to oligomeric Aβ1-42 ( Figure 4I ) , the toxic form of Aβ linked to AD pathogenesis . Fifth , oligomeric Aβ1-42 interfere with DIC interaction with LE/amphisome-associated Snapin , thus impairing the recruitment of dynein motors to LEs and amphisomes prepared from mouse brains ( Figure 4J ) . Sixth , we demonstrated a DIC-Aβ complex in mutant hAPP Tg mouse brains by immunoprecipitation ( Figure 4K ) . Thus , our findings suggest that dynein-Snapin coupling , and thus cargo-motor association , is compromised by accumulation of cytoplasmic Aβ1-42 oligomers , which could be a pathogenic mechanism of impaired dynein-driven axonal transport in AD . AD patient brains display unique autophagic stress characterized by massive AV accumulation within large swellings along dystrophic neurites ( Nixon et al . , 2005 ) , so we next examined whether AVs abnormally accumulate at nerve terminals in AD patient brains . Strikingly , LC3-II levels were robustly increased in synapse-enriched synaptosomal preparations from AD patient brains relative to those of control subjects ( 8 . 05 ± 1 . 51; p=0 . 009418 ) ( Figure 5A , B ) . The results were quantified from the experiments using four control subjects and four patient brains ( postmortem interval 7 . 08 hr – 22 . 5 hr ) ( Table 1 ) . Consistently , aberrant clustering of AVd-like structures was detected within enlarged dystrophic neurites in AD patient brains ( Figure 5C , D ) , a phenotype not readily found in control subjects . We quantified the average number of AVd-like organelles per EM field in the brains of two age-matched controls and three AD patients at different Braak stages with postmortem interval between 7 . 08 hr and 12 . 5 hr ( Table 1 ) . Compared to control subjects ( 1 . 54 ± 0 . 16; n = 76 EM fields ) , AD brains exhibited an increased number of AVd-like structures ( 8 . 67 ± 0 . 79; n = 91; p<1×10−14 ) ( Figure 5D ) . Moreover , immunoprecipitation analysis showed a marked reduction of dynein-Snapin complexes in AD patient brains ( p<1×10−6 ) ( Figure 5E , F ) ( Table 1 ) . These in vivo observations in AD patient brains further confirm that the interruption of the dynein-Snapin coupling impairs the removal of AVs from distal neurites and synaptic terminals , thus reducing AV clearance by lysosomes in the soma . 10 . 7554/eLife . 21776 . 010Figure 5 . Impaired dynein-Snapin coupling contributes to axonal autophagic stress in AD patient brains . ( A and B ) Synaptic autophagic stress in AD patient brains . Equal amounts ( 15 μg ) of synapse-enriched synaptosomal preparations ( Syn ) and post-nuclear supernatant ( PNS ) from human brains of control subjects and AD patients were sequentially immunoblotted on the same membrane after stripping between each antibody application . The purity of synaptosome fractions was confirmed by their relative enrichment of synaptic markers synaptophysin ( SYP ) and PSD95 compared to levels in PNS fractions , and by the absence of EEA1 . The synaptosome/PNS ratio in AD brains were compared to those in control subjects . Data were quantified from four independent repeats . ( C and D ) Representative TEM images ( C ) and quantitative analysis ( D ) showing abnormal retention of AVd-like organelles ( arrows ) within enlarged neurites in patient brains . Note that dystrophic/swollen neurites contain predominantly late stage AVs ( AVd ) . ( E and F ) Immunoprecipitation ( E ) and quantitative analysis ( F ) showing reduced Snapin-DIC coupling in AD patient brains . Data were quantified from three independent experiments . The average number of AV-like structures per EM field ( 10 μm × 10 μm ) was quantified ( D ) . Scale bars: 500 nm . Error bars: SEM . Student's t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 01010 . 7554/eLife . 21776 . 011Table 1 . Demographic details of postmortem brain specimens from patients with AD and subjects without AD ( specimens from the Harvard Tissue Resource Center and the Human Brain and Spinal Fluid Resource Center at UCLA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 011Case typeAge/sexPostmortem interval ( h ) Braak stage of AD brainsControl75/F20 . 10Control87/M9 . 30Control47/M12 . 50Control66/M22 . 50AD65/M11 . 6Braak IAD72/M21 . 8Braak IIAD86/M9 . 00Braak IIIAD86/M17 . 4Braak IIIAD60/F15 . 2Braak VAD86/F7 . 08Braak VI Given the observations of dynein-Snapin coupling deficits in AD patient brains , we next asked whether deleting Snapin in mice displays autophagic phenotypes similar to those of AD brains . To address this issue , we performed four lines of experiments using Snapin flox/flox conditional knockout ( cKO ) mice , in which the Snapin gene was deleted in the frontal cortex and hippocampus by Cre expression ( Cheng et al . , 2015a; Ye and Cai , 2014 ) . First , we examined the distribution pattern of CI-MPR-labeled LEs in the hippocampal CA3 regions . Deletion of Snapin leads to LE clustering in the hippocampal mossy fibers composed of axons and presynaptic terminals from granule cells in the dentate gyrus ( Figure 6A ) . The majority of these LE clusters were not distributed in the MAP2-labeled dendrites in the hippocampal regions of Snapin cKO mice . Co-localized pixels of CI-MPR with MAP2 in Snapin cKO mice were similar to those of WT littermates ( WT: 10 . 06 ± 2 . 09; Snapin cKO: 11 . 90 ± 1 . 17; p=0 . 45032 ) , suggesting that Snapin deficiency results in predominant accumulation of LEs within axons negative for MAP2 ( Figure 6—figure supplement 1A , B ) . Compared with the WT control , the mean intensity of CI-MPR fluorescence is significantly increased in Snapin cKO mouse brains ( 2 . 92 ± 0 . 12; p<1×10−16 ) ( Figure 6B ) . Consistent with our previous study using cultured neurons ( Cai et al . , 2010 ) , abnormal retention of immature lysosomes labeled by CI-MPR was also shown in the soma of the CA3 region after deletion of Snapin in mice ( Figure 6A ) . Second , we asked whether Snapin deficiency results in retention of amphisomes in distal regions . We detected a significant number of AVs co-labeled with both LC3 and CI-MPR , suggesting that they had the nature of amphisomes , the late stage of AVs after fusion with LEs ( Figure 6C ) . The LC3-labeled AVs clustered in the hippocampal mossy fibers of Snapin mutant mice ( WT: 7 . 09 ± 1 . 1; Snapin cKO: 68 . 44 ± 5 . 43; p<1×10−10 ) ( Figure 6D ) . 10 . 7554/eLife . 21776 . 012Figure 6 . Snapin-deficient mouse brains recapitulate AD-associated autophagic stress in axons . ( A and B ) LEs clustering ( arrows ) in the hippocampal mossy fibers of Snapin flox/flox conditional knockout ( cKO ) mice . The mean intensity of LE clusters in the mossy fibers ( mf ) of one-month Snapin mutant mice labeled with CI-MPR per section ( 320 μm × 320 μm ) was quantified and compared with that of WT mice . s . p . , stratum pyramidale ( C and D ) Aberrant accumulation of amphisomes in the mossy fibers of Snapin-deficient mice . Note that LC3-marked AVs were labeled with CI-MPR , suggesting that they were amphisomes in nature following fusion with LEs . The number of LC3 clusters per section ( 320 μm × 320 μm ) was quantified . ( E and F ) Immunoisolation showing reduced dynein attachment to amphisomes . Rab7-associated organelles were immunoisolated with anti-Rab7-coated Dyna magnetic beads , followed by sequential immunoblotting on the same membranes after stripping between each antibody application . Note that purified Rab7 organelles were enriched with various AV markers including LC3-II , p62 , and syntaxin 17 ( Stx17 ) in Snapin cKO mouse brains . Data were quantified from four repeats . ( G and H ) Representative TEM images ( G ) and quantitative analysis ( H ) showing retention of AVd-like organelles at presynaptic terminals in Snapin mutant mice . Arrows indicate AVd-like structures , which were not readily detected in WT mice . Scale bars: 25 μm ( A and C ) and 500 nm ( G ) . Data were quantified from a total number of imaging slice sections ( B and D ) or from a total number of electron micrographs ( H ) indicated in parentheses from three pairs of mice . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 01210 . 7554/eLife . 21776 . 013Figure 6—figure supplement 1 . Axonal autophagic stress in the hippocampal regions of Snapin-deficient mice . ( A and B ) Representative images ( A ) and quantitative analysis ( B ) showing that majority of CI-MPR-labeled LEs were not co-localized with MAP2-marked dendrites in Snapin mutant mouse brains . The percentage of co-localized pixels of CI-MPR with MAP2 ( in white ) was quantified . Data were quantified from a total number of imaging slice sections ( 320 μm × 320 μm ) indicated in parentheses . ( C and D ) Representative TEM images ( C ) and quantitative analysis ( D ) showing massive accumulation of AVd-like organelles ( arrows ) within large swellings along axonal processes surrounded by myelin in Snapin mutant mouse brains . The average number of AVd per EM field was quantified . Data were quantified from a total number of EM fields ( 10 μm × 10 μm ) indicated in parentheses . Scale bars: 25 μm ( A ) and 500 nm ( C ) . Error bars represent SEM . Student’s t test: ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 013 Third , we examined the recruitment of dynein motors to LEs/amphisomes by immunoisolation using Dyna magnetic beads coated with an anti-Rab7 antibody . When equal amounts of LEs/amphisomes--as reflected by Rab7 levels--were loaded , the normalized intensity of the dynein DIC in Snapin cKO mouse brains was significantly reduced to ~55% in comparison with that of WT littermates ( p=0 . 003992 ) ( Figure 6E , F ) , indicating a reduced loading of the dynein motors onto LEs/amphisomes . The significantly reduced but not fully abolished DIC recruitment in the Snapin cKO mouse brains may suggest ( 1 ) a compensatory role of other dynein adaptors in LE-dynein coupling , or ( 2 ) the remaining Snapin expressed in other types of cells in mouse brains . Interestingly , from the purified LEs in Snapin cKO mouse brains , we also detected increased LC3-II , and syntaxin 17 ( Stx17 ) ( LC3-II: p=0 . 0014707; Stx17: p=0 . 013641 ) ( Figure 6E , F ) , an autophagosome-targeted protein mediating the fusion with late endosomes/lysosomes by forming the SNARE fusion complex with SNAP29 and VAMP8 ( Cheng et al . , 2015a; Guo et al . , 2014; Itakura et al . , 2012; Wang et al . , 2016 ) . This study further confirms that Snapin is required for dynein motor recruitment to amphisomes , and the subsequent removal of AVs from distal axons and synapses . In addition , we performed TEM analysis to assess AV accumulation in presynaptic terminals of WT and Snapin cKO mice . Consistent with the results from immunostaining and immunoisolation assays , Snapin cKO mice exhibited a significant number of AVd-like structures at presynaptic terminals ( Figure 6G ) . These AV-like organelles were not readily observed in WT synapses ( WT: 0 . 29 ± 0 . 11; cKO: 1 . 19 ± 0 . 18; p<1×10−5 ) ( Figure 6G , H ) . Moreover , massive AV accumulation within dystrophic axonal processes was also detected in Snapin cKO mouse brains . Compared to WT controls , the average number of AVd-like organelles per EM field within the hippocampal regions of Snapin cKO mice was significantly increased ( WT: 0 . 71 ± 0 . 10; n = 79; Snapin cKO: 5 . 46 ± 1 . 0; n = 91; p<1×10−6 ) ( Figure 6—figure supplement 1C , D ) . This observation is consistent with our immunostaining results ( Figure 6C , D ) , suggesting autophagic stress in distal axons of Snapin cKO mouse brains . Thus , these morphological observations confirm that Snapin mediates recruitment of dynein motors to AVs for retrograde transport; deleting Snapin recapitulates AD-associated autophagic stress in axons . Because Snapin deficiency leads to AD-like autophagic phenotypes , we sought to reverse autophagic stress by elevating Snapin expression in AD neurons . First , we assessed AV retention in AD axons . We observed reduced density of axonal AVs in hAPP neurons expressing HA-Snapin , but not the HA-Snapin-L99K mutant defective in DIC binding , or HA vector control ( amphisomes per 100 μm length: 4 . 57 ± 0 . 34 for HA-Snapin , p<1×10−7; 8 . 92 ± 0 . 7 for HA-Snapin-L99K , p=0 . 50207; 8 . 30 ± 0 . 56 for HA vector ) ( Figure 7A , B ) . As expected , expressing HA-Snapin or HA-Snapin-L99K in mutant hAPP Tg neurons did not show significant change in the percentage of amphisomes , eliminating the possibility for defective fusion of autophagosomes with LEs by expressing Snapin ( Figure 7—figure supplement 1A ) . Second , we monitored the motility of LEs and amphisomes along axonal processes . Mutant hAPP Tg neurons expressing HA-Snapin exhibited enhanced retrograde transport of both amphisomes ( 42 . 56% ± 3 . 04%; p<1×10−8 ) and LEs ( 36 . 36% ± 2 . 53%; p<1×10−8 ) along the same axons relative to control hAPP neurons expressing HA-Snapin-L99K ( 19 . 82% ± 2 . 97% for amphisomes; 18 . 35% ± 2 . 07% for LEs ) or HA vector ( 17 . 78% ± 1 . 64% for amphisomes; 16 . 42% ± 1 . 40% for LEs ) , resulting in reduced stationary pools of both organelles ( Figure 7C–E ) . 10 . 7554/eLife . 21776 . 014Figure 7 . Elevated Snapin expression reduces axonal autophagic stress of mutant hAPP Tg neurons . ( A and B ) Images ( A ) and quantitative analysis ( B ) showing that expressing Snapin , but not the Snapin-L99K mutant , reduces the density of axonal amphisomes in mutant hAPP Tg neurons . ( C–E ) Kymographs ( C ) and quantitative analysis ( D and E ) showing that impaired amphisome retrograde transport was rescued by expressing Snapin , but not Snapin-L99K . hAPP neurons were co-transfected with GFP-LC3 and mRFP-Rab7 along with HA-Snapin , HA-Snapin-L99K , or HA vector , followed by time-lapse imaging at DIV17-19 . ( F and G ) Representative images ( F ) and quantitative analysis ( G ) showing reduced autophagic accumulation at presynaptic terminals after overexpression of Snapin , but not Snapin-L99K in hAPP neurons . Scale bars: 5 μm ( A ) and 10 μm ( C and F ) . Data were quantified from a total number of neurons ( n ) indicated on the top of bars ( B and G ) or in parentheses ( D and E ) from more than four independent experiments . Error bars: SEM . Student's t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 01410 . 7554/eLife . 21776 . 015Figure 7—figure supplement 1 . Snapin-mediated rescue effects on axonal autophagic stress under conditions of autophagy induction in AD neurons . ( A ) Expression of Snapin or Snapin-L99K had no significant effect on the fusion of autophagosomes with LEs to form amphisomes . Data were quantified from a total number of vesicles ( v ) and a total number of neurons ( n ) indicated in parentheses ( A ) from more than four independent experiments . ( B and C ) Representative images ( B ) and quantitative analysis ( C ) showing that elevated Snapin expression reduced AV retention within hAPP axons in response to autophagy induction . Mutant hAPP Tg neurons were transfected with GFP-LC3 and HA-Snapin or HA vector , followed by treatment with 5 μM 10-NCP 24 hr before imaging at DIV18-20 . ( D and E ) Quantitative analysis ( E ) and representative kymographs ( D ) showing that retrograde transport of AVs was enhanced in mutant hAPP axons expressing Snapin , but not vector control following 10-NCP treatment . Scale bars: 10 μm . Error bars represent SEM . Student’s t test: ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21776 . 015 Presynaptic terminals of mutant hAPP Tg neurons exhibit AV retention ( Figure 2C , E ) . We found that presynaptic autophagic accumulation was significantly reduced following overexpression of Snapin , but not Snapin-L99K in hAPP neurons ( Vector: 46 . 69% ± 2 . 46%; Snapin: 18 . 20% ± 1 . 44% , p<1×10−14; Snapin-L99K: 51 . 41% ± 2 . 16% ) ( Figure 7F , G ) . We next determined whether impaired retrograde transport contributes to autophagic stress in AD axons under conditions of autophagy induction . By utilizing 10-[4′- ( N-diethylamino ) butyl]-2-chlorophenoxazine ( 10-NCP ) , which induces neuronal autophagy in an mTOR-independent fashion ( Tsvetkov et al . , 2010 ) , we examined whether Snapin-mediated enhancement of AV transport rescues autophagic stress upon autophagy induction . Axonal AV density was increased in hAPP neurons treated with 10-NCP relative to untreated controls ( AVs per 100 μm length: hAPP 8 . 87 ± 0 . 4; hAPP with 10-NCP 21 . 26 ± 1 . 12; p<1×10−12 ) ( Figure 7—figure supplement 1B , C ) . 10-NCP-induced autophagy did not lead to the increase in the axonal AV density of AD neurons overexpressing Snapin ( AVs per 100 μm length: 4 . 71 ± 0 . 31; p<1×10−11 ) as a result of enhanced retrograde transport ( 10-NCP treated hAPP neurons: 11 . 3% ± 1 . 19%; hAPP neurons expressing Snapin: 43 . 57% ± 3 . 4% , p<1×10−8 ) ( Figure 7—figure supplement 1D , E ) . These results suggest that enhanced retrograde transport of AVs by elevated Snapin expression efficiently removes AVs from distal axons and presynaptic terminals , thus reducing autophagic stress in AD axons . Altogether , our study provides new mechanistic insights into AD-associated axonal autophagic stress , establishing a foundation for ameliorating axonal pathology in AD .
Defective autophagy has been implicated in AD pathogenesis ( Funderburk et al . , 2010; Nixon , 2013; Nixon and Yang , 2011 ) . The presence of massively accumulated AVs in dystrophic ( swollen ) neurites is a unique feature linked to AD pathology ( Nixon et al . , 2005 ) . Microtubule-based long-distance axonal transport is essential for autophagic clearance because autophagosomes are predominantly generated in distal axons and rely heavily on retrograde transport toward the soma for lysosomal proteolysis ( Cheng et al . , 2015a , 2015b; Lee et al . , 2011a; Maday and Holzbaur , 2016; Maday et al . , 2012 ) . Such a mechanism enables neurons to efficiently remove autophagic cargos from axons and synapses , thus reducing autophagic stress . While previous studies provide important information about autophagic flux and trafficking in healthy neurons under physiological conditions , mechanisms underlying AD-linked autophagic stress remain largely unknown . In the current study , we provide mechanistic insights into AD-associated autophagic stress in AD neurons and patient brains . First , we reveal that amphisomes predominantly accumulate in distal axons and at the presynaptic terminals of mutant hAPP Tg mouse brains ( Figure 1 and Figure 1—figure supplement 1 ) . Second , we show that retrograde transport of amphisomes is impaired in mutant hAPP Tg neurons , leading to axonal and presynaptic retention of AVs ( Figure 2 and Figure 2—figure supplement 1 ) . Third , we reveal that cytoplasmic Aβ1-42 oligomers associates with these accumulated amphisomes in the distal axons of AD mice ( Figure 3 and Figure 3—figure supplement 1 ) . Fourth , the Aβ1-42-DIC interaction interferes with dynein-Snapin coupling and reduces the recruitment of dynein motor to amphisomes , thus impairing AV transport ( Figure 4 ) . Fifth , we further confirm such impaired dynein-Snapin coupling in AD patient brains ( Figure 5 ) . Sixth , we demonstrate that Snapin deficiency in mice recapitulates axonal autophagic stress ( Figure 6 and Figure 6—figure supplement 1 ) . Seventh , we show that elevated Snapin expression in AD neurons reduces axonal AV retention by enhancing their retrograde transport ( Figure 7 and Figure 7—figure supplement 1 ) . Therefore , our study provides the first indication that defects in dynein-Snapin-driven AV retrograde transport contributes to AD-associated axonal autophagic stress . Under AD-associated pathological conditions , both endocytic and autophagic pathways are sites of APP processing and Aβ production ( Nixon , 2007 ) . It was reported that Aβ peptides oligomerize and accumulate within neuronal processes in AD mice and patient brains ( Takahashi et al . , 2004 ) . This raises a fundamental question as to whether overloaded Aβ in axons impairs dynein-driven retrograde transport of axonal cargoes such as LEs and AVs , two main organelles in the endo-lysosomal and autophagic pathways . In this study , we showed that soluble Aβ associates with amphisomes in the distal AD axons of AD neurons ( Figure 3 ) . Through direct interaction with DIC , Aβ1-42 interferes with the assembly of dynein-Snapin motor-adaptor complex , thus interrupting the recruitment of dynein motors to LEs and amphisomes for driving their retrograde transport ( Figure 4 ) . Given that AD-linked autophagic stress is uniquely associated with Aβ generation ( Benzing et al . , 1993 ) , our findings suggest that dynein is a target of Aβ-mediated toxicity , thus resulting in impaired dynein-Snapin coupling and AV retrograde transport in AD axons . Snapin mutant mouse brains exhibit a striking phenotype: AD-like axonal autophagic stress ( Figure 6 and Figure 6—figure supplement 1 ) . More importantly , elevated Snapin expression reverses AV retention by enhancing AV retrograde transport in the axons and presynaptic terminals of AD neurons ( Figure 7 and Figure 7—figure supplement 1 ) . These results are consistent with a recent study showing the in vitro and in vivo rescue effects of enhanced Snapin expression on the clearance of autophagic cargos in the axons of amyotrophic lateral sclerosis ( ALS ) -linked motor neurons ( Xie et al . , 2015 ) . Therefore , our results support a model that impaired AV retrograde transport plays a critical role in autophagic pathology in AD axons . Many studies have been focused on the mechanisms underlying defects in kinesin-mediated anterograde transport in AD ( Pigino et al . , 2009 , 2003; Stokin et al . , 2005; Tang et al . , 2012 ) . Synthetic Aβ1-42 was proposed to inhibit axonal transport through CK2 activation , which regulates kinesin-1 light chain and thus cargo attachment of kinesin-1 , the anterograde transport motors ( Pigino et al . , 2009 ) . A recent study showed that cell-derived soluble Aβ-oligomers induced early and selective diminutions in anterograde transport of synaptic cargoes ( Tang et al . , 2012 ) . These reported toxic effects on axonal transport were elicited by exogenously added Aβ , and were proposed to depend on CK2 , GSK3 , or tau ( Decker et al . , 2010; Pigino et al . , 2009 , 2003; Rui et al . , 2006; Stokin et al . , 2005; Tang et al . , 2012; Vossel et al . , 2010 ) . To our knowledge , it has not been investigated whether intracellular Aβ impairs dynein motor-driven axonal transport , in particular , whether and how intracellular Aβ disrupts Snapin-DIC interaction and thus AV retrograde transport , and whether these defects contribute to autophagic pathology in AD . We showed that soluble Aβ1-42 oligomers are enriched within distal axons and presynaptic terminals of AD mice , where they associate with accumulated amphisomes ( Figure 3—figure supplement 1 ) . In our immuno-EM analysis , we provided more direct evidence showing the presence of Aβ outside of the AVs . The immuno-gold anti-A11 antibody-labeled oligomeric Aβ is mostly present in the cytoplasm and is enriched in AD axons ( Figure 3F , G , I ) . More importantly , a majority of these cytoplasmic Aβ gold particles are located outside of AVs and associate with or surround AVs ( Figure 3H , I ) . Thus , we proposed that soluble Aβ oligomers in the cytoplasm interfere with the assembly of the DIC-Snapin complex through direct interaction with dynein DIC , thus interrupting cytoplasmic dynein motor recruitment to Snapin-associated AVs . We showed that this interruption was detected in the concentration of oligomeric Aβ1-42 as low as 0 . 2 μM in vitro in our study ( Figure 4 ) . These findings suggest that dynein-Snapin coupling , and thus cargo-motor association , is compromised in response to cytoplasmic accumulation of Aβ1-42 oligomers , which could be a pathogenic mechanism of impaired dynein-driven retrograde transport in AD . Recent studies from several groups provide consistent evidence that autophagosomes are predominantly generated in distal axons , and undergo retrograde transport toward the soma for lysosomal proteolysis ( Cheng et al . , 2015a; Fu et al . , 2014; Lee et al . , 2011a; Maday and Holzbaur , 2014; Maday et al . , 2012 ) . One recent study reported that newly generated autophagosomes in distal axon undergo robust retrograde transport toward the soma for maturation into autolysosomes ( Maday and Holzbaur , 2016 ) . They showed that these axon-generated autophagosomes enter the soma and then remain in the somatodendritic domain . Such compartmentalization facilitates degradation of axonal AVs in the soma , where mature lysosomes are mainly located . Consistently , the density of autophagosomes in the soma , but not in the axon , was increased by blocking lysosome function with Bafilomycin A1 . Their study concluded that lysosomal proteolysis in the soma is essential for the clearance of autophagic cargoes , which are mostly delivered from distal axons ( Maday and Holzbaur , 2016 ) . This study also showed that some autophagosomes are locally formed in the soma , are less motile than the axonal ones under basal condition , and do not increase in response to stress induced autophagy by nutrient deprivation . Together , this study and others consistently support our current findings that the soma is the primary site for autophagosome clearance; axonal autophagosomes undergo retrograde transport from the distal region to the soma for their degradation in order to maintain axonal homeostasis ( Cai et al . , 2010; Cheng et al . , 2015a; Lee et al . , 2011a; Maday and Holzbaur , 2014; Maday et al . , 2012 ) . We have focused on the mechanism underlying autophagic pathology in the axon , but not in the soma of AD neurons because AD brains exhibit a unique phenotype—massive accumulation of AVs within distal dystrophic neurites ( Figure 1 and Figure 5 ) . Our study addresses a new fundamental issue that retrograde transport is critical for the reduction of autophagic stress in distal axons under AD condition . ( 1 ) AD axons display predominant amphisome accumulation and reduced retrograde transport of LEs/amphisomes ( Figure 2 and Figure 2—figure supplement 1 ) ; ( 2 ) Expression Snapin , but not its dynein-binding defective mutant , significantly reduces AV retention by enhancing their retrograde transport in AD axons ( Figure 7 ) ; ( 3 ) Snapin-enhanced AV transport rescues AV accumulation in AD axons upon autophagy induction ( Figure 7—figure supplement 1 ) ; and ( 4 ) deleting Snapin recapitulates AD-associated autophagic stress in distal axons ( Figure 6 and Figure 6—figure supplement 1 ) . Given that the soma is the primary site of AV clearance ( Maday and Holzbaur , 2016 ) , our data consistently suggests that impeded AV retrograde transport contributes to autophagic pathology in AD axons . Martinez-Vicente et al . ( 2010 ) revealed an important mechanism that inefficient autophagy is attributed to cargo recognition failure in Huntington’s disease . However , there is no indication of a similar mechanism in AD ( Nixon , 2013 ) . Instead , amphisomes/autolysosmes , rather than empty autophagic organelles , are the predominant AVs accumulated in AD brains ( Nixon , 2007; Nixon et al . , 2005; Yu et al . , 2005 ) . We provide new evidence showing aberrant accumulation of AVd-like structures containing vesicles and partially digested materials accumulate in distal axons and presynaptic terminals of AD mouse model and patient brains ( Figure 1 , Figure 2 , Figure 3 , and Figure 5 ) . Moreover , these AVs contain ubiquitinated cargoes ( Figure 1—figure supplement 1 , Figure 2—figure supplement 1 , and Figure 3—figure supplement 1 ) . In addition , deleting Snapin in mice , which impedes AV transport , but does not affect cargo engulfment , exhibits similar phenotypes of autophagic accumulation ( Figure 6 and Figure 6—figure supplement 1 ) . These findings suggest that cargo recognition failure in AD is unlikely a predominant mechanism underlying AD-associated autophagic stress . Proper lysosomal function is critical for the removal of autophagic substrates . Autophagic stress has been linked to lysosomal storage diseases ( LSDs ) ( de Pablo-Latorre et al . , 2012; Sano et al . , 2009 ) . The lysosomal deficits in AD are thought to play a critical role in autophagy dysfunction , leading to the disruption of substrate proteolysis within autolysosomes ( Boland et al . , 2008; Lee et al . , 2011b; Yang et al . , 2011 ) . Defective lysosomal proteolysis produces similar neuropathology in WT mice and exacerbates amyloid and autophagy pathology in AD mouse models ( Nixon , 2013; Nixon and Yang , 2011 ) . Because neurons are particularly dependent on lysosomal degradation capacity for eliminating Aβ generated in the endocytic-autophagic pathways ( LeBlanc and Goodyer , 1999; Lee et al . , 2011a ) , impaired AV retrograde transport we observed in AD neurons could be indirectly attributed to lysosomal deficits . Thus , increased Aβ load upon lysosomal inhibition may impair the recruitment of dynein motors to AVs by competitively interfering with dynein-Snapin coupling in AD neurons . In summary , our study provides new mechanistic insights into autophagic defects under AD-linked pathogenesis , which conceptually advances current knowledge of Aβ-induced impairment of dynein-mediated retrograde transport underlying autophagic stress in axons and at synapses of AD brains . Given that synaptic retention of AVs alters synaptic structure and neurotransmission , elucidation of this pathological mechanism has a broad neurobiological impact because impaired axonal transport , autophagic stress , and synaptic dysfunction are all associated with major neurodegenerative diseases . Therefore , enhancing clearance of AVs by regulating retrograde trafficking may be a potential therapeutic strategy for AD and other major neurodegenerative diseases . This study advances our understanding of autophagic stress in AD and may have broader relevance to other neurodegenerative diseases associated with defective axonal transport and autophagy dysfunction . Further therapeutic approaches aimed at regulating the dynein-Snapin coupling may help attenuate axonal pathology in AD .
Snapin flox mice were provided by ZH Sheng ( National Institute of Neurologic Disorders and Stroke , NIH , Bethesda , MD ) . Camk2α-tTA and tet-APPswe/ind mice were obtained from H Cai ( National Institute on Aging , NIH , Bethesda , MD ) . hAPP mice ( C57BL/6J ) from line J20 ( Mucke et al . , 2000 ) and Thy1-Cre Tg mice ( Campsall et al . , 2002 ) were purchased from the Jackson Laboratory ( Bar Harbor , ME ) . Postmortem brain specimens from AD patients and age-matched control subjects were obtained from the Harvard Tissue Resource Center , and the Human Brain and Spinal Fluid Resource Center at UCLA . Specimens were from patients diagnosed with AD according to Braak criteria ( Braak and Braak , 1991 ) . The specimens were from the frontal cortex and were quick-frozen ( BA9 ) ( Table 1 ) . The EM data were from two control subjects and three AD patient brains at different Braak stages with postmortem interval between 7 . 08 hr and 12 . 5 hr . Four control subjects and four patient brains ( postmortem interval 7 . 08 hr – 22 . 5 hr ) were used for synaptosomal fraction purification . The data of immunoprecipitation assays were from three sets of human brains with postmortem interval 9 . 00 hr – 20 . 1 hr . pmRFP-Rab7 and pmRFP-Ub were from A . Helenius and N . Dantuma , respectively . The constructs encoding Snapin , Snapin-L99K , GFP-DIC , GST-DIC , GFP-LC3 , hAPP , and hAPPswe , were prepared as previously described ( Cai et al . , 2010; Ye and Cai , 2014; Zhou et al . , 2012 ) . The purified polyclonal antibody against mouse N-terminal Snapin was described previously ( Tian et al . , 2005 ) and obtained from Z . H . Sheng . Sources of other antibodies and reagents are as follows: polyclonal anti-EEA1 , anti-MAP2 , anti-syntaxin 1 , anti-TOM20 , and anti-synaptophysin antibodies ( Santa Cruz , Dallas , TX ) ; monoclonal anti-DIC , anti-GAPDH , and anti-synaptophysin antibodies , polyclonal Aβ1-42 and anti-APP c-terminal antibodies ( Millipore/CHEMICON , Billerica , MA ) ; monoclonal anti-GFP ( JL-8 ) antibody ( Clontech , Madison , WI ) ; monoclonal anti-β Amyloid ( 6E10 ) and anti-HA antibodies ( Biolegend , San Diego , CA ) ; monoclonal anti-p115 , anti-MAP2 , and anti-p150Glued antibodies ( BD Biosciences , San Jose , CA ) ; monoclonal anti-Ubiquitin and polyclonal anti-LC3 antibodies ( Cell Signaling Technology , Danvers , MA ) ; monoclonal anti-p62/SQSTM1 antibody ( Abnova , Taiwan , China ) ; monoclonal anti-Rab7 and anti-Neurofilament 200 antibodies , and polyclonal anti-LC3 , anti-Neurofilament 160 , and anti-syntaxin 17 antibodies ( Sigma , St . Louis , MO ) ; monoclonal anti-GST antibody ( Thermo scientific , Grand Island , NY ) ; monoclonal anti-PSD95 antibody ( UpstateMillipore/CHEMICONUpstate Upstate , Billerica , MA ) ; monoclonal anti-CI-MPR and anti-LAMP-1 antibodies were developed by D Messner and JT August and were obtained from Developmental Studies Hybridoma Bank ( Iowa City , Iowa ) . Polyclonal anti-Aβ ( A11 ) antibody and Alexa fluor 546- , and 633-conjugated secondary antibodies ( Invitrogen/Thermo scientific , Grand Island , NY ) . COS7 Cells were purchased from ATCC ( CRL-1651 ) ( Manassas , VA ) . The authenticity was confirmed by STR profiling . The mycoplasma contamination was tested and showed negative result prior to the experiments . Cortices were dissected from E18–19 mouse embryos as described ( Cai et al . , 2010 , 2012; Goslin and Banker , 1998 ) . Cortical neurons were dissociated by papain ( Worthington , Lakewood , NJ ) and plated at a density of 100 , 000 cells per cm2 on polyornithine- and fibronectin-coated coverslips . Neurons were grown overnight in plating medium ( 5% FBS , insulin , glutamate , G5 and B27 ) supplemented with 100 × L-glutamine in Neurobasal medium ( Invitrogen ) . Starting at DIV 2 , cultures were maintained in conditioned medium with half-feed changes of neuronal feed ( B27 in Neurobasal medium ) every 3 days . Primary hAPP Tg neurons were cultured from breeding mice of hemizygous mutant hAPPSwe/Ind Tg ( J20 line ) with WT animals ( Mucke et al . , 2000 ) . Genotyping assays were performed following culture plating to verify mouse genotypes . In our study , we examined both transgenic neurons and non-transgenic neurons derived from their littermates . Neurons were transfected with various constructs using Lipofectamine 2000 ( Invitrogen ) followed by time-lapse imaging at DIV16-20 transfection prior to qualification analysis . COS7 Cells ( ATCC ) were incubated with high glucose DMEM containing sodium pyruvate , L-glutamine , supplemented with 10% FBS and penicillin-streptomycin ( Invitrogen ) . Transient transfection COS7 was performed using Lipofectamine 2000 . 100 μl of Opti-MEM ( Invitrogen ) and 1–2 μl of Lipofectamine 2000 ( Invitrogen ) per chamber were pre-incubated at room temperature ( RT ) for 5 min and then mixed with 100 μl of Opti-MEM containing DNA constructs ( 2–3 μg per chamber ) and incubated for 20 min at RT to allow complex formation . The entire mixture was added directly to cultured cells . Following transfection , cells were cultured for an additional 1–2 days before harvesting for biochemical analysis . Animals were anaesthetized with 2 . 5% avertin ( 0 . 5 ml per mouse ) , and transcardially perfused with fixation buffer ( 4% paraformaldehyde in PBS , pH 7 . 4 ) . Brains were dissected out and postfixed in fixation buffer overnight and then placed in 30% sucrose at 4°C . 10-µm-thick coronal sections were collected consecutively to the level of the hippocampus and used to study co-localization of various markers . After incubation with blocking buffer ( 2 . 5% goat serum , 0 . 15% Triton X-100 , 1 . 5% BSA , 0 . 5% glycine in H2O ) at RT for 1 hr , the sections were incubated with primary antibodies at 4°C overnight , followed by incubating with secondary fluorescence antibodies at 1:400 dilution at RT for 1 hr . After fluorescence immunolabeling , the sections were stained with DAPI , washed three times in PBS . The sections were then mounted with anti-fading medium ( vector laboratories , H-5000 , UK ) for imaging . Confocal images were obtained using an Olympus FV1000 oil immersion 40 × objective with sequential-acquisition setting . Eight to ten sections were taken from top-to-bottom of the specimen and brightest point projections were made . Confocal images were obtained using an Olympus FV1000 oil immersion 60 × objective ( 1 . 3 numerical aperture ) with sequential-acquisition setting . For fluorescent quantification , images were acquired using the same settings below saturation at a resolution of 1024 × 1024 pixels ( 8 bit ) . Eight to ten sections were taken from top-to-bottom of the specimen and brightest point projections were made . Morphometric measurements were performed using NIH ImageJ . Measured data were imported into Excel software for analysis . The thresholds in all images were set to similar levels . Data were obtained from at least three independent experiments and the number of cells or imaging sections used for quantification is indicated in the figures . All statistical analyses were performed using the Student's t-test and are presented as mean ± SEM . For live cell imaging , cells were transferred to Tyrode's solution containing 10 mM Hepes , 10 mM glucose , 1 . 2 mM CaCl2 , 1 . 2 mM MgCl2 , 3 mM KCl and 145 mM NaCl , pH 7 . 4 . Temperature was maintained at 37°C with an air stream incubator . Cells were visualized with a 60 × oil immersion lens ( 1 . 3 numerical aperture ) on an Olympus FV1000 confocal microscope , using 488 nm excitation for GFP and 543 nm for mRFP . Time-lapse sequences of 1024 × 1024 pixels ( 8 bit ) were collected at 1–2 s intervals with 1% intensity of the argon laser to minimize laser-induced bleaching and damage to cells , and maximum pinhole opening . Dual-color time-lapse images were captured by a total of 100 frames . All recordings started 6 min after the coverslip was placed in the chamber . The stacks of representative images were imported into NIH ImageJ . A membranous organelle was considered stopped if it remained stationary for the entire recording period; a motile one was counted only if the displacement was at least 5 μm . For analyzing the motility of AVs or late endosomes in live neurons , we selected axons for time-lapse imaging and measuring motility because axons , but not dendrites , have a uniform microtubule organization and polarity . Axonal processes were selected as we previously reported ( Cai et al . , 2010 , 2012; Kang et al . , 2008 ) . Briefly , axons in live images were distinguished from dendrites based on known morphological characteristics: greater length , thin and uniform diameter , and sparse branching ( Banker and Cowan , 1979 ) . Only those that appeared to be single axons and separate from other processes in the field were chosen for recording axonal AVs or late endosomes transport . Regions where crossing or fasciculation occurred were excluded from analysis . Kymographs were used to trace axonal anterograde or retrograde movement of membranous organelles and to count stationary ones as described previously ( Kang et al . , 2008; Miller and Sheetz , 2004 ) with extra plug-ins for ImageJ ( NIH ) . Briefly , we used the ‘Straighten’ plugin to straighten curved axons and the ‘Grouped ZProjector’ to z-axially project re-sliced time-lapse images . The height of the kymographs represents recording time ( 100 s unless otherwise noted ) , while the width represents the length ( μm ) of the axon imaged . Counts were averaged from 100 frames for each time-lapse image to ensure accuracy of stationary and motile events . Measurements are presented as mean ± SEM . Statistical analyses were performed using unpaired Student's t-tests . Synaptosome preparations from WT and mutant hAPP Tg mouse brains or AD patient brains and control subjects were collected using Percoll gradient centrifugation as described in the protocol ( Leenders et al . , 2004 ) . Cortex tissues were homogenized in ice cold Sucrose Buffer [5 mM HEPES , 1 mM EDTA , 0 . 32 M sucrose and protease inhibitors ( Roche , Indianapolis , IN ) , pH 7 . 4] . Homogenates were centrifuged at 1000 × g for 10 min , the supernatant was gathered and overlaid on Percoll gradients that has 2 ml of 10% Percoll gradient layered over 15% , 23% , and 40% Percoll gradients . The gradient was then separated by centrifugation for 5 min at 32 , 500 × g . The synaptosomal fraction was collected from the 15%/23% Percoll layers , and combined with 5 ml the Sucrose buffer . The mixture was then centrifuged at 15 , 000 × g for 15 min and resuspended in the Sucrose buffer . Protein quantification was performed by BCA assay ( Pierce Chemical Co . /Thermo scientific ) . 15 µg of protein from synaptosome and post nuclear supernatant ( PNS ) homogenates were resolved by 4–12% SDS-PAGE for sequential Western blots on the same membranes after stripping between each application of antibody . For multiple detection with different antibodies , blots were first stripped in a solution of 62 . 5 mM Tris-HCl , pH 7 . 5 , 20 mM dithiothreitol and 1% SDS for 15 min at 50°C with agitation and then washed with TBS/0 . 1% Tween-20 for 2 × 15 min ( Leenders et al . , 2004; Cai et al . , 2010 ) . Brain tissues from WT or mutant hAPP Tg or Snapin cKO mice were homogenized in the buffer ( 10 mM HEPES [pH 7 . 4] , 1 mM EDTA , 0 . 25 M sucrose , and protease inhibitors ) and centrifuged at 800 × g for 10 min , and the supernatant was collected . The pellet was re-suspended in the homogenization buffer using a glass rod with 3 to 4 gentle strokes of the pestle of the 30 ml Dounce Homogenizer and re-centrifuged at 800 × g for 10 min . The combined first and second supernatants were centrifuged at 3500 × g for 10 min and then collected for high-speed centrifugation at 20 , 000 × g for 10 min . The pellet was re-suspended in the homogenization buffer using a glass rod with 3 to 4 gentle strokes of the pestle of the 30 ml Dounce Homogenizer and re-centrifuged at 20 , 000 × g for 10 min . The pellet was then re-suspended in the homogenization buffer as light membrane fraction ( LMF ) and subjected to immuno-isolation with tosylated linker-coated superparamagnetic beads ( Dynabeads M-450 Subcellular; Invitrogen ) as previously described ( Cai et al . , 2010; Ye and Cai , 2014; Zhou et al . , 2012 ) . For all subsequent steps , beads were collected with a magnetic device ( MPC; Invitrogen ) . After washing once for 5 min in PBS ( pH 7 . 4 ) with 0 . 1% BSA at 4°C , the linker-coated beads ( 1 . 4 mg ) were incubated with 1 μg anti-Rab7 mAb , or control mouse IgG overnight at 4°C on a rotator . After incubation , the beads were washed four times ( 5 min each ) in PBS [pH 7 . 4] with 0 . 1% BSA at 4°C , and then re-suspended in an incubation buffer containing PBS [pH 7 . 4] , 2 mM EDTA , 5% fetal bovine serum . Approximately 400 μg of light membrane fraction from mutant hAPP Tg or Snapin cKO mouse brains were mixed with incubation buffer containing beads ( final reaction volume 1 ml ) and incubated for 4 hr at 4°C on a rotator . After incubation , the beads were collected with a magnetic device and washed five times with the incubation buffer and three times with PBS for 10 min each and then resolved by 4–12% Bis-Tris PAGE for sequential Western blots on the same membranes after stripping between each application of the antibody . For semi-quantitative analysis , protein bands detected by ECL were scanned into Adobe Photoshop CS6 , and analyzed using NIH ImageJ . Dynein intermediate chain ( DIC ) pull-down experiments were performed as previously described 77 . In brief , GST fusion proteins were bound to glutathione-Sepharose beads ( GE Healthcare , Port Washington , NY ) in TBS buffer ( 50 mM Tris-HCl at pH 7 . 5 , 140 mM NaCl ) with 0 . 1% Triton X-100 and protease inhibitors , and incubated at 4°C for 1 hr with constant agitation , followed by washing for three time with TBS . The beads coupled with ~1 μg GST fusion protein were added to 2 mg LMF prepared from mouse brains in the presence or absence of Aβ1-42 , and then incubated overnight at 4°C . The beads were then washed three times with TBS buffer; bound proteins were processed for 4–12% Bis-Tris PAGE and immunoblotting on the same membranes after stripping between each application of the antibody . Scrambled Aβ , oligomeric Aβ1-42 , or Aβ1-40 was prepared as previously described ( Du et al . , 2010; Rui et al . , 2006; Vossel et al . , 2010 ) . Aβ1-42 or Aβ1-40 peptide ( Sigma ) was diluted in 1 , 1 , 1 , 3 , 3 , 3-hexafluoro-2-propanol to 1 mM using a glass gas-tight Hamilton syringe with a Teflon plunger . The clear solution was then aliquoted in microcentrifuge tubes , followed by evaporation in the fume hood over night at RT , and it was then dried under vacuum for 1 hr in a speedVac ( DNA Vap , Labnet , Edison , NJ ) . Peptide film was diluted in DMSO to 5 mM and sonicated for 10 min in bath sonicator . The peptide solution was resuspended in cold TBS buffer to 100 μM and immediately vortexed for 30 s; the solution containing monomeric Aβ was then incubated at 4°C for 24 hr to form oligomeric Aβ before applying to in vitro binding assays . DIC and Snapin were constructed into the GST-fusion vector pGEX-4T ( GE Healthcare ) and the His-tagged vector pET28a ( Novagen/EMD , Billerica , MA ) , respectively . Fusion proteins were prepared as crude bacterial lysates by mild sonication in PBS containing 1% Triton X-100 and protease inhibitors ( 1 mM phenylmethylsulfonyl fluoride , 10 μg/ml leupeptin , 2 μg/ml aprotinin ) . in vitro binding experiments were performed as described previously ( Cai et al . , 2010 ) . In brief , GST fusion proteins were bound to glutathione-Sepharose beads ( GE Healthcare ) in TBS buffer ( 50 mM Tris-HCl at pH 7 . 5 , 140 mM NaCl ) with 0 . 1% Triton X-100 and protease inhibitors , incubated at 4°C for 1 hr with constant agitation and washed with TBS . The beads coupled with ~1 μg GST fusion protein were added to Aβ or His-Snapin lysates , and then incubated for 3 hr at 4°C . The beads were washed three times with TBS; bound proteins were processed for 4–12% Bis-Tris PAGE and immunoblotting on the same membranes after stripping between each application of the antibody , or spotted for the dot blot analysis to detect Aβ . For immunoprecipitation , an equal amount ( 600 μg ) of COS7 cell lysates or ( 750 μg ) human brain homogenates from AD patients or control subjects or ( 750 μg ) brain homogenates from WT and mutant hAPP Tg mice were incubated with anti-GFP or anti-Snapin or anti-DIC antibody in 200 μl of TBS with 0 . 1% Triton X-100 and protease inhibitors , and incubated on a rotator at 4°C for overnight . 2 . 5 mg Protein A-Sepharose CL-4B resin ( GE Healthcare ) were added to each sample , and the incubation continued for an additional 3 hr followed by three washes with TBS/0 . 1% Triton X-100 . Immobilized protein complexes were processed for 4–12% Bis-Tris PAGE and immunoblotting on the same membranes after stripping between each application of the antibody . For semi-quantitative analysis , protein bands detected by ECL were scanned into Adobe Photoshop CS6 and analyzed using NIH ImageJ . Care was taken during exposure of the ECL film to ensure that intensity readouts were in a linear range of standard curve blot detected by the same antibody . Paired Student t-tests were carried out and results are expressed as mean ± SEM . Hippocampi from WT and mutant hAPP Tg mice , or AD patient brains were cut into small specimens ( one dimension <1 mm ) and fixed in Trumps fixative ( Electron Microscopy Sciences , Hatfield , PA ) for 2 hr at RT . The sections were then washed by 0 . 1 M Cacodylate buffer , and postfixed in 1% osmium tetroxide , followed by dehydrating in ethanol , and embedding using the EM bed 812 kit ( Electron Microscopy Sciences ) according to a stand procedure . Cultured mouse cortical neurons at DIV18-19 were fixed at RT with EM fixative ( 2% glutaraldehyde and 2% paraformaldehyde in 0 . 1 N Na+ cacodylate buffer ) for 2 hr . Samples were then stored at 4°C for overnight and then treated with osmium tetroxide , en bloc mordanted with uranyl acetate , dehydrated through a series of graded ethanol washes , and embedded in epoxy resins . Images were acquired on an electron microscope ( 100C ×; JEOL ) ( Division of Life Sciences , Rutgers University Electron Imaging Facility ) . For quantitative studies , AVs were characterized by initial AVs ( AVi ) with double-membrane structures containing organelles or vesicles , or late-stage degradative AVs ( AVd ) after fusion with late endocytic organelles containing partially degraded cytoplasmic material , small vesicles , and electron-dense material ( Cheng et al . , 2015a; Klionsky et al . , 2012 ) . Quantification analysis was performed blindly to condition . Immuno-EM were performed as described previously ( Cai et al . , 2010 ) . In brief , cultured cortical neurons at DIV19-20 were fixed with 4% paraformaldehyde and 0 . 05% glutaraldehyde in PBS for 45 min , washed , and then permeabilized and blocked with 0 . 1% saponin/5% normal goat serum in PBS for 1 hr , incubated with primary antibody for 1 hr at RT , washed , and incubated with secondary antibody conjugated to 1 . 4 nm Nanogold ( Nanoprobes , Yaphank , NY ) for 1 hr . Samples were fixed with 2% glutaraldehyde in PBS , washed and followed by silver enhancement ( Nanoprobes ) for 15 min , treated sequentially with 0 . 2% OsO4 in phosphate buffer for 30 min and 0 . 25% uranyl acetate at 4°C overnight , washed and dehydrated in ethanol and finally embedded in epoxy resins . All steps were carried out at RT unless otherwise indicated . The control for specificity of immunolabeling omitted the primary antibody . | Alzheimer’s disease is the result of protein fragments called amyloid-β peptides accumulating in the brain and forming clumps . These protein “aggregates” disrupt cellular activities and cause serious problems . To combat this process , healthy cells use a process called autophagy to destroy aggregated proteins . The aggregates are first loaded into structures called autophagosomes that then fuse with cell compartments called lysosomes , which contain enzymes that can break down the proteins . Brain cells called neurons have an unusual shape with branch-like structures and a long projection called an axon that all form off the main cell body . Autophagosomes predominantly form in the axons and need to move toward the cell body where the lysosomes are found . A motor protein called dynein drives the movement of autophagosomes by interacting with an adaptor protein known as Snapin on the surface of these structures . Autophagosomes tend to accumulate within the neurons of individuals with Alzheimer’s disease , but it is not known why . Cai et al . examined the ability of autophagosomes to move to the cell body of neurons from a mouse model of Alzheimer’s disease in which human amyloid-β peptides accumulate in the brain , and in the brains of human patients with Alzheimer’s disease . The experiments show that autophagosomes predominantly accumulate in the axons and at the ends of axons during Alzheimer’s disease . Amyloid-β aggregates associate with autophagosomes in the axons and interact with dynein motors . This disrupts the interaction between dynein and Snapin and impairs dynein binding to the autophagosomes , trapping the autophagosomes in the axons . Increasing the production of Snapin proteins inside the mouse neurons enhances dynein binding to autophagosomes and thus helps these structures move to the cell body . The next step is to investigate whether increasing the ability of autophagosomes to move to the cell body reduces the symptoms of Alzheimer’s disease in the mutant mice . This will help to build a foundation for the future development of new strategies to treat Alzheimer’s disease and other neurodegenerative disorders that are caused by protein aggregates . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2017 | Impaired retrograde transport of axonal autophagosomes contributes to autophagic stress in Alzheimer’s disease neurons |
Microdosing is the practice of regularly using low doses of psychedelic drugs . Anecdotal reports suggest that microdosing enhances well-being and cognition; however , such accounts are potentially biased by the placebo effect . This study used a ‘self-blinding’ citizen science initiative , where participants were given online instructions on how to incorporate placebo control into their microdosing routine without clinical supervision . The study was completed by 191 participants , making it the largest placebo-controlled trial on psychedelics to-date . All psychological outcomes improved significantly from baseline to after the 4 weeks long dose period for the microdose group; however , the placebo group also improved and no significant between-groups differences were observed . Acute ( emotional state , drug intensity , mood , energy , and creativity ) and post-acute ( anxiety ) scales showed small , but significant microdose vs . placebo differences; however , these results can be explained by participants breaking blind . The findings suggest that anecdotal benefits of microdosing can be explained by the placebo effect .
There is renewed interest in the medical application of psychedelic drugs , such as lysergic acid diethylamide ( LSD ) and psilocybin . Contemporary research is predominantly focusing on ‘psychedelics assisted psychotherapy’ , where a few ( one to three ) large doses of psychedelics are used as adjunct to psychotherapy . Using this paradigm , psychedelics have shown promise in the treatment of conditions such as depression , end-of-life-anxiety , addiction , and obsessive-compulsive behaviors ( Carhart-Harris and Goodwin , 2017; Nutt et al . , 2020 ) . Recently , ‘microdosing’ has emerged as an alternative paradigm of psychedelic use . Due to its underground origin , microdosing does not have a universally agreed upon definition , and inconsistencies exist in substance , dose , frequency , and duration of use ( Kuypers et al . , 2019 ) . However , microdosing can be broadly defined as the frequent use ( one to three times per week ) of low doses of psychedelics ( 10–20% of a typical ‘full’ dose , e . g . 10–15 μg LSD or 0 . 1–0 . 3 g of dried psilocybin containing mushrooms ) . Anecdotal evidence suggests that microdosing may improve well-being , creativity , and cognition ( Fadiman and Krob , 2017 ) , and recent uncontrolled , observational studies have provided some empirical support for these claims ( Anderson et al . , 2019; Polito and Stevenson , 2019; Prochazkova et al . , 2018 ) . While encouraging , these studies are vulnerable to experimental biases , including confirmation-bias and placebo effects , in particular , because microdosers are a self-selected sample with optimistic expectations about psychedelics and microdosing ( Polito and Stevenson , 2019 ) . This positivity bias , combined with the low dose and the subjective evaluation of effects , pave the way for a strong placebo response . A few recent double-blind , controlled studies have been conducted on microdosing . All studies used LSD and focused on the acute effects of a single microdose in a small number of healthy subjects ( Yanakieva et al . , 2019; Bershad et al . , 2019a; Bershad et al . , 2019b; Family et al . , 2020; Hutten et al . , 2020b ) . Studies have found large variability in LSD blood concentration after microdosing ( Family et al . , 2020 ) , along with increased BDNF blood levels ( Hutten et al . , 2020a ) . No robust evidence was found to support the positive anecdotal claims about microdosing , but some dose-related self-rated subjective effects were detected ( e . g . self-ratings of ‘feel drug’ , ‘feel high’ , and ‘like drug’ ) ( Yanakieva et al . , 2019; Bershad et al . , 2019b; Hutten et al . , 2020b ) , along with concomitant changes in brain function ( Bershad et al . , 2019b ) . Two key issues need to be considered when assessing the scientific credibility of microdosing: the lack of placebo control in uncontrolled studies and the small sample size in controlled studies . Uncontrolled , observational studies affirm the anecdotal reports , but by design , these studies cannot provide evidence for beyond placebo benefits . Lab-based , controlled studies have small samples ( Yanakieva et al . , 2019; Bershad et al . , 2019a; Bershad et al . , 2019b; Family et al . , 2020 ) due to restrictive drug policies that render randomized controlled trials prohibitively expensive , and hence may be statistically underpowered . In the present study we conceived of a novel citizen-science ( Silvertown , 2009 ) initiative as a solution to this problem , exploiting modern technology and the popularity of microdosing . The key component is a self-blinding setup procedure that enabled self-experimenters , who microdose on their own initiative using their own psychedelic , to implement placebo control and randomization without clinical supervision . To investigate potential changes over the study period , participants were directed to online self-report surveys and cognitive tasks at various timepoints . The strength of this design is that it allowed us to obtain a large sample size while implementing placebo control at minimal logistic and economic costs . The primary objective of the study was to test whether psychedelics microdosing produces superior outcomes compared to placebo on psychological state and cognitive function . We hypothesized that improvements from baseline will be positively correlated with the number of microdoses taken during the dose period and that acute/post-acute outcomes will be better under/after taking a microdose .
This study had a naturalistic design involving elements of experimental control ( self-blinding ) , prospective data collection and online citizen-science . From baseline to the final endpoint , the study was 10 weeks long ( weeks 0–9 ) , including a core 4-week microdosing period . Primary endpoint was at week 5 and there was an optional follow-up at week 9 . The self-blinding procedure randomly assigned individuals to three groups , where the groups are defined by the number of weeks taking placebos/microdoses during the dose period . The three groups were: Individuals took two microdoses during each microdose week , resulting in 0/4/8 total microdoses for the PL/HH/MD groups . Participants had equal probability ( 1/3 ) of being assigned to each group; Figure 1 illustrates the experimental timeline and the groups’ dose schedule . Outcomes can be organized into three categories capturing the effects of microdosing on different timescales . An overview of the outcomes can be found in Table 1 and a description of each measure is in Appendix 1 . See Figure 1 for the experimental timeline and assessment timepoints . A high-level overview of self-blinding is provided here; for a detailed illustration see Figure 2 . First , two sets of capsules had to be prepared using non-transparent capsules: one set with microdoses inside and another set without anything inside ( placebos ) . Next , these capsules were packaged into weekly sets , which were then placed inside envelopes together with a QR code ( Figure 2A ) . The envelopes were grouped and shuffled . Then , using a semi-random drawing process , four of them were selected ( Figure 2B ) corresponding to the 4 weeks of the dose period ( i . e . each envelope held capsules for 1 week of the dose period ) . The drawing process was constrained such that only three combinations of the envelopes were possible to draw , matching the three study groups: placebo ( four placebo weeks ) , half-half ( 2–2 placebo and microdose weeks ) , and microdose group ( four microdose weeks; Figure 2C ) . At this stage , participants were ready to start the experiment . When the dose period started , one envelope was opened per week and the capsules inside were used as scheduled ( Figure 2D ) . Additionally , the QR code from the envelope had to be scanned , which shared a numeric code with our informatics infrastructure . The decryption key ( i . e . how capsule types are encoded by the numbers ) was not shared with participants , so the numeric code allowed only us to deduce which type of capsule was taken when . In summary , the two key elements of self-blinding are to hide the active components inside opaque capsules while preparing identical looking placebos ( 1 ) and to position non human-readable QR codes along the capsules prior to randomization ( 2 ) . With the QR codes in place , it is possible for the experimenter to recover knowledge of capsule types after randomization without revealing that information to participants . Participants were allowed to use any psychedelic substance to microdose with . The microdose dose , which is the amount of substance to use as a microdose , was not defined for participants , rather they were instructed to use a microdose dose that they would use outside the study . The rationale for this direction was threefold . First , given that participants typically would source their substance from the black market , the precise microdose dose could not have been known even if instructions requested it . Second , based on community feedback , most experienced microdosers have a preferred dose that they would not have liked to change to participate in the study . Lastly , this study was not a clinical trial and therefore from a regulatory perspective not allowing for control over and/or directing about drug doses . Psychedelics users were recruited through advertisement on relevant online and offline forums . Individuals could sign up through the study’s website , https://selfblinding-microdose . org/ , where they could find information about the study , including the study manual and explainer videos , the participant information’s sheet , and procedure for declaring informed consent . Once informed consent was given , individuals were able to sign up by providing their email address and planned start date . The inclusion criteria were: >18 years of age , good understanding of English , intention to microdose with psychedelics , previous experience with psychedelics ( either micro- or macrodosing ) , no use of psychedelic drugs from a week before the start until the completion of the post-regime timepoint ( other than the study’s microdoses ) , and willingness to follow the study protocol . All the questionnaires were implemented online using the SurveyGizmo platform ( https://www . surveygizmo . com/ ) . For the online assessment of cognitive performance , the Cambridge Brain Sciences ( https://www . cambridgebrainsciences . com/ ) service was used . At each timepoint , links to each test were sent in a dedicated email via the Psychedelics Survey ( https://www . psychedelicsurvey . com/ ) service . These links had a personal ID embedded , so each test completion could be matched to individuals . Participants were asked to guess which type of capsule they had taken that day during the dose period ( for days when capsule was taken ) . This guess was a forced binary choice between microdose and placebo options . At the end of the post-acute test sessions , participants were asked separately to guess whether the current week was a microdose or a placebo week ( Figure 1A ) . In the discussion of our results , the term ‘break blind’ indicates that the participant guessed the capsule correctly for the day ( acute outcomes ) or week ( post-acute outcomes ) . No guess was collected about perceived group allocation at the end of study , because information about group structure was not shared with participants . Group differences in demographics , recreational drug use , and baseline scores of the accumulative outcomes were assessed with ANOVA and chi-square tests for continuous and categorical variables . Accumulative outcomes were analyzed with mixed-effect repeated measurement models , using the SAS PROC MIXED method with compound symmetry covariance structure . Models were constructed with change from baseline as the dependent variable , group , time and group*time interaction as factors , and individuals as experimental unit . Models were adjusted for all significant baseline covariates ( the following variables were tested as potential covariates: age , sex , education , baseline score , dose , total dose , short suggestibility scale score , expectation score , number of past psychiatric diagnosis , number of current psychiatric medications , number of lifetime macrodose experiences , and number of lifetime months microdosing ) . To accommodate dose as a potential covariate , psilocybin mushroom mass was converted to an estimated equivalent LSD dose ( 0 . 1 g of dried mushroom ~4 . 6 µg LSD; Kaplan et al . , 1994; Carbonaro et al . , 2016 ) . The following planned comparisons were made: within-group comparisons of change over time from baseline to the primary endpoint at week 5 and from baseline to the final follow-up at week 9 . Additionally , between-group comparisons were made ( PL vs . HH and PL vs . MD ) at week 5 and week 9 . To analyze acute and post-acute outcomes , mixed linear models were constructed . Models included score as dependent variable , subject ID as a random-effect , and condition as fixed-effect , where condition was a binary categorical variable ( PL/MD ) . For acute outcomes , condition was PL/MD when the score was obtained under the influence of a placebo/microdose capsule , while for post-acute outcomes condition was PL/MD when the score was obtained at the end of placebo/microdose week . Planned comparisons were made between scores obtained under PL and MD conditions . Each participant contributed four scores to these models , corresponding to the four acute/post-acute assessment timepoints during the dose period . All acute/post-acute models were adjusted for all significant baseline covariates ( same variables were tested for significance as in the case for the accumulative outcomes , except baseline score and total dose consumed ) . To better understand how guess influenced scores , a second set of models were constructed with the addition of guess ( binary categorical variable , PL/MD ) and guess*condition factors . Using these guess adjusted models , planned comparisons were made between PL and MD conditions . Finally , the two binary variables ( condition and guess ) divided the data into 2*2 = 4 strata , post-hoc comparisons were made between the following strata ( condition/guess ) : PL/PL vs . MD/PL , PL/MD vs . MD/MD , PL/PL vs . PL/MD and MD/PL vs . MD/MD . This selection was made such that condition changes while guess remains fixed in the first two comparisons , and guess changes while condition remains fixed in the last two comparisons . The study only engaged people who planned to microdose through their own initiative with their own psychedelic substance , but who consented to incorporate placebo control to make their self-experimentation compatible with our study . Investigators did not endorse any use of psychedelics , and no financial compensation was offered to participants . Email addresses were the only personally identifiable data collected . The email address was retained after study completion if permission was given ( checkbox ) by the participant to receive information regarding future studies , discarded otherwise . The study was approved by Imperial College Research Ethics Committee and the Joint Research Compliance Office at Imperial College London ( ICREC reference number 18IC4518 ) .
A total of 1630 participants signed-up , 240 started , and 191 participants completed the study . The optional follow-up at week 9 was completed by 159 individuals . No statistically significant differences were found between the groups in any demographic , recreational drug use or baseline measures , confirming efficiency of the randomization ( see Supplementary file 1 for details on demographics , Supplementary file 2 for recreational drug use , and Supplementary file 3 for statistical analysis of baseline variables ) . Completion rate was highly similar across the three groups ( χ2 ( 12 , N = 240 ) =0 . 64 , p=0 . 99 ) , see Figure 3 . For the most part , the sample consisted of educated , middle-age ( 33 . 5 ± 9 . 4 ) , healthy males ( 70% male , 19% female , 1% other ) from western countries . As expected , most participants had a positive attitude toward psychedelic drugs , in particular toward medical use: 74% and 90% either agreed or strongly agreed with the statements 'I am an active advocate of psychedelic drug-use' and 'I am an active advocate of the therapeutic use of psychedelics' , respectively . See Appendix for details on the sample’s expectations/attitude about microdosing and psychedelics . The sample consisted of healthy individuals for the most part: 33% of participants reported to have had at least one psychiatric diagnosis in the past , the most frequent past diagnoses were: anxiety disorder ( 13% ) , depression ( 13% ) , and PTSD ( 7% ) . Only 7% of the sample had current mental diagnosis . Most participants microdosed with LSD ( n = 147; 61% ) /LSD analogue ( n = 33; 14% ) , followed by psilocybin containing mushrooms ( n = 57; 24% ) and three individuals used other psychedelics ( LSA: n = 1; DOB: n = 2 ) . The average reported dose for LSD/LSD analogues was 13 ± 5 . 5 µg , while for psilocybin mushroom it was 0 . 2 ± 0 . 12 g , see Appendix 1—figure 3 for further details . Accumulative outcomes were first collected at baseline , then at week 5 ( i . e . after the completion of the 4 weeks long dose period ) and at the optional long-term follow-up timepoint at week 9 . The following two sets of pre-planned comparisons were made: within group comparisons of baseline vs . week 5 , baseline vs . week 9 ( changes over time ) and between-group comparisons at the week 5 and week 9 timepoints . Sample sizes were n = 240/191/159 at baseline , week 5 and week 9 , respectively . Data was also analyzed separately for LSD/LSD-analogues and psilocybin microdoses , the results from both subgroups matched the results of the combined analysis presented here . For the within group ( change over time ) comparison of baseline vs . week 5 , all self-reported psychological outcomes improved significantly in the MD group: well-being ( RPWB ) increased with 4 . 2 ± 3 . 9 ( adjusted mean estimate ±95% CI; p=0 . 04* ) , mindfulness ( CAMS ) increased with 2 . 4 ± 1 . 1 ( p<0 . 001*** ) , life satisfaction ( SWL ) increased with 1 . 2 ± 1 . 2 ( p=0 . 04* ) , and paranoia ( GPTS ) decreased with −5 . 0 ± 1 . 7 ( p<0 . 001*** ) . Personality structure ( B5 ) showed reduced neuroticism trait score ( −1 . 3 ± 0 . 9 , p<0 . 01** ) and increased openness ( 0 . 9 ± 0 . 8 , p=0 . 03* ) . Significant changes over the same period ( from baseline to week 5 ) were also observed in the PL and HH groups for mindfulness ( PL: 1 . 6 ± 1 . 1 , p<0 . 01**; HH: 1 . 3 ± 1 . 2 , p=0 . 02* ) and paranoia ( PL: −3 . 4 ± 1 . 7 p<0 . 001***; HH: −4 . 9 ± 1 . 9 p<0 . 001*** ) , but not for well-being or life satisfaction . Neuroticism also decreased in the PL group ( −1 . 0 ± 1 . 0 , p=0 . 04* ) . Changes in mindfulness and paranoia were sustained at the week 9 follow-up timepoint for all groups , while decreased neuroticism only prolonged in the MD group , see Supplementary file 5 for details . CPS did not change in the MD group ( from baseline to week 5 ) , but significantly decreased in the HH group ( −0 . 16 ± 0 . 14 , p=0 . 03* ) . Among individual cognitive tests over the same period , rotations ( 0 . 34 ± 0 . 28 , p=0 . 02* ) and odd one out ( 0 . 52 ± 0 . 31 , p=0 . 001** ) increased significantly in the MD group , while spatial span ( −0 . 49 ± 0 . 30 , p=0 . 02* ) and paired associates ( −0 . 51 ± 0 . 30 , p=0 . 02* ) decreased in the HH group . The increased rotations score in the MD group was sustained at the follow-up ( 0 . 45 ± 0 . 46 , p<0 . 01** ) , but not the other task scores . Planned comparisons revealed no significant between-group differences at either the week 5 or week 9 follow-up timepoints , including all subscales , except that in the HH group the paired associates scores decreased ( PL vs HH adjusted treatment difference: −0 . 55 ± 0 . 43 , p<0 . 01** ) . Time course of the adjusted mean estimates is summarized in Figure 4 . See Supplementary file 4 for descriptive statistics , including subscale and individual cognitive test scores , adjusted over time and between group differences ( Supplementary file 5 ) , and model parameters ( Supplementary file 6 ) . As secondary analysis to further examine the role of placebo-like expectation effects in the accumulative outcomes , we performed a post-hoc adjustment by adding the ‘number of times microdose capsule was guessed’ variable as a covariate to the models ( irrespective whether the guess was correct or not ) . This variable was significant for some models ( RPWB: p<0 . 01**; CAMS: p=0 . 02*; B5 agreeableness: p=0 . 02*; B5 openness: p=0 . 03* ) and further decreased the already small between-group differences on self-reported scales , while it did not affect cognitive outcomes . Specifically , the adjusted treatment difference ( ±95% CI ) at the week 5 timepoint between PL and MD groups without/with the number of MD guesses covariate was: well-being ( RPWB ) 2 . 5 ± 5 . 6 ( p=0 . 37 ) /0 . 9 ± 5 . 7 ( p=0 . 76 ) , mindfulness ( CAMS ) : 0 . 8 ± 1 . 5 ( p=0 . 32 ) /0 . 4 ± 1 . 5 ( p=0 . 65 ) , paranoia ( GPTS ) : −1 . 6 ± 2 . 5 ( p=0 . 21 ) /−1 . 2 ± 2 . 5 ( p=0 . 36 ) , life satisfaction ( SWL ) 0 . 4 ± 1 . 7 ( p=0 . 67 ) /0 . 2 ± 1 . 8 ( p=0 . 83 ) , B5 intellect: −0 . 2 ± 1 . 2 ( p=0 . 80 ) /−0 . 2 ± 1 . 2 ( p=0 . 71 ) , B5 openness: 0 . 3 ± 1 . 2 ( p=0 . 57 ) /0 . 0 ± 1 . 2 ( p=0 . 97 ) , B5 neuroticism: −0 . 3 ± 1 . 4 ( p=0 . 70 ) /−0 . 1 ± 1 . 4 ( p=0 . 87 ) , B5 extraversion: −0 . 2 ± 1 . 2 ( p=0 . 81 ) /−0 . 4 ± 1 . 3 ( p=0 . 52 ) , B5 agreeableness: 0 . 5 ± 1 . 1 ( p=0 . 37 ) /0 . 2 ± 1 . 1 ( p=0 . 75 ) , and B5 consciousness: 0 . 8 ± 1 . 3 ( p=0 . 24 ) /0 . 5 ± 1 . 3 ( p=0 . 44 ) . First , outcomes are described without considering the guess component , which is discussed in the next section . Acute outcomes were measured during the dose period while the potential microdose was still active , while post-acute outcomes were measured every Sunday , when no capsule was taken , 48–72 hr after the last placebo/microdose capsule . For psychological measures the average sample size was 857 ( between 849 and 884 due to partial completions; participants contributed four scores corresponding to the four acute timepoints , see Materials and methods for details ) , while for cognitive performance it was 684 ( between 678 and 689 ) . Data was also analyzed separately for LSD/LSD-analogues and psilocybin microdoses , and the results from both subgroups matched the results of the combined analysis presented here . Among acute measures , condition ( PL vs . MD ) was significant for acute emotional state ( PANAS ) ( adjusted mean estimate ±95% CI: 2 . 2 ± 1 . 4 , p<0 . 01** ) and the acute drug intensity ( 12 . 5 ± 3 . 0 , p<0 . 001*** ) , mood ( 4 . 6 ± 2 . 9 , p<0 . 001*** ) , energy ( 5 . 3 ± 2 . 7 , p<0 . 001*** ) , and creativity ( 4 . 7 ± 2 . 6 , p<0 . 001*** ) VASs , meaning that scores collected on days when a microdose was taken were significantly higher compared to scores collected on placebo days . Effect sizes , as quantified by Cohen’s d , remained small ( d < 0 . 3 ) on all scales , with the exception of the drug intensity VAS ( d = 0 . 58 ) . Among post-acute measures , condition was significant only on the anxiety measure ( STAIT; −1 . 4 ± 1 . 3 , p=0 . 03* ) , meaning that anxiety was reduced at the end of microdose weeks compared with placebo weeks , see Table 2 for details on both acute and post-acute outcomes . Next , the acute and post-acute results were re-analyzed with the addition of guess into the models . Condition ( PL vs . MD ) was no longer significant for any scale , except for acute drug intensity VAS ( adjusted mean difference ±95% CI: 3 . 4 ± 2 . 0; p<0 . 001*** ) , which increased under MD ( Table 2 ) . The guess*condition interaction term was non-significant for all scales , except for drug intensity ( p<0 . 01** ) . To better understand the role of guess , the data was further analyzed by comparing the 2*2 = 4 strata formed by the two binary variables , condition ( PL/MD ) , and guess ( PL/MD ) , in the models . For self-reported outcomes , no significant differences were found between microdose and placebo conditions with fixed guess ( condition/guess: PL/PL vs . MD/PL and PL/MD vs . MD/MD comparisons ) , except for acute drug intensity visual analogue scale , which was higher when microdose was taken ( adj . mean difference ±95% CI; 7 . 3 ± 3 . 1 , p<0 . 001*** ) . Conversely , when drug condition was fixed ( condition/guess: PL/PL vs . PL/MD and MD/PL vs . MD/MD comparisons ) , significant differences were found in 21 of the 22 comparisons ( =2*conditions* ( 4*post-acute+7*acute scales ) ) , all favoring MD guess . These findings suggest that scores are significantly better when the participant believed they had taken a microdose irrespective of what was actually taken . Taking an actual microdose was only associated with a significant difference in the drug intensity scale . Figure 5 shows the stratified distribution of selected outcomes , see Supplementary file 8 for all comparisons . Break blind rate , defined as the proportion of correct capsule guesses ( see section Blind breaking and collection of guess data for details ) , was 0 . 72 ± 0 . 18 ( M ± SD ) . Specificity ( true negative rate: ratio of true placebo guesses to all placebo guesses ) was 0 . 82 ± 0 . 16 , noticeably higher than sensitivity ( true positive rate: ratio of true microdose guesses to all microdose guesses ) 0 . 45 ± 0 . 30 , meaning that placebo capsules were guessed correctly at a higher rate than microdoses . Based on knowledge of the ratio of PL/MD capsules ( 3/1 ) in the envelopes , which is evident to participants when they prepare the capsules , a ‘random guesser’ would have a break blind rate of 0 . 62 with 0 . 75 specificity and 0 . 25 sensitivity . The high sensitivity exhibited by participants ( 0 . 46 vs . the random guesser’s 0 . 25 ) suggests that the higher than random break blind rate is mostly due to superior ability to identify microdoses , see Appendix 1—table 1 for details . Break blind rate was positively associated with reported microdose dose ( F ( 1 , 237 ) =7 . 4 , p<0 . 01** ) , meaning that the higher the dose was , the more likely participants guessed their daily condition correctly . For this analysis psilocybin mushroom doses were converted to estimated LSD dose equivalent , see Statistical analysis in Materials and methods for details . The estimated ‘detection threshold’ , that is , the dose above which participants guess significantly better than random , was 12 µg .
It is our view that the present part-controlled , part-observational design yields data superior to conventional observational data ( inclusion of placebo control ) , but inferior to controlled clinical trial data ( incomplete control over recruitment , screening , assessment , drug administration , etc . ) . This study does , however , have greater ecological validity than would a fully controlled lab study . A key limitation of the present study is the lack of verification of the nature , purity , and dosage of the psychedelic substance used for microdosing . Psilocybin-containing mushrooms were used by 23% of the sample , 14% used legal LSD analogues ( such as 1P-LSD ) , whereas 62% sourced their substance from the black market , mostly LSD ( 61% ) . According to the Energy Control's drug checking service ( Barcelona ) , LSD blotter adulteration rates were low during the period when our study was running: in both 2018 and 2019 blotters sold as LSD contained LSD only in 90% ( n = 735 ) of tested samples [personal communication with M . Ventrua from EC , June 2020] . The exact quantity of active ingredient within a given microdose cannot be known with certainty; however , the positive relationship between dose and blind breaking ( Figure 4 ) and that the threshold dose for psychoactivity was consistent with a recent controlled study ( 12 µg vs 13 µg; Bershad et al . , 2019a ) provide some reassurance . Nonetheless , our results should be not understood as clinical evidence , rather they are representative of ‘real life microdosing’ . We could not confirm whether participants followed accurately the self-blinding procedure . Three individuals reported following an invalid sequence of weeks , but these individuals did their setups together , all committing the same mistake ( 1 . 3% error rate ) . Furthermore , we had no way of confirming whether the capsules were taken as instructed during the dose period . Instructions emphasized not to complete assessments planned on dosing days in case the dose schedule could not be followed for any reason , but we could not confirm whether participants adhered to this rule . Our stratification analysis does not allow for a strict determination of a causal relationship between guess and outcome , because guess was recorded after completion of assessments , guess was last question during test sessions . After closing the study , a survey was conducted among participants , where 86% ( n = 166 ) responded that "I was thinking about whether I took a microdose or placebo even before I was asked to guess" ( opposed to "I was not thinking about whether I took a microdose or placebo , except when I was asked to guess" ) , making a causal interpretation more likely . We note that the order we chose is consistent with previous work in psychiatric studies ( Baethge et al . , 2013; Chen et al . , 2011; Rabkin et al . , 1986 ) ; had the guesses been requested prior to the assessments , it could have primed responses . Also , we cannot rule out that performance during the assessments influenced the guess . However , the lack of any feedback from the assessments mitigates this risk . Most participants reported to break blind due to body and perceptual sensations , rather than improved outcomes , see Blind breaking cues in the Appendix for details . We cannot rule out the possibility that a study in a clinical population would yield more promising results . In the present healthy sample , where well-being scores are high at baseline , there is less scope for potential improvements , which could have prevented the observation of placebo-microdose differences . Most study participants reported not to have any history of mental health problems; only 7% reported having a current psychiatric diagnosis , and 33% reported to have had a psychiatric diagnosis in the past ( Supplementary file 1 ) . We conducted two post-hoc analysis for two selective pseudo-depression subsamples: participants with the lowest 25% baseline well-being scores and those with the highest 25% baseline neuroticism scores ( Ryff and Keyes , 1995; Wood and Joseph , 2010 ) . Results in these subsamples were entirely consistent with those from the complete sample: there were no significant differences between conditions for any of the accumulative outcomes ( adjusted treatment difference ±95% CI of PL vs MD at week 5 for the lowest 25% baseline well-being subsample: well-being ( RPWB ) −1 . 6 ± 13 . 6 ( p=0 . 81 ) , mindfulness ( CAMS ) 0 . 3 ± 3 . 3 ( p=0 . 85 ) , paranoia ( GPTS ) −5 . 1 ± 6 . 8 ( p=0 . 14 ) , life satisfaction ( SWL ) 0 . 3 ± 4 . 5 ( p=0 . 87 ) , cognition ( CPS ) 0 . 1 ± 0 . 55 ( p=0 . 71 ) ; same measures for the highest 25% baseline neuroticism subsample: well-being ( RPWB ) 4 . 8 ± 14 . 3 ( p=0 . 50 ) , mindfulness ( CAMS ) 1 . 3 ± 3 . 7 ( p=0 . 49 ) , paranoia ( GPTS ) −3 . 1 ± 8 ( p=0 . 43 ) , life satisfaction ( SWL ) −1 . 4 ± 4 . 6 ( p=0 . 53 ) , cognition ( CPS ) 0 . 04 ± 0 . 67 ( p=0 . 90 ) ) . Thus , although not designed as a clinical study , data from this opportunistic naturalistic study do not provide support for clinical effects of microdosing . Although this was the largest placebo-controlled psychedelic research study published to-date , we note that one could argue that the study was still underpowered to detect a true effect based on the fact that the MD group did improve more than the PL group on all scales ( from baseline to week 5 ) , but just not to a statistically significant extent ( Figure 4 ) . On the well-being scale ( RPWB ) , the adjusted PL vs . MD group difference was 2 . 5 ± 5 . 6 points . To illustrate this difference in practice , this scale consists of 42 statements that participants rate on a 6-point Likert scale ( Strongly disagree - Strongly agree ) , thus , the full range of scores is thus 0–252 , so the 2 . 5 point mean difference is 1% of the total scale . This difference is equivalent to scoring one item , for example ‘I like most aspects of my personality’ , Strongly agree instead of Slightly agree or Slightly disagree , while responding the same to the remaining 41 items . Based on our data , we calculated that the sample size ( 90% power and alpha of 0 . 05 ) required to observe a true between-group difference would be: 1508 for well-being ( RPWB ) , 1638 for mindfulness ( CAMS ) , 4918 for life satisfaction ( SWL ) , 1392 for paranoia ( GPTS ) , and 366 for cognitive performance ( CPS ) . These differences therefore are not clinically meaningful or sufficient to justify the cost of intervention . The successful execution of this initiative here may inspire similar initiatives throughout the world in a broad range of scientific and medical contexts . Controlling for placebo effects is important for trending phenomena , such as cannabidiol ( CBD ) oils , nootropics , and nutrition , where social-pressure , expectancy , positive-test strategies , and confirmation bias can lead to false-positive findings . Self-blinding citizen-science initiatives could be employed in these areas as a cost-efficient screening tool prior to conducting expensive clinical studies . An important feature of the self-blinding methodology is the low cost; we estimate that the current study’s costs were about 0 . 5–1% of an equivalent clinical study . Since the research team is not providing the study drug/placebo and on-site staffing is not required , expenses are similar to a conventional observational study , yet still with incorporation of randomization and placebo control . Important lessons can be taken from the current study for the design of future microdosing trials . The combination of the lack of detected efficacy in this study and an association between self-reported doses and ability to break blind ( see Figure 4 ) suggest that selecting dosage is fraught with difficulties: if a low microdose is chosen , efficacy is unlikely if we extrapolate current results , whereas a high microdose could jeopardize the blinding . Randomization to microdose versus an active placebo conditions ( e . g . niacin , which has been employed in macrodosing studies Ross et al . , 2016 ) and careful assessment of blinding could , in principle , alleviate some of these concerns . The present study also has implications for full/‘macrodose’ psychedelic studies , where blinding is impossible due to the intense nature of the experience . It can be hypothesized that the intense hallucinations are essential for therapeutic outcome ( Griffiths et al . , 2011; Roseman et al . , 2017 ) , questioning the suitability of placebo-controlled trials in this context . The fact that one may be unable to fully extricate belief , or ‘context’ more broadly , from the direct ( e . g . pharmacological ) action of a given intervention , raises interesting philosophical and ethical question with implications for drug development and regulation . One might also hypothesize that the action of microdosing and psychedelics relies on prior and continuously updating belief combining ( perhaps synergistically ) with a direct drug effect ( Carhart-Harris et al . , 2015; Carhart-Harris and Friston , 2019 ) . Such a positive interaction could , in theory , be tested ( Carhart-Harris et al . , 2018 ) , and if endorsed , this could be interpreted as implying that belief is an active component of the psychedelic treatment model , rather than a problematic confound . In summary , here we created a novel , cost-effective , self-blinding , citizen-science methodology that enabled us to conduct the largest placebo-controlled study on psychedelics to-date and the first placebo-controlled examination of repeated psychedelic microdosing . Our findings confirm the anecdotal benefits of microdosing ( improvements in a broad range of psychological measures ) ; however , the results also suggest that the improvements are not due to the pharmacological action of microdosing , but are rather explained by the placebo effect ( lack of significant between-groups effect ) . | Psychedelic psychotherapy , therapy enhanced with psychedelic drugs such as LSD or psilocybin ( the active ingredient of ‘magic mushrooms’ ) , has been suggested to improve psychological well-being . For this reason , trials on psychedelic therapy for the treatment of depression , addiction and other conditions are ongoing . Recently , ‘microdosing’ – a way of administering psychedelics that involves taking about 10% of a recreational dose two or three times per week – has gained popularity . Unlike taking large doses of psychedelics , microdosing does not induce hallucinations , but anecdotal reports suggest that it yields similar benefits as psychedelic therapy . A key feature of modern medicine are ‘placebo control’ studies that compare two groups of patients: one that takes a drug and another that takes inactive pills , known as placebos . Crucially , neither group knows whether they are taking drug or placebo . This control ensures that observed effects are due to the drug itself and not to unrelated psychological causes . For example , in trials of mood medicines , participants often expect to feel happier , which in itself improves their mood even when taking a placebo . This is known as the placebo effect . Restrictive drug policies make placebo-controlled studies on psychedelics difficult and expensive , in particular for microdosing , which involves taking psychedelics over a longer time period . To overcome this problem , Szigeti et al . developed a new citizen-science approach , where microdosers implemented their own placebo control based on online instructions . The advantages are the low cost and the ability to recruit participants globally . The experiment was completed by 191 microdosers , making it the largest placebo-controlled study on psychedelics to-date , for a fraction of the cost of an equivalent clinical study . The trial examined whether psychedelic microdosing can improve cognitive function and psychological well-being . The team found that microdosing significantly increased a number of psychological measures , such as well-being and life satisfaction . However , participants taking placebo also improved: there were no significant differences between the two groups . The findings confirmed positive anecdotes about microdosing improving people’s moods , but at the same time show that taking empty capsules , knowing they might be microdoses , have the same benefits . This result suggests that the observed benefits are not caused by the microdose , but rather by psychological expectations . The study’s innovative ‘do-it-yourself’ approach to placebo control may serve as a template for future citizen science studies on other popular phenomena where positive expectations and social factors could play a role , such as cannabidiol ( CBD ) oils , nootropics and nutrition . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"medicine",
"neuroscience"
] | 2021 | Self-blinding citizen science to explore psychedelic microdosing |
Viral infection is usually studied at the population level by averaging over millions of cells . However , infection at the single-cell level is highly heterogeneous , with most infected cells giving rise to no or few viral progeny while some cells produce thousands . Analysis of Herpes Simplex virus 1 ( HSV-1 ) infection by population-averaged measurements has taught us a lot about the course of viral infection , but has also produced contradictory results , such as the concurrent activation and inhibition of type I interferon signaling during infection . Here , we combine live-cell imaging and single-cell RNA sequencing to characterize viral and host transcriptional heterogeneity during HSV-1 infection of primary human cells . We find extreme variability in the level of viral gene expression among individually infected cells and show that these cells cluster into transcriptionally distinct sub-populations . We find that anti-viral signaling is initiated in a rare group of abortively infected cells , while highly infected cells undergo cellular reprogramming to an embryonic-like transcriptional state . This reprogramming involves the recruitment of β-catenin to the host nucleus and viral replication compartments , and is required for late viral gene expression and progeny production . These findings uncover the transcriptional differences in cells with variable infection outcomes and shed new light on the manipulation of host pathways by HSV-1 .
Viruses are obligatory intracellular parasites that rely on the biochemical functions of their hosts to carry out infection . Although usually studied at the level of cell populations , viral infection is inherently a single-cell problem , where the outcome of infection can dramatically differ between genetically identical cells . For example , early studies in the 1940 s investigated the burst size of individually infected bacteria and concluded that it both spans three orders of magnitude and cannot be solely attributed to differences in bacteria size ( Delbrück , 1945 ) . A later study measured the burst size from individual HeLa cells infected with Herpes Simplex virus 1 ( HSV-1 ) and found that many of the infected cells did not release viral progeny , that the variability between individual cells was high and that it did not correlate with the multiplicity of infection ( MOI ) used ( Wildy et al . , 1959 ) . More recently , technological improvements have allowed the quantification of burst sizes and determination of the infection kinetics of different mammalian viruses , pointing to a high degree of cell-to-cell variability in infection ( Zhu et al . , 2009; Timm and Yin , 2012; Schulte and Andino , 2014; Combe et al . , 2015; Heldt et al . , 2015; Cohen and Kobiler , 2016; Guo et al . , 2017; Drayman et al . , 2017 ) . One well-known source of this variability is the random distribution of the number of viruses that individual cells encounter ( Parker , 1938; Smith , 1968; Cohen and Kobiler , 2016 ) . Another source is genetic variability in the virus population , with some virus particles being unable or less fit to establish infection ( Huang and Baltimore , 1970; Lauring et al . , 2013; Stern et al . , 2014 ) . It is becoming clear that a third source of variability among individually infected cells is the host cell state at the time of infection ( Snijder et al . , 2009; Snijder et al . , 2012; Drayman et al . , 2017 ) . Variability in the host cell state can arise from both deterministic processes , such as the cell-cycle , and stochastic processes , such as mRNA transcription and protein translation ( Elowitz et al . , 2002; Cohen et al . , 2008; Tay et al . , 2010; Loewer and Lahav , 2011; Kellogg and Tay , 2015 ) . Recently , the advent of single-cell RNA-sequencing ( scRNA-seq ) has allowed researchers to examine virus–host interactions in multiple systems , mainly those of RNA viruses , highlighting the extreme cell-to-cell variability during viral infection ( Steuerman et al . , 2018; Russell et al . , 2018a; Xin et al . , 2018; Zanini et al . , 2018; Shnayder et al . , 2018 ) . Thus , it is clear that a better understanding of viral infection requires studies at the single-cell level . Although scRNA-seq is providing a wealth of new information , it is currently limited to the characterization of highly abundant transcripts and can be augmented by other approaches such as live-cell imaging and RNA-sequencing of defined cell populations . HSV-1 is a common human pathogen that belongs to the herpesviridae family and serves as the prototypic virus for studying alpha herpesviruses infection . De novo HSV-1 infection has both lytic and latent phases . In the lytic phase , the virus infects epithelial cells at the site of contact , replicates withing these host cells , before destroying them and releasing viral progeny . The latent phase is restricted to neurons , in which the virus remains silent throughout the host life with occasional reactivation . Here , we focus on the lytic part of the virus life cycle . Lytic infection is usually asymptomatic , but in some cases ( particularly in immune-compromised individuals and infants ) , it can result in life-threatening conditions such as meningitis and encephalitis . To initiate infection , HSV-1 must bind to its receptors , enter the cytoplasm , travel to the nuclear pore and inject its linear double-stranded DNA into the host nucleus ( Kobiler et al . , 2012 ) . Once in the nucleus , viral gene expression proceeds in a temporal cascade involving three classes of viral genes: immediate-early ( IE ) , early ( E ) and late ( L ) ( Honess and Roizman , 1974; Honess and Roizman , 1975; Harkness et al . , 2014 ) . Transcription of IE genes is initiated by VP16 ( Weir , 2001 ) , whereas transcription of the E and L genes is activated by the IE protein ICP4 ( Dixon and Schaffer , 1980; Watson and Clements , 1980; DeLuca et al . , 1985 ) . ICP0 , another IE gene , is a multifunctional E3-ubiquitn ligase that counteracts some of the host anti-viral systems ( Lanfranca et al . , 2014 ) and interferes with the action of transcriptional repressors ( Gu et al . , 2005; Lomonte et al . , 2004; Lutz et al . , 2017 ) . Viral mutants that lack ICP0 expression ( ΔICP0 ) are highly attenuated in a variety of cell types ( Stow and Stow , 1986 ) . Viral DNA replication occurs in sub-nuclear structures , called replication compartments ( RCs ) , that aggregate the seven essential replication proteins as well as other viral and host proteins ( de Bruyn Kops and Knipe , 1988; Liptak et al . , 1996; Weller and Coen , 2012; Dembowski and DeLuca , 2015; Dembowski et al . , 2017; Reyes et al . , 2017; Dembowski and DeLuca , 2018 ) . Upon viral DNA replication , ICP4 is predominantly localized in the RCs , with some diffuse nuclear and cytoplasmic localization ( Knipe et al . , 1987; Zhu and Schaffer , 1995 ) . Several studies applied high-throughput technologies to analyze the cellular response to HSV-1 infection at the population level . RNA sequencing revealed a widespread deregulation of host transcription , including the disruption of transcription termination ( Rutkowski et al . , 2015; Hennig et al . , 2018 ) , activation of anti-sense transcription ( Wyler et al . , 2017 ) , depletion of RNA-polymerase II from the majority of host genes ( Abrisch et al . , 2015; Birkenheuer et al . , 2018 ) and changes in splicing and polyadenylation ( Hu et al . , 2016 ) . Although most cellular genes are downregulated by infection , some genes have been reported to be upregulated , including some anti-viral genes and genes encoding host transcription factors ( Pasieka et al . , 2006; Taddeo et al . , 2002; Hu et al . , 2016 ) . Proteomics studies have defined the different stages and protein complexes that are present during HSV-1 replication ( Dembowski and DeLuca , 2015; Suk and Knipe , 2015; Dembowski et al . , 2017; Reyes et al . , 2017; Dembowski and DeLuca , 2018 ) , as well as the cellular protein response to infection ( Kulej et al . , 2017; Lum et al . , 2018 ) . Although incredibly informative , population-level analyses suffer in that they average over all the cells in the population . In the case of virus-infected cells , the population is far from homogenous and could in fact contain opposite phenotypes , such as highly- infected and abortively infected cells , leading to contradictory results . One such example is the seemingly complex relation between HSV-1 infection and type I interferon ( IFN ) signaling . The picture that emerges from population-level measurements is paradoxical , with wildtype HSV-1 infection both clearly activating ( Gianni et al . , 2013; Hu et al . , 2016; Liu et al . , 2016; Reinert et al . , 2016 ) and clearly repressing ( Lin et al . , 2004; Johnson et al . , 2008; Kew et al . , 2013; Johnson and Knipe , 2010; Su et al . , 2016; Christensen et al . , 2016; Manivanh et al . , 2017; Yuan et al . , 2018; Chiang et al . , 2018 ) the type I IFN pathway . Such discrepancies might be resolved with the use of single-cell measurements . Here , we apply a combination of live-cell time-lapse fluorescent imaging , scRNA-seq and the sequencing of sorted cell populations to explore HSV-1 infection at the single-cell level . We find that single cells that are infected by the virus show variability in all aspects of infection , starting from the initial phenotype ( abortive infection vs . successful initiation of viral gene expression ) , through the timing and rate of viral gene expression , and ending with the host cellular response . This study resolves the apparent discrepancy in the literature regarding type I IFN induction and shows that this induction is restricted to a rare sub-population of abortively infected cells . Surprisingly , we find that the main transcriptional response in highly infected cells is the reprogramming of the cell to an embryonic-like state . We focus on the viral activation of the WNT/β-catenin pathway and find that β-catenin is recruited to the cell nucleus and the viral RCs , and is required for viral gene expression and progeny production .
We began by studying the temporal variability in viral gene expression initiation . To do so , we employed a wildtype HSV-1 ( strain 17 ) that was genetically modified to express ICP4-YFP ( Everett et al . , 2003 ) . Primary human fibroblasts ( HDFn ) were infected and monitored by time-lapse fluorescent microscopy ( Figure 1A , Figure 1—video 1 ) . HDFn were infected at an MOI of 2 ( calculated on the basis of virus titration on Vero cells , which are ~2-fold more susceptible to HSV-1 infection than HDFn ) . This MOI was chosen because we found empirically that it resulted in ~50% of HDFn becoming ICP4-positive during primary infection . Note that we determined the genome:PFU ratio for our viral stock and found it to be 36 ± 4 , suggesting that all the cells in the culture have probably encountered numerous virus particles . Initiation of ICP4 expression was observed to occur mostly between 1 hr and 4 hr post-infection ( Figure 1B ) . Almost no new infections were observed between 4 hr and 6 hr post-infection , but two infection peaks were later seen at 8 and 11 hr . These peaks are probably the result of secondary infections , as new viral progeny can be detected in infected cells starting at 6 hr post-infection ( Pomeranz and Blaho , 2000; Ikeda et al . , 2011; Drayman et al . , 2017 ) . Given that the majority of infected cells have initiated viral gene expression by 5 hr , we chose this time point for further analyses . HDFn were infected as described above , fixed and stained with DAPI at 5 hr post-infection ( to allow automated cell segmentation and quantification ) . This allowed the distinction of two cellular populations: cells that successfully initiated viral gene expression ( ICP4+ ) and cells that did not ( ICP4– ) . Of 1814 cells exposed to HSV-1 , 996 cells ( 55% ) were ICP4+ and 818 ( 45% ) were ICP4– . Cells were classified as ICP4-negative or -positive on the basis of a threshold calculated from mock-infected cells ( mean +3 standard deviations ) . Among the ICP4+ cells , nuclear levels of ICP4 varied by ~100 fold , ranging from 7 × 104 to 9 × 106 AU ( Figure 1C ) . Infected cells showed three distinct phenotypes related to ICP4 localization ( Figure 1D ) . Upon its expression , ICP4 is initially diffuse throughout the cell nucleus . As its level increases , ICP4 forms discrete foci in the nucleus . These are the viral RCs , where viral DNA replication takes place . Later , the levels of cytoplasmic ICP4 increase and interspersed foci can be seen in the cytoplasm . These phenotypes are temporally linked and delineate the progression through infection . As evident by time-lapse microscopy ( Figure 1—Video 1 ) , individual cells show a high degree of variability , not only in the timing of initial gene expression but also in the rate of infection progression . Taken together , we find that not all cells that are exposed to HSV-1 successfully initiate viral gene expression under these experimental conditions . Those that do initiate viral gene expression show variation in the timing of initial gene expression , in the rate of infection progression and in the level and localization of the immediate-early protein ICP4 . These results prompted us to explore cellular heterogeneity on a larger scale by applying single-cell RNA-sequencing ( scRNA-seq ) to infected cells . HDFn were mock-infected or infected with wildtype HSV-1 ( MOI 2 , equivalent to ~70 genomes per cell , see 'Materials and methods' ) or a ΔICP0 HSV-1 mutant ( MOI 0 . 5 , equivalent to ~700 genomes per cell ) and harvested for scRNA-seq at 5 hr post-infection . We chose to include the ΔICP0 mutant because it results in a relatively high number of abortive infections and in a robust activation of anti-viral responses . For scRNA-seq , we applied the Drop-seq protocol ( see 'Materials and methods' and Macosko et al . , 2015 ) . Briefly , a microfluidic device was used to encapsulate individual cells in a water-in-oil droplet in which cell lysis , mRNA-capture and barcoding took place . The barcoded mRNA was then recovered from the droplets , reverse-transcribed , amplified and sequenced . As each cDNA was barcoded with a cell and transcript ID , the sequencing data allow the reliable quantification of the number of transcripts in individual cells . Technical data on sequencing depth and filtering criteria are presented in Figure 1—figure supplement 1 . Only 0 . 4% of mock-infected cells had any reads aligned to the HSV-1 genome , with a maximal expression of two viral gene counts ( 0 . 05% of transcripts ) . Cells that were infected with either wt or ΔICP0 HSV-1 showed extreme cell-to-cell variability in the amount of viral transcripts that they expressed , ranging from 0–36% ( Figure 1E ) . The viral gene expression distribution was highly skewed , with most cells expressing low levels of viral transcripts and some cells expressing much higher levels ( Figure 1E ) . A joint analysis of mock and wt-infected cells showed that highly infected cells clustered separately from the mock-infected and low-infected cells , which were intermingled ( Figure 1—figure supplement 2 ) . The Gini coefficient , a measurement of population inequality ranging from zero ( complete equality ) to one ( complete inequality ) , was used to evaluate the distribution of viral gene expression among individual cells . The Gini coefficients were 0 . 8 for wildtype infection and 0 . 77 for ΔICP0 infection , higher than that reported for viral gene expression by an influenza virus ( 0 . 64; Russell et al . , 2018a ) . When wildtype viral gene expression is visualized in two-dimensions ( using the tSNE dimensionality reduction technique; van der Maaten and Hinton , 2008 ) , two clusters of cells can be seen , which are distinguished by the amount of viral gene expression ( less or more than ~1% , Figure 1F ) . A similar distribution was seen for ΔICP0-infected cells , although there were significantly fewer cells in the ‘highly infected’ cluster in this case ( Figure 1—figure supplement 3 ) . To further explore cell-to-cell variability in viral gene expression , we analyzed the relative expression of the four temporal groups of viral transcripts: immediate-early ( IE ) , early ( E ) and late ( subdivided into early-late ( γ1 ) and true-late ( γ2 ) ) . We focused on the group of highly infected cells , because the lowly infected cells had too few viral gene counts for accurate analysis . Figure 1G shows the relative expression of the viral gene classes in single cells , ordered from low to high viral gene expression . The fraction of late genes increases as total viral gene expression increases , at the expense of IE and E genes . The correlations between viral gene expression and the four classes of viral transcripts are shown in Figure 1H–K . Similar observations were made for ΔICP0-infected cells ( Figure 1—figure supplement 3 ) . Our scRNA-seq data indicate a wide and uneven distribution of viral gene expression during HSV-1 infection , with most cells expressing no or low levels of viral gene transcripts and a smaller group expressing much higher levels ( in agreement with the ICP4 expression levels presented above ) . The vast majority of cells exposed to HSV-1 , either wildtype or ΔICP0 , had some level of viral gene expression , suggesting that the fraction of lowly expressing cells ( and the ICP4– population noted above ) are indeed abortively infected cells , rather than cells that did not encounter a virus . We note that significant cell-to-cell differences are seen even within the group of highly infected cells , with viral gene expression ranging from 1% to >30% , and that this ‘viral expression load’ is correlated with late gene expression . We have previously shown that the cell-cycle stage at the time of infection is a cellular determinant of successful HSV-1 infection in the H1299 cell line ( Drayman et al . , 2017 ) . In that study , we found that cells that were infected in the G2 phase were less likely to initiate viral gene expression and that cellular escape from viral-enforced mitotic arrest is highly detrimental to HSV-1 infection . To evaluate the effect of the cell-cycle in HDFn infection , we calculated a cell-cycle score for each cell in our dataset ( see 'Materials and methods' section and Tirosh et al . , 2016 ) and measured the correlation between HSV-1 gene expression and the cell-cycle score ( Figure 2—figure supplement 1 ) . We found that viral gene expression was negatively correlated with the cell-cycle score , with cells in the later parts of the cell-cycle expressing ~10-fold fewer viral genes than those in the early part of the cycle , in agreement with our previous finding in the H1299 cell-line . As the cell-cycle is both a major source of cell-to-cell variability and negatively correlated with viral gene expression , it was crucial to regress out the cell-cycle effect before analyzing the host response ( Figure 2—figure supplement 1 ) . We could now turn to analyzing the host genes that are differentially expressed among HSV-1-infected cells , starting with the anti-viral response . Previous population-level studies reported the activation of anti-viral genes during wildtype HSV-1 infection , so we hypothesized that highly infected cells ( Figure 2A , B , cluster 1 ) should be enriched for anti-viral genes . To our surprise , differential gene expression analysis of the two clusters did not indicate upregulation of the anti-viral response in cluster 1 ( Supplementary file 1a ) . In fact , canonical anti-viral genes such as IFIT2 and IFIT3 were only detected in 2–3% of the cells from both clusters 1 and 2 ( Figure 2C ) . When comparing a larger panel of interferon-stimulated genes ( ISGs ) in high- vs low-infected cells ( Figure 2—figure supplement 2 ) , most ISGs are in fact more highly expressed in cells with low HSV-1 gene expression . One possible explanation is that anti-viral genes are indeed expressed in highly infected cells but were not detected by scRNA-seq because of technical limitations . To investigate this , infected cells were sorted by fluorescence-activated cell sorting ( FACS ) into two populations based on ICP4-YFP expression ( ICP4+ and ICP4– ) , and each population was sequenced . In agreement with the scRNA-seq data , the expression of canonical anti-viral genes was not significantly different between mock-infected and ICP4+ cells . Rather , our analysis indicated that a small group of genes , including the anti-viral genes IFIT1 and IFIT2 , were specifically upregulated in the ICP4– population ( Figure 2D , Supplementary file 2a ) . The Gene Ontology ( GO ) biological processes associated with these upregulated genes included terms such as ‘response to type I interferon’ and ‘immune response’ ( Supplementary file 2b ) . Validation of selected transcripts by quantitative PCR ( QPCR ) is shown in Figure 2E . ISGs are usually expressed after interferon stimulation , but we failed to detect the activation of any of the interferon genes in either the scRNA-seq or the sequencing of the sorted cell populations . To pinpoint the origin of the anti-viral response , cells were stained for IRF3 , because IRF3 translocation from the cytoplasm to the nucleus is one of the first steps in type I interferon induction . In ICP4+ cells , IRF3 was blocked from entering the nucleus and concentrated in the nuclear periphery ( Figure 2F ) , while a rare subset of ICP4– cells ( <1% ) showed nuclear localization of IRF3 . We thus conclude that wildtype HSV-1 infection efficiently blocks the induction of the anti-viral response and that the activation of anti-viral genes in population-averaged measurements is the result of anti-viral signaling that is elicited in a rare population of abortively infected cells . We next evaluated the anti-viral response in cells infected by ΔICP0 HSV-1 . ICP0 is a multifunctional viral protein , which blocks IRF3 signaling ( Lin et al . , 2004 ) . Cells infected with ΔICP0 clustered into four groups ( Figure 3A ) . Cluster 1 consists of cells with very few viral transcripts . Clusters 2 and 3 have slightly higher viral gene expression , and the small cluster 4 consists of highly infected cells ( Figure 3B , D ) . Although the magnitude of the anti-viral response in ΔICP0-infected cells was much greater than that in the wildtype-infected cells ( Figure 2 ) , it was still only observed in a small population of cells , with ~8% of the cells expressing IFIT1 , MX2 or OASL ( compared to none of the mock-infected cells ) . These cells had low viral gene expression levels and mostly belonged to clusters 1–3 ( Figure 3C , D ) . Anti-viral signaling was not seen in the highly infected cells of cluster 4 , with the exception of two cells expressing IFIT1 ( Figure 3C , D ) . As described above for the wildtype infected cells , we also compared a larger panel of ISGs in highly vs lowly infected cells ( Figure 3—figure supplement 1 ) , with similar results . RNA-sequencing of sorted cells that were infected with ΔICP0 identified ~80 genes that are significantly upregulated in ICP4– cells compared to mock and ICP4+ cells ( Figure 3E , Supplementary file 2c ) . These genes were enriched for functional annotations of anti-viral signaling ( Figure 3F , Supplementary file 2d ) and for binding sites of the transcription factors IRF1 , IRF7 and STAT5 ( Figure 3G , Supplementary file 2e ) . An important difference from wildtype infection is that , although enriched in the ICP4– population , these anti-viral genes are also activated in the ICP4+ population , albeit to a lesser extent ( Figure 3H ) . We confirmed this observation through immunofluorescent staining of IRF3 ( Figure 3I ) . The staining showed that a higher proportion of ΔICP0-infected cells showed nuclear IRF3 localization when compared to wildtype-infected cells . The majority of cells with nuclear IRF3 were ICP4– but some ICP4+ cells also showed nuclear IRF3 staining . These ICP4+ cells showed diffuse nuclear localization of ICP4 , indicating that their infection was aborted prior to the generation of replication compartments ( Figure 1D ) . The few cells that were able to proceed to the later stages of infection ( as indicated by the appearance of replication compartments and cytoplasmic ICP4 foci ) showed the same peri-nuclear aggregation of IRF3 as wildtype-infected cells ( Figure 3I and Figure 2F ) . Altogether , the sequencing of both single cells and sorted cell populations , as well as immuno-fluorescence staining of IRF3 , suggests that the anti-viral program is initiated in a small subset of abortively infected cells but is blocked in highly infected cells , even in the absence of ICP0 . This behavior explains the apparent discrepancy in previous population-level measurements that showed both activation and inhibition of type I interferon signaling during HSV-1 infection . We next focused on genes that are upregulated during HSV-1 infection , in either the scRNA-seq or sorted cell population experiments . A total of 977 genes were significantly upregulated in wildtype ICP4+ cells as compared to both ICP4– and mock-infected cells in the bulk RNA-sequencing ( Figure 4A , Supplementary file 3a ) . In the single-cell RNA-sequencing , 21 genes were significantly upregulated in highly infected single cells ( Figure 4B , Supplementary file 1a ) . Remarkably , we found that a major portion of these upregulated genes are associated with GO terms that concern the regulation of RNA transcription and developmental processes ( Figure 4C , D and Supplementary files 1b and 3b ) . Similar results were observed in cells infected with ΔICP0 ( Figure 4—figure supplement 1 and Supplementary files 1d-f and 3d-f ) . The promoters of these upregulated genes are enriched for the binding sites of several transcription factors , including LEF1 , TCF3 and MYC ( Figure 4E , F and Supplementary files 1c and 3c ) . 23% of the promoters of the upregulated genes in ICP4+ cells contained a binding site for the TCF/LEF transcription factors , which are activated by the WNT/β-catenin pathway ( Sokol , 2011 ) ( Figure 4G ) . Note that LEF1 itself is a WNT target gene that is upregulated during HSV-1 infection ( Figure 5A ) . Examples of upregulated genes are shown in Figure 5A and include canonical WNT target genes such as AXIN2 , key developmental genes belonging to the SOX , HOX and HES families , stem-cell associated transcripts such as LGR5 , and a multitude of extra-cellular ligands of various developmental pathways , including the WNT , Notch , Hedgehog and TGFβ signaling pathways . In agreement with the less-efficient infection by ΔICP0 , most of these transcripts are also upregulated in ΔICP0-infected ICP4+ cells , but to a lesser degree than in wildtype-infected cells . Concomitant with the establishment of this embryonic-like transcriptional program , we observed a reduction in the levels of key fibroblast-marker genes , such as those encoding α1 ( III ) collagen and fibronectin ( Figure 5B ) . We conclude that cells that are highly infected by HSV-1 undergo de-repression of embryonic and developmental genetic programs , including the WNT/β-catenin pathway . As many of the upregulated genes in infected cells are known WNT target genes and/or contain LEF/TCF-binding sites in their promoters , we investigated the state of β-catenin in these cells . Infected cells were fixed and stained for β-catenin at 5 hr post-infection ( Figure 6A ) . As expected , β-catenin was mainly cytoplasmic/membrane-bound in mock-infected cells . In HSV-1-infected cells , β-catenin showed three distinct localization patterns: un-perturbed ( cytoplasmic ) , diffuse nuclear or aggregated in nuclear foci ( Figure 6A , B ) . Similar results were obtained for cells infected by ΔICP0 ( Figure 6—figure supplement 1 ) . At 5 hr post-infection , 37% of the cells were ICP4 negative , 14% were at the earliest stage of infection ( diffuse nuclear ICP4 ) , 31% had assembled viral replication compartments and 18% had progressed to show cytoplasmic foci of ICP4 ( Figure 6C ) . ICP4 levels increased from one group to the next , in accordance with the temporal progression of infection ( Figure 6D ) . β-catenin localization was linked to the temporal progression of infection . β-catenin remained cytoplasmic in both ICP4–cells and cells with diffused nuclear ICP4 , translocated to the nucleus upon the generation of the replication compartments and subsequently co-localized with the RCs , but only in cells showing cytoplasmic foci of ICP4 ( Figure 6B , E ) . A similar phenotype was seen in two epithelial cell-lines: A549 , a lung cancer cell-line , and Mel624 , a patient-derived melanoma cell-line ( Figure 6F ) . This analysis indicates that β-catenin is indeed co-opted by HSV-1 . Cell-to-cell variability in the progression of infection results in heterogeneity of β-catenin localization , with recruitment of β-catenin to the viral replication compartments occurring during the later stages of infection . As β-catenin target genes are activated by the virus , and because β-catenin is recruited to the viral replication centers , we hypothesized that β-catenin activity is required for the completion of the viral life cycle . We infected cells treated with an inhibitor of β-catenin activity , iCRT14 ( Gonsalves et al . , 2011 ) , and measured viral gene expression at 5 hr post-infection ( Figure 7A–D ) . Our results indicate that β-catenin inhibition had no or minimal impact on immediate-early gene expression ( Figure 7A ) but significantly inhibited early and late gene expression ( Figure 7B–D ) . These observations are in agreement with the late recruitment of β-catenin to the viral RCs described above . To measure the impact of β-catenin inhibition on viral progeny formation , we treated the cells with iCRT14 , infected the cells for 24 hr and harvested and titrated the resulting viral progeny by plaque assay ( Figure 7E ) . In accordance with its impact on late viral gene expression , β-catenin inhibition significantly reduced viral progeny formation ( Figure 7F ) . This is in agreement with a recent report by Zhu and Jones ( 2018 ) showing iCRT14-reduced plaque formation in two other cell types ( HLF and Vero ) . Similar results were obtained when β-catenin was silenced using siRNA ( Figure 7G , H ) . Taken together , our data show that HSV-1 reprograms the cell to an embryonic-like state , in part through the co-option of β-catenin , which is needed for late viral gene expression and progeny production .
In this study , we applied a combination of time-lapse fluorescent imaging , scRNA-seq and sequencing of sorted cell populations to understand HSV-1 infection at the single-cell level . We find that single cells tha are infected by the virus show variability in all aspects of infection , starting from the initial phenotype ( abortive infection vs . successful initiation of viral gene expression ) , through the timing and rate of viral gene expression and ending with the cellular response of the host cell . Such heterogeneity in the population of infected cells is detrimental when performing population-averaged measurements but can be untangled through single-cell analyses , which can provide new insights into biological processes . With regard to IFN response , we find that two opposite phenotypes exist in the population of infected cells , explaining the discrepancy in the literature . Surprisingly , we find that IFN induction is limited to a small group of abortively infected cells , even in cells infected by the ΔICP0 mutant ( Figures 2 and 3 ) . This rare activation of anti-viral signaling seems to be widespread , as single-cell investigations into RNA viruses such as West-Nile virus ( O’Neal et al . , 2019 ) and influenza ( Russell et al . , 2018a; Russell et al . , 2018b ) have reported similar findings . Why only a subset of cells activate the anti-viral program is an intriguing question and several hypotheses come to mind . For example , these cells could be poised for IFN induction due to stochastic variability in expression of the signaling pathway components ( Zhao et al . , 2012; Patil et al . , 2015 ) or IFN induction might be linked to the number of viral particles that a cell encounters . We plan to pursue and further characterize these rare cells in future studies . We further found that highly infected cells undergo transcriptional reprogramming and activate multiple developmental pathways . This is exemplified by changes in β-catenin localization that correspond to distinct stages in HSV-1 infection ( Figure 6 ) : β-catenin is first recruited to the cell nucleus and then later to the viral replication compartments . These findings augment a growing body of literature that shows a link between viral infection and the β-catenin pathway ( reviewed in van Zuylen et al . , 2016 ) , such as during infections with murine cytomegalovirus ( MCMV ) ( Juranic Lisnic et al . , 2013 ) , influenza ( More et al . , 2018 ) , hepatitis B virus ( HBV ) ( Daud et al . , 2017 ) and Rift Valley Fever virus ( Harmon et al . , 2016 ) . How HSV-1 infection causes this massive reprogramming of the host cell state is currently unknown , although we can rule out a direct involvement of ICP0 because this reprogramming also occurs during infection with ΔICP0 , albeit to a lesser extent . β-catenin activation is certainly one part of this , but it is likely that the expression of epigenetic regulators is also important . Indeed , a proteomics study of host chromatin during HSV-1 infection has identified widespread changes in the epigenomic landscape of infected cells ( Kulej et al . , 2017 ) . Although we describe a positive role for β-catenin activation during HSV-1 infection here , a recent report described an inhibitory role for the germline transcription factor DUX4 during HSV-1 infection ( Full et al . , 2019 ) . Thus , future studies will need to tease apart the differential contributions and effects of different developmental pathways that are activated during infection . β-catenin and other developmental pathways are often found to be mutated or dysregulated during tumorigenesis ( Reya and Clevers , 2005; Krausova and Korinek , 2014; Duchartre et al . , 2016 ) and are considered to be promising targets for cancer treatment ( Takebe et al . , 2015 ) . In this context , it is interesting to speculate as to the possible impact of β-catenin activation by HSV-1 on its use as an oncolytic agent ( Sanchala et al . , 2017; Watanabe and Goshima , 2018 ) . At present , the first-line treatment for late-stage melanoma is the use of immune checkpoint inhibitors ( Tracey and Vij , 2019 ) . These inhibitors revolutionized melanoma treatment , but not all patients respond to them . This heterogeneity was shown to be associated with β-catenin activity in the tumor , where high β-catenin levels negatively correlate with treatment success ( Spranger and Gajewski , 2015; Spranger et al . , 2015 ) . Given that an HSV-1 -based oncolytic therapy has been FDA approved for late-stage melanoma ( Pol et al . , 2015 ) , it is tempting to speculate that the high level of β-catenin in melanomas that are resistant to checkpoint inhibitors would serve to augment oncolytic HSV-1 replication and anti-tumor effects , although this of course would have to be assessed carefully in separate studies .
Primary neonatal human dermal fibroblasts ( HDFn ) were purchased from Cascade Biologics ( cat #C0045C ) , grown and maintained in medium 106 ( Cascade Biologics , cat #M106500 ) supplemented with Low Serum Growth Supplement ( Cascade Biologics , cat #S00310 ) . Cells were maintained for upto eight passages , and experiments were performed on cells between passages 4 and 7 . A549 cells were purchased from Sigma-Aldrich and maintained in DMEM supplemented with 10% fetal bovine serum . Mel624 , a patient-derived melanoma cell-line , was obtained from the lab of Professor Thomas Gajewski at the Univeristy of Chicago and maintained in RPMI supplemented with HEPES , NEAA , Pen/Strep and 10% fetal bovine serum . Vero and U2OS cells ( obtained from the laboratory of Matthew D . Weitzman , University of Pennsylvania ) were grown in DMEM supplemented with 10% fetal bovine serum and were used for viral propagation and titration . All of the cells that were used were routinely subjected to mycoplasma testing by PCR and found negative . Wildtype and ΔICP0 HSV-1 ( strain 17 ) viruses expressing ICP4-YFP were generated by Roger Everett ( Everett et al . , 2003 ) and were a kind gift from Matthew D . Weitzman . Viral stocks were prepared by infecting Vero cells ( for wildtype virus ) or U2OS cells ( for ΔICP0 ) at an MOI of 0 . 01 . Viral progeny were harvested 2–3 days later using three cycles of freezing and thawing . Viral stocks were titrated by plaque assays on Vero cells , aliquoted and stored at −80°C . iCRT14 , a β-catenin inhibitor , was purchased from Sigma-Aldrich ( cat #SML0203 ) and dissolved in DMSO to make a 20 mM stock solution . iCRT14 stock solution ( or DMSO alone as a control ) was diluted 1:1000 in growth medium for cell treatment ( 20 µM final concentration ) . We determined the amount of viral genomes in our viral stocks using digital droplet PCR ( ddPCR ) , a method that allows absolute quantification of nucleic acids . 10 µl of viral stock was combined with 90 µl of lysis solution ( 0 . 6% SDS , 400 µg/ml Proteinase K ) and incubated over night at 37°C . The solution was then boiled ( at 95°C ) for 10 min and 10-fold serial dilutions were made in H2O . Three primer sets ( detecting the viral DNA of the TK , gB or UL36 genes ) were used to quantify the amount of viral DNA . PFU were counted by plaque assay on Vero cells . These measurements revealed a genomes:PFU ratio of 36 ± 4 for wildtype HSV-1 and 1 , 422 ± 34 for ΔICP0 . The MOI for experiments was determined empirically , to achieve ~50% ICP4+ cells at 5 hr post infection of HDFn . This corresponded to an MOI of 2 for wild-type virus and an MOI of 0 . 5 for ΔICP0 . Note that , assuming a Poisson distribution , it is unlikely that any of the cells in our experiment did not encounter at least one viral genome ( p=4×10−31 for wildtype and 1 × 10−304 for ΔICP0 ) . HDFn cells were seeded on 6-well plates and allowed to attach and grow for one day . On the day of the experiment , cells were counted and infected with HSV-1 at an MOI of 2 . Cells were washed once with 106 medium without supplements , and virus was added in the same serum-free media at a final volume of 300 μl per well . Virus was allowed to adsorb to cells for one hour at 37°C with occasional agitation to avoid cell drying . The inoculum was aspirated and 2 ml of full- growth medium was added: this point was considered as ‘time zero' . Cells were imaged on a Nikon Ti-Eclipse , which was equipped with a humidity and temperature control chamber . Images were acquired every 15 min for 24 hr from multiple fields of view . Image analysis was performed with ImageJ and MATLAB . HDFn infected with wildtype HSV-1 at an MOI of 2 or ΔICP0 at an MOI of 0 . 5 were harvested at 5 hr post-infection and washed three times in PBS containing 0 . 01% BSA . Cells were counted and processed according to the Drop-seq protocol ( Macosko et al . , 2015 ) in the Genomics facility core at the University of Chicago . Sequencing was performed on the Illumina NextSeq500 platform . Preliminary data analysis ( quality control , trimming of adaptor sequences , UMI and cell barcode extraction ) was performed on a Linux platform using the Drop-seq Tools ( Version 1 . 13 ) and the Drop-seq Alignment Cookbook ( Version 1 . 2 ) , which are available at https://github . com/broadinstitute/Drop-seq/releases . Alignment of reads was performed using the STAR aligner ( Version 2 . 5 . 4b ) ( Dobin et al . , 2013 ) to a concatenated version of the human GRCh38 primary assembly ( Gencode release 27 ) and HSV-1 genomes ( Genbank accession: JN555585 ) . The HSV-1 genome annotation file was kindly provided by Moriah Szpara ( Pennsylvannia State University ) . Following the generation of the DGE ( digital gene expression ) file , further analyses were performed in MATLAB , these included quality control , cell clustering , correlation and differential gene expression analyses and data visualization . All the of the scripts used for data analysis have been deposited in Github ( Drayman , 2019; copy archived at https://github . com/elifesciences-publications/single-cell-RNAseq-HSV1 ) . Key points in the analysis are expanded on below . HDFn cells were mock infected or HSV-1 infected as described above , trypsinized , washed and re-suspended in full growth media . Cells were filtered through a 100 μm mesh into FACS sorting tubes and kept on ice . HSV-1 infected cells were sorted into two populations based on their ICP4-YFP expression . 0 . 5 million cells were collected from each population . Mock-infected cells were similarly sorted . ICP4-negative cells had the same level of YFP fluorescence at mock-infected cells . For ICP4-positive cells , we collected cells that were in the top 30% of YFP expression . The two populations were clearly separated from each other . Sorting was performed on an AriaFusion FACS machine ( BD ) at the University of Chicago flow-cytometry core facility . Total RNA was extracted from cells using the RNeasy Plus Mini Kit ( QIAGEN ) and submitted to The University of Chicago Genomics core for library preparation and sequencing on a HiSeq4000 platform ( Illumina ) . Reads were mapped to a concatenated version of the human and HSV-1 genomes with STAR aligner ( see single-cell RNA-sequencing above for details ) . Reads were counted using the featureCounts command , which is a part of the Subread package ( Liao et al . , 2013 ) . Further analyses were performed in MATLAB and these included differential gene expression analyses and data visualization . All sequencing data have been deposited in the Gene Expression Omnibus ( GEO ) under accession number GSE126042 . All of the scripts used for data analysis and visualization are available through GitHub at: https://github . com/nirdrayman/single-cell-RNAseq-HSV1 . git . HDFn were seeded in 24-well plates and allowed to attach and grow for one day . Cells were infected as described above and fixed using a 4% paraformaldehyde solution at 5 hr post-infection . Cells were fixed for 15 min at room temperature and washed , blocked and permeabilized with a 10% BSA , 0 . 5% Triton-X solution in PBS for one hour . Cells were then incubated with primary antibodies in a staining solution ( 2% BSA , 0 . 1% Triton-X in PBS ) overnight at 4°C . Cells were washed three times with PBS , incubated with secondary antibodies in staining solution for 1 hr at room temperature , washed three times with PBS and covered with 1 ml PBS containing a 1:10 , 000 dilution of Hoechst 33342 ( Invitrogen , cat #H3570 ) . Cells were imaged on a Nikon Ti-Eclipse inverted epi-fluorescent microscope . Primary antibodies were mouse monoclonal anti-β-catenin ( R and D systems , cat #MAB13291 , used at 1:200 dilution ) and rabbit monoclonal anti-IRF3 ( Cell Signaling Technologies , Cat #11904S , used at 1:400 dilution ) . Secondary antibodies were AlexaFluor 555 conjugated anti-mouse and anti-rabbit F ( ab’ ) two fragments ( Cell Signaling Technologies , cat #4409S , #4413S , used at 1:1000 dilution ) . 5 × 105 HDFn cells were washed once in PBS and nucleofected with 1 µM siRNA against β-catenin ( Dharmacon , siGENOME Human CTNNB1 , cat #M-003482-00-0005 ) or with a scrambled siRNA control ( Dharmacon siGENOME Non-Targeting siRNA Pool #1 , cat #D-001206-13-05 ) using the Human Dermal Fibroblast Nucleofector Kit ( Lonza , cat #VPD-1001 ) . β-catenin expression was assayed 3 days later by Q-PCR . | Herpes simplex virus 1 , or HSV-1 , is a virus that infects most of the human population . In many people , the virus stays dormant in nerve cells , but in some individuals , it can ‘wake up’ regularly and cause painful facial lesions known as cold sores . In very few cases , the virus can enter the brain and become life threatening . When HSV-1 encounters a human cell , there are three possible outcomes . The virus can either enter the cell and then replicate uncontrollably , get inside the cell but not multiply , or fail to enter the cell altogether . However , during experiments , researchers do not usually look at individual cells but instead consider whole populations . This makes it hard to understand the exact mechanisms that contribute to a cell resisting or succumbing to the virus . New approaches are now making it possible to study individual cells over time . Here , Drayman et al . harnessed these methods to understand how individual human cells respond to HSV-1 . The experiments show that most cells are actually able to resist the infection . Amongst those , a small fraction managed to stop the virus replicating by initiating a built-in ‘antivirus program’ . However , a minority of cells did become highly infected , shutting down the signaling process that fends off the virus . In these cells a different set of genes were switched on , making them more similar to the cells found in embryos . In the process , the virus recruited a protein called β-catenin to help with its multiplication . There are efforts to develop drugs to interfere with β-catenin , as this protein is also produced differently in people with cancer . Such drugs , if identified and safe in humans , could potentially serve to treat HSV-1 infections . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | HSV-1 single-cell analysis reveals the activation of anti-viral and developmental programs in distinct sub-populations |
Antigenic variation in the human malaria parasite Plasmodium falciparum involves sequential and mutually exclusive expression of members of the var multi-gene family and appears to follow a non-random pattern . In this study , using a detailed in vitro gene transcription analysis of the culture-adapted HB3 strain of P . falciparum , we show that antigenic switching is governed by a global activation hierarchy favouring short and highly diverse genes in central chromosomal location . Longer and more conserved genes , which have previously been associated with severe infection in immunologically naive hosts , are rarely activated , however , implying an in vivo fitness advantage possibly through adhesion-dependent survival rates . We further show that a gene’s activation rate is positively associated sequence diversity , which could offer important new insights into the evolution and maintenance of antigenic diversity in P . falciparum malaria .
Acquired antibody-mediated immunity against Plasmodium falciparum is directed against the disease-causing , intra-erythrocytic life-stage of the parasite’s life-cycle . Continual exposure to P . falciparum infection can lead to a form of semi-immunity , whereby protection against life-threatening disease is acquired after a few infections only ( Gupta et al . , 1999 ) whereas sterile and long-lasting immunity against infection is never achieved through natural exposure ( Greenwood et al . , 1987; Marsh , 1992; Tran et al . , 2013 ) . The process of natural acquired immunity therefore coincides with a transition from acute and severe infection in young children to asymptomatic carriage in older individuals ( Marsh and Snow , 1997 ) . At the core of this variable outcome of infection and poor development of immunity lies the major parasite virulence factor and variant antigen , P . falciparum erythrocyte membrane protein 1 ( PfEMP1 ) . Members of this family of proteins are expressed on the surface of infected red blood cells and are involved in the binding of parasitised cells to other host cells and tissues ( for an overview see Kraemer and Smith , ( 2006 ) ) . The conferred binding phenotypes are believed to enhance parasite survival by avoiding splenic clearance ( Boone and Watters , 1995; Buffet et al . , 2009 ) and also contribute to malaria pathology through parasite sequestration in the deep vasculature ( Ockenhouse et al . , 1991; Pongponratn et al . , 1991 ) . Mutually exclusive transcriptional switching between PfEMP1 variants during the course of an infection thus affords the parasite an immune evasion strategy and a means to exploit different host tissues ( Roberts et al . , 1992; Smith et al . , 1995 ) . Each haploid parasite genome contains approximately 60 members of the var gene family ( Gardner et al . , 2002 ) encoding distinct PfEMP1 variants , with a high degree of divergence between the antigenic repertoires of any two parasites . The high diversity of var genes and var gene repertoires is predominantly generated and maintained by frequent recombination and gene conversion events ( Deitsch et al . , 1999; Ward et al . , 1999; Freitas-Junior et al . , 2000; Taylor et al . , 2000; Kraemer et al . , 2007; Frank et al . , 2008 ) , which has also resulted in a high degree of mosaicism within these genes ( Ward et al . , 1999; Taylor et al . , 2000; Bull et al . , 2008 ) . The modular and highly polymorphic , extracellular portion of PfEMP1 is composed of a variable number of Duffy-binding-like domains ( DBL; at least two and up to seven per gene ) and cysteine-rich interdomain regions ( CIDR; up to two per gene ) ( Kraemer et al . , 2007 ) . Despite the overall diversity of var genes at the sequence level , distinct associations with severe malaria have been identified for several such genes ( Salanti et al . , 2003; Jensen et al . , 2004 ) , with the involvement of var2csa in pregnancy-associated malaria being the clearest example to date ( Salanti et al . , 2004 ) , as well as associations with particular DBL domains ( Warimwe et al . , 2009 ) and combinations of DBL and CIDR ( Avril et al . , 2012; Claessens et al . , 2012; Lavstsen et al . , 2012 ) . Var genes can also be classified according to a conserved upstream promoter sequence into four types termed UpsA , UpsB , UpsC , and UpsE ( Lavstsen et al . , 2003 ) . Of particular interest are the UpsA-type genes , which appear to have diverged from the other groups ( Kraemer et al . , 2007; Bull et al . , 2008 ) and have been shown to be upregulated during severe infections in young hosts ( Jensen et al . , 2004; Bull et al . , 2005; Kaestli et al . , 2006; Kyriacou et al . , 2006; Rottmann et al . , 2006; Warimwe et al . , 2009 ) . As a possible consequence , acquisition of anti-PfEMP1 immunity appears to follow a particular order , with UpsA variants generally being the first to be broadly recognised in older children ( Warimwe et al . , 2009; Cham et al . , 2010 ) . Nevertheless , clear associations between the expression of particular genes or Ups-groups , host age and severity of disease are still missing and could in fact be strain- or location-specific . Understanding the underlying pattern of antigenic switching between var genes is therefore important to explain not only the mechanisms and dynamics of persistent infections but also the relationship between immune-mediated expression of particular PfEMP1 sub-types and infection outcomes . Various studies have investigated var gene expression during natural or experimental infections ( Peters et al . , 2002; Lavstsen et al . , 2005; Wunderlich et al . , 2005; Blomqvist et al . , 2010 ) . However , as the diversity and order of PfEMP1 variants in the human host is influenced by immune responses and other host factors ( Kyriacou et al . , 2006; Warimwe et al . , 2009 ) , attempts to elucidate the underlying patterns of antigenic change and estimate related switching parameters are commonly based on in vitro cultured parasites in the absence of selection pressure . In this setting , Horrocks et al . ( Horrocks et al . , 2004 ) found that the rates at which var genes activate or deactivate are non-random , gene-specific and highly dissimilar . It has furthermore been suggested that var genes occupying subtelomeric loci tend to switch off faster than those positioned in central chromosomal regions ( Frank et al . , 2007 ) , offering a possible explanation for the rapid decline in the transcription of subtelomeric genes in parasites during culture adaptation ( Peters et al . , 2007; Zhang et al . , 2011 ) . More recently , Recker et al . ( 2011 ) proposed that var gene transcriptional change involves a highly structured pathway that has evolved in response to a trade-off between the within-host and the between-host level fitness . However , most of these previous studies were limited in analytical depth due to restrictions in the number of variants considered and/or time points at which gene transcription levels were measured , and could therefore describe only small fragments of the whole switching network . In this study , we used a novel approach to provide the first detailed characterisation of antigenic switching in the HB3 isolate of P . falciparum . We analysed pooled var gene transcription data from verified quantitative real time PCR measurements of a diverse set of clonal parasite cultures using a statistically rigorous method of parameter estimation . This revealed a global hierarchy in var gene activation favouring a highly diverse set of genes in central chromosomal locations . Our results further suggest a role of active gene transcription in the generation of antigenic diversity and will have important implications for understanding the age- and exposure-dependent pathology of P . falciparum malaria .
The proportional transcript levels of the different starter genes declined at notably different rates over the course of the experiment ( Figure 1 , Figure 1—figure supplements 1 and 2 ) . For example , the centrally located , UpsC starter gene var29 in culture #8 accounted for more than 90% of the total transcript level at the first measurement and remained the dominant variant over the following 60 generations . In contrast , the transcript level of the var35 starter gene of culture #11 ( subtelomeric , UpsA promoter type ) was already below 50% of the total transcript level at the point of the first observation ( approximately 30 generations post cloning ) and continued to decline during the time course of culture , to be eventually replaced by other genes as dominant transcripts . Similar patterns were seen in cultures #3 and #7 , which both had centrally located , UpsC starter genes , whereas the remaining five starter genes showed intermediate rates of decline . 10 . 7554/eLife . 01074 . 004Figure 1 . Proportional var transcript levels for six in vitro cultures . The parasite cultures initially expressed a variety of dominant ‘starter genes’ , which switched off at notably different rates . Nevertheless , most cultures converged towards high level transcription of two centrally located genes var27 and var29 ( red and blue lines , respectively ) , whereas most other gene transcripts ( grey lines ) remained at relatively low levels . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 00410 . 7554/eLife . 01074 . 005Figure 1—figure supplement 1 . Proportional var transcript levels and model output for cultures 1–6 . The relative transcript levels of activated var genes were measured by qRT-PCR at various time points during in vitro culture for 11 clones with different starter genes , plus the parent culture . Columns one and two show the measured transcript levels at linear and logarithmic scale , respectively; columns three and four show the model fit . Error bars show approximate 95% confidence intervals estimated from qRT-PCR replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 00510 . 7554/eLife . 01074 . 006Figure 1—figure supplement 2 . Proportional var transcript levels and model output for cultures 7–11 . The relative transcript levels of activated var genes were measured by qRT-PCR at various time points during in vitro culture for 11 clones with different starter genes , plus the parent culture . Columns one and two show the measured transcript levels at linear and logarithmic scale , respectively; columns three and four show the model fit . Error bars show approximate 95% confidence intervals estimated from qRT-PCR replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 006 Of particular interest was the apparent convergence of most cultures towards high transcription of var27 and/or var29 ( Figure 1 , Figure 1—figure supplements 1 and 2 ) , which are both situated in central chromosomal location but have different promoter types ( UpsB and UpsC , respectively ) . In light of the very different initial conditions , in terms of var gene transcription levels at which these cultures were established , this indicates that antigenic switching in P . falciparum might be governed by an activation hierarchy . To fully understand the patterns of antigenic switching underlying the observed transcriptional changes in our parasite cultures , we estimated relevant switch parameters using a previously described Markov chain Monte Carlo ( MCMC ) method ( Noble and Recker , 2012 ) ( ‘Materials and methods’ ) . We have previously proposed that var gene switching can be fully described dynamically by two sets of parameters ( ‘Materials and methods’ and [Recker et al . , 2011] ) . First , each gene i has an off-rate , ωi , which is the per generation probability of switching from active to silenced . Second , each gene has a switch bias , βij , which is the probability that when gene i switches off , gene j becomes activated . We applied our algorithm to the full 38-gene data set and separately to a reduced set of the 16 most transcribed genes , and both models generally gave good and similar fits between the data and predicted transcript levels ( Figure 1—figure supplements 1 and 2 ) , with deviations resembling stochastic measurement errors . Our method provided estimated switch parameters for the entire var repertoire , which are illustrated in Figure 2 for the 16 most highly transcribed genes . Due to higher signal to noise ratios , parameter estimates of the starter genes ( indicated in red ) were more accurate and resulted in much narrower credible intervals , as indicated by less fuzzy rings that represent respective switch probabilities . As clearly demonstrated in Figure 2 , there was wide distribution in switch biases and off-rates , and we next analysed for specific associations between these parameters and other gene characteristics . 10 . 7554/eLife . 01074 . 007Figure 2 . Estimation of switch parameters of highly transcribed genes . Parameter estimations for the 16 most transcribed var genes are represented as a switch matrix and an off-rate vector . The diameter of a circle in the ith row and jth column of the matrix is proportional to the switch bias βij from gene i to gene j , and the diameter of a circle in the off-rate vector is proportional to the rate ωi at which gene i becomes silenced . The fuzziness indicates uncertainty in the estimate , such that the darkness of each concentric ring is proportional to the likelihood that the parameter is within the corresponding range ( Noble and Recker , 2012 ) . Starter gene parameters , which are more precisely estimated , are in red , and genes are arranged from left to right , and top to bottom , in order of average transcript level . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 00710 . 7554/eLife . 01074 . 008Figure 2—source data 1 . Estimated switch biases and 95% credible intervals ( CI ) . Mean switch biases from gene i ( row ) to gene j ( column ) for the 16 most transcribed var genes , together with the 95% credible intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 00810 . 7554/eLife . 01074 . 009Figure 2—figure supplement 1 . Validation of estimated switching parameters . Our method was able to estimate switching parameters not only for starter genes , which were initially transcribed in our cultures , but also for other var genes . To test the accuracy of the latter estimates , we used a cross-validation technique . For example , to generate an alternative set of parameter estimates for var29 , we ran the MCMC algorithm without the data for culture #8 , which began as a clone transcribing var29 . The matrix shown here is a composite of estimates derived in this way . The similarity to Figure 2 confirms that our method was reliable . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 009 Estimated off-rates ranged from 0 . 3% to more than 5% per generation , with a mean of approximately 2 . 9% ( Table 2 ) . Notably , the two most transcribed genes , var29 and var27 ( both centrally located on chromosome 4 ) , had exceptionally low off-rates of approximately 0 . 3% and 0 . 8% , respectively , whereas the other starter genes switched off at rates between 1 . 4% and 3 . 5% per generation . In general we found off-rates of centrally located genes to be significantly lower than of genes in subtelomeric regions ( Welch’s t = 3 . 3 , p=0 . 002 ) , as shown in Figure 3 . This result remained significant ( p=0 . 035 ) when we accounted for uncertainty in our estimates ( ‘Materials and methods’ ) . 10 . 7554/eLife . 01074 . 010Table 2 . Parameter estimates for the HB3 var repertoire with 95% credible intervalsDOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 010GeneLocationDomains*UpsOff-rate ( 95% CI ) Activation bias ( 95% CI ) var29Central4C0 . 3% ( 0 . 1 , 0 . 5 ) 27% ( 22 , 32 ) var27Central4B0 . 8% ( 0 . 5 , 1 . 2 ) 23% ( 19 , 28 ) var35Central4C3 . 3% ( 2 . 2 , 4 . 7 ) 8 . 0% ( 5 . 8 , 11 ) var28Central4C2 . 2% ( 1 . 4 , 3 . 1 ) 6 . 9% ( 5 . 2 , 9 . 1 ) var32Central4C4 . 7% ( 2 . 5 , 5 . 9 ) 2 . 5% ( 1 . 5 , 3 . 3 ) var36Central4C1 . 8% ( 0 . 2 , 4 . 2 ) 2 . 5% ( 1 . 5 , 4 . 1 ) var31Central4C1 . 4% ( 0 . 9 , 2 . 0 ) 2 . 2% ( 1 . 7 , 2 . 9 ) var34Central6C1 . 3% ( 0 . 1 , 3 . 4 ) 2 . 1% ( 1 . 4 , 3 . 6 ) var13Subtelomeric4B1 . 9% ( 1 . 5 , 2 . 4 ) 2 . 1% ( 1 . 6 , 2 . 8 ) var30Central4C1 . 9% ( 1 . 3 , 2 . 7 ) 1 . 9% ( 1 . 4 , 2 . 5 ) var10Subtelomeric5B3 . 4% ( 1 . 1 , 5 . 7 ) 1 . 7% ( 0 . 94 , 2 . 7 ) var17Central6B3 . 0% ( 0 . 8 , 5 . 6 ) 1 . 7% ( 0 . 90 , 2 . 8 ) var24Central6B1 . 9% ( 0 . 1 , 5 . 2 ) 1 . 6% ( 0 . 85 , 3 . 1 ) var7Subtelomeric7B2 . 2% ( 0 . 3 , 5 . 1 ) 1 . 5% ( 0 . 83 , 2 . 7 ) var19Central4B1 . 3% ( 0 . 1 , 3 . 2 ) 1 . 4% ( 0 . 93 , 2 . 5 ) var26Central4B1 . 5% ( 0 . 2 , 3 . 4 ) 1 . 4% ( 0 . 90 , 2 . 3 ) var25Central6B2 . 7% ( 0 . 4 , 5 . 6 ) 1 . 2% ( 0 . 60 , 2 . 1 ) var1csaSubtelomeric8A5 . 5% ( 4 . 4 , 6 . 0 ) 1 . 2% ( 0 . 89 , 1 . 5 ) var33Central4C2 . 7% ( 0 . 5 , 5 . 8 ) 1 . 1% ( 0 . 57 , 1 . 9 ) var16Subtelomeric4B2 . 8% ( 1 . 0 , 5 . 3 ) 0 . 91% ( 0 . 52 , 1 . 5 ) var4Subtelomeric8A4 . 3% ( 2 . 1 , 5 . 9 ) 0 . 86% ( 0 . 52 , 1 . 3 ) var14Subtelomeric4B4 . 0% ( 2 . 2 , 5 . 9 ) 0 . 73% ( 0 . 45 , 1 . 1 ) var22Central7B1 . 9% ( 0 . 2 , 5 . 1 ) 0 . 72% ( 0 . 42 , 1 . 3 ) var21Central7B4 . 7% ( 2 . 7 , 5 . 9 ) 0 . 70% ( 0 . 46 , 0 . 94 ) var8Subtelomeric7B3 . 7% ( 1 . 8 , 5 . 8 ) 0 . 66% ( 0 . 39 , 1 . 0 ) var11Subtelomeric5B1 . 5% ( 0 . 9 , 2 . 2 ) 0 . 60% ( 0 . 45 , 0 . 80 ) var18Subtelomeric4B5 . 2% ( 3 . 8 , 6 . 0 ) 0 . 60% ( 0 . 42 , 0 . 79 ) var5Subtelomeric6A3 . 5% ( 2 . 7 , 4 . 3 ) 0 . 53% ( 0 . 39 , 0 . 69 ) var39p†Subtelomeric2B4 . 7% ( 2 . 7 , 6 . 0 ) 0 . 48% ( 0 . 30 , 0 . 66 ) var50Ψ‡Central6C4 . 4% ( 2 . 0 , 5 . 9 ) 0 . 46% ( 0 . 26 , 0 . 67 ) var23Central6B0 . 6% ( 0 . 0 , 2 . 0 ) 0 . 38% ( 0 . 26 , 0 . 55 ) var9Subtelomeric6B2 . 8% ( 0 . 7 , 5 . 3 ) 0 . 36% ( 0 . 20 , 0 . 58 ) var2csaSubtelomeric6E4 . 7% ( 2 . 3 , 5 . 9 ) 0 . 28% ( 0 . 16 , 0 . 40 ) var2Subtelomeric6A1 . 4% ( 0 . 2 , 4 . 2 ) 0 . 27% ( 0 . 17 , 0 . 50 ) var12Subtelomeric4B3 . 8% ( 1 . 5 , 5 . 8 ) 0 . 26% ( 0 . 15 , 0 . 40 ) var20Subtelomeric4B4 . 1% ( 1 . 8 , 5 . 9 ) 0 . 25% ( 0 . 14 , 0 . 37 ) var1Subtelomeric7A5 . 5% ( 4 . 2 , 6 . 0 ) 0 . 024% ( 0 . 018 , 0 . 031 ) var6Central8A2 . 9% ( 0 . 4 , 5 . 9 ) 0 . 018% ( 0 . 0086 , 0 . 032 ) *Number of encoded DBL ( Duffy binding-like ) and CIDR ( Cys rich inter-domain region ) . †Partial gene . ‡Psuedogene . 10 . 7554/eLife . 01074 . 011Figure 3 . Off-rate estimates show strong dependency on chromosomal location . The estimated rates at which active var genes become silenced are significantly lower for centrally located var genes than for those in subtelomeric locations . There was no significant effect of upstream promoter type when differentially testing off-rates of genes in subtelomeric location ( UpsA vs non-UpsA ) or in central location ( UpsB vs UpsC ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 011 Previous studies reported that UpsA var genes from patient isolates were down-regulated faster than other var genes during adaptation to culture ( Peters et al . , 2007; Zhang et al . , 2011 ) . The HB3 repertoire contains seven UpsA genes , all but one of which are subtelomeric . We compared the off-rates of the subtelomeric UpsA genes with those of other subtelomeric genes and found no significant difference ( Welch’s t = 0 . 65 , p=0 . 54; accounting for uncertainty p=0 . 55 ) , as shown in Figure 3 . Among centrally located genes we compared UpsB with UpsC and again found no significant difference ( Figure 3; Welch’s t = 0 . 44 , p=0 . 44; accounting for uncertainty p=0 . 57 ) . This lends support to a previous study suggesting that the rate at which transcriptionally active var genes become silenced is less dependent on promoter type than on chromosomal location ( Frank et al . , 2007 ) . In each of our cultures , gene transcription levels spanned several orders of magnitude , and only a minority of genes were ever highly transcribed during the experiment . As mentioned earlier , despite the diversity of the starter genes , the relative transcript levels of the activated genes appeared to converge towards similar levels across all cultures ( Figure 1 , Figure 1—figure supplements 1 and 2 ) . To test the degree of similarity in gene activation , we excluded the nine starter genes and ranked the rest of the repertoire according to average transcript levels at later time points ( more than 65 generations ) for each culture . Pair-wise comparison showed that the ranking was highly correlated between the cultures ( Spearman’s ρ > 0 . 8 , p=10−5 for all pairs ) , suggesting that the outcome of antigenic switching in P . falciparum , or at least within the HB3 isolate , is largely governed by an activation hierarchy that is independent of the currently activated gene . Consistent with this hypothesis , the estimated switch biases between genes , βij , were remarkably similar , so that βij≈βkj for each i , j and k . ( Figure 2 , Figure 2—source data 1 , Figure 2—figure supplement 1 ) . All genes were found to have high switch biases towards the same few genes , which explain the observed convergence in var transcription in our cultures . As expected , the highest biases across the repertoire were found towards var29 and var27 , with some genes also frequently switching towards var35 and/or var28 ( both UpsC-types in central chromosomal location ) . To better describe the observed activation hierarchy , we used our MCMC method to estimate the average switch bias towards each gene , which we hereafter refer to as the gene’s activation bias . That is , the activation bias of gene j is defined as the average of the switch biases βij for all deactivating genes i . It can be understood as a gene specific , per switch activation probability and is therefore distinct from the previously defined on-rate ( Horrocks et al . , 2004 ) , in that it is invariant to transcript levels . As shown in Table 2 , there was a clear hierarchy in gene activation biases spanning over three orders of magnitude . The ranking also appeared to be non-random so that , for example , most genes within central chromosomal location were situated in the top half of the table . Analysing for possible associations with genetic attributes confirmed that activation biases of centrally located var genes are significantly higher than those of subtelomeric genes ( Figure 4 , Welch’s t = 2 . 8 , p=0 . 009; accounting for uncertainty p=0 . 011 ) . In fact , of the highest four activation biases , three belonged to neighbouring genes on chromosome 4: var29 , var27 and var28 . Importantly , no significant and independent effect of promoter type was found , which again points towards a strong influence of chromosomal location on var gene activation . 10 . 7554/eLife . 01074 . 012Figure 4 . Associations between activation biases and genetic attributes . The mean activation bias is higher for centrally located var genes . Within the set of centrally located genes , those encoding only four binding domains have higher activation biases than longer genes . Together , short , central genes have activation biases ≈ 10 times higher than the rest of the repertoire . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 012 Additionally , we considered whether gene architecture had an effect on the observed activation biases , as recent research suggests that gene length , in terms of the number of binding domains , is another important var gene characteristic besides chromosomal location and promoter type ( Buckee and Recker , 2012 ) . Grouping genes depending on whether they encoded only four domains ( short genes ) or more than four domains ( long genes ) showed that activation biases differed significantly depending on gene length ( Figure 4 , F = 9 . 7 , p=0 . 004 ) as well as chromosomal location ( F = 8 . 1 , p=0 . 007 ) and the interaction of these two factors ( F = 5 . 0 , p=0 . 032 ) . As such , central var genes encoding only four binding domains ( also known as Type 1 var genes [Gardner et al . , 2002] ) had on average a 10 times higher activation bias than the rest of the repertoire . Results were similar when we accounted for uncertainty in our estimates ( p=0 . 005 , p=0 . 0096 , p=0 . 037 , respectively ) . We next tested explicitly whether the currently active gene has an influence on the direction of switching , that is if in general βij≠βkj . We used an optimisation algorithm ( ‘Materials and methods’ ) to fit two alternative models to the data for the 16 most highly transcribed var genes: ( i ) a general model including off-rates and individual switch biases ( 240 independent parameters ) , and ( ii ) a simpler model , in which all sets of switch biases were required to be identical , that is where βij=βj for all i and j ( 31 independent parameters ) . A likelihood ratio test indicated that the model with individual switch biases was significantly more likely than the second model ( F = 1 . 84 , p<0 . 0001 ) , confirming that antigenic switching in P . falciparum is not only controlled by an activation hierarchy but also influenced by the currently active gene . We again tested for associations with chromosomal location and promoter type ( ‘Materials and methods’ ) , which indicated that switch biases were higher between var genes with matching chromosomal locations than between differently located genes ( Figure 5; Welch’s t = 2 . 2 , p=0 . 029 ) , whereas no significant association was found between switch biases and matching Ups types or gene lengths . This result needs to be confirmed with larger data sets , however , as statistical significance was lost when we accounted for uncertainty in our parameter estimates . 10 . 7554/eLife . 01074 . 013Figure 5 . Variation in switch biases ( deviation from the mean ) . Antigenic switching between var genes is more frequent between genes with matching chromosomal locations , for example from central to central or from subtelomeric to subtelomeric . No significant associations between switch bias and matching Ups type or gene length are found . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 013 Var gene sequences can be described using a set of 628 conserved homology blocks , with an average length of 19 amino acids ( Rask et al . , 2010 ) . Using homology blocks found only within the DBL1 and CIDR1 domains , which are present in all HB3 var genes except var1csa , var2csaA , var2csaB and the partial gene var39 , we constructed a var gene homology network , in which edges between genes are defined and weighted by the number of shared homology blocks ( Figure 6—figure supplement 1 ) . We then calculated the centrality of each gene within this network . This is a distance measure between all pairs of nodes ( i . e . , genes ) and can therefore be used as a proxy for the relatedness of a gene to the rest of the repertoire . The inclusion of network centrality as an explanatory variable in the previous ANOVA model confirmed that the degree by which var genes share sequence blocks is positively correlated with activation bias , independently of chromosomal location and gene length ( Figure 6A; F = 8 . 8 , p=0 . 006 for the model comparison; accounting for uncertainty p=0 . 006 ) . 10 . 7554/eLife . 01074 . 014Figure 6 . Effect of var gene activation on sequence evolution . ( A ) Var gene activation biases are significantly correlated with each gene’s relatedness to the rest of the HB3 repertoire , as measured by the gene’s centrality within a shared homology block network , independently of chromosomal location or gene length . Trend lines show the fit of a linear regression model . ( B ) Simulation of var gene evolution by gene conversion , whereby homology blocks are swapped among pairs of genes chosen at random according to their activation biases , shows a similar correlation between the genes’ activation biases and their relatedness to the rest of the repertoire . The red points show the outcome of one simulation , and the dashed line is the smoothed average of 50 runs . ( C ) Activation biases are negatively correlated with the gene’s average domain sequence conservation , as a measure of population-level diversity , independently of chromosomal location and gene length . Trend lines show the fit of a linear regression model . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 01410 . 7554/eLife . 01074 . 015Figure 6—figure supplement 1 . Relatedness network of the HB3 repertoire . Each node represents a var gene , colour coded according to the chromosomal location and the number of binding domains . Node size indicates estimated activation bias . The strength of a connection between two nodes is proportional to the number of shared homology blocks in the DBL1 and CIDR1 domains . Stronger connections are shown thicker and darker , and the layout is force-directed . DOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 015 We hypothesised that this could indicate a possible role of activation bias , or gene transcription , in var gene sequence evolution . Under the assumption that gene transcription facilitates recombination we would expect more frequently activated genes to share more sequence blocks with other genes , simply because they more often act as recombination sites . To test this we explicitly simulated gene conversion , in which pairs of genes were chosen at random , weighted by their activation bias and homology blocks copied between them ( ‘Materials and methods ) . As expected , we found a strong positive correlation between activation bias and our measure of relatedness in terms of network centrality ( Figure 6B ) . Under the same assumption we would also expect genes with higher activation biases to be overall more diverse . Indeed , using the average sequence conservation of all domains as a measure of population level diversity ( Buckee and Recker , 2012 ) , we again found this to be significantly and positively correlated , independently of chromosomal location and gene length ( Figure 6C; F = 13 . 6 , p=0 . 0009 for the model comparison; accounting for uncertainty p=0 . 0006 ) , suggesting a role of active gene transcription in generating antigenic diversity in P . falciparum malaria .
Several studies have noted that var genes located near the centres of chromosomes tend to be more highly transcribed in vitro than those in subtelomeric location ( Frank et al . , 2007; Peters et al . , 2007; Enderes et al . , 2011; Zhang et al . , 2011; Fastman et al . , 2012 ) , and it has been suggested that this might be due to low deactivation rates ( Frank et al . , 2007 ) . In this study , we found that high transcription levels are confined to a subset of central genes that encode only four PfEMP1 binding domains ( termed Type 1 var genes [Gardner et al . , 2002] ) and , importantly , are the result of high activation biases rather than low off-rates . That is , even though some of the most highly transcribed genes had very low off-rates , the total sum of other genes switching off with high bias towards these genes significantly outweighs the effect of off-rates . This is consistent with our previous study using artificial transcription data , which also showed that off-rates , at least within a biologically plausible range , can have only a modest effect on the total transcript levels within parasite populations ( Noble and Recker , 2012 ) . Reanalysis of data from a recent study on the NF54 parasite strain by Fastman et al . ( 2012 ) supports these findings . Consistent with our results for HB3 , centrally located var genes encoding only four binding domains were more highly transcribed than the rest of the repertoire ( Welch’s t = 2 . 3 , p=0 . 02 ) , with PF3D7_0809100 , PF3D7_0421300 and PF3D7_0800100 being the most highly transcribed genes . According to our model , these are predicted to have high activation bias , and indeed , PF3D7_0809100 has previously been reported as the most transcribed var gene in NF54 bulk culture , whereas PF3D7_0421300 and PF3D7_0800100 were found to be the two genes most commonly activated by var gene switching in previous studies ( Dzikowski et al . , 2006; Frank et al . , 2007; Enderes et al . , 2011 ) . Also in agreement with our results , NF54 gene transcription levels were positively correlated with relatedness to the rest of the repertoire in terms of shared DBL1 and CIDR1 homology blocks ( Spearman’s ρ = 0 . 3 , p=0 . 044 ) . Our results further indicate that the direction of transcriptional switching between var genes depends not only on intrinsic activation biases but also on the particular preferences of each deactivating gene . Even though these variations were found to have only a modest effect on in vitro switching , they are expected to play a much greater role during infection , where factors such as immune responses and binding affinities can significantly alter the var gene expression profiles . Furthermore , the combination of an activation hierarchy and these specific variations could explain our previously proposed source-sink model underlying antigenic switching in P . falciparum ( Recker et al . , 2011 ) , in which genes with high in-degrees , i . e . , those most commonly switched to , correspond to the genes here identified as having high activation biases . Intriguingly , also , our results suggest that switch biases may be stronger between genes in similar chromosomal locations: central genes prefer to switch to other central genes , and subtelomeric genes prefer to switch to other subtelomeric genes . More data are needed to fully establish the factors influencing variation in switch biases , however , especially for less activated genes , where parameter estimations are more prone to uncertainties . Despite recent advances in elucidating the genetics and epigenetics of var gene activation and silencing , not much is known about how mutually exclusive switching between the 60 var genes is controlled at the molecular level . The discovery of intrinsic var gene activation biases and their associations with chromosomal locations and other genetic features could therefore have important implications for future studies . These factors are clearly correlated , however , and it is uncertain how they are linked by cause and effect . For example , in a recent study it was found that gene length is positively correlated with sequence conservation ( Buckee and Recker , 2012 ) , and long var genes are more often positioned in subtelomeric regions . Nevertheless , the correlation between activation bias and relatedness , which was found to be independent of chromosomal location , gene length and upstream promoter , suggests a role of gene transcription in facilitating recombination and therefore in generating and maintaining antigenic diversity in P . falciparum . Transcription-associated recombination has been observed in other eukaryotes and may result from increased accessibility to recombination proteins or from stalled replication forks ( Prado and Aguilera , 2005; Gottipati and Helleday , 2009 ) . Importantly , for P . falciparum this phenomenon could help to explain why different groups of var genes within the repertoire display very different levels of diversity within the population ( Buckee et al . , 2009; Buckee and Recker , 2012 ) . It is further possible that some var genes are under balancing selection , whereby the relative sequence conservation of longer genes could be due to optimised functionality , for example in terms of binding avidity and/or affinity . This assumption would also help to reconcile apparent contradictions between our observations and var gene transcription in vivo . That is , various in vivo studies have reported UpsA var genes to be over-expressed during infections in individuals with little previous exposure ( Jensen et al . , 2004; Bull et al . , 2005; Lavstsen et al . , 2005; Kyriacou et al . , 2006; Warimwe et al . , 2009 ) , despite their low activation rates in vitro . Each var repertoire consists of a set of relatively long and conserved genes , such as those predominantly found with in the UpsA group , and another , short and diverse set of genes , such as those predominantly found within the UpsB and UpsC groups ( Buckee and Recker , 2012 ) . Under the assumption that more conserved genes have an in vivo survival advantage , possibly due to higher binding affinities ( Avril et al . , 2012; Claessens et al . , 2012; Lavstsen et al . , 2012 ) , these variants will dominate early infections in naive hosts , regardless of activation rates . Their lower sequence diversity , however , will then lead to a more rapid acquisition of protective immunity , causing a transition towards the expression of more diverse variants , facilitated by their inherent activation biases . In light of this we would also speculate that the proposed order in which individuals acquire immunity ( anti-UpsA followed by anti-UpsB/C ) , as well as the proposed associations between the expression of UpsA var genes and severe disease outcome , especially in young children , is less determined by the Ups group itself and is more a consequence of within-host selection of long , conserved genes , which are often but not exclusively found within the UpsA group . Together , these results demonstrate how selection pressures operating on multiple ecological scales have shaped the phenotypic plasticity embodied within the antigenic repertoire of P . falciparum . Within this setup , var genes and the molecular mechanisms that underlie their sequential activation and silencing have co-evolved to allow the parasite to express the most advantageous phenotype in response to its current environment .
Parasites were cultured and sorbitol-synchronised using standard techniques ( Trager and Jensen , 1976; Lambros and Vanderberg , 1979 ) , and RNA was extracted from saponin lysed , mid-to-late ring stage parasites with Trizol as previously described ( Kyes et al . , 2000 ) . For cDNA preparation , 5 μg RNA was treated in 20 μl reactions with TurboFree DNase ( Ambion®; following manufacturer’s recommendations except incubation was at 16°C for 40 min ) . All DNase treated RNA samples were tested by PCR with primers to fructose-bisphosphate aldolase ( PF14_0425 ) ( Salanti et al . , 2003 ) to ensure complete removal of gDNA . Duplicate 20 μl reactions consisting of 6 μl RNA were reversed transcribed with 100 ng random hexamers ( Invitrogen ) and Bioscript reverse transcriptase ( Bioline; following manufacturer’s instructions ) at 40°C for 40 min . The resulting cDNA was stored at −20°C in single-use aliquots and diluted 1:6 for real time PCR analysis . cDNA was synthesised in triplicate ( 84% of all data points ) or duplicate , and real time PCR was done in duplicate for each primer pair on each cDNA . Primers specific for the HB3 var exon 1 repertoire were optimised to meet several basic criteria: that they distinguished between the known var gene sequences within the HB3 genome; that they were specific for HB3 var genes with no cross-reactivity to 3D7 var genes ( for future application to studies on 3D7 × HB3 cross progeny ) ; and that they reflected measurement of full-length var transcript . Primer pairs were designed using Primer3plus ( Kraemer et al . , 2007; Untergasser et al . , 2007 ) , with parameters manually set for product size range ( 100–150 nt ) , primer length ( 20 nt ) , Tm ( 60°C ) , GC clamp ( 1 nt ) , salt correction formula and thermodynamic parameters ( SantaLucia 1998 ) ; all other parameters were default settings ( see Table 3 for primer sequences ) . To avoid general cross-detection of all var types , the first kilobase of exon 1 containing the conserved DBLα domain was avoided . The final kilobase was also excluded to avoid sterile transcripts ( Su et al . , 1995; Kyes et al . , 2007 ) . Each primer pair was specific for a single gene except , unavoidably , the var3 primer pair and the two pairs designed to the highly similar copies of var2csa . Each HB3 primer pair was tested five times on gDNA; pairs that varied more than ±1 . 5 threshold cycles ( Ct ) values from the median or had efficiency value ( Amplification ) less than 1 . 85 were redesigned . Primer pairs that amplified 3D7 genomic DNA were discarded and redesigned so that they were specific for HB3 var genes only . 10 . 7554/eLife . 01074 . 016Table 3 . HB3 qPCR primers and cross-referenced var gene identifiersDOI: http://dx . doi . org/10 . 7554/eLife . 01074 . 016Gene nameContig nameBroad locus nameF oligonucleotideR oligonucleotideHB3 var16HB3-1000-1PFHG_03232 . 1ccctgtccacaaccatcagccgtcgtcgtcatcagtgtccHB3 var1HB3-1000-2PFHG_03234 . 1ccaaaggagaaggcaccaccacctatggcacccctctcacHB3 var12HB3-1040PFHG_03416 . 1gatgctacaaccaccccaccgtgttaccactcgcccactcHB3 var27HB3-1704_1PFHG_03476 . 1gctcccaaccaccacgttccgcttcctgctggtggctgtcHB3 var28HB3-1074-2PFHG_03478 . 1gatggcacaaaagttggcggtgttctgggtcgacctcctcHB3 var29HB3-1074-3PFHG_03480 . 1aagaagatggcgacgaaggctccggtgatccctcttctggHB3 var13HB3-1107PFHG_03516 . 1tggtaaatgcaagggtgatacaggtgcatcgttatcactcaccagcHB3 var1csaHB3-1108PFHG_03521 . 1cgcaatatgcaactaatgacacttggcaatattctgaacgHB3 var2HB3-1210PFHG_03840 . 1cgaggacaccacggaggaggttggtgctgctggttgtggcHB3 var5HB3-1235PFHG_03671 . 1aggtctgctccttcagatgcgtgtgttttccctaccatgacaaggatgccHB3 var36HB3-1296PFHG_04012 . 1atggacaaatgatggtaaggtaggagtaggtgttgcgttcHB3 var17HB3-1296-2PFHG_04014 . 1agatggcgacaaaggccaagttgggtttggcaccactagcHB3 var35HB3-1296-3PFHG_04015 . 1aaacggaaaacctggcctcctcgtcttggcctttggcttcHB3 var14HB3-1308PFHG_04035 . 1ggtggtggtgccgatcccgcctgtgacgcctccgtcttagtggcccHB3 var8HB3-1334PFHG_04081 . 1ggcggtgtctgtattccaccgctgcctcaccacctgttagHB3 var9HB3-1408PFHG_04057 . 1tgctatgacgtgtaatgccccacttacatgagtcccatctggtgHB3 var32HB3-1459FbadfggaaaccgcggtggactcacacttgtgggtgctttggggcHB3 var11HB3-1499PFHG_04491 . 1attggatgatgcctgtcgccggcacctggtttagtggtggHB3 var19HB3-1514PFHG_04620 . 1aaactgacaatggccccgacgttgttgagggggtcttcggHB3 var18HB3-1523PFHG_04593 . 1gcggctcacccgacatcttcgccgcctcgtcttcttcgtcHB3 var10HB3-1587PFHG_04749 . 1accactcgtgccaccacctcgagtttgtacctggcacccccHB3 var7HB3-1604-1PFHG_04769 . 1agcgagtggtactcaggagggatggaccacgagatgtgccHB3 var20HB3-1604-2PFHG_04770 . 1acgaagaagacgatgccaccgaagtcttcggagcgaccacHB3 var4HB3-1703PFHG_04861 . 1tggtgccaaagacccctcccggccactcgctgtgtctgtgHB3 var2csaAHB3-1727PFHG_05046 . 1gggggaaatgtggggtgccggggggataccccacactcattaccagHB3 var3HB3-1737PFHG_05052 . 1aaagtgcgaagcacctcccccgccactgcagggattagctgHB3 var2csaBHB3-1817PFHG_05155 . 1tggtacagctgatggtggtacttccgtgtgcccgctttacggtttcgHB3 var39p*HB3-2007efbeagccattacgtgcgaagctggagcggcacatcggcatttttgHB3 var34HB3-209PFHG_00592 . 1agtggtgctgtagagccaaaagaccctgcggcggtgctgtaaggHB3 var23HB3-699-1PFHG_02272 . 1gaaccccttgacgacgacacctcaacacacgtcaaaggcgHB3 var30HB3-699-2PFHG_02273 . 1aagacgacaaacctggcaccgtcgttgcttttggcttcggHB3 var6HB3-699-3PFHG_02274 . 1attcacagcactgaaagtcctcacaatcattaaaagcatccHB3 var26HB3-699-4PFHG_02276 . 1aagcagctgatggaacggactggttgttgtgggtcttggcHB3 var31HB3-699-5PFHG_02277 . 1cgcgaagacgaaaacgtcacgtttcatccggaccgtcctcHB3 var50Ψ†HB3-752-1PFHG_02419 . 1tggtaatgatgaagatgacggaattggcttcactttgttcHB3 var24HB3-752-2PFHG_02421 . 1gctcgctctttaccacccgcttccgtctcctccttcgccgHB3 var25HB3-752-3PFHG_02423 . 1agtggtgccaaaactgtcggaccacaaaagtcgcttccccHB3 var22HB3-752-4PFHG_02425 . 1ccaccacaaaacccctccagtccgcttgtggttcgtcttcHB3 var33HB3-752-5PFHG_02429 . 1acagaaagttggacaggatgatggttgttttgagaattgc*Partial gene . †Pseudogene . Quantitative real time PCR ( qRT-PCR ) reactions were prepared with a Corbett CAS1200 liquid handling system and run on a Corbett Rotor-Gene 6000 . Each 10 μl reaction contained 1 μl template , 5 μl SensiMix SYBR ( Bioline ) , 3 μl water and 1 μl primer mix ( 1 . 6 μM of each primer ) . The cycling conditions were as follows: hold , 95°C/10 min; 45 cycles ( 95°C/25 s , 58°C/25 s , 68°C/30 s ) ; melt , 55°C−99°C . Amplicons were checked for a single product of the expected size by agarose gel electrophoresis . Melt analysis confirmed single products . Transcription levels for each replicate were calculated by the comparative quantification method of Pfaffl ( 2001 ) . Comparative quantification results are given as expression levels relative to a control ( housekeeping ) gene , seryl-tRNA synthetase ( Seryl-tRNAF agctacctcagaacaaccattatgtgc , Seryl-tRNAR atcctttccatgtgcccctgc ) , which is arbitrarily set at 1 . 0 . Minor differences in Take Off values between primer pairs were accounted for using a correction factor derived from five replicates on gDNA . Expression was determined using AT . O . −CF , where A is the average amplification , T . O . is the average Ct and CF is the correction factor . To express results relative to the control gene , the following formula was used: ArefTO−CF/AxTO−CF . Expression levels were converted to proportions before calculating geometric means of the replicates . Replicates were excluded as outliers ( 1 . 1% ) if they deviated from the mean by a factor of more than 101 . 5 or had an amplification efficiency below 1 . 5 ( most excluded measurements met both criteria ) . Dominant var gene expression detected by qRT-PCR was verified as both full-length and dominant by northern blot analysis , comparing specific probes for each dominant gene vs a generic exon 2 probe ( data not shown ) . The primer pairs for var genes 5 , 11 , 13 , 27 , 28 , 29 , 31 and 35 were thus verified as specific for detecting full-length transcripts . The var3 primer pair , which was known to also detect var5 and pseudogene var46 , indicated dominant var3 alongside var5 in culture #4 . However , we were unable to detect var3 expression by northern blot . As this sample was positive by northern blot probed with var5 , and it was the only one with ‘dominant’ var3 expression by qRT-PCR , we discarded all var3 qRT-PCR data as ambiguous . We also discarded all the data from one of the two var2csaA/var2csaB primer pairs . Following Recker et al . ( 2011 ) , we assumed that var transcription profiles can be described by a time-discrete model:vi , c , t+1= ( 1−ωi ) vi , c , t+∑j≠iωjβjivj , c , t , where vi , c , t is the relative transcript level of variant i in culture c at time t , ωi is the variant specific off-rate and βji is the switch bias ( probability of a switch ) from variant j to variant i . We assumed a maximum off-rate of 6% per generation , which is consistent with previous studies and theory ( Roberts et al . , 1992; Horrocks et al . , 2004; Frank et al . , 2007; Recker et al . , 2011 ) . We used a Markov chain Monte Carlo ( MCMC ) method to obtain posterior distributions of likely values for the switch biases and off-rates , using the pooled data from all observed cultures as input . For model comparisons we used a simulated annealing algorithm to find maximum likelihood parameter sets . These methods have previously been described in detail and tested using diverse artificial data sets ( Noble and Recker , 2012 ) . For some of our analyses it was desirable to use a model of reduced dimension . Previously we have shown that reliable parameter estimates can be obtained from data for the 16 most highly transcribed var genes ( Noble and Recker , 2012 ) , which in this case comprised the 9 starter genes and the 7 others with highest average transcript levels . These 16 genes together accounted , on average , for more than 92% of the sum of transcript levels in each culture ( excluding the culture’s starter gene ) , whereas none of the excluded genes accounted for more than 1% . To estimate activation biases , we modified the model so that the direction of switching was independent of the previously active gene ( i . e . , βji=βki for all i , j and k ) . Before conducting statistical tests , we transformed switch biases using the logit function logit ( x ) =log ( x/ ( 1−x ) ) to map from the range ( 0 , 1 ) to the range ( −∞ , ∞ ) . Additionally , to account for uncertainty in our parameter estimates , we performed tests on the collected output of our MCMC method . This method samples from the posterior distribution , which means it selects numerous parameter sets such that the probability of each being chosen depends on how likely it is to explain the data . We performed our statistical tests on all the selected parameter sets and calculated mean p-values . We used the same parameter sets to obtain credible intervals for our estimates , as shown in Table 2 . We found that all var genes had similar sets of switch biases , such that βij≈βkj for each i , j and k . A model comparison indicated that these sets were not identical but varied depending on the deactivating gene . To estimate these variations we took a subset of the MCMC switch bias estimates and standardised each set {β1j , β2j , … , βnj} by subtracting the mean and dividing by the standard deviation ( after rescaling each set to compensate for βjj=0 ) . Negative values then corresponded to biases that were smaller than average and positive values to those that were larger . We used the subset of switch biases from the 9 starter genes to the 25 most transcribed genes , omitting the diagonal terms . We constructed a network in which each node represented a gene and each edge was weighted by the number of shared variable DBL1 and CIDR1 homology blocks . Each gene’s centrality within the network was calculated using a weighted closeness measure ( Newman , 2001 ) implemented by the tnet package for R ( Opsahl , 2009 ) . We applied variance stabilising transformations to the data prior to analysis . Specifically , the centrality measure was rescaled and squared , and to each switch bias we applied the logit transformation , added five and then square-rooted the magnitude . These transformations made the data conform better to the assumptions of our statistical model; test results would have been more significant had we instead used the untransformed data . We simulated gene conversion in a repertoire of 30 genes , each comprising 40 homology blocks , with 30 possible variants for each homology block . Genes were initially constructed of randomly chosen homology blocks and were assigned activation biases according to a geometric sequence . For 30 , 000 iterations , pairs of genes were selected at random , with the probability of selection being proportional to activation bias . A randomly chosen homology block was copied from the first of the chosen genes to replace the corresponding block in the second gene . Homology blocks in all genes also underwent random mutation at a rate 10 times lower than the gene conversion rate . We constructed a relatedness network and calculated centrality in the same way as for the HB3 data . Our switching model required a set of initial conditions . Assuming that each culture , excluding the parent , began with a single parasite and that populations expanded approximately fivefold per generation , we deduced that the first transcriptional switch most likely occurred , on average , two or three generations after cloning . Therefore we assumed that each of these cultures remained clonal for the first three generations . For the parent culture , we used the first observation as the initial condition . If a switch had occurred in one or more of the initially clonal cultures within the first few generations , when the population was very small , then our assumed initial conditions would be inaccurate . Therefore we also ran our algorithms using the first observations as the initial conditions for all cultures . The parameter estimates were highly consistent with those obtained using clonal initial conditions . The simplest way to estimate switch biases and off-rates for a single var gene is to obtain a clone transcribing the gene , grow it in culture and analyse changes in transcript levels over time . A particular strength of our method was that it also estimated parameters for genes we did not clone . To assess the accuracy of results obtained in this way , we performed cross-validation by sequentially excluding data sets . For example , to derive alternative estimates for var29 we ran the MCMC algorithm without the data for culture #8 ( which initially expressed var29 ) , and to derive alternative estimates for var27 we excluded data for cultures #5 and #6 . The results , shown in Figure 2—figure supplement 1 , confirmed our method’s ability to estimate switch biases of non-starter genes with sufficient accuracy to determine the general switching pattern . Off-rate estimates were however not always accurate , especially for genes with very low transcript levels . We used a similar method to test the accuracy of our method for estimating variations in switch biases . Cultures #2 and #3 initially expressed the same starter gene var13 , and both cultures #5 and #6 initially expressed var27 . By excluding the data from culture #2 we derived estimates for the var13 switch biases mostly dependent on the culture #3 data , and vice versa . Similarly we excluded data from each of cultures #5 and #6 to obtain two sets of estimates for the var27 switch biases . For each of the two different data sets we calculated the deviations from the mean switch biases and found these to be positively correlated ( Pearson’s ρ = 0 . 43 , p=0 . 015 ) , confirming the validity of our method . Our likelihood function , used for parameter estimation and model comparisons , assumed that input data were subject to measurement errors following a log-normal distribution . We have previously shown that our MCMC algorithm reliably determines network structures when the measurement error distribution is log-normal ( base 2 ) with standard deviation σ≤1 , and that σ can be reliably estimated using our simulated annealing algorithm ( Noble and Recker , 2012 ) . In the 16-gene model , σ≈0 . 73 , which was well within the required range . To test whether the measurement errors obeyed a log-normal distribution , we analysed the distribution of the deviations between the data and the simulated annealing output , after applying log transformations . The distribution for the 16-gene model was unimodal and only slightly asymmetrical ( skewness 0 . 4 ) but was more peaked than a normal distribution ( excess kurtosis 2 . 4 ) , so that it somewhat resembled a Laplace distribution . We tried running the simulated annealing algorithm with an error function that assumed log-Laplace-distributed noise , and this made very little difference to the results . The deviations for the constrained 38-gene model were similarly distributed , excluding a very small minority of outliers . Unusually large deviations occurred at 13 of the 64 observed time points , 8 of which were between generations 48 and 52 and therefore belonged to the same round of measurements . Transcript levels deviated more for longer genes than for shorter genes , and northern blotting of the samples taken at the affected time points showed unusually low levels of long , for example var length , RNA , confirming that the cause was not PCR error but most likely an as yet unidentified artefact of RNA preparation . Excluding the 13 data points affected by this phenomenon made very little difference to our parameter estimates and statistical test results . Analysis of the qRT-PCR replicates revealed that measurements of lower transcript levels had higher standard errors , but the average standard error increased only threefold for each 1000-fold difference in transcript level . It has been suggested that var gene transcript levels in culture may be affected by small differences in parasite growth rates ( Horrocks et al . , 2004 ) , although no evidence has been found to support this hypothesis . To test whether such differences might affect our estimates we added to our model an excess growth rate parameter ϕi for each gene i:vi , c , t+1= ( 1−ωi ) ( 1+ϕi ) vi , c , t+∑j≠iωjβji ( 1+ϕj ) vj , c , t . When we allowed growth rates to vary by up to 0 . 5% per generation ( equivalent to approximately 50% over the duration of the experiment ) , the extended model was not significantly more likely than the model with only switch biases and off-rates . When we increased the upper bound on the excess growth rates to 1% per generation , the extended model became significantly more likely . However , such large differences in growth rates in vitro are biologically implausible and have never been observed . Importantly , the inclusion of growth rate differences had very little effect on our switch bias and off-rate estimates . | Our ability to acquire immunity to a disease depends on our immune system learning to recognise foreign molecules—called antigens—that are specific to the disease-causing virus , bacterium or parasite . However , some pathogens , such as the malaria-causing parasite Plasmodium falciparum , get around this defence through a process called antigenic variation . This involves the parasite switching between different antigens over the course of an infection , preventing the host immune system from learning to recognise them and leading to infections that last many weeks or even months . The main antigen in P . falciparum is a protein called PfEMP1 , which is encoded by a family of genes called var ( ‘variable’ ) . Var genes have evolved to be highly diverse , and different parasites have different repertoires of around 50–60 var genes . This ensures that there are a huge number of distinct variants of the PfEMP1 antigen available within the population , allowing the malaria parasite to maintain long-lasting infections and also to infect the same individuals again and again . Previous work has shown that the expression of var genes is not random , but it is not clear what determines which genes are expressed at any given time . Now , Noble et al . have performed a detailed investigation of antigenic switching in P . falciparum . Using clonal parasites , they closely monitored the expression of the entire var gene repertoire during many generations of parasite culture . They observed that although different cultures initially expressed distinct var genes , most of them ended up expressing two particular genes—var27 and var29—at high levels , indicating a hard-wired gene ‘activation hierarchy’ . Noble et al . found that whenever the parasites switched antigens , var genes that were centrally located on chromosomes—such as var27 and var29—were more likely to be activated than those at the ends of chromosomes . Moreover , var genes that were highly diverse were more likely to be activated than more conserved genes: this is the first evidence linking var gene evolution with gene activation probabilities . Together , these factors gave rise to the proposed activation hierarchy , which favours genes optimised for immune evasion and aids their continued evolution and diversification . Further work is now needed to identify the molecular mechanisms that control antigenic switching and to determine whether these could represent new therapeutic targets for malaria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2013 | The antigenic switching network of Plasmodium falciparum and its implications for the immuno-epidemiology of malaria |
Competition among sperm to fertilize oocytes is a ubiquitous feature of sexual reproduction as well as a profoundly important aspect of sexual selection . However , little is known about the cellular mechanisms sperm use to gain competitive advantage or how these mechanisms are regulated genetically . In this study , we utilize a forward genetic screen in Caenorhabditis elegans to identify a gene , comp-1 , whose function is specifically required in competitive contexts . We show that comp-1 functions in sperm to modulate their migration through and localization within the reproductive tract , thereby promoting their access to oocytes . Contrary to previously described models , comp-1 mutant sperm show no defects in size or velocity , thereby defining a novel pathway for preferential usage . Our results indicate not only that sperm functional traits can influence the outcome of sperm competition , but also that these traits can be modulated in a context-dependent manner depending on the presence of competing sperm .
Sexual selection operates at the level of reproductive success to promote traits that improve offspring production ( Darwin , 1871 ) . It thus influences a wide array of processes that affect not only the likelihood of mating , but also the probability that gametes will interact within a female to form a viable zygote . In many species , a female can mate with multiple males , resulting in competition between male ejaculates , known as sperm competition ( Parker , 1970 ) . In addition , having multiple mates provides opportunities for a female to influence the outcome , known as cryptic female choice ( Eberhard , 1996 ) . These post-copulatory forms of sexual selection have driven the diversification of sperm and reproductive tract morphologies as well as the divergence of reproductive proteins , and have likely contributed to reproductive isolation and speciation ( Ritchie , 2007; Howard et al . , 2008; Manier et al . , 2013 ) . Sperm competition is a widespread phenomenon that occurs in species utilizing a wide range of reproductive strategies , and a variety of different patterns of preferential usage , generally referred to as precedence , have been observed . ( Smith , 1984; Birkhead and Møller , 1992 , 1998 ) . For example , in some species , the first male to mate may show precedence , while in others , the last mate's sperm may win , and the strength of such effects varies widely . By their nature , events in the reproductive tract that determine the outcome of competition are difficult to study , so in most cases the mechanistic basis for a particular precedence pattern is poorly understood . When sperm competition is intense , males often respond by production and transfer of numerous smaller sperm ( Gomendio et al . , 1998; Simmons , 2001 ) . However , in some cases , sperm may gain an advantage by modulating functional traits , for example , by increasing migration velocity , promoting retention , or blocking subsequent access to the site of fertilization ( Wigby and Chapman , 2004; Gomendio and Roldan , 2008; Pizzari and Parker , 2009 ) . Due to the difficulty of distinguishing sperm from different ejaculates or of observing sperm directly within the selective environment , indirect assays have often been employed to measure sperm usage . The cell behaviors underlying sperm competition have only been investigated in a few species that are amenable to such analyses , and little is known about the genetic basis for differences in competitive ability among cells . However , in vivo imaging studies have recently begun to reveal the cellular mechanisms of sperm behavior in competitive contexts , where multiple males have mated with a female ( e . g . , Civetta , 1999; Manier et al . , 2010; Marie-Orleach et al . , 2014 ) . For example , in Drosophila , analyses of genetically labeled fluorescent sperm have revealed that stored sperm are highly motile and that modulation of sperm storage , release , and ejection by the female contribute strongly to second-male precedence in that organism ( Manier et al . , 2010; Lupold et al . , 2012 ) . Some genetic loci that affect male reproductive success have recently been identified in Drosophila and in mammals ( e . g . , Fiumera et al . , 2005; Sutton et al . , 2008; Yeh et al . , 2012; Civetta and Finn , 2014 ) . Specific seminal fluid components have been shown to play an important role in male competitive advantage by affecting sperm motility and storage , as well as female responses ( e . g . , Mueller et al . , 2008; reviewed in Avila et al . , 2011; Simmons and Fitzpatrick , 2012 ) . However , very few examples are known of genes that function in sperm to control characteristics directly involved in sperm competition . An open question is whether genes exist that specifically regulate competition , without affecting core sperm functions , or whether competitive advantage is always gained by modulating the activity of genes involved in other processes . The nematode Caenorhabditis elegans provides a model system to address the cellular behaviors and molecular pathways that mediate sperm competition . C . elegans is a male-hermaphrodite species in which hermaphrodites produce their own self sperm but also can be inseminated by males . In a self-fertilizing context , hermaphrodite self sperm reside in the spermathecae , sperm storage organs where fertilization occur , and are used with very high efficiency . Typically , more than 99% of sperm go on to fertilize an oocyte ( Ward and Carrel , 1979 ) . However , if mating occurs , male sperm migrate through the uterus to the spermathecae , where they encounter and must compete with stored self sperm . Importantly , during male-hermaphrodite sperm competition , male sperm are used preferentially ( Ward and Carrel , 1979; LaMunyon and Ward , 1995 ) . Male precedence is very robust , and many crosses result in male sperm exclusively fertilizing oocytes . Simple numerical advantage , seminal fluid factors , and the order of introduction into the reproductive tract have been ruled out as potential causes ( Ward and Carrel , 1979; LaMunyon and Ward , 1994 , 1995 ) . Instead , the competitive advantage of C . elegans male sperm has been shown to rely on intrinsic differences between male and hermaphrodite sperm cells . While the form of male and hermaphrodite sperm is the same , male sperm are generally larger than hermaphrodite sperm ( LaMunyon and Ward , 1999 ) . Consistent with the idea that this is significant , experimental evolution under crossing conditions has been shown to lead to increased sperm size ( LaMunyon and Ward , 2002 ) . Like those of other nematodes , C . elegans sperm move by crawling using a pseudopod , and this motility is required for precedence ( Nelson et al . , 1982; Singson et al . , 1999 ) . Larger sperm crawl faster in vitro ( LaMunyon and Ward , 1998 ) , and male sperm displace self sperm from the walls of the spermathecae ( Ward and Carrel , 1979 ) . However , male sperm need not fertilize oocytes to outcompete hermaphrodite sperm; mutant males whose sperm are motile , but fertilization-defective , block self progeny production even though their sperm cannot be used ( Singson et al . , 1999 ) . These data suggest a model for male precedence in which the presence of larger , faster , male sperm leads to the exclusion of self sperm from the fertilization process ( LaMunyon and Ward , 1998; reviewed in Ellis and Stanfield , 2014 ) . Differences in the migration behaviors of male and hermaphrodite sperm could affect the processes of sperm migration towards , retention in , or localization within the spermathecae , where there could be sites especially favorable for sperm–egg interaction ( Han et al . , 2009 ) . Although many mutants defective for spermatogenesis and/or fertilization have been identified in genetic screens , most mutations affect both male and hermaphrodite sperm equally and none specifically affect male precedence ( reviewed in Nishimura and L'Hernault , 2010 ) . Thus , the underlying mechanisms , in terms of either cellular behaviors or genetic controls , remain unclear . In this study , we report the use of a genetic screen in C . elegans to identify a sperm competition gene . While sperm lacking comp-1 activity are used efficiently in the absence of competition , comp-1 sperm are outcompeted by wild-type sperm from either hermaphrodites or males , resulting in reduced reproductive success for both comp-1 mutant males and the hermaphrodites that mate with them . Strikingly , comp-1 sperm are normal in size . However , they show defects in sperm motility and storage in vivo , coupled with context-dependent defects in pseudopodial extension in vitro . These results suggest a model in which comp-1 functions in sperm to coordinate environmental signals that influence motility-related functions required for sperm to compete with one another . Our findings provide key insight into the genetic regulation of sperm competition and suggest that in C . elegans , sperm gain advantage by modulating their motility and storage depending on their competitive milieu .
We took advantage of the male-hermaphrodite reproductive system and robust sperm precedence order of C . elegans to perform a forward genetic screen for males with less-competitive sperm . After a wild-type male mates with and transfers sperm to a hermaphrodite , his sperm rapidly migrates to the spermathecae and begins fertilizing oocytes , and in ideal conditions , most crosses result in more than 90% cross progeny ( Ward and Carrel , 1979; LaMunyon and Ward , 1995 ) . However , the underlying mating and sperm transfer behaviors are variable in efficiency , so that in practice a wide range of cross-progeny frequencies are often observed , and some crosses fail altogether ( Ward and Carrel , 1979 and unpublished observations ) . Thus , for our screen , we developed a sperm competition assay , using spe-8; dpy-4 hermaphrodite recipients , that allowed us to exclude crosses for which cross progeny numbers were decreased due to behavioral defects ( Figure 1A , ‘Materials and methods’ ) . spe-8 hermaphrodites are self-sterile due to a defect in the ability to activate their self sperm to become motile ( L'Hernault et al . , 1988 ) . In the absence of mating , they produce no offspring . However , if a male mates with and transfers seminal fluid to a spe-8 recipient , both the male and self sperm are activated to become motile and fertilization-competent , and since male sperm are superior , they fertilize the vast majority of oocytes ( LaMunyon and Ward , 1995 ) . The dpy-4 mutation is recessive and allows discrimination of self progeny from cross progeny on the basis of the Dumpy phenotype . For our assay , we established mating conditions in which most crosses were successful and fewer than five total Dumpy self progeny were produced in the vast majority of cases , providing a readily scored cutoff for candidate mutants ( data not shown ) . 10 . 7554/eLife . 05423 . 003Figure 1 . Isolation of the male precedence mutant me69 in a genetic screen . ( A ) Screening assay for mutants with reduced male precedence , showing outcomes for mating failure , mating by wild-type males , and mating by males with less-competitive sperm . ( B ) me69 males have decreased precedence in the screen assay . Males were mated to spe-8 ( hc53 ) ; dpy-4 hermaphrodites , and offspring were scored as Dumpy ( self ) or non-Dumpy ( cross ) progeny . ( C ) me69 and gk1149 mutant hermaphrodites have normal self fertility . Total self progeny of unmated hermaphrodites were counted . Error bars , 95% confidence intervals; p > 0 . 05 ( Student's t test ) . ( D ) Mutants for comp-1 have defects in male precedence . Males were mated to dpy-4 hermaphrodites , and offspring were scored as Dumpy ( self ) or non-Dumpy ( cross ) . ( E ) Schematic of the F37E3 . 3 gene showing the kinase-like domain ( green ) predicted by the Conserved Domains Database ( Marchler-Bauer et al . , 2011 ) and the locations of the me69 and gk1149 alleles . Boxes , exons; black , coding regions; gray , 5′ and 3′UTRs . ( F ) Expression of the F37E3 . 3 gene in sperm rescues the comp-1 male precedence defect . Male precedence was assayed for comp-1 ( gk1149 ) ; jnSi168[Ppeel::comp-1] and control strains as in Figure 1D . ( B , D , F ) Each point represents the result of an individual cross; lines indicate medians . *** , p < 0 . 001; ** , p < 0 . 01; ns , not significant ( Kolmogorov–Smirnov test; all comparisons are to wild-type ) . In addition to the genotypes shown , all males were homozygous ( B , C , F ) or heterozygous ( D ) for him-5 ( ok1896 ) , and control strains in ( F ) harbored oxSi221 . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 00310 . 7554/eLife . 05423 . 004Figure 1—figure supplement 1 . COMP-1 is highly conserved within the Caenorhabditis genus and present in related parasitic species . Alignment of C . elegans COMP-1 with its orthologs from other nematode species . Yellow highlighting represents amino acids conserved with the C . elegans protein . Bars above the sequence indicate the positions of the divergent kinase-like ( black ) and SH2-like ( gray ) domains , as predicted in the Conserved Domains Database ( Marchler-Bauer et al . , 2011 ) ; an alternate prediction for the kinase-like domain includes amino acids 123–394 ( Manning , 2005 ) . Positions of the me69 and gk1149 alleles are shown in red and blue as in Figure 1E . For C . japonica , C . sinica , and N . americanus , the predicted COMP-1 proteins present in current databases were incomplete or had regions with poor similarity . In each case , examination of sequence from the comp-1 region allowed us to generate new predicted proteins showing similarity across the full-length Ce-COMP-1 and the other orthologs . For C . sinica , we obtained sequence from the comp-1 region , which revealed a minor assembly error in Csp5 . Scaffold_00675 . g14294 . tt ( PRJNA194557 ) . For C . japonica , previous annotation ( WormBase ) of the comp-1 region predicted two overlapping gene models encoding proteins similar to the N and C terminus of COMP-1; our new predicted Cja-COMP-1 fuses JA64544 and CJA40432 . For N . americanus , we were able to identify additional exons upstream of the comp-1 gene model predicted from NECAME_15795 . Other accession numbers: Acey_s0303 . g1899 . t2 ( PRJNA231479 ) , Cang_2012_03_13_00293 . g8644 . t2 ( PRJNA51225 ) , CBP27128 , RP29840 , Csp5 . Scaffold_00675 . g14294 . tt ( PRJNA194557 ) , Csp11 . Scaffold630 . g17815 . ti ( PRJNA53597 ) , HCOI01934100 . t1 ( PRJEB506 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 00410 . 7554/eLife . 05423 . 005Figure 1—figure supplement 2 . COMP-1 transgenes rescue the male precedence defects of comp-1 mutants . ( A , B ) The jnSi109[Pcomp-1::COMP-1] transgene , which contains a 3 . 9 kb region surrounding F37E3 . 3 , rescues the precedence defect of ( A ) comp-1 ( me69 ) and ( B ) comp-1 ( gk1149 ) males in crosses to dpy-4 hermaphrodites . ( C ) Expression of COMP-1::GFP rescues the precedence defect . comp-1 ( gk1149 ) ; jnSi171[Pcomp-1::COMP-1::GFP] males have a wild-type precedence pattern in crosses to dpy-4 hermaphrodites . Precedence assays were performed as in Figure 1D . *** , p < 0 . 001; ** , p < 0 . 01; ns , not significant ( Kolmogorov–Smirnov test ) . Lines indicate medians . In addition to the indicated genotypes , control strains contained the transgene oxSi221[Peft-3 ::GFP] . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 005 We performed EMS mutagenesis on a male-producing him-5 strain ( Hodgkin et al . , 1979 ) , established lines from individual F2 hermaphrodites , and tested F3 males from each line in the sperm competition assay . We identified one mutant , me69 , which showed reduced male precedence as compared to the wild type ( Figure 1B; GMS , unpublished data ) . The percentage of cross progeny that resulted from mating with me69 mutant males was rarely comparable to that of wild-type crosses . However , me69 mutant hermaphrodites produced a normal number of offspring ( Figure 1C ) , setting the me69 phenotype apart from those of previously identified spe mutants , most of which were isolated based on the reduction of hermaphrodite fertility but usually affect sperm production in males as well ( Nishimura and L'Hernault , 2010 ) . While the use of spe-8 recipients was critical for our screen , their immotile self spermatids cannot maintain proper positioning within the reproductive tract , resulting in mislocalization and gradual loss of the sperm to the external environment ( L'Hernault et al . , 1988 ) . To assess precedence of me69 males in a more natural competitive context , we performed crosses to dpy-4 hermaphrodites , whose sperm localize appropriately to the spermathecae ( data not shown ) . We placed individual L4 males and dpy-4 hermaphrodites together for 40 hr and quantified self and cross progeny generated during this time period . Under these conditions , most matings with wild-type control males resulted in at least some cross progeny , and most successful males sired a high fraction of offspring ( Figure 1D ) . However , matings with me69 males resulted in few or no cross progeny during the time frame of this assay . We confirmed that me69 males were capable of mating and transferring sperm to these hermaphrodites at a high frequency ( 48–85% of crosses were successful , as compared to 74–100% for wild-type ) , so their poor reproductive success was not simply due to behavioral defects . Rather , me69 mutant males show post-copulatory defects in sperm usage consistent with a defect in male precedence . We used meiotic mapping to localize me69 to a 6 . 7-Mb interval on chromosome I ( Davis et al . , 2005 , Tables 1 and 2 ) . Whole-genome sequencing of the me69 strain revealed a likely candidate for the causal mutation as a G to A transition in the coding region of F37E3 . 3 , an uncharacterized gene that we have renamed comp-1 for sperm competition defective ( Figure 1E ) . Based on global expression analyses , comp-1 is expressed in the germ line during time periods that coincide with sperm production: the L4 larval stage in hermaphrodites and in both L4 and adult males ( WormBase; Reinke et al . , 2000 , 2004; Ortiz et al . , 2014 ) . 10 . 7554/eLife . 05423 . 006Table 1 . me69 is linked to chromosome IDOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 006Marker*Genetic position*Genomic position†Haw/+ frequency‡WBVar00240399I:0 . 91I:63508031/16WBVar00172772II:0 . 12II:67892088/16WBVar00067953III:−0 . 31III:831864010/16WBVar00188750IV:1IV:46253173/16WBVar00240687V:0 . 88V:81775209/16*Wicks et al . ( 2001 ) . †WormBase WS243 ( accessed 30 August 2014 ) . ‡me69; him-5 males were crossed to CB4856 Hawaiian hermaphrodites , F1 males were crossed back to me69; him-5 hermaphrodites , and F2 males were assayed for precedence defects in crosses to spe-8; dpy-4 hermaphrodites . Animals scoring as mutant ( me69 homozygotes ) were scored by PCR and restriction digest for centrally-located SNPs on each chromosome . Animals lacking Hawaiian alleles at all loci tested were considered self progeny and excluded from analysis . 10 . 7554/eLife . 05423 . 007Table 2 . Mapping of me69 on chromosome IDOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 007No . F2s*WBVar 00240 394†WBVar 00240 397WBVar 00240 399WBVar 00155 231WBVar 00240 416WBVar 00240 407WBVar 00159 097WBVar 00240 414WBVar 00161 629825026548253163518038646304106146901147209312433167130663811415488916B/BB/BB/B6H/BB/BB/B2H/BB/BB/BB/BB/BB/BB/BH/BH/B3B/BB/BB/BB/BB/BB/BB/BH/BH/B1H/BH/BH/BB/BB/BB/BB/BB/BB/BF2 males from the cross described in Table 1 were scored for SNPs across chromosome I . Animals were either homozygous Bristol ( B/B ) or heterozygous for the Hawaiian allele ( H/B ) at each SNP . *Number of F2 males showing each pattern . †SNP designation and genomic position on chromosome I . Wicks et al . ( 2001 ) ; WormBase . The COMP-1 protein contains divergent SH2 and protein kinase-like domains and has been classified within a ‘unique’ subset of C . elegans kinases that do not fall clearly within defined families ( Manning , 2005 ) ; it also lacks closely related paralogs within the C . elegans genome . It is missing three highly conserved core motifs present in active kinases , including the VAIK motif in the N lobe , the HRD motif in the catalytic loop , and the DFG motif within the activation loop , though it does contain the tripeptide APE motif located within the activation segment ( Figure 1—figure supplement 1 ) ( Hanks et al . , 1988; Hanks and Hunter , 1995; Manning et al . , 2002; Nolen et al . , 2004; Marchler-Bauer et al . , 2011 ) . The absence of these features suggests that the protein is unlikely to have catalytic activity . The me69 allele is predicted to result in a glycine to arginine change in a residue that is conserved in all other orthologs identified to date . COMP-1 orthologs are present in other Caenorhabditis species as well as in the parasites Haemonchus contortus , Ancylostoma ceylanicum , and Necator americanus ( Figure 1—figure supplement 1 ) ( WormBase; Laing et al . , 2013; Schwarz et al . , 2013; Tang et al . , 2014 ) . Although COMP-1 appears to be absent from more distant species ( WormBase , and unpublished data ) , it is present in nematodes that utilize male-female as well as male-hermaphrodite reproductive modes . We obtained a comp-1 deletion allele , gk1149 , from the C . elegans Deletion Mutant Consortium ( C . elegans Deletion Mutant Consortium , 2012 ) . gk1149 eliminates a large region of the coding sequence and is likely a null allele . To test if the me69 and gk1149 alleles result in a similar male precedence defect , we crossed gk1149 males to dpy-4 hermaphrodites and found that gk1149 mutant males indeed showed a reduction in male precedence as compared to the wild type ( Figure 1D ) . Like me69 , gk1149 is recessive; crosses with heterozygous gk1149/+ males showed a wild-type precedence pattern . However , me69/gk1149 heterozygotes had male precedence defects , indicating the two mutations failed to complement one another . To confirm that loss of comp-1 function is responsible for the male precedence defect , we performed rescue experiments . We generated animals harboring a Mos-mediated single copy insertion ( MosSCI ) transgene ( Frøkjær-Jensen et al . , 2008 , 2012 ) encompassing a 3 . 9-kb genomic fragment surrounding F37E3 . 3 ( Tables 3 and 4 ) . This transgene rescued the male precedence defect of both me69 and gk1149 males ( Figure 1—figure supplement 2 ) , confirming that comp-1 is the gene affected in these mutants . 10 . 7554/eLife . 05423 . 008Table 3 . Construction of entry plasmids used to generate targeting constructsDOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 008Fragment descriptionFragment lengthForward primerReverse primerVectorPlasmid namecomp-1 promoter712GGGACAACTTTGTATAGAAAAGTTGCCAGTTCCTCGCCTAGCTTTCGGGACTGCTTTTTTGTACAAACTTGATGCTTTTGATTCGATAGATGATCCpDONR P4-P1rpJMH1comp-1 coding region1921GGGGACAAGTTTGTACAAAAAAGCAGGCTCAATGACGTTGGTCGAATCGAAACGGGACCACTTTGTACAAGAAAGCTGGGTCTTATTTGCGCTGGAATTGATCpDONR 221pJMH2comp-1 coding region without stop codon1918GGGGACAAGTTTGTACAAAAAAGCAGGCTCAATGACGTTGTCGAATCGAAACGGGACCACTTTGTACAAGAAAGCTGGGTATTTGCGCTGGAATTGATCpDONR 221pJMH3comp-1 3’ region561GGGGACAGCTTTCTTGTACAAAGTGGAAGAACTTACGGAAGAATATGGGGGACAACTTTGTATAATAAAGTTGATGCGTTCTCATCAGGCTTCpDONR P2r-P3pJMH4peel-1 coding region* without stop codon3279GGGGACAAGTTTGTACAAAAAAGCAGGCTGCTTAATGCGCTTTGGTAAGGGGGACCACTTTGTACAAGAAAGCTGGGTCTGGATTTTCAACACTTGGATCpDONR 221pJMH20*Seidel et al . ( 2011 ) . 10 . 7554/eLife . 05423 . 009Table 4 . Description of targeting constructs used to generate transgenic worm strainsDOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 009ConstructPosition 1 pDONR P4-P1rPosition 2 pDONR 221Position 3 pDONR P2r-P3Destination vectorLocusPcomp-1::comp-1::comp-1 3’ regionpJMH1pJMH2pJMH4pCFJ150*ttTi5605Pcomp-1::H2B::GFP::comp-1 3’ regionpJMH1pCM1 . 35†pJMH4pCFJ150ttTi5605Ppeel-1::comp-1::tbb-2 3’ regionPpeel-1 [4-1]‡pJMH2pCM1 . 36†pCFJ150ttTi5605Ppeel-1::comp-1::GFP::unc-54 3’ regionPpeel-1 [4-1]pJMH3pGH50§pCFJ150ttTi5605Ppeel-1::comp-1::mCherry::unc-54 3’ regionPpeel-1 [4-1]pJMH3mCherry::unc-54 3’ region§pCFJ150ttTi5605Ppeel-1::peel-1::GFP::unc-54 3’ regionPpeel-1 [4-1]pJMH20pGH50pCFJ212*cxTi10816*Frøkjær-Jensen et al . ( 2008 ) . †Merritt et al . ( 2008 ) . ‡Seidel et al . ( 2011 ) . §Liu et al . ( 2009 ) . To test if comp-1 function is required in sperm cells , we generated MosSCI transgenes to express it specifically in sperm , using the promoter for the peel-1 gene ( Seidel et al . , 2011 ) . We observed full rescue of the male precedence defect in comp-1 ( gk1149 ) ; Ppeel-1::comp-1 males ( Figure 1F ) , indicating that expression of comp-1 in sperm is indeed sufficient to rescue the male precedence defect . Thus , comp-1 acts in sperm to promote their preferential usage . Since COMP-1 is highly conserved in both male-hermaphrodite and male-female species ( Figure 1—figure supplement 1 ) , we hypothesized that comp-1 might function in male–male sperm competition . In the standard laboratory strain of C . elegans ( N2 ) , sequential male matings normally show no precedence pattern , i . e . , the first and second males to transfer sperm are equally likely to sire offspring ( Ward and Carrel , 1979; LaMunyon and Ward , 1998 ) . However , sequential matings of males from different wild-type strains can show preferential sperm usage patterns ( LaMunyon and Ward , 1998; Murray et al . , 2011 ) , indicating that differences in competitive ability can occur among males in this species . To determine if comp-1 function influences sperm competition in a male vs male context , we performed sequential matings of wild-type and/or comp-1 males to fog-2 mutant hermaphrodites , which fail to produce self sperm and are essentially female ( Schedl and Kimble , 1988 ) . To facilitate assignment of paternity , we used strains containing a GFP transgene , mIs11 , for either the first or second sets of crosses , and scored offspring for the presence or absence of fluorescence . In control crosses , in which two wild-type males were sequentially mated to hermaphrodites , progeny numbers from the first and second male were variable , but no consistent bias was observed , other than a weak trend in which non-mIs11 males seemed to be slightly favored over mIs11-containing males ( Figure 2A , Figure 2—figure supplement 1 ) . Similarly , in sequential matings of two comp-1 males , no precedence order was observed . However , sequential matings of wild-type and comp-1 males resulted in strong precedence for the wild-type sperm , regardless of whether wild-type males were the first or second mates . Notably , comp-1 males showed full fertility in crosses to fog-2 hermaphrodites , which lack their own sperm ( Figure 2B ) . These data indicate that comp-1 males transfer normal numbers of functional sperm , which can be used efficiently when they do not need to compete . However , when other sperm are present , comp-1 sperm show poor usage . Furthermore , the reduced usage of comp-1 sperm is unrelated to the order of their introduction into the hermaphrodite reproductive tract . Rather , male sperm lacking comp-1 function appear to have an intrinsic disadvantage as compared to wild-type sperm . 10 . 7554/eLife . 05423 . 010Figure 2 . The comp-1 mutant has defects in male–male sperm competition . ( A ) comp-1 male sperm are outcompeted by wild-type male sperm . Wild-type and/or comp-1 ( gk1149 ) males were mated sequentially to fog-2 hermaphrodites; second-mated males harbored the transgene mIs11 ( GFP+ ) . Offspring were scored for GFP , and the percentage of GFP-positive progeny produced 0–16 hr after second-male mating is shown . ( B ) comp-1 mutant males have wild-type levels of fertility in the absence of competition . Males were crossed to fog-2 hermaphrodites and total progeny were counted . ( A , B ) Individual data points are shown; lines indicate medians . * , p < 0 . 05; *** , p < 0 . 001; ns , not significant ( Kolmogorov–Smirnov test; comparisons are to wild-type unless indicated by a line linking the two data sets ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 01010 . 7554/eLife . 05423 . 011Figure 2—figure supplement 1 . The comp-1 mutant has defects in male–male sperm competition . comp-1 male sperm are outcompeted by wild-type male sperm . Wild-type and/or comp-1 ( gk1149 ) males were mated sequentially to fog-2 hermaphrodites; first-mated males harbored the transgene mIs11 ( GFP+ ) . Offspring were scored for GFP , and the percentage of GFP-positive progeny produced 0–16 hr after second-male mating is shown . GFP-marked males show an apparent slight disadvantage , which is observed consistently but is not statistically significant . Lines indicate medians . *** , p < 0 . 001; ns , not significant ( Kolmogorov–Smirnov test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 011 To investigate the importance of comp-1 activity for male reproductive success , we sought to determine the nature of the competitive defect of comp-1 mutant sperm . In particular , we wished to know if comp-1 sperm were lost or remained active within the gonads of hermaphrodites . To address this question , we assayed the long-term kinetics of usage of comp-1 male sperm within hermaphrodites . We crossed wild-type or comp-1 males to dpy-4 hermaphrodites for 16 hr , transferred the recipients at 12-hr intervals until they ceased egg laying , and counted the total number of self and cross progeny at each time point . Wild-type male sperm usage increased rapidly after mating ( Figure 3A ) , consistent with previous evidence that male sperm are used preferentially over hermaphrodite self sperm ( Ward , 1977; Ward and Carrel , 1979; LaMunyon and Ward , 1995 ) . However , comp-1 mutant males sired almost no progeny until late in the hermaphrodite lifespan ( Figure 3A , B ) . Furthermore , while mating with wild-type males suppressed usage of self sperm , mating with comp-1 males had no effect on self-progeny production ( Figure 3C ) . Thus , comp-1 males show severe long-term defects in their ability to produce offspring after mating , and hermaphrodites that mate with comp-1 males produce a decreased number of total offspring ( Figure 3D ) . However , although they are initially unsuccessful in fertilizing eggs , at least some comp-1 sperm are eventually used , indicating that they can remain in the reproductive tract . 10 . 7554/eLife . 05423 . 012Figure 3 . comp-1 male sperm have long-term precedence defects . ( A ) Crosses with comp-1 males result in a low percentage of cross progeny . ( B ) The number of cross progeny sired by comp-1 increases at late time points . ( C ) Crosses with comp-1 males do not suppress production of self progeny . Purple line indicates self progeny of unmated hermaphrodites . ( D ) Crosses with comp-1 males result in decreased progeny numbers as compared to those with wild-type males . ( A–D ) Males were crossed to dpy-4 hermaphrodites for 16 hr ( gray line ) ; progeny were collected throughout the recipients' reproductive lifespans and scored as self or cross progeny . All graphs are from a single data set that is representative of three repeats . For B–D , cumulative progeny numbers are shown . ( E ) comp-1 male sperm are used at wild-type levels in crosses to sperm-depleted hermaphrodites . Males were crossed to staged dpy-4 recipients for 24 hr and progeny generated during the mating period were scored as self or cross progeny . ‘No . remaining sperm’ indicates the number of self sperm present within recipients at each stage , inferred from brood counts of unmated dpy-4 hermaphrodites performed in parallel . Data points indicate averages; error bars , 95% confidence intervals . ** , p < 0 . 01; *** , p < 0 . 001; ns , not significant ( Kolmogorov–Smirnov test , comparing wild-type to each comp-1 mutant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 012 Since hermaphrodites make their entire store of self sperm prior to oocyte production , they gradually run out of self sperm during adulthood . The onset of comp-1 sperm usage correlated with the depletion of stored hermaphrodite self sperm , suggesting that although comp-1 sperm remained present from matings that occurred at an earlier time , they were only used once fewer self sperm were present to compete with ( Figure 3B , C; compare time points at 52 and 76 hr ) . To test if comp-1 sperm can be used more rapidly when fewer self sperm are present , we aged hermaphrodites until they had used up part or all of their self sperm reservoir , then crossed them to males and assessed the short-term usage of male sperm . In crosses to 12 , 24 , and 36 hr post-L4 recipients , which retain moderate levels of self sperm ( see Figure 3E; ‛No . remaining sperm’ ) , the number of offspring sired by comp-1 males increased proportionally to the age of the hermaphrodite , but success was always reduced as compared to wild-type males ( Figure 3E ) . However , in crosses performed with 48 hr post-L4 hermaphrodites , which have nearly run out of self sperm , comp-1 males produced as many offspring as the wild type . Thus , regardless of the length of time they have been resident in the reproductive tract , comp-1 sperm are unsuccessful specifically in situations where other sperm are present , but can be used in the absence of competition . Global expression studies suggested that comp-1 is expressed not only in males , but also in hermaphrodites ( Reinke et al . , 2000 , 2004 ) . We sought to determine if COMP-1 is indeed present in hermaphrodite sperm and if it shows any differences in localization as compared to male sperm . We first generated transgenic animals carrying a Pcomp-1::GFP::H2B transcriptional reporter , in which GFP localizes to the nuclei of comp-1-expressing cells . Expression was visible in developing spermatocytes and spermatids in both males and hermaphrodites , and we observed no obvious differences in abundance between the two sexes ( Figure 4A–D and data not shown ) . 10 . 7554/eLife . 05423 . 013Figure 4 . COMP-1 is expressed and functions in sperm of both males and hermaphrodites . ( A–D ) Images of jnSi118[Pcomp-1::GFP::H2B] adult males ( A , B ) and hermaphrodites ( C , D ) , which express the comp-1 reporter in developing sperm . int , intestinal autofluorescence . Scale bar ( A–D ) , 30 μm . ( E ) Schematic of structural organization of spermatozoa . ( F–H ) COMP-1 does not colocalize with GSP-3/4 , which is in the pseudopod . Images of jnSi171[COMP-1::GFP] male spermatozoa fixed and stained with α-GSP-3/4 antibody ( red ) and DAPI ( blue ) . Scale bar ( F–P ) , 5 μm . ( I–L ) COMP-1 does not colocalize with mitochondria . Images of jnSi171[COMP-1::GFP] male spermatozoa stained with Mitotracker . ( M–P ) COMP-1 does not colocalize with PEEL-1::GFP , which is at the membranous organelles . Images of jnSi143[COMP-1::mCherry]; jnSi177[PEEL-1::GFP] male spermatozoa . ( Q ) comp-1 functions in both male and hermaphrodite sperm . Wild-type and comp-1 ( gk1149 ) males were tested against wild-type and comp-1 ( gk1149 ) hermaphrodites in the short-term precedence assay . Lines indicate medians . * , p < 0 . 05; *** , p < 0 . 001; ns , not significant , Kolmogorov–Smirnov test . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 01310 . 7554/eLife . 05423 . 014Figure 4—figure supplement 1 . A COMP-1::GFP transgene rescues the male precedence defects of comp-1 mutants . Expression of COMP-1::GFP rescues the precedence defect . comp-1 ( gk1149 ) ; jnSi171[Pcomp-1::COMP-1::GFP] males have a wild-type precedence pattern in crosses to dpy-4 hermaphrodites . Precedence assays were performed as in Figure 1D . *** , p < 0 . 001; ** , p < 0 . 01; ns , not significant ( Kolmogorov–Smirnov test ) . Lines indicate medians . In addition to the indicated genotypes , control strains contained the transgene oxSi221[Peft-3 ::GFP] . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 014 To determine the localization of COMP-1 , we generated worm strains expressing transgenes that contained the full-length comp-1 coding region fused to either mCherry or GFP . Worms carrying the GFP fusion showed rescue of the male precedence defect , suggesting that the fluorescent tags did not interfere with protein function or localization ( Figure 4—figure supplement 1 and data not shown ) . The COMP-1 fusion proteins displayed a punctate pattern in the cytoplasm of both developing spermatids and mature sperm , where they were restricted to the cell body region ( Figure 4E–F and data not shown ) . These punctae were visible in sperm from both males and hermaphrodites ( Figure 4 and data not shown ) . To determine whether the COMP-1 protein was localized to a specific subcellular location , we performed co-labeling experiments with the vital dye Mitotracker , a marker of mitochondria , and PEEL-1::GFP , which labels the sperm-specific membranous organelles ( MOs ) ( Chen et al . , 2000; Seidel et al . , 2011 ) . We also examined the phosphatase GSP-3/4 , which is involved in cytoskeletal dynamics and shows polarized localization within the pseudopod ( Wu et al . , 2011 ) . COMP-1 did not colocalize with any of these markers of sperm structure or with the sperm nucleus ( Figure 4E–P ) , and its absence from the pseudopod suggests that it is not involved directly with cellular locomotion , at least by modulating cytoskeletal dynamics . The absence of obvious differences between males and hermaphrodites in the expression and localization of COMP-1 raised the question of a potential role for comp-1 in hermaphrodite self sperm . We thus assayed precedence of wild-type and comp-1 mutant males in crosses to comp-1 hermaphrodites . Matings of wild-type males to comp-1 hermaphrodites resulted in even higher levels of cross progeny production than those seen in crosses to wild-type hermaphrodites , consistent with comp-1 hermaphrodite sperm having reduced ability to compete ( Figure 4Q ) . Interestingly , when comp-1 males were mated to comp-1 hermaphrodites , mutant male sperm usage was indistinguishable from that of wild-type , suggesting that the male precedence order is regained when comp-1 sperm compete against each other . These results indicate that comp-1 functions to promote sperm usage not only in males , but also in hermaphrodites . In addition , factors other than comp-1 must influence the outcome of competition , as a strong precedence effect can be observed in the absence of its activity in both competing populations of sperm . One potential explanation for the male precedence defect in comp-1 mutants was that sperm do not undergo proper spermatogenesis or spermiogenesis necessary to mature into functional sperm . Loss of function of spe or fer genes required for spermatogenesis generally leads to hermaphrodite self sterility and male infertility , and reduction of gene function can result in partial fertility ( Kadandale and Singson , 2004; Nishimura and L'Hernault , 2010 ) . We thus examined available markers of sperm morphology to determine if comp-1 sperm harbor any general defects . Males and hermaphrodites both produce immotile , spherical spermatids that must be activated to become mature , pseudopod-bearing sperm competent for motility and fertility ( Wolf et al . , 1978 ) . comp-1 mutant spermatids and sperm appear grossly normal by light microscopy ( Figure 5A , C; Figure 5E , G and data not shown ) . In addition , several markers of sperm structures localized appropriately in the comp-1 mutant . As in wild-type sperm , mitochondria and membranous organelles were restricted to cell bodies ( Figure 5A–H ) , and the GSP-3/4 phosphatase was polarized within pseudopodia ( Figure 5I–L ) . The presence of properly polarized sperm structures in mutant sperm indicates that comp-1 is not required to complete spermatogenesis , nor is it necessary for proper localization of sperm structures in mature sperm cells . These findings are consistent with the absence of fertility or sperm usage defects in comp-1 animals in the absence of competition . 10 . 7554/eLife . 05423 . 015Figure 5 . comp-1 sperm have normal organization and size . ( A–D ) Wild-type ( A , B ) and comp-1 ( gk1149 ) ( C , D ) spermatozoa stained for mitochondria using Mitotracker . ( E–H ) Wild-type ( E , F ) and comp-1 ( G , H ) spermatozoa expressing PEEL-1::GFP ( membranous organelles ) . ( I–L ) Wild-type ( I , J ) and comp-1 ( gk1149 ) ( K , L ) spermatozoa fixed and stained with α-GSP-3/4 antibody ( green ) . ( A–L ) Scale bar , 5 μm . ( M ) comp-1 male spermatid size is not significantly different from wild-type . Cross sectional areas through the center of spermatids were measured . Error bars , 95% confidence interval; p = 0 . 41 , Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 015 In C . elegans , a key factor in conferring male precedence is thought to be the differential size of male and hermaphrodite sperm cells . Male sperm are generally larger than hermaphrodite sperm , correlating with faster crawling speeds in vitro , and growth under conditions with a high risk of sperm competition has been shown to result in increased sperm size ( LaMunyon and Ward , 1998 , 2002 ) . Thus , we investigated the possibility that the precedence defects of comp-1 males might be due to a reduction in the size of mutant sperm . To assay cell size , we measured spermatids , which are spherical , by obtaining a cross-sectional area through the center of each cell ( LaMunyon and Ward , 1998 ) . comp-1 mutant spermatids were variable in size , but the average and distribution of their sizes were indistinguishable from those of wild-type spermatids ( Figure 5M ) . Therefore , we conclude that loss of comp-1 does not reduce competitive ability by affecting cell size . Furthermore , C . elegans sperm can achieve precedence by a size-independent mechanism . In C . elegans , sperm are stored and fertilization occurs within the spermathecae ( Ward and Carrel , 1979 ) . Transferred male sperm must migrate through the uterus and into the spermathecae to be eligible to fertilize oocytes , and male sperm have been observed to displace hermaphrodite sperm from the walls of these structures ( Ward and Carrel , 1979 ) . We thus examined the ability of comp-1 mutant sperm to migrate toward and access the spermathecae . We crossed unlabeled hermaphrodites to males either labeled with Mitotracker dye ( Kubagawa et al . , 2006; Stanfield and Villeneuve , 2006 ) or expressing a sperm H2B::GFP reporter , and then examined male sperm positioning at different time points after transfer to the hermaphrodite reproductive tract . Similar to previously reported analyses of sperm migration ( Kubagawa et al . , 2006 ) , we divided each proximal gonad arm into four regions: zone 1 , near the sperm entry point at the vulva; zone 2 , within the uterus; zone 3 , the region near the spermatheca; and the spermatheca itself ( Figure 6A ) . 10 . 7554/eLife . 05423 . 016Figure 6 . comp-1 sperm have defects in migration and spermathecal accumulation . ( A ) Schematic of hermaphrodite gonad arm showing zones used to quantify sperm position . ( B , C ) Localization of wild-type ( B ) and comp-1 ( gk1149 ) ( C ) Mitotracker-labeled male sperm 1–1 . 5 hr after transfer to hermaphrodites . Percentage of total male sperm is shown . ( D , E ) Localization of jnSi118[GFP::H2B] male sperm 12 hr ( D ) and 24 hr ( E ) after transfer to hermaphrodites . Percentage of male sperm in the focal plane with maximum sperm in the spermatheca is shown . ( F ) Localization of hermaphrodite self sperm in 24 hr post-L4 unmated hermaphrodites . Animals were stained with DAPI to facilitate counting of sperm cells . Percentage of total hermaphrodite sperm is shown . Error bars , 95% confidence intervals . * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001; Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 016 By 1–1 . 5 hr after transfer , a majority of wild-type male sperm had migrated to zone 3 and the spermatheca ( Figure 6B ) . Some crosses with comp-1 males also showed this pattern . However , in many cases , a large percentage of comp-1 sperm remained in zone 1 and/or zone 2 , and accumulation in zone 3 and the spermatheca was reduced ( Figure 6C ) . Importantly , wild-type and comp-1 sperm were present in similar , high numbers ( an average of 222 . 4 ± 78 . 6 for wild-type , n = 13; an average of 191 . 1 ± 63 . 3 for comp-1 , n = 14 ) , and there was no obvious correlation between the number of sperm transferred and their migration efficiency ( data not shown ) . Sperm-specific expression of comp-1 rescued the migration defect , confirming that the altered migration was due to loss of comp-1 ( data not shown ) . By 12 hr after transfer , both wild-type and mutant sperm were rarely found in zones 1 and 2 , instead localizing to zone 3 and/or the spermatheca ( Figure 6D; see below ) . Thus , mutant male sperm show a delay in reaching the spermathecal region . However , their ability to accumulate near the spermathecae at later time points indicates that they are competent to respond to directional cues . In addition to this delay in migration , we observed a significant decrease in residency of comp-1 sperm within the spermathecae . At 12 hr after transfer , when wild-type sperm consistently occupied the spermathecae , very few comp-1 sperm localized there , even though they were present in zone 3 ( Figure 6D ) . Interestingly , by 24 hr post-mating , mutant male sperm numbers increased within the spermathecae and there was little , if any , difference between wild-type and comp-1 sperm positions ( Figure 6E ) . This later time point corresponded to 48 hr post-L4 adult hermaphrodites , in which self sperm numbers are largely depleted and comp-1 male sperm start to show increased usage ( Figure 3B , E ) . Taken together , these results suggest that mutant male sperm are not used because they are present at lower numbers in the spermathecae during periods when these structures are occupied with large numbers of self sperm . Since fertilization can occur only within these structures , this defect is likely the primary reason for the reduction in the competitive ability of comp-1 sperm . Since comp-1 functions in both male and hermaphrodite sperm , we also analyzed self sperm in comp-1 hermaphrodites to assess whether localization defects might still be present in a non-competitive context . We quantified the position of sperm in different zones in DAPI-stained 24 hr adult hermaphrodites . In wild-type hermaphrodites , most of the sperm resided in zone 3 , tightly concentrated just outside of the spermathecae ( Figure 6F and data not shown ) ; a smaller number was present within the spermathecae . In comp-1 hermaphrodites , while the majority of sperm were localized within zone 3 , fewer sperm resided within the spermathecae as compared to wild-type . In addition , some comp-1 self sperm were mislocalized to zone 2 and occasionally zone 1 . Thus , comp-1 hermaphrodite sperm have minor defects in localization and spermathecal residency that are similar to those of comp-1 male sperm . However , these defects do not result in reduced fertility . To probe the cellular basis for the localization defects of comp-1 sperm , we analyzed their motility using established in vivo and in vitro assays ( Geldziler et al . , 2011 ) . Measured immediately after transfer , the migration velocities of comp-1 sperm within the hermaphrodite uterus were indistinguishable from those of wild-type ( Figure 7A ) . Furthermore , migrating comp-1 sperm showed highly directional movement through the uterus towards the spermathecae ( Figure 7A ) and a low reversal frequency , consistent with guided migration ( among cells analyzed for motility , only 3/28 wild-type cells and 1/25 comp-1 cells showed one or more reversals during the assay period ) . The ability of comp-1 sperm to migrate rapidly in vivo suggests that basal motility is not affected in the mutant . However , the difference between wild-type and comp-1 mutant sperm migration patterns could be due to aspects of other migratory behaviors , such as the amount of time individual sperm spend actively migrating through the reproductive tract . 10 . 7554/eLife . 05423 . 017Figure 7 . comp-1 male sperm can migrate normally but have context-dependent defects in cell morphology . ( A ) comp-1 ( gk1149 ) sperm can migrate in vivo at speeds equivalent to wild-type sperm . Mitotracker-labeled males were crossed to N2 hermaphrodites and time-lapse images of sperm migrating through zone 2 were collected . Velocity and directional velocity toward the spermatheca were measured using ImageJ . ( B–F ) comp-1 ( gk1149 ) spermatids show reduced pseudopodial extension after activation by Pronase . ( B ) Quantification of aspect ratios of wild-type and comp-1 ( gk1149 ) sperm treated with either TEA or Pronase . ( C–F ) Representative images of wild-type ( C , E ) and comp-1 ( gk1149 ) ( D , F ) sperm treated with TEA ( C , D ) or Pronase ( E , F ) . Error bars , 95% confidence interval; *** , p < 0 . 001 , Kolmogorov–Smirnov test . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 01710 . 7554/eLife . 05423 . 018Figure 7—figure supplement 1 . comp-1 sperm can crawl and be activated in vitro . ( A ) comp-1 sperm can crawl at wild-type velocities in vitro . Wild-type and comp-1 ( gk1149 ) spermatids were treated with TEA for 30 min and velocity was obtained from time-lapse images collected every 30 s . As we observed a high level of variability among different samples for each genotype , the range of observed values is shown using a box-and-whiskers plot . For each genotype , n = 5–6 samples , 65–130 cells . ( B ) comp-1 sperm activate in Pronase . Wild-type and comp-1 ( gk1149 ) spermatids were treated with Pronase for 30 min and scored for activation based on the presence or absence of a pseudopod . Error bars , 95% confidence intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 018 To further analyze the motility of comp-1 mutant sperm , we dissected spermatids , treated them with the known in vitro activators TEA ( triethanolamine , a weak base ) or Pronase ( a protease mixture ) ( Ward et al . , 1983; Shakes and Ward , 1989 ) , and sought to measure the velocities of cells crawling on glass slides ( Nelson et al . , 1982 ) . comp-1 sperm activated in TEA had extended pseudopods and were capable of crawling at speeds similar to those of wild-type cells ( Figure 7B–D , Figure 7—figure supplement 1A ) . comp-1 sperm treated with Pronase activated at rates similar to the wild type ( Figure 7—figure supplement 1B ) , based on the presence of a pseudopod in the majority of cells . However , the shapes of comp-1 cells were markedly different from wild-type ( Figure 7E–F and data not shown ) . Quantification of pseudopod length , using an aspect ratio measurement to normalize for variation in cell size ( Batchelder et al . , 2011 ) , confirmed that Pronase-treated comp-1 cells were significantly shorter than either wild-type or TEA-treated comp-1 cells ( Figure 7B ) . Since Pronase-treated comp-1 cells contained distinct cell body and pseudopod regions , with normal localization of organelles ( Figure 5 , Figure 7 , and data not shown ) , it is likely that these cells were polarized but failed to extend their pseudopods appropriately . Similar to other amoeboid cells , locomotion of nematode sperm depends on protrusion of the lamellipodium-like pseudopod , adhesion to substrate , and retraction of the cell body ( Roberts and Stewart , 2000; Bottino et al . , 2002 ) . Pseudopod extension defects would be expected to result in altered locomotion and/or interactions with the hermaphrodite reproductive tract , which in turn should affect migration to and occupation of the spermathecae .
For males , reproductive success depends on several aspects of sperm function . To fertilize oocytes , sperm must be transferred , become motile , and migrate to the site of fertilization in response to guidance signals ( Ward and Carrel , 1979; Kubagawa et al . , 2006 ) . Overall fecundity depends on the number of sperm that accomplish these behaviors as well as their ability to be stored so as to ensure long-term usage ( Murray and Cutter , 2011 ) . We have found that comp-1 sperm are transferred at rates comparable to the wild type , so they achieve initial entry into the reproductive tract , but they then show varying defects in the ensuing steps ( Figure 8 ) . Although comp-1 sperm show delays in migration toward the spermathecae , at least some sperm migrate rapidly and directionally , arguing against a defect in locomotion per se . Large numbers eventually accumulate in the spermathecal region , suggesting that they respond to directional cues , but they are generally found outside the spermathecal valve . Once self-sperm stores are depleted , comp-1 sperm concomitantly gain residency in the spermathecae and begin to fertilize oocytes . Since fertilization occurs only in these structures , it is likely that this localization defect underlies the reduced competitive ability and generally poor reproductive success of comp-1 males . 10 . 7554/eLife . 05423 . 019Figure 8 . Model: comp-1 sperm have localization defects that result in failure to compete with wild-type sperm . ( A ) Wild-type male sperm ( blue ) migrate to the region of the spermathecae , where they displace hermaphrodite self sperm ( pink ) and preferentially fertilize oocytes . Oocytes fertilized by male sperm are shown in blue; oocytes fertilized by self sperm are shown in pink . ( B ) comp-1 mutant male sperm ( light blue ) migrate to the spermathecae , but remain outside while wild-type sperm ( pink ) are present , and are thus excluded from opportunities to fertilize oocytes . They also show delayed migration to the spermathecal region and increased localization in the periphery of the female reproductive tract . DOI: http://dx . doi . org/10 . 7554/eLife . 05423 . 019 Spermathecal occupancy depends on the balance between the rate of entry due to migration and the rate of loss due to displacement by oocytes , which rearrange and even expel a subset of stored sperm as they pass through during ovulation ( Ward , 1977; Ward and Carrel , 1979 ) . Male sperm could thus increase their numbers in the spermathecae either by resisting removal , for example , by increasing adhesion to the spermathecal walls , and/or by migrating quickly back into the spermathecae , for example , by increasing their crawling velocity . Defects in pseudopodial extension like those observed in vitro for comp-1 could affect either of these processes , allowing wild-type sperm to preferentially associate . Future studies will be necessary to differentiate between the two models , as well as to characterize the dynamics of sperm behavior in storage . A unique feature of the comp-1 phenotype is its dependence on the presence of wild-type sperm in recipient hermaphrodites . Do comp-1 sperm defects occur because wild-type sperm behavior is superior , leading to their physical displacement , or does the presence of wild-type sperm make comp-1 sperm inferior , through an indirect mechanism such as signaling ? In some organisms , sperm may cooperate by associating with one another to promote fertility or by providing different functions within an ejaculate ( reviewed in Higginson and Pitnick , 2010 ) , but neither cooperative nor detrimental interactions between sperm have been described for C . elegans . Some defects in localization of comp-1 sperm are observed in the absence of competing cells; a few sperm can be found scattered throughout the uterus , though this mislocalization apparently does not lead to significant reduction in usage or loss from the reproductive tract . These findings are consistent with the migration and localization defects we observed in competitive contexts , and they indicate that comp-1 defects are not solely induced by the presence of wild-type sperm . However , we cannot exclude the possibility that sperm–sperm interactions influence the outcome of competition between comp-1 and wild-type cells . Several studies have demonstrated a strong association between precedence and cell size in C . elegans ( LaMunyon and Ward , 1998 , 1999; Murray et al . , 2011 ) . However , loss of comp-1 has no effect on cell size . Furthermore , comp-1 activity appears to override the contribution of size , since large comp-1 male sperm are completely outcompeted by small wild-type hermaphrodite sperm . Interestingly , the normal male precedence order is restored when both male and hermaphrodite sperm lack comp-1 function , consistent with the idea that the size effect again predominates . Our data thus suggest that multiple activities contribute to precedence and can be independently modulated to affect sperm competitive ability . We note that a mechanism involving altering the activity of COMP-1 is likely to be less costly than production of larger sperm , which is associated with a reduced rate of sperm production ( LaMunyon and Ward , 1998; Murray et al . , 2011 ) . How does COMP-1 function in sperm to alter motility-related behaviors ? For crawling cells , locomotion and interaction with substrate are dependent on maintenance of polarity and extension of the lamellipodium or the pseudopod in the case of nematode sperm ( reviewed in Lammermann and Sixt , 2009; Reig et al . , 2014 ) . C . elegans sperm are stably polarized , though the shape and size of their pseudopods is dynamically regulated ( Nelson et al . , 1982 ) . Markers of the cell body and pseudopod are appropriately localized in comp-1 sperm , suggesting polarity is not disrupted . However , treatment with Pronase in vitro , which is thought to mimic the endogenous male activator ( Smith and Stanfield , 2011 ) , generates activated cells with severely shortened pseudopods . The sperm cytoskeleton lacks actin and instead consists of Major Sperm Protein ( MSP ) , which generates a network of fibers that drives cell protrusion via its expansion and contraction ( Italiano et al . , 1999; Roberts and Stewart , 2012 ) . In the related nematode Ascaris , MSP filament assembly is mediated by MPOP , a pH-dependent phosphoprotein that is active at the leading edge ( LeClaire et al . , 2003 ) and the soluble proteins MFP1 and MFP2 ( Buttery et al . , 2003 ) . MSP dynamics are also governed in part by the PP1 phosphatase GSP-3/4 , which localizes to the proximal pseudopod near the cell body ( Wu et al . , 2011 ) . Since COMP-1 localizes to the cell body , it seems unlikely to interact directly with the MSP cytoskeleton , but rather might function upstream of locomotion per se . COMP-1 contains a protein kinase-like domain , which might suggest a role in signal transduction . Like many other reproductive proteins , it represents a divergent member of its family , and its primary sequence suggests that it is unlikely to be catalytically active . However , in spite of lacking or having reduced enzymatic activity , pseudokinases have been shown to play important roles in cell signaling via interactions with active kinases or their substrates , scaffolding or tethering of signaling complexes , and other mechanisms ( reviewed in Reiterer et al . , 2014 ) . The punctate localization of COMP-1 within the sperm cell body is intriguing in this context . Our finding that comp-1 sperm have reduced pseudopod lengths in an in vitro assay fits with their altered patterns of localization in vivo . However , measurements of cell velocity indicate that cells lacking comp-1 are capable of wild-type crawling speeds and they eventually accumulate near their appropriate target . Therefore , it is probable that the cellular defects of comp-1 sperm in vivo are less severe than those of Pronase-treated comp-1 sperm , which have severely shortened pseudopods and should be nearly incapable of movement ( Nelson et al . , 1982; LaMunyon and Ward , 1999 ) . The comp-1 phenotype is also distinct from that caused by lack of prostaglandin cues involved in guidance toward oocytes , which leads to a severe reduction in crawling velocity along with loss of directionality ( Kubagawa et al . , 2006; Edmonds et al . , 2010 ) . Thus , comp-1 sperm are capable of directional migration , though some aspect of sensing or responding to prostaglandins could be impaired . Alternatively , the altered localization of comp-1 sperm could stem from decreased adhesion to the substrate , leading to a reduced ability to crawl directionally and/or maintain position within the spermathecae . Overall , the context dependence of comp-1 sperm usage suggests that cellular defects may be limited to a subset of sperm cells or may be manifested only some of the time , for example during interaction with particular substrates within the reproductive tract . Sperm migrate across a variety of tissues including uterine and spermathecal cells and fertilized eggs , each of which could be more or less permissive for migration of comp-1 mutant cells due to effects on either adhesion or signaling . The role of comp-1 in C . elegans is evident by the reduction in reproductive success for both sexes in crosses to comp-1 males . Wild-type males who mate successfully can produce hundreds ( up to thousands ) of offspring ( Wegewitz et al . , 2008 ) , but comp-1 males produce very few cross progeny , and these are delayed until other sperm are no longer available . Hermaphrodites mated to wild-type males increase their overall progeny production , but this increase is significantly lower in crosses to comp-1 males , and few cross progeny are generated . Even in crosses between comp-1 males and comp-1 hermaphrodites , where the male precedence order is largely restored , males show reduced success as compared with wild-type to wild-type matings . Thus , sperm with comp-1 function should be highly selected for usage when competing with sperm without comp-1 . In male-female species , we expect that comp-1 may have a similar function in improving male reproductive success , depending on the rate of polyandry in a given population . Although self fertilization allows C . elegans to propagate without the need to mate and eliminates the cost of producing males , it also leads to reduced genetic variation ( discussed in Anderson et al . , 2010 ) . The rate of outcrossing in wild populations is estimated to be low , yet males exist , suggesting that some outcrossing may be selected for , or alternatively , that androdioecy has arisen sufficiently recently that the specialized developmental and behavioral characteristics of males have not had time to degrade . Selective pressure has been shown to increase the rate of outcrossing in C . elegans in several experimental schemes ( Lopes et al . , 2008; Morran et al . , 2009a; Morran et al . , 2009b; Anderson et al . , 2010 ) . By promoting the preferential usage of male sperm , COMP-1 should function to increase the genetic diversity of offspring and thus may confer a fitness benefit in situations where adaptation is beneficial ( Carvalho et al . , 2014 ) . The outcome of sperm competition depends on the arena in which it occurs , which depends on the specialized reproductive biology and anatomy of the species in question . In particular , differences in the capacity of the sperm storage organ ( s ) , functional characteristics of sperm and seminal fluid , and the degree of sperm mixing lead to distinct patterns of sperm usage ( Parker and Pizzari , 2010 ) . In C . elegans , the spermathecae are somewhat limited as storage sites , which likely reduces the incentive for males to produce and transfer vast numbers of sperm . Instead , the arms race between the sexes leads to males producing sperm that are functionally superior . Once male sperm reach the spermathecae , they are immediately used even though they lack numerical superiority ( GMS , unpublished data ) . Interactions between competing ejaculates can be divided into offense , the ability to displace previous sperm , and defense , the ability to block subsequent sperm . Observations of the processes of ovulation , sperm migration , and fertilization in wild-type C . elegans , as well as the ability of fertilization-incompetent sperm to sterilize hermaphrodites , suggest that wild-type male sperm most likely block the access of self sperm to the site of fertilization ( Ward and Carrel , 1979; Singson et al . , 1999 ) . However , they fail to block the sperm of another male , as no precedence order is observed in sequential wild-type matings ( Ward and Carrel , 1979; LaMunyon and Ward , 1998 ) . comp-1 male sperm lack the ability to suppress self progeny production , and they also show severe defects in male–male competition whether they are the first or a subsequent mate . Thus , they appear to totally lack the offensive capabilities of normal sperm and also show defects in defense against new rivals . COMP-1 is present in both male-hermaphrodite and male-female species of nematodes . Since the male–female reproductive mode is ancestral ( Kiontke et al . , 2004; Cutter et al . , 2008 ) , the function of COMP-1 in sperm competition most likely originated in male–male competition and has been retained in androdioecious species , such as C . elegans , where it remains necessary for both male–male and male-hermaphrodite sperm competition . Our results thus establish that C . elegans provides a general model to study the molecular mechanisms that underlie sperm competition as well as the interplay between the cell biology of sperm and the forces of sexual selection .
C . elegans strains were grown at 20°C , except where noted , and fed with OP50 Escherichia coli bacteria as previously described ( Brenner , 1974 ) . All strains were derived from the N2 Bristol wild-type strain , with the exception of the CB4856 Hawaiian strain used for mapping . For experiments involving males , him-5 strains were used as our wild-type: him-5 ( e1490 ) was used for the genetic screen and him-5 ( ok1896 ) was present in all other strains from which males were obtained ( Hodgkin et al . , 1979 ) , unless explicitly noted . comp-1 ( me69 ) was identified in this study and the comp-1 ( gk1149 ) allele was generated by the C . elegans Deletion Mutant Consortium ( 2012 ) . Other alleles used for experiments were spe-8 ( hc40 , hc53 ) I , mIs11[myo-2::GFP , pes-10::GFP and gut::GFP] , ttTi5605 II , oxSi221[Peft-3::GFP] II , unc-119 ( ed3 ) III , fem-3 ( q20gf ) IV , dpy-4 ( e1166 ) IV , cxTi10816 IV , fog-2 ( q71 ) V , and him-5 ( e1490 , ok1896 ) V ( Wood and the Community of C . elegans Researchers , 1988; Maduro and Pilgrim , 1995; Frøkjær-Jensen et al . , 2008 , 2012; Meneely et al . , 2012 ) . To generate transgenic strains , Mos-mediated Single Copy Insertion ( MosSCI ) was used to integrate transgenes at the ttTi5605 II and cxTi10816 IV loci ( Frøkjær-Jensen et al . , 2008 , 2012 ) . The me69 mutant was isolated in a screen for males with reduced sperm precedence or fertility . him-5 ( e1490 ) hermaphrodites were mutagenized using ethyl methanesulfonate ( EMS ) mutagenesis as described in Wood and The Community of C . elegans Researchers , 1988 . Groups of 7–8 P0 hermaphrodites were allowed to self-fertilize; L4 F1 hermaphrodites were picked ( 25 per plate ) ; and individual L4 F2s were used to establish lines potentially homozygous for newly induced mutations . To assay male precedence , from each viable line 4–5 L4 males were mated to one spe-8 ( hc40 ) ; dpy-4 hermaphrodite for approximately 48 hr , at which time the cross was terminated by removing the hermaphrodite . When all progeny reached at least the L4 stage , mating plates were examined . If at least 5 Dumpy ( self ) progeny were present , the number of Dumpy ( self ) and non-Dumpy ( cross ) progeny were counted . Such lines were retested using the same precedence assay as before . Approximately 3400 mutagenized lines were tested and 16 lines were recovered as homozygous mutants . Of the 16 lines , six lines had normal gonadal and sperm morphology , consistent with a precedence-specific defect . The me69 mutant was among those 6 lines . To map me69 , CB4856 ( Hawaiian ) males were crossed to me69; him-5 hermaphrodites , F1 males were crossed back to me69; him-5 hermaphrodites , and individual F2 males were tested for the male precedence defect . Each male was recovered into lysis buffer and males scoring as mutant were assayed for a centrally located SNP on each chromosome ( Wicks et al . , 2001 ) . Linkage was detected to chromosome I and additional SNPs were scored in individual males to narrow me69 to a 6 . 7-Mb region between WBVar00240399 and WBVar00240414 ( Tables 1 and 2 ) ( WormBase; Jakubowski and Kornfeld , 1999 ) . To identify the gene affected in me69 , whole genome sequence was obtained from the strain isolated in our genetic screen . Of 45 variations in the me69 region , 24 were consistent with EMS , seven affected coding regions , and only one affected a gene ( F37E3 . 3 ) showing sperm-enriched gene expression . Molecular biology was performed according to standard protocols . The Multisite Gateway Three-Fragment Vector Construction Kit ( Life Technologies , Grand Island , NY ) was used to construct donor plasmids . Fragments were then recombined into the MosSCI destination vectors pCFJ150 or pCFJ212 ( Frøkjær-Jensen et al . , 2008 , 2012 ) . For constructs in which two fragments were ligated by PCR , fusion PCR was performed as in Hobert ( 2002 ) . Primers used for generating constructs are listed in Table 3 , and plasmid construction strategies are summarized in Table 4 . To measure hermaphrodite fertility , L4 hermaphrodites were individually placed on a freshly seeded lawn and moved to a new plate every 24 hr until eggs were no longer laid . To measure male fertility , L4 males were crossed in a 1:1 ratio to L4 fog-2 females for 24 hr . The males were then removed and the females were transferred every 24 hr until egg laying ceased . Progeny were counted after reaching at least the L4 stage . The variability in cross progeny number observed in these experiments is typical of this assay and is generally attributed to variation in mating , sperm transfer , and/or sperm loss ( Murray et al . , 2011 ) . To test short-term male precedence , L4 males and spe-8 ( hc53 ) ; dpy-4 or dpy-4 L4 hermaphrodites were placed together in a 1:1 ratio onto plates with freshly seeded lawns . After 40 hr , both parents were removed . Upon reaching adulthood , offspring were scored as either Dumpy ( self ) or non-Dumpy ( cross ) progeny and counted . To test long-term male precedence , animals were allowed to mate for 16 hr , hermaphrodites were transferred to fresh plates every 12 hr , and self and cross progeny were scored as described above . To test the effect of hermaphrodite age on male precedence , 12 , 24 , 36 , or 48 hr post-L4 hermaphrodites were crossed to 24 hr post-L4 males for 24 hr , both parents were removed , and the number of self and cross progeny were scored as described above . To estimate the number of self sperm remaining in the hermaphrodite reproductive tract at each time point , the number of progeny from unmated hermaphrodites picked in parallel was counted . Male–male competition assays were performed by placing 24 hr post-L4 adult males ( ‘first’ males ) with 24 hr post-L4 adult fog-2 ( q71 ) hermaphrodites for 3 hr in an 8:6 ratio of males to hermaphrodites . The hermaphrodites were allowed to recover for 1 hr , and those lacking visible embryos in their uteri were removed from the plate . The ‘second’ males , 28 hr post-L4 , were then placed with the hermaphrodites and allowed to mate for 3 hr . Individual hermaphrodites were then moved to fresh plates , allowed to lay eggs for 16 hr , then transferred . To distinguish progeny of first and second mates from one another , second-male strains harbored an integrated GFP transgene , mIs11; to control for possible marker-specific effects , experiments were repeated with the mIs11 strains as first males . Progeny generated 0–16 hr after mating were quantified . Subsequent progeny were scored for GFP , and only plates that contained both GFP-positive and GFP-negative offspring were included in analyses . Male mating efficiency was assessed based on two of the assays described above . The frequency of observing successful sperm transfer and/or offspring production in crosses to spe-8; dpy-4 hermaphrodites was 74–88% for wild type and 48–61% for comp-1 . The frequency of offspring production in crosses to fog-2 females was 85–100% for wild type , 60–85% for comp-1 ( me69 ) , and 55–90% for comp-1 ( gk1149 ) . For all experiments involving measurements of progeny numbers , wild-type and mutant animals were tested in parallel to control for variations in temperature and/or media quality that can affect mating and fertility . Each experiment was repeated 2–4 times , and figures show representative results . To release spermatids , adults were dissected in a drop of sperm medium ( SM; 50 mM HEPES pH7 . 8 , 50 mM NaCl , 25 mM KCl , 1 mM MgSO4 , 5 mM CaCl2 , and 10 mM dextrose ) . Virgin 48 hr post-L4 males grown at 20°C were used . Where necessary , spermatids were incubated in SM containing 60 mM TEA or 200 µg/ml Pronase to induce activation into motile sperm ( Shakes and Ward , 1989 ) . Antibody staining followed a protocol similar to that in Wu et al . ( 2011 ) . Briefly , an equal volume of 4% paraformaldehyde in SM was added to the dissected cells . The slides were then incubated in a humid chamber for 5 min , freeze-cracked on a metal block placed in liquid nitrogen , incubated in 95% ethanol for 1 min , and washed with PBST ( phosphate-buffered saline pH 7 . 2 , 0 . 5% Triton X-100 , 1 mM EDTA ) . Antibody incubations were performed for 16 hr at 4°C with rabbit anti-GSP-3/4 ( rb1496 , 1:500 ) ( Wu et al . , 2011 ) and 1 µg/mL DAPI ( 4’ , 6-diamidino-2-phenylindole ) and for 2 hr at 4°C in goat anti-rabbit AlexaFluor 488- or AlexaFluor 568-labeled IgG ( Life Technologies ) at 1:1000; antibodies were diluted and washes were performed in PBST with 1% BSA ( bovine serum albumin ) . Slides were mounted with VectaShield ( Vector Laboratories , Burlingame , CA ) . Confocal images were acquired using an Olympus FV1000 confocal microscope . To analyze localization of male sperm up to 2 . 5 hr after transfer , Mitotracker Red CMXRos ( Life Technologies ) was used to label male sperm as in Stanfield and Villeneuve , 2006 . To analyze sperm localization more than 2 . 5 hr after transfer , virgin 24 hr post-L4 males carrying the Pcomp-1::GFP::H2B transcriptional reporter were mated for 45 min to 24 hr post-L4 N2 hermaphrodites anesthetized in 0 . 1% tricaine and 0 . 01% tetramisole ( McCarter et al . , 1997 ) , and males were then removed . At 12 hr or 24 hr post-mating , images of each recipient were captured in multiple focal planes to capture an entire gonad arm . Analysis of sperm position was performed as in Edmonds et al . ( 2010 ) . Depending on the experiment , either all GFP-positive male sperm in a gonad arm were counted , or those in the focal plane that had the most sperm in the spermatheca were counted . To analyze localization of self sperm , 24 hr post-L4 hermaphrodites were fixed with Carnoy's fixative ( Ellis and Horvitz , 1986 ) and stained with DAPI at 1 µg/ml in M9 . Image collection and data analysis were performed as for male sperm . To measure in vivo velocity , images of migrating cells were collected as in Kubagawa et al . ( 2006 ) using an AxioImager M1 microscope , Axiocam camera , and Axiovision software ( Zeiss , Germany ) . Cells within Zone 2 that moved for at least four consecutive frames were analyzed using the plugins Manual Tracking and Chemotaxis and Migration Tool ( Ibidi , Germany ) in ImageJ ( Schneider et al . , 2012 ) . Sperm were activated in vitro as described previously and DIC ( differential interference contrast ) images were captured every 60 s for 30 min ( Shakes and Ward , 1989; Fenker et al . , 2014 ) . To quantify activation , sperm were scored for the presence of either spikes or a pseudopod at 30 min after adding Pronase . Non-activated sperm from control slides lacking activator were used to measure spermatid size . Aspect ratio was measured by dividing the total length of the pseudopod and cell body by the width of the cell body . The center of the cell body was determined by fitting a circle or ellipse around the cell body and finding the center of that object . The length was then determined by drawing a line from the tip of the pseudopod to an edge of the cell body , with the line dissecting the center of the cell body , and the width of the cell was measured by drawing a line perpendicular to the length and dissecting the center of the cell body . Velocity was measured in TEA-activated sperm that moved for at least 3 consecutive frames . Measurements were obtained using ImageJ ( Schneider et al . , 2012 ) . | The ornate feathers of a peacock and the antlers of a stag are both traits that have evolved because they help a male to outcompete his rivals and mate with more females . Similarly , in species where each female mates with multiple males , a male can improve his reproductive success if his sperm outcompetes the other males' sperm and fertilizes the female's eggs . One way that males try to gain a competitive edge is by producing large quantities of sperm . However , it is also possible that males could compete by generating higher-quality sperm , for example , cells that are better at migrating through the reproductive tract . A microscopic worm called Caenorhabditis elegans is often used to investigate sperm competition . Each worm is either a hermaphrodite or a male; and hermaphrodites store their sperm within a storage structure and use it later to fertilize their own eggs . However , if a hermaphrodite mates with a male , the male's sperm displaces the hermaphrodite's stored sperm and fertilizes the eggs instead . The male's sperm cells were thought to be more competitive because they are larger and faster than the hermaphrodite's sperm , but recent findings suggest a more complex scenario . Hansen et al . identified a genetic mutation that causes male C . elegans sperm cells to lose their competitive advantage . Male worms with a mutation in a gene called comp-1 produce sperm cells that are normal in size , but that cannot outcompete sperm from a non-mutant hermaphrodite . Although sperm cells from a comp-1 mutant male can migrate through the reproductive tract and fertilize eggs when other sperm are not present , in competitive situations the mutant sperm cells have difficulties migrating and are often absent from the hermaphrodite's sperm storage structure . Thus , they no longer come into contact with the eggs that they seek to fertilize . Sperm cells from hermaphrodite comp-1 mutants have similar defects; but when a comp-1 mutant male mates with a comp-1 mutant hermaphrodite , the mutant male sperm regains its fertilization advantage . The identification of the comp-1 gene provides a preview into a complex network of environmental cues and genetically encoded traits that influence which sperm cells are most likely to fertilize an egg cell , and thus live on in the next generation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | COMP-1 promotes competitive advantage of nematode sperm |
Animals move by adaptively coordinating the sequential activation of muscles . The circuit mechanisms underlying coordinated locomotion are poorly understood . Here , we report on a novel circuit for the propagation of waves of muscle contraction , using the peristaltic locomotion of Drosophila larvae as a model system . We found an intersegmental chain of synaptically connected neurons , alternating excitatory and inhibitory , necessary for wave propagation and active in phase with the wave . The excitatory neurons ( A27h ) are premotor and necessary only for forward locomotion , and are modulated by stretch receptors and descending inputs . The inhibitory neurons ( GDL ) are necessary for both forward and backward locomotion , suggestive of different yet coupled central pattern generators , and its inhibition is necessary for wave propagation . The circuit structure and functional imaging indicated that the commands to contract one segment promote the relaxation of the next segment , revealing a mechanism for wave propagation in peristaltic locomotion .
Animal locomotion is generated by coordinated activation of muscles throughout the body ( Grillner , 2003; Marder and Calabrese , 1996; Mulloney et al . , 1998 ) . For example , during axial locomotion such as lamprey swimming and Drosophila larval crawling , muscles present in each segment are sequentially activated along the body axis in a stereotypic temporal and spatial pattern ( Grillner , 2003 ) . How neural networks , including those underlying central pattern generators ( CPGs ) and sensory feedback circuits , orchestrate the precisely timed activation of motor and premotor neurons in multiple body segments remains poorly understood . Previous studies have identified functional connectivity among neurons that are important for rhythmic movements and intersegmental coordination , using electrophysiology in leech ( Kristan et al . , 2005 ) , lamprey ( Grillner , 2003 ) and crayfish ( Smarandache-Wellmann and Gratsch , 2014; Smarandache-Wellmann et al . , 2014; Smarandache et al . , 2009 ) among others . Recent studies in mouse ( Goetz et al . , 2015; Talpalar et al . , 2013 ) , zebrafish ( Kimura et al . , 2013 ) and worm ( Wen et al . , 2012 ) revealed the roles played by different classes of interneurons in the regulation of motor coordination . A complete wiring diagram with synaptic resolution of motor circuits spanning the entire nervous system would contextualize current knowledge and facilitate advancing our understanding of motor pattern generation . Larval Drosophila has recently emerged as a powerful model system for studying the neural regulation of locomotion ( Heckscher et al . , 2012; Kohsaka et al . , 2014; Landgraf et al . , 1997 ) . Its primary locomotor pattern consists of wave-like muscular contractions that propagate either from posterior to anterior segments ( forward movement ) or from anterior to posterior ( backward movement ) segments ( Heckscher et al . , 2012 ) . This sequential activation of segmental musculature is generated by segmentally interconnected circuits in the ventral nerve cord ( VNC ) . The basic pattern of motor activity can be observed as fictive locomotion in dissected larvae or in isolated nerve cords , to which localized optogenetic manipulation can be applied ( Fox et al . , 2006; Kohsaka et al . , 2014; Pulver et al . , 2015 ) . Furthermore , the larva also is capable of a variety of other locomotive patterns and can adjust to changes in environmental conditions ( Godoy-Herrera , 1994; Hwang et al . , 2007; Ohyama et al . , 2015; Vogelstein et al . , 2014 ) . Powerful genetic tools , including a resource of GAL4 drivers ( Pfeiffer et al . , 2010 ) , allow for the manipulation of the activity of uniquely identified neurons in this simple nervous system ( Li et al . , 2014; Manning et al . , 2012 ) . These genetic tools enable optogenetic manipulation and the monitoring of neural activity in larvae in the context of mapped circuitry thanks to novel circuit mapping tools ( Saalfeld et al . , 2009 ) and an electron microscopy volume of the complete central nervous system of the larva ( Ohyama et al . , 2015 ) . Here , we report a novel circuit and mechanism for mediating wave propagation in peristaltic locomotion . We screened GAL4 driver lines and identified neurons that are active with the peristaltic wave of larval muscle contraction . We then mapped the circuits with synaptic resolution in which these neurons are embedded , and we found a repeating modular circuit formed by an inhibitory ( GDL ) and an excitatory neuron ( A27h ) in each hemisegment , connected in a chain across consecutive segments . Using optogenetics and functional imaging , we determined that the inhibitory neuron GDL is necessary for both forward and backward locomotion , but the excitatory neuron A27h is necessary only for forward locomotion , suggesting underlying coupled circuits . Body-wide activation of GDL led to the paralysis of the abdominal segments , but its localized activation in a few consecutive segments was sufficient to arrest the wave of propagation . Taken together , our findings define a mechanism for wave propagation in which the contraction of one segment is concomitant with the relaxation of the adjacent anterior segment , and the cessation of contraction is in turn coupled with the stimulation of contraction in next anterior segment . The logic of this network allows for additional models for coordinated muscle contraction that incorporate feedback from stretch receptors and also descending neurons from the subesophageal zone ( SEZ ) .
To identify interneurons that are involved in the regulation of larval locomotion , we screened for interneurons that exhibit an activity pattern correlated with larval locomotion . In previous studies , we reported on two classes of putative premotor interneurons ( PMSIs and GVLIs ) that inhibit motor neurons via glutamatergic transmission ( Itakura et al . , 2015; Kohsaka et al . , 2014 ) . In this study , we selected for GABA-positive and rhythmically active neurons within sparsely expressing GAL4 lines and identified a class of interneurons , which we named GABAergic dorsolateral neurons ( GDLs , also annotated as A27j2 ) . GDLs are a pair of neurons present in each abdominal segment , and were identified in 9-20-GAL4 ( Hughes and Thomas , 2007 ) . This line drives expression not only in GDLs but also in a subset of mechanoreceptors ( the chordotonals ) and a small number of cells in the brain and SEZ ( Figure 1 and Figure 1—figure supplement 1A–C ) . Since expression in the mechanoreceptors would complicate anatomical and functional analyses of GDLs , we used the promoter of the inactive ( iav ) gene , which is known to be specifically expressed in the mechanoreceptors ( Kwon et al . , 2010 ) , to generate iav-GAL80 ( see Materials and methods ) . When combined with 9-20-GAL4 , iav-GAL80 suppressed the GAL4-driven expression in the mechanoreceptors without affecting the expression in GDLs ( Figure 1—figure supplement 1D , E ) . We used the combined line ( 9-20-GAL4 , iav-GAL80; hereafter referred to as GDL-GAL4 ) as a driver for GDLs in the following experiments . 10 . 7554/eLife . 13253 . 003Figure 1 . Morphology of GDLs . The GDL-GAL4 ( 9-20-GAL4 , iav-GAL80 ) drives expression in GDLs and a small number of cells in the brain and SEZ . All panels show dissected third instar larval CNS . ( A–C ) Morphology of GDLs was visualized with 10xUAS-IVS-myr::GFP reporter expressed by GDL-GAL4 . Anti-GFP ( green ) and anti-FasII ( magenta ) antibodies were used . ( A ) A low magnification view showing GDL-GAL4 expression in a GDL ( arrow ) and in a small number of cells in the brain and SEZ ( arrowheads ) . ( B ) A cross sectional view of an abdominal segment . White arrow denotes the cell body of a GDL in a dorsolateral area of the VNC . Yellow arrow denotes the presynaptic terminals of a GDL . ( C ) A dorsal view showing segmentally repeated GDLs in the VNC . Each GDL extends its neurites locally within the segment . Anterior is to the left and posterior is to the right . ( D ) An image of a fluorescently labelled single-cell clone of GDL ( courtesy of James W . Truman , HHMI Janelia Research Campus ) . GDL is also annotated as A27j2 . ( E ) UAS-syt::GFP was used as a reporter to visualize presynaptic terminals of GDLs ( yellow arrows ) . Signals seen in a medial region ( arrowhead ) are likely presynaptic terminals of descending neurons from the brain or SEZ ( Figure 7—figure supplement 1B ) . ( F ) The cell body of GDLs was positive for GABA . ( G , H ) Schematic drawings of GDLs . Scale bar represents 80 μm in ( A ) , 30 μm in ( C , E ) , 20 μm in ( B ) , 10 μm in ( D ) and 5 μm in ( F ) . ( See also Figure 1—figure supplement 1 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 00310 . 7554/eLife . 13253 . 004Figure 1—figure supplement 1 . Expression driven by 9-20-GAL4 and inactive ( iav ) -GAL80 . ( A , B ) Expression pattern of the driver line ( 9-20-GAL4 ) was assessed with UAS-mCD8::GFP . ( A ) Expression was seen in the axonal projections of the chordotonal neurons ( white arrow ) and the soma of a GDL ( yellow arrow ) . ( B ) Expression in the cell body of chordotonal neurons on the body wall . The sensory cilia in chordotonal neurons were observed ( white arrow ) . ( C ) Double-staining for GFP and GABA showing that GDLs were GABAergic ( yellow arrows ) . ( D , E ) Expression of 9-20-GAL4 with iav-GAL80 . Flies were raised at 25°C ( D ) and 29°C ( E ) , respectively . No expression was seen in chordotonal neurons ( dashed arrows ) whereas expression in GDLs was retained ( yellow arrows ) . Scale bar represents 50 μm in ( B ) , 30 μm in ( A , D , E ) and 10 μm in ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 004 We studied the morphology of GDLs with GDL-GAL4 driving the expression of myr-GFP ( Pfeiffer et al . , 2010 ) . GDLs project their axons to a lateral area of the neuropile under the DL tract ( FasII landmark system; [Landgraf et al . , 2003] ) and present their dendrites in the motor domain ( Figure 1A–C ) . Clonal analyses further confirmed the projection pattern ( Figure 1D ) . We labeled the axon terminals with the presynaptic marker syt::GFP ( Figure 1E ) and also determined GDLs as GABAergic with antibody staining ( Figure 1F ) . To summarize , GDLs are segmental pairs of GABAergic interneurons local to each segment in the abdominal VNC ( Figure 1G , H ) . The isolated central nervous system ( CNS ) presents fictive , rhythmic motor patterns , which facilitates experimentation ( Fox et al . , 2006; Pulver et al . , 2015 ) . We monitored the activity of GDLs during fictive motor patterns of the dissected CNS by the targeted expression of GCaMP6m ( Chen et al . , 2013 ) . We observed bilaterally symmetric propagation of calcium signals that travel along the segments both in forward and backward directions ( Pulver et al . , 2015 ) ( Figure 2A and Video 1 ) . We validated these observations with GCaMP6m imaging in a semi-intact preparation where we observed that GDLs are active simultaneously with muscle contractions in the adjacent segments ( Figure 2—figure supplement 1 and Video 1 ) . 10 . 7554/eLife . 13253 . 005Figure 2 . Wave-like activities of GDLs and their phase difference to motor neurons . ( A ) High-resolution calcium imaging of GDL activity in an isolated CNS preparation ( GDL-GAL4>20xUAS-IVS-GCaMP6m ) . The increase in the calcium signal in the presynaptic terminals of GDLs propagated from posterior to anterior segments ( arrowheads ) . ( B ) ( B1 ) Regions of interest ( ROI ) used for the simultaneous calcium imaging of GDLs and aCC motor neurons . We compared the activities between the cell bodies of aCC motor neurons and the dendrites of GDLs ( GDL-GAL4 , eve-GAL4>20xUAS-IVS-GCaMP6m ) . ( B2 ) Dendrites of GDLs ( arrows ) can be clearly distinguished from the neurites and cell bodies of aCC motor neurons ( GDL-GAL4 , eve-GAL4>10xUAS-IVS-myr::GFP ) . ( B3 ) Temporal correlation between the activity of GDLs and aCC motor neurons . Note that activation of GDLs ( green ) occurs at a similar timing as that of aCC motor neurons in the next posterior segment ( arrows , n = 10 ) . Scale bar represents 15 μm in ( A , B ) . ( See also Figure 2—figure supplement 1 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 00510 . 7554/eLife . 13253 . 006Figure 2—figure supplement 1 . Simultaneous imaging of GDLs activity and peristaltic waves . ( A ) GCaMP-based calcium imaging of GDLs during forward movements ( GDL-GAL4>20xUAS-IVS-GCaMP6m ) . ( A1 , A2 ) Representative fluorescence change of GCaMP6m in GDLs in a posterior ( ROI1 , blue ) and an anterior ( ROI2 , red ) region of the VNC and autofluorescence changes of muscular contractions in a posterior ( ROI3 , orange ) and an anterior ( ROI4 , green ) region of the larva are plotted . The signals of GDLs and segmental muscle contraction propagate at a similar timing . Since the waves of muscle contractions rippled the VNC up and down , the adverse signals can be detected at the posterior VNC ( A1 , arrows ) . The amplitude of calcium signals was smoothed ( moving average of 30 points ) . Scale bar represents 250 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 00610 . 7554/eLife . 13253 . 007Video 1 . Calcium imaging of GDLs . GCaMP6m was expressed in GDLs ( GDL-GAL4>20xUAS-GCaMP6m ) . An isolated CNS preparation or semi-intact preparation from third instar larva . Double-speed or Quad-speed . ( Related to Figure 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 007 To further examine the coordinated activity of GDLs and muscles , we imaged the activity of GDLs and anterior corner cell ( aCC ) motor neurons ( with eve-GAL4 , [Fujioka et al . , 2003] ) . We performed calcium imaging focusing on the dorsomedial region of the VNC where the dendrite of GDLs and the cell bodies of aCCs can be uniquely identified in the same focal plane ( Figure 2B1 , 2 ) . We found GDLs in each segment were activated earlier than aCCs in the same segment and at a similar time as aCCs in the next posterior segment ( Figure 2B3 and Video 2 ) . Thus , the activity of GDLs propagates along the segments ahead of the wave of motor neuron activity during forward locomotion . This suggests a role for GDLs in relaxing and resetting anterior segments prior to the arrival of the contraction wave . 10 . 7554/eLife . 13253 . 008Video 2 . Simultaneous calcium imaging of GDLs and aCC motor neurons . GCaMP6m was expressed in GDLs and subsets of motor neurons ( GDL-GAL4 , eve-GAL4>20xUAS-GCaMP6m ) . An isolated CNS preparation from third instar larva . Quad-speed . ( Related to Figure 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 008 To address the role of GDLs in larval locomotion , we first disrupted synaptic transmission in GDLs with GDL-GAL4 driving the expression of tetanus toxin light chain ( TNT ) ( Sweeney et al . , 1995 ) . We observed a significant decrease in the speed of larval locomotion ( ~30% slower than control , p<0 . 001; Figure 3A , B and Video 3 ) . We also found a significant increase in the wave duration ( ~40% longer than control , p<0 . 001; Figure 3C ) and a decrease in the number of forward peristaltic waves ( ~1/5th the normal frequency , Figure 3D ) . The GDL-GAL4 is also expressed in a small subset of cells in the brain and SEZ . To test whether inhibition of GDLs alone is responsible for the observed TNT phenotype , we suppressed GAL4 activity in the VNC with tsh-GAL80 ( Clyne and Miesenbock , 2008 ) . We analyzed the resulting expression pattern by combining GDL-GAL4 and tsh-GAL80 and did not observe GAL4 activity in GDLs , however expression in the brain and SEZ remained intact ( Figure 7—figure supplement 1A , B ) . The TNT phenotype was rescued with tsh-GAL80 , indicating that GDLs , the only GDL-GAL4–expressing neurons in the VNC , were solely responsible for the phenotype ( p<0 . 001; Figure 3C ) . As further proof , we specifically disrupted synaptic transmission in GDLs by disrupting GABA synthesis with RNAi directed against the Glutamic acid decarboxylase 1 ( Gad1 ) gene , since other neurons in the GDL-GAL4 pattern are not GABAergic ( data not shown ) . RNAi knock-down of Gad1 using two independent constructs that target different portions of the Gad1 mRNA resulted in a similar increase in wave duration as we observed for TNT-expressing larvae ( ~40% longer than control , p<0 . 001; Figure 3E , F and Video 3 ) . These results show that the activity of GDLs is required for larvae to crawl at a normal speed and for normal muscle contraction wave frequency . Therefore , GABAergic transmission is critical for the function of GDLs and larval locomotion . 10 . 7554/eLife . 13253 . 009Figure 3 . Inhibition of GDLs transmission reduced the speed and frequency of larval peristalsis . ( A ) The path taken by third instar larvae undergoing locomotion for 3 min is shown ( left: w1118 , right: GDL-GAL4>UAS-TNT ) . ( B ) Inhibition of GDLs with TNT decreased the speed of larval locomotion ( Locomotion speed , 0 . 69 ± 0 . 03 mm/sec [GDL-GAL4>UAS-TNT] compared to 1 . 00 ± 0 . 04 mm/sec [w1118] , 0 . 97 ± 0 . 03 mm/sec [iav-GAL80>UAS-TNT] and 1 . 06 ± 0 . 02 mm/sec [GDL-GAL4>UAS-IMP ( imperfect ) TNT]; p<0 . 001 ) . ( C ) Expression of TNT in GDL-GAL4 greatly increased the wave duration and the phenotype was rescued by tsh-GAL80 ( Wave duration , 1 . 40 ± 0 . 23 sec [GDL-GAL4>UAS-TNT] compared to 0 . 95 ± 0 . 08 sec [w1118] , 0 . 84 ± 0 . 12 sec [iav-GAL80>UAS-TNT] , 0 . 80 ± 0 . 07 sec [GDL-GAL4>UAS-IMP ( imperfect ) TNT] and 0 . 90 ± 0 . 14 sec [GDL-GAL4>tsh-GAL80 , UAS-TNT]; p<0 . 001 ) . ( D ) TNT-mediated inhibition also caused a significant decrease in the frequency of larval locomotion ( Number of forward waves , 13 . 0 waves/min [GDL-GAL4>UAS-TNT] compared to 46 . 0 waves/min [w1118] , 57 . 8 waves/min [iav-GAL80>UAS-TNT] , 59 . 1 waves/min [GDL-GAL4>UAS-IMP ( imperfect ) TNT] and 45 . 3 waves/min [GDL-GAL4 , tsh-GAL80>UAS-TNT] ) . ( E ) Expression of two independent Gad1-RNAi transgenes in GDLs also increased the wave duration ( Wave duration , 1 . 27 ± 0 . 1 sec [GDL-GAL4>Gad1-RNAi ( VALIUM10 ) , Dicer-2] compared to 0 . 84 ± 0 . 08 sec [GDL-GAL4>w1118 , Dicer-2] and 0 . 98 ± 0 . 09 sec [iav-GAL80>Gad1-RNAi ( VALIUM10 ) , Dicer-2] , 1 . 41 ± 0 . 05 sec [GDL-GAL4>Gad1-RNAi ( VALIUM20 ) ] compared to 0 . 98 ± 0 . 03 sec [GDL-GAL4>w1118] and 0 . 96 ± 0 . 02 sec [iav-GAL80>Gad1-RNAi ( VALIUM20 ) ]; p<0 . 001 ) . ( F ) GDL-GAL4;10xUAS-IVS-myr::GFP driving Gad1-RNAi ( VALIUM20 ) showed a significant reduction of GABA immunoreactivity of GDLs . Box plots in ( C and E ) indicate the median value ( horizontal line inside the box ) , 25–75% quartiles ( box ) , and the data range ( whiskers ) . Statistical significance was determined by Student t-test or one-way ANOVA followed by Tukey’s test for multiple comparisons ( ***p<0 . 001 ) . For all conditions in each figure , n = 20 in ( B ) and n = 10 in ( C , D , E ) . Scale bar represents 15 mm in ( A ) and 5 μm in ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 00910 . 7554/eLife . 13253 . 010Video 3 . Slow and uncoordinated locomotion in the third instar larvae expressing TNT or Gad1-RNAi in GDLs ( GDL-GAL4>UAS-TNT , GDL-GAL4>Gad1-RNAi ) . ( Related to Figure 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 010 Having identified GDLs as necessary for propagating peristaltic waves , we then studied the neural circuit basis for GDL function . First , we determined that GDLs do not synapse directly onto motor neurons by using GRASP ( Feinberg et al . , 2008; Gordon and Scott , 2009 ) , expressing each half of the GFP protein in GDLs and motor neurons , respectively ( Figure 4—figure supplement 1 ) . To confirm this , we then identified GDLs in an electron microscopy ( EM ) volume comprising the entire larval CNS ( Figure 4A ) and reconstructed all neurons synaptically connected to GDLs in segment A1 , none of which were motor neurons ( Figure 4—figure supplement 2 , 3 ) . We also found that no strongly connected GDL partners synapse with each other , suggesting that GDLs act as hub neurons , with the potential to orchestrate activity patterns of postsynaptic neurons ( Figure 4B ) . One of the top synaptic GDL partner cell types ( by number of synapses ) , connected both presynaptically ( “upstream” ) and postsynaptically ( “downstream” ) , is the segmentally repeated premotor interneuron A27h ( Figure 4C , D and Figure 5—figure supplement 1A ) . Interestingly , though all the downstream premotor interneurons were found in the same segment as GDLs , all the upstream premotor interneurons were located in the next posterior segment ( Figure 4D ) . Furthermore , GDLs receive the inputs from somatosensory neurons ( vdaA and vdaC class II dendritic arborization neurons; Figure 4D ) that likely mediate gentle touch ( Tsubouchi et al . , 2012 ) . Taken together , this arrangement configures a feed-forward circuit in which premotor interneurons of one segment not only drive motor neurons in the same segment but also transmit an inhibitory signal to their own homologs in the adjacent anterior segment via GDLs ( Figure 4E ) , in parallel with a synaptic pathway for sensory feedback that also regulates transmission of the peristaltic wave ( see Discussion ) . 10 . 7554/eLife . 13253 . 011Figure 4 . Circuit diagram around GDLs . ( A ) Comparing the confocal images ( left ) and EM reconstruction ( right ) of a GDL ( top: cross-sectional view , bottom: dorsal view ) . Postsynaptic sites ( cyan ) and presynaptic sites ( red ) are shown in the EM images ( right ) . Scale bar represents 20 μm ( upper left ) , 10 μm ( bottom ) . ( B ) The EM circuit graph of GDLs and their postsynaptic neurons . Hexagonal shape denotes a group of left-right homolog neurons . Connections with less than 6 synapses are not included ( green: premotor neurons , yellow: others ) . ( C ) Major postsynaptic ( “downstream” ) targets of a GDL in the abdominal segment A1 . A27h is the strongest synaptic partner of a GDL . Numbers on the directed arrows indicate the number of synapse . ( D ) Major presynaptic ( “upstream” ) targets of a GDL include two dendritic arborization ( da ) sensory neurons ( blue ) . All other presynaptic targets identified are premotor neurons in the posterior segments ( green ) . ( E ) A circuit model around GDLs . From the wiring diagram , a GDL has connections with several premotor neurons at both upstream and downstream . The symbols: NMJ ( arrowheads ) , putative excitatory synapse ( circles ) , and inhibitory synapse ( bars ) . Thickness corresponds synaptic strength . ( See also Figure 4—figure supplement 1–3 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01110 . 7554/eLife . 13253 . 012Figure 4—figure supplement 1 . No GRASP signal was detected with motor neurons ( Related to Figure 4 ) . ( A–D ) GRASP experiments . ( A , B ) Expression pattern of the driver lines 9-20-GAL4 and OK6-LexA were assessed with 10xUAS-IVS-mCD8::RFP , 13xLexAop2-mCD8::GFP and immunostaining with anti-GFP ( green ) , anti-DsRed ( yellow ) , anti-FasII ( magenta ) antibodies . Arrows denote GDLs in ( B ) . ( C ) Results of GRASP ( 9-20-GAL4 , OK6-LexA>OK6-LexA , LexAop-CD4::spGFP11;UAS-CD4::spGFP1-10 ) visualized by immunolabeling with a monoclonal antibody against GFP ( Sigma ) . No GRASP signal was detected . Dashed box indicates the positions of the presynaptic sites of GDLs . ( D ) Syt::HA was coexpressed with GRASP reporters to assess the precise location of GDL presynaptic terminals ( UAS-syt::HA;9-20-GAL4 , OK6-LexA>OK6-LexA , LexAop-CD4::spGFP11;UAS-CD4::spGFP1-10 ) . However , no GRASP signals were observed at the terminals ( arrows ) . Scale bar represents 30 μm in ( A–C ) and 20 μm in ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01210 . 7554/eLife . 13253 . 013Figure 4—figure supplement 2 . Morphology of the major presynaptic and postsynaptic neurons of GDL . ( A ) An aCC motor neuron in A1 segment is shown in the comprehensive EM dataset . We use the position of the aCC for reference in ( B ) and ( C ) . ( B ) Seven major presynaptic targets of a GDL ( A1: vdaA , vdaC , A2: A02j , A03a4 , A27h , A27k , A3: A02j ) . ( C ) Seven major postsynaptic targets of a GDL ( A02d , A05k , A08e3 , A27a , A27h , A31d , A31x: all of the neurons exist at A1 segment ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01310 . 7554/eLife . 13253 . 014Figure 4—figure supplement 3 . Adjacency matrix for GDL circuits . Each row and column represents a neuron's’ pre- and post- synaptic contacts , respectively . The number in the matrix is the synapse number between the target neurons and their partners . Neurons are grouped by segments . GDLs are annotated as “A27j2_a1” in this figure . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 014 The A27h neuron , which is the strongest GDL synaptic partner , arborizes in the motor domain , potentially driving motor neurons ( Figure 5—figure supplement 1B , C ) . To determine which motor neurons A27h connects , we reconstructed the postsynaptic partners of A27h in an EM volume of the whole CNS . We found that A27h synapses bilaterally onto two identified motor neurons , aCC and RP5 ( Figure 5—figure supplement 2A ) , which innervate longitudinal muscles via the intersegmental nerve ( ISN ) , and also additional ISN motor neurons . We validated these findings by reconstructing these neurons in an EM volume of a second larva ( Figure 5—figure supplement 2B ) . We tested whether A27h excites motor neurons by doing paired whole-cell recordings . In current clamp , we injected current into A27h to induce action potentials and recorded the membrane potential of an aCC motor neuron within the same segment . We found that the aCC motor neuron was efficiently depolarized in response to action potential generation in A27h ( Figure 5A–D ) . The depolarizing response was with a very short delay ( ≤5 ms ) , consistent with the direct synaptic connection shown by the EM reconstruction ( Figure 5—figure supplement 2 ) . The efficiency with which A27h is capable of driving aCC correlates with the position of A27h presynaptic terminals , near the proximal portion of the aCC axon ( Figure 5—figure supplement 2C ) , which is the presumptive spike initiation zone ( Gunay et al . , 2015 ) . We also recorded the intrinsic activity of A27h and aCC and found that they were synchronized ( Figure 5E , F ) . In order to determine the neurotransmitter used A27h , we first asked whether the expression pattern of R36G02-GAL4 includes A27h neurons by driving the expression of myr-GFP ( Figure 5G ) . We then used a photoactivatable GFP ( Ruta et al . , 2010 ) and identified the A27h axon within the R36G02-GAL4 expression pattern by comparing it to the EM reconstructions ( Figure 5H ) . Then , we labeled the presynaptic sites by driving synaptotagmin-HA and confirmed they were cholinergic using anti-ChAT antibody staining ( Figure 5I , J ) . Acetylcholine is known to excite motor neurons in Drosophila larva ( Baines and Bate , 1998; Rohrbough and Broadie , 2002 ) . 10 . 7554/eLife . 13253 . 015Figure 5 . A27h is an excitatory premotor interneuron . ( A ) An example of a paired recording of an aCC motor neuron ( asterisk ) and a presynaptic A27h ( arrowhead ) dye-filled with Alexa 568 in the intracellular recording solution . Recording electrodes are indicated with chevrons . ( B ) EM reconstructions of aCC ( magenta ) and A27h ( green ) . Input synapses are labeled in cyan , output synapses in red . ( C ) A current command ( 50 pA ) results in A27h firing action potentials ( see zoomed-in view in inset , scale bar indicates 10 ms , 0 . 5 mV ) , which efficiently drives the postsynaptic aCC motor neuron ( magenta trace depicts mean of 100 trials ± SEM ) . ( D ) The maximum voltage response in aCC to presynaptic stimulation . Each point indicates the mean response of 100 trials of current injection in a different cell . ( E ) Endogenous activity patterns of these two cells , with each burst corresponding to a peristaltic wave . ( F ) Phase plot describing the coherency between the two cells , with magnitude of coherence depicted as the distance from the center , and the phase shift as deviation from 0° ( with aCC at 0° ) . Dashed line indicates α = 0 . 05 for coherence magnitude statistically deviating from 0 . ( G ) Expression driven by R36G02-GAL4 . Assessed with the 10xUAS-IVS-myr::GFP reporter and immunostaining with anti-GFP ( green ) and anti-FasII ( magenta ) antibodies . Strong expression was seen in A27h ( arrows ) and a small number of other cells in the VNC . ( H ) Photolabeling of A27h neurons . A flash of near-UV light ( ~405 nm ) was applied to a dorso-lateral region of the VNC dissected from a R36G02-GAL4>UAS-C3PA larva , to label A27h and neighboring cells and their axonal arborization . The cell body of A27h can be uniquely identified for its stereotypic relative position to other cells ( arrows ) ; white arrowheads , axons of A27h; yellow arrowhead , an axon of a different cell . ( I , J ) A27h presynaptic terminals ( arrowheads ) express ChAT . Triple labeling for membrane-GFP ( green ) , presynaptic marker ( red ) and ChAT ( blue ) ( in R36G02-GAL4>UAS-syt::HA;10xUAS-IVS-myr::GFP ) . Dorsal ( I ) and cross-sectional ( J ) view are shown . Scale bar represents 30 μm in ( G ) , 20 μm in ( A , H ) , 10 μm in ( I ) and 5 μm in ( J ) . ( See also Figure 5—figure supplement 1 , 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01510 . 7554/eLife . 13253 . 016Figure 5—figure supplement 1 . The connectivity of premotor neurons ( Related to Figure 5 ) . ( A ) Some presynaptic ( “upstream” ) and postsynaptic ( “downstream” ) neurons are premotor neurons . ( B ) Morphological features of A27h . The cell body of A27h is in the most dorsal part of the VNC and the axon extends to the anterior midline . ( C ) GDL ( orange ) and A27h ( blue ) are connected to each other in a lateral-medial part of the VNC . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01610 . 7554/eLife . 13253 . 017Figure 5—figure supplement 2 . Bilateral A27h connection to motor neurons was confirmed in two independent EM volumes . ( A , B ) The bilateral connectivity between A27h , aCC and RP5 was confirmed in two different EM volumes ( left: a first larva containing whole CNS , right: a second larva containing 1 . 5 abdominal segments ) . ( C ) Both left and right A27h neurons were connected to near-soma positions of the left ( or right ) aCC motor neuron . Arrows denote the synaptic connections between A27h neurons ( green ) and an aCC motor neuron ( magenta ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 017 Taken together , these results suggest that the neuron A27h induces muscle contraction . To test this , we targeted the expression of ChR2 ( T159C ) to A27h using R36G02-GAL4 and applied localized light to two segments in dissected larvae while monitoring muscle contractions along the body wall ( see Materials and methods ) . We found that upon localized stimulation , muscles in the corresponding body wall segments contracted ( Video 4 ) . Although involvement of other neurons included in the R36G02-GAL4 expression pattern cannot formally be excluded as being involved in this light activated muscle contraction response , these results provide strong support for the notion that A27h activates motor neurons and induces muscle contraction . 10 . 7554/eLife . 13253 . 018Video 4 . Localized photoactivation of A27h neurons induced muscle contractions . ChR2-T159C was expressed in A27h neurons ( 36G02-GAL4>UAS-ChR2-T159C ) . A semi-intact larva preparation from third instar larva . Double-speed . ( Related to Figure 5 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 018 The sequential intersegmental connections between the inhibitory GDL and the excitatory A27h neurons ( Figure 4E ) suggest that A27h may be active synchronously with the peristaltic wave of motor neuron activity that propagates locomotion . To test this hypothesis , we monitored the activity of A27h neurons and aCC motor neurons during fictive locomotion ( Figure 6 ) . We observed a wave-like activity that propagates from posterior to the anterior segments ( Figure 6A ) . Interestingly , unlike GDLs that are active during both forward and backward locomotion ( Video 5 ) , A27h was activated only during forward locomotion ( Figure 6B ) . This suggests that though GDL participates in both forward and backward locomotion , the excitatory neuron A27h is specialized in forward locomotion . We postulate a different premotor neuron acts during backward locomotion and we found a possible candidate for which a genetic driver line does not exist ( Figure 8—figure supplement 1A and see Discussion ) . 10 . 7554/eLife . 13253 . 019Figure 6 . A27h participates in forward motor activity . ( A ) Calcium imaging of A27h ( in R36G02-GAL4>20xUAS-IVS-GCaMP6m ) . Arrows denote the cell bodies of A27h neurons and arrowheads axons of A27h neurons . ( B ) Simultaneous imaging of the activity of A27h neurons ( green ) and aCC motor neurons ( magenta ) ( in R36G02-GAL4 , eve-GAL4>20xUAS-IVS-GCaMP6m ) . The top panel shows the region of interests ( ROI ) used for the analyses . ( B1 , 2 ) Dashed arrows denote the directions of motor activity . A27h was activated only during forward movement ( B1 ) but not backward movement ( B2 ) . Scale bar represents 30 μm in ( B ) and 15 μm in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 01910 . 7554/eLife . 13253 . 020Video 5 . Simultaneous calcium imaging of A27h neurons and aCC motor neurons . GCaMP6m was expressed in A27h neurons and subsets of motor neurons ( 36G02-GAL4 , eve-GAL4>20xUAS-GCaMP6m ) . A27h neurons are indicated by arrows . An isolated CNS preparation from third instar larva . Double-speed . ( Related to Figure 6 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 020 The segmentally linked connections between inhibitory GDL neurons and excitatory A27h neurons in the next anterior segment ( Figure 4E ) suggest a mechanistic explanation for wave propagation in peristaltic locomotion . We hypothesize that the commands to contract one segment also promote the relaxation in the next anterior segment , and that contraction termination is coupled with circuit activity that enables contraction of the next segment . To test this hypothesis , we monitored the peristaltic waves following GDL activity perturbation during larval locomotion . First , we observed that coordinated GDL activity is necessary for locomotion . We activated GDLs in all segments simultaneously by driving ChR2 ( T159C ) ( Berndt et al . , 2011 ) with GDL-GAL4 . All individual larvae stopped moving upon presentation of blue light ( 10 out of 10; [Figure 7A and Video 6] ) . Larval abdominal segments were paralyzed but , interestingly , they could still move their thoracic segments , which do not participate in peristaltic wave propagation . To control for a potential startle response to blue light ( Xiang et al . , 2010 ) , we confirmed these findings using thermogenetics and dTRPA1 ( Pulver et al . , 2009 ) . Larvae showed very slow and uncoordinated locomotion at a restrictive temperature at which dTRAPA1 expression is driven ( 32°C; p<0 . 001; Figure 7B–D ) . To determine the nature of this locomotion blockage , we activated all GDLs by ChR2 ( T159C ) in a semi-intact preparation where we could monitor muscle contractions using mhc::GFP ( Hughes and Thomas , 2007 ) . We found that muscles relaxed when all GDLs were active ( Figure 7E ) , contrary to the whole-body contraction ( hunch ) normally observed as part of the startle response elicited by blue light ( Ohyama et al . , 2013; Vogelstein et al . , 2014 ) . To exclude that neurons in the GDL-GAL4 expression pattern other than GDLs played a role in this muscle relaxation , we used tsh-GAL80 to suppress expression in abdominal segments , and this rescued the immobilization phenotype ( Video 6 ) . These results were confirmed using optogenetic CsChrimson-mediated activation of GDLs and a different driver line , R15C11-LexA; this resulted in similar phenotypes ( Figure 7—figure supplement 1 and Video 6 ) . 10 . 7554/eLife . 13253 . 021Figure 7 . Optical perturbation of the activity of GDLs disrupts the peristalsis . ( A ) Behavioral responses induced by optogenetic activation of GDL-GAL4-expressing neurons ( A1 ) A wild-type larva . Illumination with blue light ( ~480 nm ) induced light-avoidance behaviors such as backward movement and head turning . ( A2 ) Channelrhodopsin-2 ( ChR2 ) -mediated activation of GDLs completely immobilized the abdominal segments of the larva ( yellow dashed circles; 10 of 10 larvae [GDL-GAL4>UAS-ChR2-T159C] compared to 0 of 8 cases in the control larvae [w1118>UAS-ChR2-T159C] ) . ( B–D ) Larvae expressing dTRPA1 in GDL-GAL4 showed locomotion defects at a restrictive temperature . Traces ( B ) , locomotion speed ( C ) and wave duration ( D ) at permissive and restrictive temperatures are shown ( C; Locomotion speed , 0 . 85 ± 0 . 05 mm/sec compared to 1 . 17 ± 0 . 05 mm/sec in the control larvae , the larvae with the same genotype at a permissive temperature ( 22°C ) , D; Wave duration , 1 . 42 ± 0 . 43 sec [GDL-GAL4>UAS-dTRPA1] compared to 0 . 79 ± 0 . 06 sec [w1118] and 0 . 74 ± 0 . 05 sec [iav-GAL80>UAS-dTRPA1]; p<0 . 001 . Note that larvae normally crawl faster at 32°C than at 22°C ) . For all conditions in each figure , n = 20 in ( C ) and n = 10 in ( D ) . ( E ) A dissected larva expressing ChR2-T159C in GDL-GAL4 and mhc::GFP in muscles ( GDL-GAL4>mhc::GFP , UAS-ChR2-T159C ) . When blue light was applied during peristalsis , contracted muscles became relaxed ( n = 12 ) . ( F ) ( F1 ) Localized photostimulation was applied to an anterior portion of the VNC ( around A3-A5 , yellow arrow ) during peristalsis . Arrowhead denotes the contracting segments at this moment . ( F2 ) Muscular movement was examined by using the scattered light changes . The light intensity change in muscles in A3-A5 is plotted . In this example , the peristaltic wave halted at A3 ( dashed circle with arrowheads ) . Statistical significance was determined by Student’s t-test or one-way ANOVA followed by Tukey’s test for multiple comparisons ( ***p<0 . 001 ) . Scale bar represents 15 mm in ( C ) , 9 mm in ( A ) , 250 μm in ( F ) and 200 μm in ( B ) . ( See also Figure 7—figure supplement 1 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 02110 . 7554/eLife . 13253 . 022Figure 7—figure supplement 1 . Confirmation of the expression of ChR2 in GDLs . Expression of ChR2 reporters , ChR2 ( T159C ) ::YFP ( A , B , C , E , F ) driven by GDL-GAL4 , and CsChrimson::mVenus driven by R15C11-LexA ( D ) , in GDLs were confirmed . ( A , B ) tsh-GAL80 specifically eliminates GDL-GAL4-mediated expression in the VNC , without affecting the expression in cells in the brain , SEZ and the terminal ( arrowheads ) . ( C–F ) Yellow and white arrows denote presynaptic terminals and cell bodies of GDLs , respectively . ( A–D ) Third instar , ( E , F ) First instar . ( F ) is a high magnification image of ( E ) . Scale bar represents 80 μm in ( A , B ) , 30 μm in ( C , D ) , 20 μm in ( E ) and 10 μm in ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 02210 . 7554/eLife . 13253 . 023Video 6 . Optogenetic activation of GDLs induced locomotion defects . Behavior of first or third instar larvae expressing ChR2-T159C ( GDL-GAL4>UAS-ChR2-T159C ) or CsChrimson ( R15C11-LexA>LexAop2-CsChrimson ) in GDLs , upon light application . ( Related to Figure 8 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 023 Then , we determined that the suppression of GDL activity is indeed necessary for the propagation of the peristaltic wave . In a semi-intact preparation , we restricted blue light illumination to a window comprising two to three consecutive abdominal segments to excite GDLs for a few seconds using ChR2 ( T159C ) . This localized stimulation induced muscles relaxation in the corresponding body-wall segments and the disappearance of peristaltic waves ( 72% , 18/25 trials ) only when the segments were illuminated at the front of the muscle contraction wave ( Figure 7F and Video 7 ) . Furthermore , upon removal of light , the wave sometimes resumed at the illuminated segments ( 16% , 4/25 trials ) ( Video 7 ) . Illuminating segments more anterior to the front of the wave did not prevent the wave from propagating across them , but the wave appeared slower ( 12% , 3/25 trials ) ( Video 7 ) . These results show that local GDL activation in a few segments at the front of the wave is sufficient to arrest the peristaltic wave . 10 . 7554/eLife . 13253 . 024Video 7 . Localized activation of GDLs affected larval peristalsis . ChR2-T159C was expressed in GDLs ( GDL-GAL4>UAS-ChR2-T159C ) . ( I ) Localized photoactivation of GDLs in a portion of VNC during peristalsis halted the peristaltic wave at the corresponding region in the body wall . ( II ) The wave sometimes resumed at the illuminated segments . ( III ) Illuminating segments more anterior to the front of the wave did not prevent the wave from propagating . A semi-intact larva preparation from third instar larva . Double-speed . ( Related to Figure 8 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 024 Taken together , our results support a model of peristaltic wave propagation consisting of co-activation ( e . g . A27h ) of the motor neurons in one segment with the inhibitory neurons ( e . g . GDL ) that suppress activity of the homologous excitatory neurons ( A27h ) in the next segment ( Figure 8 ) . 10 . 7554/eLife . 13253 . 025Figure 8 . Summary of the GDL circuit . The information flow in the GDL-A27h premotor circuit . At a time point during forward peristalsis when A27h in segment N is active and driving motor activity in the segment , GDL in the next anterior segment N-1 is active and inhibits the activity of A27h and the downstream motor activity in the segment . As the motor wave propagates anteriorly and motor activity in segment N declines , so does the GDL in segment N-1 , thus releasing the target A27h from its inhibition ( gray: inactive , other colors: active ) . ( See also Figure 8—figure supplement 1 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 02510 . 7554/eLife . 13253 . 026Figure 8—figure supplement 1 . A proposed circuit mechanism for moderating peristaltic locomotion . ( A , B , D ) EM reconstructions . ( A ) A candidate neuron for backward peristalsis . T01x3 is a homolog of A01x3 in thoracic segment . ( B ) A proprioceptive ( vpda ) and two other sensory neurons ( vdaA and vdaC ) synapse axo-dendritically onto A27h and axo-axonically onto a GDL of their own segment . ( C ) A proposed model of sensory feedback ( per hemisegment ) . Note that the feedback have two simultaneous effects: promote the contraction of its own segment ahead of the wave , and then relaxation ( or stretch ) of the next anterior segment . ( D ) A descending neuron from the SEZ has connections with A27h neurons at each segment . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 02610 . 7554/eLife . 13253 . 027Figure 8—figure supplement 2 . Synaptic relations of GDL and A27h with known larval interneurons . ( A ) One of the PMSI neurons ( glutamatergic neurons involved in the speed regulation; [Kohsaka et al . , 2014] ) , named A02j , relates GDLs to each other across abdominal segments . In particular , A02j synapses onto the GDLs of the two segments anterior to its own segment , potentially starting the excitatory drive over GDLs to promote the relaxation of segments anterior to the muscle contraction wave . Additionally , A02j synapses directly onto some motor neurons of the segments anterior to its own segment ( not shown ) , with potentially an inhibitory effect as shown in ( Kohsaka et al . , 2014 ) . Interestingly , A02j in one segment might promote the disinhibition of its anterior homologs , given than GDL , a GABAergic neuron , synapses onto a segment-local putatively GABAergic neuron ( A31d; similar morphology and belonging to the same lineage as the GABAergic neurons A31b and A31k [Schneider-Mizell et al . , in press] ) . ( B ) The GDL-A27h circuit interacts with neurons known to affect the speed of locomotion ( PMSIs and GVLIs ) . A02d is a PMSI neuron ( Kohsaka et al . , 2014 ) that receives inputs from GDL and in turn synapses onto the A08a neuron ( a GVLI; [Itakura et al . , 2015] ) . A08a , in turn , synapses onto two putatively GABAergic neurons ( A31d and A31x ) . This circuit suggests that GDL prevents A02d from driving A08a , which could potentially underlie the observed activation pattern of A08a , which is active two segments posterior to the forward-moving peristaltic wave ( Itakura et al . , 2015 ) . We did not observe synapses between A08a and motor neurons . Furthermore , GDL might provide inhibition ipsilaterally to the contralaterally projecting , Eve-Skipped+ neuron A08e3 , which is necessary for maintaining bilaterally symmetric muscle contraction amplitude ( Heckscher et al . , 2015 ) . This suggests a role for GDL in the regulation of posture adjustment , and therefore a close relationship between the circuits for wave propagation and the circuits for ensuring symmetrical muscle contractions in forward locomotion . ( C ) The connectivity matrix between proprioceptive sensory neurons and A02d ( one of the PMSIs ) , A08a ( GVLIs ) , and A08e3 ( one of the Even-Skipped+ neurons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13253 . 027
We discovered a circuit whose structure and function provides a mechanism for understanding forward wave propagation in peristaltic locomotion . This circuit consists of a chain of alternating excitatory and inhibitory neurons spanning all abdominal segments . The core elements of the chain include just one excitatory and one inhibitory neuron per hemisegment . We demonstrate here that the inhibitory neuron ( GDL ) is sufficient to halt the peristalsis and to relax muscles in all segments , suggesting it is a point of coordination between forward and backward locomotion . We further demonstrate that the excitatory neuron ( A27h ) is active during forward but not backward peristalsis , suggesting the existence of another excitatory circuit component critical for backward peristalsis among the synaptic partners of the GDL inhibitory neuron . This circuit defines a backbone of repeating , connected , modules for excitation and inhibition similar to those postulated in a computational model for peristalsis ( Gjorgjieva et al . , 2013 ) on the basis of behavioral observations that predicted the existence of central pattern generators ( Suster and Bate , 2002 ) . We found that the excitatory neuron ( A27h ) is premotor , directly synapsing onto motor neurons of its own segment only and that control both dorsal and ventral longitudinal muscles . This suggests an explanation for the observation that in forward crawling , dorsal and ventral longitudinal muscles contract simultaneously ( Heckscher et al . , 2012 ) . In backward peristalsis , however , a phase gap has been observed in the timing of dorsal and ventral muscle contraction ( Heckscher et al . , 2012 ) . This decoupling could require a more complex circuit structure for backward wave propagation , and therefore suggests an explanation for the lack of an equivalent excitatory neuron in the circuit chain for backward peristalsis . We found , however , neurons postsynaptic to the inhibitory neuron ( GDL ) whose anatomy and position in the circuit suggest a role in backward peristalsis ( Figure 8—figure supplement 1A ) . In contrast , the inhibitory neuron ( GDL ) itself does not synapse onto motor neurons , and therefore occupies a higher-order position in the circuit that allows its participation in both forward and backward wave propagation in peristalsis . Furthermore , the GDL axon targets the intermediate lateral neuropil , which is neither in the domain of motor neuron dendrites nor in the somatosensory domain , suggestive of a role higher-order motor coordination . Relevant for forward peristalsis , GDL disinhibits the excitation of its anterior homologs , by removing inhibition from a glutamatergic interneurons ( A02j ) implicated in the regulation of peristaltic speed ( one of the PMSIs; [Kohsaka et al . , 2014] ) . A02j is presynaptic to GDLs in anterior segments ( Figure 4D and Figure 8—figure supplement 2A ) . A model of peristaltic locomotion must consider the coordination of left and right hemisegments ( Gjorgjieva et al . , 2013 ) . Though we found that the chain of alternating inhibitory and excitatory neurons runs independently on the left and right sides of the body , the excitatory neuron ( A27h ) presents a bilateral arbor and drives motor neurons bilaterally . Our wiring diagram best supports a model of left-right coordination where excitatory neurons communicate with each other ( Gjorgjieva et al . , 2013 ) , but with the caveat that this synergy takes place by the simultaneous co-activation of the target motor neurons rather than reciprocal excitation . This model has been shown to support longer contraction episodes at the front of the wave ( Gjorgjieva et al . , 2013 ) , consistent with observations of muscle contraction in peristalsis ( Heckscher et al . , 2012 ) . Independently of the timing , the fine-tuning in the intensity of left-right contractions has been shown to be under control of Even-skipped+ evolutionarily conserved neurons , which integrate both proprioceptive inputs and motor commands ( Heckscher et al . , 2015 ) . The dissected larval CNS undergoes spontaneous waves of motor neuron activation at about 1/10th the normal speed ( Fox et al . , 2006; Pulver et al . , 2015 ) . These waves occur in the absence of sensory feedback , indicating the presence of CPGs and also suggesting a role for sensory feedback in speeding up the peristaltic wave ( Suster and Bate , 2002 ) . The circuit chain of excitatory and inhibitory neurons described here could be a part of the CPG , and we additionally found these neurons are modulated by proprioceptive inputs ( from vpda class I dendritic arborization neuron; Figure 8—figure supplement 1B ) . Given that the vpda is a stretch receptor ( Cheng et al . , 2010; Tamarkin and Levine , 1996 ) , it would be active in the segment ahead of the wave of contraction , which is being stretched by the pull exerted by the contracting segment ( Figure 8—figure supplement 1B ) . Proprioceptive feedback action onto the excitatory neuron of the circuit chain could then have two simultaneous effects: promotion of the contraction in the segment ahead of the wave ( via activation of A27h ) , and relaxation of the segment twice removed ( via activation of GDL , which acts on the segment anterior to it; Figure 8—figure supplement 1C ) . We also found two somatosensory neurons ( vdaA and vdaC ) synapse axo-dendritically onto the premotor excitatory neuron ( A27h ) and axo-axonically onto the inhibitory neuron ( GDL ) in their own segment ( Figure 8—figure supplement 1B ) . Although the function of these two sensory neurons remains unclear , we speculate that this axo-axonic , likely depolarizing , connection onto GDL reduces the membrane action potential of its axon , reducing synaptic release of GABA onto A27h in the same segment ( Burrows and Matheson , 1994 ) . Our model refines a previous model where the proprioceptive feedback was thought to signal the successful contraction of a segment ( Hughes and Thomas , 2007 ) . We suggest that , in addition , at least some of the proprioceptive feedback ( vpda ) facilitates wave propagation and , therefore , may underlie the reduction in speed observed in fictive crawling ( Fox et al . , 2006; Pulver et al . , 2009 ) . In addition to the excitatory premotor interneuron A27h , we found two other interneurons that receive direct synaptic inputs from a GDL ( A02d and A08e3 ) and that , like A27h , also integrate inputs from stretch receptors ( vpda , dbd and vbd; Figure 8—figure supplement 2B , C ) . One interneuron ( A08e3 ) is an Even-Skipped+ neuron that maintains left-right symmetric muscle contraction amplitude ( Heckscher et al . , 2015 ) . The other ( A02d ) is a glutamatergic interneuron that belongs to a lineage of neurons thought to mediate speed of locomotion ( one of the PMSIs; [Kohsaka et al . , 2014] ) . While A02d is a segment-local interneuron , proprioceptive axons span multiple segments ( Merritt and Whitington , 1995; Schneider-Mizell et al . , in press ) , suggesting that a GDL can suppresses the effect of proprioceptive feedback specifically within its own segment without affecting the relay of proprioception to adjacent segments . Furthermore , A02d synapses onto a glutamatergic interneuron ( A08a ) thought to contribute to muscle relaxation in the wake of the peristaltic wave ( Itakura et al . , 2015 ) , which could be mediated via putative GABAergic premotor neurons ( A31d; Figure 8—figure supplement 2B ) . Taken together , we suggest that one of the functions of the inhibitory neuron GDL is to gate proprioceptive feedback within its segment which has implications for the control of both speed and posture ( Heckscher et al . , 2015 ) . Finally , we observed a descending neuron from the SEZ that synapses onto the excitatory neuron ( A27h ) of the circuit chain in all segments ( Figure 8—figure supplement 1D ) . This motif has been observed and modeled in the leech and crayfish , where it enables the modulation of wave propagation speed ( Acevedo et al . , 1994; Cacciatore et al . , 2000; Smarandache et al . , 2009; Stein , 1971; Wiersma and Ikeda , 1964 ) . The brain and SEZ have been deemed non-essential for wave propagation ( Berni et al . , 2012 ) . Speed of wave propagation , therefore , may be controlled in at least two ways: by proprioceptive feedback and by descending inputs . The existence of a circuit chain formed by excitatory and inhibitory neurons might be all that remains when both sensory feedback and the brain are absent , explaining the existence of wave propagation in decerebrated animals ( Berni et al . , 2012 ) , and even for a small set of isolated abdominal segments ( Pulver et al . , 2015 ) .
The following fly strains were used: w1118 ( Bloomington stock number: #6326 ) ( Hoskins et al . , 2001 ) , 9-20-GAL4 ( Hughes and Thomas , 2007 ) , eve ( RRa ) -GAL4 ( Fujioka et al . , 2003 ) , R36G02-GAL4 ( #49939 ) , OK6-LexA ( Kohsaka et al . , 2014 ) , R15C11-LexA ( #52492 ) , UAS-mCD8::GFP ( #5137 ) ( Lee and Luo , 1999 ) , 10xUAS-IVS-mCD8::GFP ( #32185 , #32186 ) ( Pfeiffer et al . , 2010 ) , 10xUAS-IVS-myr::GFP ( #32197 , #32198 ) ( Pfeiffer et al . , 2010 ) , 10xUAS-IVS-mCD8::RFP , 13xLexAop2-mCD8::GFP ( #32229 ) ( Liu et al . , 2012 ) , UAS-CD4::spGFP1-10 ( Gordon and Scott , 2009 ) , LexAop-CD4::spGFP11 ( Gordon and Scott , 2009 ) , UAS-syt::GFP ( Zhang et al . , 2002 ) , UAS-syt::HA ( Robinson et al . , 2002 ) , 20xUAS-IVS-GCaMP6m ( #42748 , #42750 ) ( Chen et al . , 2013 ) , UAS-dTRPA1 ( Pulver et al . , 2009 ) , UAS-TNT ( #28838 ) ( Sweeney et al . , 1995 ) , UAS-IMPTNT ( V1 ) ( #28840 ) ( Sweeney et al . , 1995 ) , 13xLexAop2-IVS-CsChrimson-mVenus ( #55139 ) , UAS-C3PA-GFP ( Ruta et al . , 2010 ) , mhc::GFP/Cyo ( Hughes and Thomas , 2007 ) , tsh-GAL80/Cyo ( Clyne and Miesenbock , 2008 ) , Gad1-RNAi ( #28079 ( VALIUM10 ) , #51794 ( VALIUM20 ) ) and Dicer-2 ( #24650 , #24651 ) . Flies were raised on conventional cornmeal agar medium at 25°C except the following: in order to enhance RNAi potency , the transgenic fly ( UAS-Gad1-RNAi ( VALIUM10 ) ) was combined with Dicer-2 and reared at a higher temperature ( 29°C ) . To generate iav ( inactive ) -GAL80 transgenic line , we excised the GAL80 sequence from pBPGAL80Uw-6 ( Pfeiffer et al . , 2010 ) using BamHI and XbaI and subcloned the DNA between BamHI and StuI ( blunt-ended ) sites of iav-GAL4 ( Kwon et al . , 2010 ) . The resulting construct was used to transform w1118 embryos using standard Drosophila micro-injection techniques ( BestGene Inc ) . To generate UAS-ChR2 ( T159C ) transgenic line , we first introduced KpnI and AgeI sites between the SwaI ( 12079 ) and PmeI ( 12095 ) sites of pJFRC2-INS ( Plasmid #26215 ) . We excised the sequence between the HindIII ( 6488 ) and XbaI ( 8490 ) sites of pJFRC2-INS ( Plasmid #26215 ) and replaced with the sites of pJFRC7-20XUAS-IVS-mCD8::GFP ( Plasmid #26220 , XbaI ( 8740 ) and HindIII ( 6488 ) , 4*5xGAL4_DBD ) . Then , we replaced the sites between XhoI ( 7341 ) and XbaI ( 8740 ) with Drosophila codon-optimized ChR2 ( T159C ) ::YFP synthesized by Biobasic inc . We next excised the sequence between the HindIII and PacI sites of the plasmid and amplified by PCR using primers 5-AgeI ( CATGCGCACCGGTGGCCAGGGCCGCAAG ) and 3-KpnI ( CACTTGGTACCTGGCCATTAATTAAGGCCGGCC ) . The resulting construct was used to transform y[1] w[67c23]; P ( CaryP ) attP40 or attP2 sites as described above . Dissected larvae were fixed in phosphate buffered saline ( PBS , NaCl 137 mM , KCl 2 . 7 mM , Na2HPO4 8 . 1 mM , KH2PO4 1 . 5 mM , pH7 . 3 ) containing 4% paraformaldehyde for 30 min at room temperature . After two 15 min washes with 0 . 2% Triton X-100 in PBS ( PBT ) , the larvae were incubated with 5% normal goat serum in PBT for 30 min . The larvae were then incubated overnight at 4°C with the primary antibody . After two 15 min washes , the larvae were incubated overnight at 4°C with the secondary antibody . Images were acquired using a confocal microscope ( FV1000 , Olympus , Japan ) . Primary antibodies were used at the following dilutions: rabbit anti-GFP ( cat# Af2020 , Frontier Institute; 1:1000 ) , mouse anti-GFP ( cat# G6539 , Sigma; 1:100 ) , guinea pig anti-GFP ( cat# Af1180 , Frontier Institute; 1:1000 ) , rabbit anti-HA ( cat# C29F4 , Cell Signaling Technology; 1:1000 ) , rabbit anti-DsRed ( cat# 632496 , Clontech; 1:1000 ) , mouse anti-FasII ( mAB-1D4 , Hybridoma Bank , University of Iowa; 1:10 ) , rabbit anti-GABA ( A2052 , Sigma; 1:100 ) , mouse anti-ChAT ( mAB-4B1 , Hybridoma Bank , University of Iowa; 1:50 ) . Secondary antibodies were used at the following dilutions: Alexa Fluor 488 or Cy3-conjugated goat anti-rabbit IgG ( A-11034 or A-10520 , Invitrogen Molecular Probes; 1:300 ) , Alexa Fluor 555 or Cy5-conjugated goat anti-mouse IgG ( A-21424 or A-10524 , Invitrogen Molecular Probes; 1:300 ) , and Alexa Fluor 488-conjugated goat anti-guinea pig IgG ( A-11073 , Invitrogen Molecular Probes; 1:300 ) . We conducted two locomotion assays . One is automated tracking of the trajectory of larval behavior and the other is manually measuring the duration of each peristaltic wave . For automated tracking , wandering third instar larvae were picked up and then transferred to an agar plate ( 90 mm in diameter ) for acclimation ( 3 min ) . The larvae were then videotaped using a digital camera ( GE60 , Library , Japan ) and tracked using the open-source ImageJ plugin wrMTrck ( http://www . phage . dk/plugins/wrmtrck . html ) . Each video containing 20 larvae was recorded five times at 30 frames/sec for 3 min . The average speed of larval locomotion was calculated by dividing the total path length of the larvae by time . For manual analysis , wandering third instar larvae were gently washed in deionized water and then placed on an agar plate . After acclimation ( 3 min ) , the movements of the larvae were videotaped under a microscope ( SZX16 , Olympus , Japan ) using an XCD-V60 CCD camera ( 30 frames/sec for 30 s ) and the movies were downloaded into VFS-42 ( Vision Freezer , Chori imaging ) . The wave duration , which is elapsed time between the landing of the posterior end and elongation of the head , was manually measured in the movies using Fiji ( 10 waves per larva ) . The frequency of larval locomotion ( number of forward waves ) was also manually calculated by dividing the total number of forward waves of each larva by the total time . Two types of microscopy were used for the measurement of neural activity , one for low magnification and the other for high-magnification imaging . Low-magnification imaging was performed on semi-intact preparation of wandering third instar larvae , in order to observe both the propagation of muscular contraction and calcium signals in the CNS . The larvae were pinned on a sylgard-coated dish ( Silpot 184 , Dow Corning Toray ) and dissected in an external saline ( NaCl 135 mM , KCl 5 mM , MgCl2 · 6H2O 4 mM , CaCl2 · 2H2O 2 mM , TES 5 mM , Sucrose 36 mM ( pH7 . 1 ) ) ( Marley and Baines , 2011 ) . The internal organs were removed without scratching the ventral nerve cord ( VNC ) and axons . To fix the position of the VNC , a pin was placed between the brain and the mouth hook . Imaging was performed on a fluorescence microscope ( MVX10 , Olympus , Japan ) equipped with a CCD camera ( XCD-V60 , Sony , Japan ) and 1x~4x objective lens . The images were acquired and downloaded into VFS-42 ( Vision Freezer , Chori imaging ) at 30 frames/sec , 640 x 480 pixels . High magnification imaging was performed on isolated CNS preparation . The third instar larvae were dissected in the external saline described above and the peripheral nerves were cut carefully to isolate the CNS . The isolated CNS was adhered to a double-sided tape ( NW-K15 , Nichiban , Japan ) on a clean glass slide in the saline . Imaging was performed on an upright microscope ( Axioskop2 FS , Zeiss , Germany ) equipped with a spinning disk confocal unit ( CSU21 , Yokogawa , Japan ) , an EMCCD camera ( iXon , Andor Technology , Germany ) and a 40x or a 63x water objective lens . The images were acquired at 20 frames/sec . Fiji was used for image analyses and pseudocolored images . Parental flies were reared in an egg collection cup with an agar plate with yeast paste at 25°C . Eggs were laid for 1 hr and transferred to another agar plate with yeast paste containing 1 mM all-trans retinal ( R2500 , Sigma ) . The larvae were picked up and gently washed in deionized water . Then , they were placed on an apple agar plate and stimulated with blue light ( for ChR2 ( T159C ) ; band-pass filtered at 460–490 nm , ~400 μW/mm2 ) or yellow light ( for CsChrimson; band-pass filtered at 540–580 nm , ~1 mW/mm2 ) using a conventional Hg arc lamp under a fluorescence microscope ( SZX16 , Olympus , Japan ) . The larvae were videotaped before and after stimulation using an XCD-V60 CCD camera ( 30 frames/sec for 1 min ) . Localized photostimulation was performed as described previously ( Matsunaga et al . , 2013 ) . Briefly , the VNC was exposed from the larvae ( without scratching the axons as described above ) and Argon laser ( 488 nm ) was applied to a few segments of the VNC under a confocal microscope ( FV1000 , Olympus , Japan ) . The movement of the dissected larva was videotaped using a XCD-V60 CCD camera ( 30 frames/sec for 5 min ) . Third instar larvae were picked up and gently washed in deionized water . For the conditional activation assay using dTRPA1 , the larvae were transferred from an agar plate at the permissive temperature ( PT , 22°C ) to a new agar plate at a restrictive temperature ( RT , 32°C ) on a heat plate ( Thermo Plate , Tokai Hit , Japan ) . The larvae were videotaped at PT or RT conditions using an XCD-V60 CCD camera ( 30 frames/sec for 1 min ) . To label the neurons expressing photoactivatable green fluorescent protein ( PA-GFP ) , we used a conventional confocal microscope ( FV1000 , Olympus , Japan ) equipped with 63x water objective lens and 405 nm violet ( near-UV ) laser . In order to fix the sample , we used an isolated CNS preparation , which was adhered to a double-sided tape on clean glass slide with the saline . We then defined the region of interest ( ROI: the size 100x100 pixels ) and stimulated 10 s . After 5 min ( for stable photoactivation ) , cells were imaged with the same confocal microscope under 488 nm excitation . Larvae were dissected and central neurons accessed as described previously ( Baines and Bate , 1998 ) . Briefly , the larval CNS was removed and pinned onto a sylgard-coated dish using fine wire ( “0 . 001 Tungsten 99 . 95% wire” , California Fine Wire Company ) . A small section of the glial sheath surrounding the VNC between segments A2-A4 was ruptured using protease ( 0 . 1–1% Protease XIV , Sigma-Aldrich ) dissolved in external saline ( the same as above [Marley and Baines , 2011] ) , to expose cell bodies underneath . The preparation was viewed with a 60x/1NA water-dipping objective on a microscope ( BX51WI , Olympus , Japan ) . GFP-expression mediated by R36G02-GAL4 , 10xUAS-IVS-myr::GFP was used to identify A27h , and bright-field microscopy to identify aCC , with post hoc confirmation of cell identity by filling with 100 µM Alexa Fluor 568 hydrazide ( Invitrogen Molecular Probes ) , which was included in the internal saline ( MgCl2 · 6H2O 2 mM , EGTA 2 mM , KCl 5 mM , HEPES 20 mM , K-D-Gluconic acid 140 mM ) . Whole-cell recordings were performed using standard thick-walled borosilicate electrodes ( GC100TF-10; Harvard ) , fire-polished to resistances of 8–12 MΩ . Recordings were made using an Axon Multiclamp 700B amplifier with two CV-7B headstages , and digitized using a Digidata 1550 . Traces were recorded using pClamp 10 ( all from Molecular Devices ) , digitized at 20 kHz and filtered at 2 kHz . Data were analyzed using Clampfit 10 ( Molecular Devices ) and Spike2 ( Cambridge Electronic Design ) . To determine the phase relationship between periodic signals in paired whole-cell recording experiments , we used direct multi-taper estimates of power spectra and coherency , as described before ( Pulver et al . , 2015 ) . Briefly , we determined the dominant frequency of activity in aCC by examining its power spectrum , and then estimated coherence between signals in aCC and A27h . All spectral calculations were carried out using custom scripts written in MATLAB , now freely available online ( https://github . com/JaneliaSciComp/Groundswell ) . EM reconstruction was performed as described previously ( Ohyama et al . , 2015; Schneider-Mizell et al . , in press ) using a modified version of CATMAID ( Saalfeld et al . , 2009 ) . We manually traced the axonal and dendritic processes of GDLs or A27h neurons and identified the location of the pre- and post-synapses . We then reconstructed the presynaptic and postsynaptic neurons from the synaptic sites . A genetic driver line such as a GAL4 line drives expression in a specific subset of neurons . Expression patterns of interest are generally sufficiently sparse that individual neurons can be located relative to gross landmarks ( see for example [Li et al . , 2014] ) such as the entry points of nerves or lineages into the neuropile , which are highly stereotyped ( Cardona et al . , 2010 ) . Each lineage in the Drosophila larval nerve cord about 10 to 15 neurons , each with a distinctive arbor . In the EM , we locate the entry point into the neuropile of the lineage bundle and then swiftly reconstruct the low-order branches ( the “backbone” containing continuous microtubule; [Schneider-Mizell et al . , in press] ) . Then these partial reconstructions are compared to the light- microscopy images of GAL4 expression patterns , and by a process of elimination the neuron of interest is easily found . These identified neurons are then reconstructed in full , and the position of the presynaptic varicosities is compared to those observed in the light microscopy volumes , to further confirm their identification . Then , each identified neurons is used as a starting point to reconstruct all their presynaptic and postsynaptic partner neurons . These additional neurons are then readability available for comparisons with light microscopy volumes or with other segment in the nerve cord . We analyzed the data using Student’s t test and one-way analysis of variance ( ANOVA ) followed by Tukey's tests for multiple comparisons . Statistical significance is denoted by asterisks: ***p<0 . 001; **p<0 . 01; *p<0 . 05; n . s . , not significant . All statistical tests were performed using R-project software ( http://www . r-project . org ) . The results are stated as mean ± s . d . , unless otherwise noted . | Rhythmic movements such as walking and swimming require the coordinated contraction of many different muscles . Throughout the animal kingdom , from insects to mammals , animals possess specialized circuits of neurons that are responsible for producing these patterns of muscle contraction . These circuits are known as ‘central pattern generators’ . Central pattern generators are made up of multiple types of neurons that exchange information . However , it is unclear how neurons controlling the movement of one part of the body relay information to neurons controlling the movement of other parts . To answer this question , Fushiki et al . used larvae from the fruit fly Drosophila melanogaster as a model , and combined techniques such as electrophysiology and electron microscopy with measures of the insect’s behavior . Fruit fly larvae have bodies that are made of segments , and they can contract and relax these segments in a sequence to propel themselves forwards or backwards . The contraction of one segment is accompanied by relaxation of the segment immediately in front . Fushiki et al . found that each body segment contains a copy of the same basic neuronal circuit . This circuit is made up of excitatory and inhibitory neurons . Both types of neurons regulate movement , but the inhibitory neurons must be suppressed for movement to occur . The experiments also showed that each circuit receives both long-range input from the brain and local sensory feedback . This combination of inputs ensures that the segments contract and relax in the correct order . Future challenges are to determine how the brain controls larval movement via its long-range projections to the body . A key step will be to map these circuits at the level of the individual neurons and the connections between them . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | A circuit mechanism for the propagation of waves of muscle contraction in Drosophila |
Release factors RF1 and RF2 promote hydrolysis of peptidyl-tRNA during translation termination . The GTPase RF3 promotes recycling of RF1 and RF2 . Using single molecule FRET and biochemical assays , we show that ribosome termination complexes that carry two factors , RF1–RF3 or RF2–RF3 , are dynamic and fluctuate between non-rotated and rotated states , whereas each factor alone has its distinct signature on ribosome dynamics and conformation . Dissociation of RF1 depends on peptide release and the presence of RF3 , whereas RF2 can dissociate spontaneously . RF3 binds in the GTP-bound state and can rapidly dissociate without GTP hydrolysis from termination complex carrying RF1 . In the absence of RF1 , RF3 is stalled on ribosomes if GTP hydrolysis is blocked . Our data suggest how the assembly of the ribosome–RF1–RF3–GTP complex , peptide release , and ribosome fluctuations promote termination of protein synthesis and recycling of the release factors .
Termination of protein synthesis occurs when a translating ribosome encounters one of the three universally conserved stop codons UAA , UAG or UGA . In bacteria , the release of the nascent peptide is promoted by release factors RF1 and RF2 which recognize the stop codons in the A site and hydrolyze the ester bond in the peptidyl-tRNA bound to the P site , allowing the nascent peptide to leave the ribosome through the polypeptide exit tunnel ( Dunkle and Cate , 2010; Nakamura et al . , 1996 ) . RF1 and RF2 bind to the ribosome in the space between the small and large ribosomal subunits . RF1 and RF2 differ in their stop codon specificity: RF1 utilizes a conserved PET motif to recognize UAG and UAA codons , whereas RF2 uses an SPF motif to recognize UGA and UAA codons . Both RF1 and RF2 have a universally conserved GGQ motif which promotes the catalysis of peptidyl-tRNA hydrolysis ( Seit-Nebi et al . , 2001 ) ; mutations of the GGQ motif to GAQ or GGA inhibit peptide release ( Frolova et al . , 1999; Mora et al . , 2003; Shaw and Green , 2007; Zavialov et al . , 2002 ) . After peptide release , RF1 and RF2 dissociate from the post-termination complex to allow for the next steps of translation . The dissociation is accelerated by RF3 , a translational GTPase that binds and hydrolyses GTP in the course of termination ( Freistroffer et al . , 1997; Koutmou et al . , 2014; Zavialov et al . , 2002 ) . In addition to canonical termination , RF2 takes part in non-canonical termination events such as post-peptidyl transfer quality control ( Zaher and Green , 2009 ) and ribosome rescue on truncated mRNAs ( Kurita et al . , 2014 ) . There are two different models concerning the sequence of events during termination , including the timing of peptide release , the order of RF1 , RF2 and RF3 binding and dissociation , and the role of nucleotide exchange in RF3 and GTP hydrolysis . The first model of translation termination was proposed by Ehrenberg and colleagues ( Zavialov et al . , 2001; Zavialov et al . , 2002 ) . Based on nitrocellulose filtration experiments , the authors reported that free RF3 has a much higher affinity for GDP ( Kd = 5 . 5 nM ) than for GTP ( Kd = 2 . 5 µM ) or GDPNP ( Kd = 8 . 5 µM ) ( Zavialov et al . , 2001 ) , which would imply that at cellular GTP/GDP concentrations RF3 is expected to be predominantly in the GDP form . The exchange of GDP for GTP occurs only when RF3–GDP binds to the ribosome in complex with RF1 or RF2 ( Zavialov et al . , 2001 ) . In the absence of the nucleotide , RF3-dependent RF1/2 recycling is slow , which has been interpreted as an indication for a high-affinity complex of apo-RF3 to the ribosome–RF1/2 complex ( Zavialov et al . , 2001 ) . Furthermore , because RF3-dependent turnover GTPase activity was stimulated by peptidyl-tRNA hydrolysis , the authors suggested that RF3 binds to the ribosome–RF1 complex only after the peptide is released . Based on these results , Ehrenberg et al . suggested the following sequential mechanism of termination: RF1/RF2 bind to the ribosome and hydrolyze peptidyl-tRNA , allowing RF3–GDP to enter the ribosome occupied by RF1 or RF2 to form an unstable encounter complex . Dissociation of GDP leads to a stable high-affinity complex with RF3 in the nucleotide-free state . The subsequent binding of GTP by RF3 promotes RF1/RF2 dissociation . In the final step , RF3 hydrolyses GTP and as a result dissociates from the ribosome ( Zavialov et al . , 2001; Zavialov et al . , 2002 ) . An alternative model was proposed based on the kinetic and thermodynamic analysis of GTP/GDP binding to RF3 by ensemble kinetics and equilibrium methods . The results of those experiments indicated that the affinity of RF3 to GDP and GTP is on the same order of magnitude ( 5 nM and 20 nM , respectively [Koutmou et al . , 2014; Peske et al . , 2014] ) . As the cellular GTP concentration is at least 10 times higher than the GDP concentration ( Bennett et al . , 2009 ) , these affinities imply that nucleotide exchange in RF3 can occur spontaneously , off the ribosome , and thus RF3 could enter the ribosome in either the GTP- or GDP-bound form . Consistent with previous findings ( Zavialov et al . , 2001; Zavialov et al . , 2002 ) , ribosome–RF1 or ribosome–RF2 complexes accelerate nucleotide exchange in RF3 ( Koutmou et al . , 2014; Peske et al . , 2014 ) ; however , this effect is independent of peptide release , because also a catalytically inactive RF2 mutant activates nucleotide exchange in RF3 ( Peske et al . , 2014; Zavialov et al . , 2002 ) . Binding of GTP to RF3 in complex with the ribosome and RF2 is rapid ( 130 s−1 ) ( Peske et al . , 2014 ) , and thus the lifetime of the apo-RF3 state would be too short to assume a tentative physiological role . Peptide release results in the stabilization of the RF3–GTP–ribosome complex , thereby promoting the dissociation of RF1/2 , followed by GTP hydrolysis and dissociation of RF3–GDP from the ribosome ( Peske et al . , 2014 ) . Efficient translation termination not only requires the coordinated action of the release factors , but also entails conformational dynamics of the factors and the ribosome . The key conformational motions of the ribosome during termination and in general in all phases of translation include the rotation of ribosomal subunits relative to each other , the swiveling motion of the body and head domains of the small ribosomal subunit , the movement of the ribosomal protein L1 toward or away from the E-site tRNA , and the movement of tRNAs between classic and hybrid conformation . These motions are loosely coupled and gated by ligands of the ribosome such as translation factors and tRNAs ( Adio et al . , 2015; Chen et al . , 2011; Cornish et al . , 2008; Horan and Noller , 2007; Sharma et al . , 2016; Shi and Joseph , 2016; Valle et al . , 2003; Wasserman et al . , 2016 ) . Crystal structures show that termination complexes with RF1 or RF2 are predominantly in the non-rotated ( N ) state . The P-site tRNA in the complexes is in the classical state and the L1 stalk in an open conformation ( Jin et al . , 2010; Korostelev et al . , 2008; Laurberg et al . , 2008; Weixlbaumer et al . , 2008 ) . A single molecule fluorescence resonance energy transfer ( smFRET ) study showed that binding of RF1 to termination complexes stabilizes the open conformation of the L1 stalk , whereas in the absence of RF1 termination complexes make reversible transitions between the open and closed state ( Sternberg et al . , 2009 ) ; the rotation of the ribosomal subunits was not investigated directly in that study . The high sequence similarity between RF1 and RF2 suggests that the two factors interact with the ribosome in the same manner and promote peptide release by a similar mechanism ( Freistroffer et al . , 1997; Zavialov et al . , 2001 ) . However , structures of RF1 or RF2 bound to termination complexes reveal differences regarding the interaction with the L11 region of the 50S subunit ( Korostelev et al . , 2008; Laurberg et al . , 2008; Petry et al . , 2005; Pierson et al . , 2016; Rawat et al . , 2006; Rawat et al . , 2003; Weixlbaumer et al . , 2008 ) . Thus , it is not clear whether RF1 and RF2 follow the same mechanism and whether they respond in the same way to the recruitment of RF3 to termination complexes . In the absence of RF1/RF2 , binding of RF3 with a non-hydrolyzable GTP analog to the ribosomes where the nascent peptide has been released induces formation of the rotated ( R ) state of the ribosome , with the tRNA in the P/E hybrid state and the closed conformation of the L1 stalk ( Gao et al . , 2007; Jin et al . , 2011; Zhou et al . , 2012 ) . A very similar effect of RF3–GDPNP was found by smFRET ( Sternberg et al . , 2009 ) . However , it is much less clear what happens when RF1/RF2 and RF3 bind to the ribosome together . Modeling of the atomic structures of RF1 and RF2 into the cryo-EM structure of RF3-bound post-release complex suggests that the RF3-induced ribosome rearrangements break the interactions of RF1/RF2 with both the decoding center and the L11 region of the ribosome , leading to the release of RF1/RF2 ( Gao et al . , 2007 ) . In this model , stable binding of RF1 and RF3 is mutually exclusive . On the other hand , a cryo-EM structure of ribosomes in complex with a deacylated tRNA in the P site , RF1 , and RF3 in the apo form , that is , in the absence of added nucleotide , suggest that both factors can bind simultaneously to the ribosome ( Pallesen et al . , 2013 ) . smFRET measurements carried out with post-release complexes in the presence of excess RF1 showed that the addition of RF3–GTP induced short-lived transitions from the L1-open to the L1-closed state which were not observed in the absence of RF3 . This suggests that the two factors can bind to the ribosome simultaneously ( Sternberg et al . , 2009 ) . No structural studies are available on the interaction of RF3 with RF2-bound complexes . The interaction of RF3 with the ribosomes prior to peptide release has not been studied . Here , we use TIRF microscopy to monitor smFRET signals reporting on subunit rotation to follow changes in ribosome conformation in response to RF1 , RF2 and RF3 and the binding of each individual release factor to the ribosome during termination . Our results demonstrate how the recruitment of release factors change the ribosome conformation in termination complexes , how the dissociation of the factors is achieved , show differences in the function of RF1 and RF2 , and explain the importance of GTP binding and hydrolysis by RF3 .
To monitor the rotation of the ribosomal subunits during termination , we utilized ribosomes with fluorescent labels attached to the small subunit protein S6 and the large subunit protein L9 , S6-Cy5 and L9-Cy3 , respectively . This FRET pair has been extensively characterized in both smFRET and ensemble kinetics experiments and reports on the formation of the non-rotated ( N ) or the rotated ( R ) state of the ribosome ( Cornish et al . , 2008; Ermolenko et al . , 2013; Sharma et al . , 2016 ) . We prepared termination complexes on an mRNA which is translated up to the stop codon UAA recognized by both RF1 and RF2 . The complexes contain a peptidyl-tRNA in the P site and have a stop codon in the A site; those complexes are denoted as pre-hydrolysis complexes ( PreHC ) . In the absence of termination factors , PreHC is found predominantly in a state with the FRET efficiency of 0 . 73 ± 0 . 01 ( denoted as 0 . 7 FRET state in the following ) ( Figure 1A , Figure 1—figure supplement 1; Supplementary file 1 ) . Previous work has shown that this state corresponds to the N state of the ribosome ( Cornish et al . , 2008; Qin et al . , 2014; Sharma et al . , 2016 ) . A small fraction of complexes shows a FRET state with an efficiency of 0 . 52 ± 0 . 02 ( 0 . 5 FRET state ) , which corresponds to the R state of the ribosome . While peptidyl-tRNA generally favors the N state , the ability of ribosomes with peptidyl-tRNA in the P site to adopt the R state at room temperature has been demonstrated previously by smFRET and cryo-EM ( Cornish et al . , 2008; Fischer et al . , 2010; Ling and Ermolenko , 2015 ) . The distribution of FRET efficiencies and thus the ratio between N and R conformations of PreHC is independent of the tRNA in the P site and of the presence of a single N-terminal amino acid ( fMet ) or a dipeptide ( fMetPhe , fMetVal or fMetLys ) at the P-site tRNA ( Figure 1—figure supplement 1; Supplementary file 1 ) . This finding has prompted us to use the PreHC with fMet-tRNAfMet in the P site and a stop codon in the A site as a minimal model system , following previous publications which used this approach to study termination ( Casy et al . , 2018; Jin et al . , 2010; Koutmou et al . , 2014; Kuhlenkoetter et al . , 2011; Pallesen et al . , 2013; Pierson et al . , 2016; Shi and Joseph , 2016; Sternberg et al . , 2009 ) . To probe the effect of RF1/RF2 binding on subunit rotation , we mixed PreHC with RF1 or RF2 at saturating concentrations of the factors ( Zavialov et al . , 2002 ) . Peptide release was avoided by using RF1 ( GAQ ) or RF2 ( GAQ ) mutants which are catalytically deficient ( Frolova et al . , 1999; Zavialov et al . , 2002 ) ( Figure 1—figure supplement 2A ) . Binding of RF1 ( GAQ ) to PreHC stabilizes the N state ( Figure 1B ) . The fraction of the PreHC in the R state , albeit small , is somewhat higher with RF2 ( GAQ ) than with RF1 ( GAQ ) ( Figure 1B , C ) . The hydrolysis of the ester bond between the tRNA and the nascent peptide in PreHC leads to the formation of post-hydrolysis complex ( PostHC ) . To prepare PostHC without the use of termination factors , we released nascent peptides with the help of puromycin , an analog of the A-site aminoacyl-tRNA that reacts with the peptidyl-tRNA in the P site to form peptidyl-puromycin ( which then dissociates from the ribosome ) and a deacylated tRNA in the P site . These complexes are denoted as PostHC* . The FRET histogram of PostHC* in the absence of the factors indicate the presence of two states , the 0 . 5 FRET ( R ) state and the 0 . 7 FRET ( N ) state ( Figure 1D , Supplementary file 1 ) . FRET time courses of individual ribosomes show reversible transitions between the N and R states ( Figure 1—figure supplement 1B ) . The exact distribution of states depends on the P-site tRNA ( Figure 1—figure supplement 1; Supplementary file 1 ) ( Cornish et al . , 2008 ) with tRNAfMet behaving similarly to tRNAVal , thus underscoring the suitability of the minimal model system . To test the effect of RF1 and RF2 on subunit rotation of PostHC , we added saturating concentrations of the wild type RF1 or RF2 to PreHC to allow peptide release . RF1 halts PostHC in the N state ( Figure 1E , Figure 1—figure supplement 2B , D ) , in agreement with the previous smFRET study where RF1 binding stabilizes the L1 stalk in the open state ( Sternberg et al . , 2009 ) . Binding of RF2 to PostHC shifts the equilibrium toward the N state , but not to the same extent as RF1 ( Figure 1F , Figure 1—figure supplement 2C , E ) . Complexes with RF2 make occasional N to R transitions , in particular with RF2 ( GAQ ) bound to PostHC ( Figure 1—figure supplement 2C ) . These initial observations suggest that although both factors favor the N state , RF1 appears more efficient than RF2 . To further probe the potential differences between RF1 and RF2 , we monitored subunit rotation in response to factor binding in real time . We injected catalytic amounts of release factors to PreHC and PostHC and recorded the time courses of FRET signal changes ( Figure 1—figure supplement 3 ) . RF1 ( GAQ ) binding to PreHC does not change the FRET efficiency appreciably , as the complex is predominantly in the N state with or without the factor ( Figure 1—figure supplement 3A ) . Also the binding of wild-type RF1 to PreHC with subsequent peptide release does not change the FRET efficiency ( Figure 1—figure supplement 3B ) , supporting the notion that stabilization of the N state by RF1 is independent of peptide release ( Figure 1B , E ) . PostHC without factor fluctuates between the N and R state; binding of RF1 to PostHC halts fluctuating ribosomes in the N state and prevents excursions to the R state ( Figure 1—figure supplement 3C ) . With RF2 the picture is somewhat different . PreHC–RF2 ( GAQ ) is predominantly in the N state ( Figure 1—figure supplement 3D ) . However , binding of wild type RF2 and peptide release shift the distribution toward the R state ( Figure 1—figure supplement 3E ) . The resulting PostHC fluctuates between N and R states as shown by synchronization of FRET traces to the first N to R transition . PostHC obtained by puromycin treatment also shows reversible N to R transitions which remain undisturbed by the addition of RF2 ( Figure 1—figure supplement 3F ) . Although the binding of the factors is not directly monitored in these experiments , the differences in the rotation pattern suggest that RF1 and RF2 have distinct effects on ribosome dynamics . Such differences may result from a shorter residence time of RF2 compared to RF1 on the ribosome , which we tested in the following experiments . To measure how long the factors remain bound to the ribosome , we prepared Cy5-labeled RF1 and RF2 , as well as the respective RF1/2 ( GAQ ) mutants and ribosomes containing Cy3-labeled protein L11 ( Adio et al . , 2015; Chen et al . , 2011; Geggier et al . , 2010; Holmberg and Noller , 1999; Stöffler et al . , 1980 ) ( Figure 2—figure supplement 1A ) . The biochemical activity of labeled release factors was indistinguishable from that of the unlabeled counterparts ( Figure 2—figure supplement 1B , C ) and the factors were fully methylated ( Figure 2—figure supplement 2 ) . L11 constitutes part of the factor binding site ( Pallesen et al . , 2013; Petry et al . , 2005; Rawat et al . , 2006; Rawat et al . , 2003 ) . Recruitment of the factors to the ribosome is expected to result in high FRET efficiency . Binding of RF1 or RF1 ( GAQ ) to either PreHC or PostHC results in a single FRET population centered at 0 . 72 ± 0 . 02 ( 0 . 7 FRET ) ( Figure 2A–C , Supplementary file 1 ) . RF1 and RF1 ( GAQ ) are stably bound to the ribosome , in agreement with previous biochemical reports on dissociation rates of RF1 and RF1 ( GAQ ) ( 0 . 005–0 . 1 s−1; [Koutmou et al . , 2014; Shi and Joseph , 2016] ) . The estimated upper limit of the dissociation rate in our experiments is 0 . 2 s−1 ( Supplementary file 1 ) , defined by the photobleaching rate of the FRET dye pair with kphotobleaching = 0 . 07–0 . 19 s−1 at the given imaging conditions ( Materials and methods ) . Binding of RF2 or RF2 ( GAQ ) to PreHC or PostHC leads to single FRET populations with efficiencies between 0 . 6 and 0 . 7 ( Figure 2D–F , Supplementary file 1 ) . However , the residence time of RF2 is much shorter compared to RF1 or RF1 ( GAQ ) , with the koff values in the range from 0 . 8 ± 0 . 1 s−1 to 1 . 3 ± 0 . 2 s−1 ( Figure 2D–F , Supplementary file 1 ) . Peptide hydrolysis has no visible effect on RF1 and only a minor effect on RF2 dissociation ( in the absence of RF3 ) . The difference in the dissociation rates of RF1 and RF2 implies that RF1 needs an auxiliary factor , RF3 , to help it to dissociate from the ribosome , whereas RF2 may be able to recycle independent of RF3 . This notion is consistent with previous reports ( Petropoulos et al . , 2014; Zavialov et al . , 2002 ) and is further supported by our peptide hydrolysis turnover assay ( Figure 2—figure supplement 3A ) . With catalytic amounts of RF1 in the absence of RF3 , that is , when RF1 turnover depends on its intrinsic dissociation rate from the ribosome , termination is essentially blocked , whereas in the presence of RF3 RF1-mediated peptide release is very efficient . In contrast , even catalytic amounts of RF2 are sufficient to complete peptide release from PreHC , although RF3 accelerates the reaction by a factor of 10 . Thus , RF3 is essential for RF1 , but not for RF2 recycling . Addition of RF3 to PreHC–RF2 ( GAQ ) or PostHC–RF2 complexes makes the complexes more dynamic ( Figure 2—figure supplement 3B–E ) . Our results demonstrate that during canonical termination RF1 and RF2 interact with termination complexes in somewhat different ways , as they have different residence times on the ribosome and respond differently to the presence of RF3 . Next , we studied how RF3 affects ribosome dynamics and promotes the dissociation of RF1/RF2 . To investigate the effect of RF3 on subunit rotation in the absence of RF1 or RF2 , we added saturating concentrations of RF3 to S6/L9-labeled PreHC ( Figure 3A , Figure 2—figure supplement 1 ) . Binding of RF3 to PreHC , which in the absence of the factor is in the N state , strongly shifts the equilibrium toward the R state ( Figure 3A ) , that is , RF3 has the opposite effect on subunit rotation than RF1 or RF2 . The traces are now highly dynamic and show reversible N to R transitions ( Supplementary file 1 ) . This finding seems unexpected as ribosomes with peptidyl-tRNA in the P site favor the N state . However , previous cryo-EM and smFRET studies have indicated that those complexes can in fact adopt the R state ( Cornish et al . , 2008; Fischer et al . , 2010; Ling and Ermolenko , 2015 ) . Thus , RF3 appears to bias spontaneous fluctuations of peptidyl-tRNA in the PreHC , rather than induce a previously disallowed conformation . To further characterize the conformational changes of the ribosome induced by RF3 , we probed the position of the P-site tRNA relative to protein L1 . We used a FRET pair with the donor label at the tRNA ( fMet-tRNAfMet-Cy3 ) and the acceptor label on ribosomal protein L1 ( L1-Cy5 ) . The two labels are close together and give a high FRET signal when ribosomes are in the L1-closed state and move apart to give a low FRET signal when ribosomes are in the L1-open state ( Fei et al . , 2009; Fei et al . , 2008; Munro et al . , 2010a; Munro et al . , 2010b; Munro et al . , 2010c; Sternberg et al . , 2009 ) . FRET histograms of PreHC in the absence of RF3 are dominated by a low-FRET population ( 0 . 32 ± 0 . 01 ) and do not show transitions to other states ( Figure 3—figure supplement 1A , Supplementary file 1 ) . This indicates that ribosomes are predominantly in the L1-open state with fMet-tRNAfMet in the classic P/P state , in agreement with previous studies ( Cornish et al . , 2009; Fei et al . , 2008; Sternberg et al . , 2009 ) . RF3 induces dynamic transitions from the low FRET state to a high FRET state ( 0 . 74 ± 0 . 02 ) , which suggests that ribosomes transiently sample the L1-closed state with fMet-tRNAfMet in a hybrid-like P/E state ( Figure 3B ) . This state is short lived ( kclosed→open = 6 . 0 ± 0 . 8 s−1; Supplementary file 1 ) . The transition rate is faster than subunit rotation ( kR→N = 2 . 2 ± 0 . 4 s−1 ) suggesting that the two processes are not tightly coupled , consistent with the previous smFRET work ( Munro et al . , 2010a; Wasserman et al . , 2016 ) and cryo-EM reconstructions ( Fischer et al . , 2010 ) . We then monitored the dissociation of RF3 from PreHC using FRET between RF3–Cy5 and L11–Cy3 ( Figure 3C ) . Labeling of RF3 did not change its catalytic properties ( Figure 2—figure supplement 1A , D ) . The dissociation rate of RF3 from PreHC is koff = 5 . 9 ± 1 . 1 s−1 ( Figure 3C; Supplementary file 1 ) . Thus , RF3-GTP can bind to PreHC and alter its conformation as shown by the rotation of subunits and movement of the peptidyl-tRNA into a P/E-like state , but the residence time of the factor on the ribosome is short . To test whether the interaction of RF3 with termination complexes depends on peptide release , we then studied the effect of RF3 on subunit rotation of PostHC prepared by puromycin treatment ( PostHC* ) ( Figure 3D–F ) . S6/L9-labeled PostHC* fluctuates between 0 . 5 and 0 . 7 FRET states ( Figure 1—figure supplement 1 ) . RF3 binding shifts the distribution toward the 0 . 5 FRET state , indicating that the R state is stabilized ( Figure 3D ) . The L1–tRNA FRET pair shows an enrichment of the high FRET state corresponding to the P/E state of the tRNA ( Figure 3E ) . While subunits are stabilized in the R state and do not fluctuate toward N state , the L1-tRNA label shows reversible transitions between P/E and P/P conformations ( Supplementary file 1 ) . This suggests that also in PostHC subunit rotation and the formation of a hybrid-like state are not tightly coupled . Dissociation of RF3 from PostHC is as rapid as from PreHC , koff = 5 . 4 ± 1 . 3 s−1 ( Figure 3F , Supplementary file 1 ) . Our results show that RF3 facilitates the formation of the R state with the tRNA in a P/E-like orientation on both PreHC and PostHC . RF3 dissociation is not directly coupled to subunit rotation , as the rate of R to N transitions is lower than that of RF3 dissociation ( Figure 3 , Supplementary file 1 ) . The residence time of RF3 on the ribosome is nearly identical on Pre- and PostHC which indicates that the presence of RF3 on the ribosome is not regulated by peptide release . We also note that the observed RF3 dissociation rates are much higher than the rate of GTP hydrolysis by RF3 ( Peske et al . , 2014; Shi and Joseph , 2016; Zavialov et al . , 2001 ) . This implies that rapid RF3 dissociation is independent of GTP hydrolysis . Next , we studied the interplay between RF1 , RF3 and ribosomes during termination . We compared three different termination conditions including PreHC , PostHC* prepared by puromycin treatment , and PreHC which was converted to PostHC in situ upon the interaction with RF1 . For each condition , we monitored ( i ) subunit rotation in the presence of saturating concentrations of both RF1 and RF3; ( ii ) RF1-Cy5 binding to the ribosome at saturating concentrations of unlabeled RF3; and ( iii ) RF3-Cy5 binding to the ribosome at saturating concentrations of unlabeled RF1 ( Figure 4 ) . To follow the interactions of RF1 and RF3 with PreHC ( Figure 4A , B , C ) , we again used the RF1 ( GAQ ) mutant , which ensures that peptidyl-tRNA in PreHC is not hydrolyzed . While RF1 ( GAQ ) alone stabilizes the N state ( grey line in Figure 4A; Figure 1B ) and RF3 alone induces transitions from the N to the R state ( Figure 3A ) , in the presence of saturating amounts of RF1 ( GAQ ) and RF3 together the N and R states are almost equally populated ( Figure 4A ) . Ribosomes show rapid reversible N to R transitions indicating that RF3 can promote subunit rotation even when RF1 is present . RF1 ( GAQ ) binding to PreHC–RF3 results in a single FRET population centered at a FRET efficiency of 0 . 67 ± 0 . 02 , similar to the 0 . 7 FRET when RF1 binds to the ribosome in the absence of RF3 . The RF1 ( GAQ ) dissociation rate is low , <0 . 3 s−1 ( Figure 4B , D; Supplementary file 1 ) , in agreement with previous reports ( 0 . 14 ± 0 . 02 s−1 , [Koutmou et al . , 2014] ) . RF3 binds to PreHC–RF1 ( Figure 4C ) , but the FRET efficiency for the RF3-L11 pair is reduced compared to the complex in the absence of RF1 ( 0 . 51 ± 0 . 03 and 0 . 62 ± 0 . 02 in the presence and absence of RF1 , respectively; Supplementary file 1 ) . Thus , the orientation of RF3 on the PreHC is shifted by RF1 , whereas the position of RF1 appears unchanged , at least with respect to L11 . The rate of RF3 dissociation in the presence of RF1 ( GAQ ) is 1 . 3 ± 0 . 1 s−1 , which is higher than the dissociation rate of RF1 ( GAQ ) , but about fivefold slower than that of RF3 in the absence of RF1 ( Figure 3C; Figure 4D; Supplementary file 1 ) , indicating that RF1 stabilizes the binding of RF3 to PreHC . Dwell time distributions for N or R state in the presence of RF1 ( GAQ ) and RF3 are biphasic , suggesting the presence of two populations of each complex . The majority of ribosomes display rapid transitions ( >70% , kN→R = 5 . 9 ± 0 . 6 s−1 , kR→N = 2 . 9 ± 0 . 4 s−1; Figure 4D; Supplementary file 1 ) that are faster than RF1 or RF3 dissociation , indicating that subunits can rotate while both factors are bound to the ribosome . Low rotation rates ( kN→R = 1 . 30 ± 0 . 07 s−1 , kR→N = 0 . 80 ± 0 . 05 s−1; <30% of ribosomes ) are also observed with RF3 alone and thus may represent subunit rotation after RF1 dissociation ( Supplementary file 1 ) . The observed shift of PreHC–RF1 from the predominantly N to a fluctuating ensemble of N and R states upon RF3 addition , together with the altered RF3 position and the decreased RF3 dissociation rate when the two factors are bound suggest that the complex undergoes conformational adjustments when RF1 and RF3–GTP are bound simultaneously . Next , we monitored subunit rotation in PostHC . For better comparison with the results obtained with PreHC and RF1 ( GAQ ) , we first prepared PostHC* by puromycin treatment of PreHC and studied the interactions with RF1 ( GAQ ) and RF3–GTP ( Figure 4E , F , G , H ) . In the presence of RF1 and RF3 , the majority of complexes undergo rapid N to R transitions and the equilibrium is shifted toward the R state ( Figure 4E ) . The mean FRET efficiency for RF1 ( GAQ ) binding to PostHC*–RF3 changes to 0 . 67 ± 0 . 02 compared to 0 . 50 ± 0 . 03 for RF1 ( GAQ ) binding to PreHC–RF3 ( Figure 4B , F ) or 0 . 71 ± 0 . 01 for binding to complexes in the absence of RF3 ( Figure 2A–C ) . The decrease in FRET efficiency suggests that peptide release allows a rearrangement of the complex which alters the position of RF1 relative to L11 . The FRET efficiency for RF3 binding to either PreHC–RF1 ( GAQ ) or PostHC*–RF1 ( GAQ ) is 0 . 51 ± 0 . 03 ( Figure 4C , G ) , as compared to 0 . 62 and 0 . 64 , respectively , for binding to PreHC or PostHC in the absence of RF1 ( Figure 3C , F ) . This suggests that the position of RF3 on PreHC and PostHC is affected by the presence of RF1 , but not by peptide release ( Figures 3F and 4G ) . The dissociation rates are 1 . 3 ± 0 . 2 s−1 and 1 . 3 ± 0 . 1 s−1 for RF1 ( GAQ ) and RF3 , respectively ( Figure 4F–H; Supplementary file 1 ) . A small fraction ( 8% ) of complexes that release RF1 ( GAQ ) slowly ( koff = 0 . 12 ± 0 . 07 s−1 ) is likely due to incomplete peptide hydrolysis by puromycin . The rotation rates ( kN→R = 4 . 20 ± 0 . 08 s−1 , kR→N = 2 . 50 ± 0 . 03 s−1 ) are somewhat higher than RF1 and RF3 dissociation rates , but the most prominent effect of peptide release is the acceleration of RF1 dissociation from <0 . 3 s−1 to 1 . 3 ± 0 . 2 s−1 ( Figure 4D , H; Supplementary file 1 ) . Similar effects are observed when instead of puromycin we used wild-type RF1 to convert PreHC to PostHC ( Figure 4I–L ) : at saturating concentrations of RF1 and RF3 the R state of PostHC is enriched and the complexes show reversible N to R transitions ( Figure 4I; Supplementary file 1 ) . RF1 and RF3 are bound to PostHC in the 0 . 5 FRET state ( Figure 4J , K; Supplementary file 1 ) . The dissociation rates are 1 . 2 ± 0 . 4 s−1 for RF1 ( >70% of ribosomes; Figure 4J; Supplementary file 1 ) and 1 . 3 ± 0 . 2 s−1for RF3 ( Figure 4K; Supplementary file 1 ) . Thus , RF1 stabilizes the binding of RF3 on PreHC or PostHC , whereas RF3 destabilizes RF1 binding , but only after peptide release . Peptide release also allows an adjustment in the positions of both factors relative to L11 . Thus , peptide release is a major determinant for RF1 , but not RF3 , dissociation . Because the kinetics of subunit rotation is faster than RF3 and RF1 dissociation , it remains unclear from which state , N or R , the factors dissociate . To test whether R state formation is required for RF3 dissociation , we used the antimicrobial peptide apidaecin 137 ( Api ) as a tool to trap RF1 on termination complexes . Api binds into the exit tunnel of PostHC and prevents RF1/RF2 dissociation ( Florin et al . , 2017 ) . When we monitor subunit rotation in the presence of saturating concentrations of RF1 , RF3 and Api , the PostHC–RF1–RF3–Api complex is stalled in the N state ( Figure 5A ) . In the absence of RF1 Api does not alter the relative fraction of N and R states induced by RF3 ( Figure 5B ) . In the PostHC–RF1–Api–RF3 complex , RF1 is stably bound in the 0 . 7 FRET state ( Figure 5C ) . RF3 is bound in 0 . 5 FRET state and dissociates with the rate of 1 . 2 ± 0 . 1 s−1 ( Figure 5D ) . These data suggest that RF3 can dissociate independent of subunit rotation from termination complexes that are exclusively in the N state as well as from termination complexes that show mixed N and R populations . By analogy with other GTPases , GTP hydrolysis by RF3 is expected to regulate the dissociation of RF3 from the ribosome . In contrast to all other GTPases , RF3 was suggested to bind to the PostHC-RF1 complex in the GDP-bound form; the ribosome-induced rapid release of GDP should stabilize RF3 binding , while subsequent GTP binding induces a conformational change of the ribosome and the release of RF1 ( Sternberg et al . , 2009; Zavialov et al . , 2002 ) . We first tested these models using a biochemical turnover peptidyl-tRNA hydrolysis assay and compared the effect of different nucleotides on factor recycling ( Figure 6A , B ) . When both RF1 and RF3 are sub-stoichiometric to PreHC , such that 10 cycles of RF1 and RF3 turnover are required to convert all PreHC to PostHC , peptide release is only observed in the presence of GTP ( Figure 6A ) . In excess of RF3 , when only RF1 has to turnover , efficient peptide release is observed with wild type RF3 in the presence of GTP , GTPγS or GDPNP ( Figure 6B ) . Also RF3 ( H92A ) –GTP , a RF3 mutant deficient in GTP hydrolysis , induces efficient recycling of RF1 , contrary to previous reports ( Gao et al . , 2007 ) , but consistent with a recent kinetic study ( Shi and Joseph , 2016 ) . Apo-RF3 has no activity , again consistent with previous reports ( Shi and Joseph , 2016; Zavialov et al . , 2002 ) . The low activity in the presence of GDP is most likely due to a minor contamination with GTP . Thus , GTP hydrolysis is not required for RF1 recycling but is necessary to ensure recycling of RF3 , while the apo and GDPforms of RF3 appear inactive . Next , we sought to understand how different nucleotides affect the interaction of RF3 with termination complexes . RF3–GTP promotes R state formation , which can be used as readout for the ribosome interaction with RF3 in complex with different nucleotides ( Figure 6—figure supplement 1 ) . PreHC in the presence of excess RF3–GDP or RF3 in the apo form are predominantly in the N state and do not show transitions to the R state ( Figure 6—figure supplement 1A , B ) ; the ratio of N and R states is identical to that in the PreHC in the absence of RF3 ( Figure 1A ) . Also RF1-bound PostHC in the presence of excess RF3–GDP or apo-RF3 are predominantly in the N state and the distribution of states is very similar to that in RF1-bound termination complexes ( Figure 6—figure supplement 1C , D and Figure 1E , respectively ) . Together , these experiments suggest that RF3-GDP and apo-RF3 are not able to induce the R state in termination complexes . By analogy , smFRET experiments monitoring the position of the L1 stalk show that addition of RF3-GDP or apo-RF3 does not change the ribosome conformation ( Sternberg et al . , 2009 ) . For a more direct observation of RF3-GDP or apo-RF3 binding to the ribosome , we made an attempt to follow FRET between RF3-Cy5 and termination complexes labeled at protein L11 with Cy3 . However , we did not find any FRET events indicative of RF3 binding in the presence of GDP or with apo-RF3 ( data not shown ) . These observations suggest that although RF3-GDP or apo-RF3 must bind to PostHC–RF1 in some way , because this interaction accelerates nucleotide exchange in RF3 ( Koutmou et al . , 2014; Peske et al . , 2014; Shi and Joseph , 2016; Zavialov et al . , 2001 ) the interaction must be transient and does not engage the factor at its binding site at L11 unless GTP is bound . We then asked whether GTP hydrolysis by RF3 is required to induce subunit rotation . We replaced GTP with a non-hydrolysable analog , GDPNP , which is extensively used in structural studies . RF3–GDPNP can bind to the PreHC or PostHC obtained by addition of RF1 and induces formation of the R state , albeit not to the same extent as RF3–GTP and with fewer transitions between N and R states ( Figure 6C , D ) . The same tendencies are observed with RF3 ( H92A ) –GTP or RF3–GTPγS ( Figure 6—figure supplement 1E , F , G ) . The exact fraction of the R state and dynamic ribosomes depends on the choice of nucleotide , which may indicate that the ability of RF3–GDPNP or RF3–GTPγS to form a stable complex with the ribosome is reduced compared to RF3–GTP . We then tested whether GTP hydrolysis is required for RF3 dissociation from the ribosome . The dissociation rate of RF3–GDPNP from PostHC* in the absence of RF1 is koff = 0 . 34 ± 0 . 04 s−1 , much lower than with GTP ( Figure 6E and Supplementary file 1 ) . In contrast , in the presence of saturating RF1 concentrations dissociation of RF3–GDPNP from PostHC-RF1 is as rapid as with GTP ( Figure 6F and Supplementary file 1 ) , indicating that GTP hydrolysis is not essential when RF1 is present . Experiments with RF3–GTPγS gave very similar results ( Figure 6—figure supplement 2 ) . At saturating RF3 concentrations , dissociation of RF1 from PostHC is independent of GTP hydrolysis ( Figure 6—figure supplement 2 ) , consistent with the biochemical data ( Figure 6B ) . In the simplest model , these findings can be interpreted as an indication for the role of GTP hydrolysis in RF3 dissociation from termination complexes in the absence of RF1 . They also explain why RF1 turnover is impaired at sub-stoichiometric RF3 concentrations when GTP hydrolysis is blocked ( Figure 6A ) : those RF3 molecules that bind to ribosomes lacking RF1 remain stalled if GTP is not hydrolyzed , thereby depleting the pool of RF3 which has to turnover to stimulate RF1 dissociation . Thus , the only reaction where GTP hydrolysis or an authentic GTP conformation appears to play an essential role is the dissociation of RF3 from termination complexes in the absence of RF1 .
Our experiments show how release factors navigate through the landscape of possible ribosome conformations during translation termination ( Figure 7A ) . Release factors not only change the ratio between the N and R states , but also alter the fraction of the ribosomes that make transient fluctuations between the states . Each factor alone has its distinct signature on ribosome conformation and dynamics . Binding of RF1 to either PreHC or PostHC favors the static N state; protein L1 adopts an open conformation , which correlates with a classical state of the P-site tRNA . The N state of the ribosome–RF1 complex has been also captured by structural studies ( James et al . , 2016; Korostelev et al . , 2008; Laurberg et al . , 2008; Petry et al . , 2005; Weixlbaumer et al . , 2008 ) . Surprisingly , we find that PreHC–RF2 is more dynamic , and has a higher fraction of the R states than the complex with RF1 . Furthermore , RF2 can dissociate equally well from the PreHC and PostHC and is less dependent on the action of RF3 than RF1 ( Figure 2—figure supplement 3A , Figure 7B ) . With its high dissociation rate , RF2 action may depend on the ratio between the rate of peptide release and factor dissociation , for example , if the rate of peptidyl-tRNA hydrolysis is about 10 s−1 ( Indrisiunaite et al . , 2015; Kuhlenkoetter et al . , 2011 ) and the rate of RF2 dissociation is ~1 s−1 ( this paper ) , the factor can achieve efficient peptide release before dissociating . Thus , RF1 and RF2 – albeit fulfilling a similar function during canonical termination – differ in their ability to affect ribosome dynamics . Binding of RF3–GTP to termination complexes shifts the conformational distribution toward the R state ( Figure 7A ) . The PreHC–RF3 complex is dynamic , whereas the PostHC–RF3 is stabilized in the R state , consistent with the previous smFRET work ( Sternberg et al . , 2009 ) and structural studies ( Gao et al . , 2007; Jin et al . , 2011; Zhou et al . , 2012 ) . After peptide release , RF1 and RF3 or RF2 and RF3 together shift the distribution of ribosome conformations towards the middle of the dynamic spectrum ( Figure 7A ) . The rates of ribosome fluctuations are in the range of 2–6 s−1 , somewhat faster than in the absence of the factors , 0 . 5–2 . 6 s−1 ( Supplementary file 1 ) . One open question is what drives the dissociation of RF1 and RF3 from the ribosome ( Figure 7B ) . Dissociation of RF1 from the static N state is very slow . RF3 accelerates the dissociation , which correlates with increased ribosome dynamics and frequent transitions from N to R state . However , dynamic transitions alone are not sufficient to induce RF1 dissociation from the ribosome , because peptide release is crucial to allow RF1 to dissociate rapidly . Peptide release leads to a change in the orientation of RF1 with respect to L11 . On the other hand , peptide release alone is not sufficient , as the dissociation rate of RF1 from the PostHC is slow in the absence of RF3 . Thus , RF1 dissociation is promoted by the concerted action of RF3 , which stimulates subunit rotation and may directly displace RF1 from its original binding site , and by peptide release , which allows a conformational adjustment of RF1 . RF3 dissociation is independent of peptide release or the ribosome dynamics , but is affected by the presence of RF1 or RF2 , which stabilize RF3 binding to the ribosome and change conformation/position of RF3 relative to L11 . In the presence of RF1 , RF3 efficiently dissociates from the N state even in the absence of GTP hydrolysis ( this paper and [Shi and Joseph , 2016] ) . The order of RF1 and RF3 dissociation appears random , because the rates of factor release are quite similar and the exact sequence depends on experimental conditions ( this paper; [Koutmou et al . , 2014; Shi and Joseph , 2016] ) . In those cases where RF1 happens to dissociate before RF3 has left the ribosome , GTP hydrolysis completes RF3 recycling . In summary , subunit rotation , peptide release , conformational changes of the factors , and GTP hydrolysis together drive dissociation of RF1 and RF3 . However , kinetically these movements are not directly coupled , that is the dissociation rates of the factors and the rates of subunit rotation are independent of each other but are individually defined by the dynamic properties of the complex . Thus , translation termination is a stochastic process that utilizes loosely coupled motions of its players to complete protein synthesis and release the newly synthesized nascent chain toward its cellular destination . Our results lead to the following model of translation termination for RF1 ( Figure 7C ) . Among all possible reaction routes , two appear most likely , either via RF1 binding to PreHC , followed by peptide release and RF3–GTP recruitment , or through simultaneous binding of RF1 and RF3–GTP to PreHC followed by peptide release . The resulting complex PostHC–RF1–RF3–GTP can make rapid transitions between the N and R states . RF1 and RF3 change their relative positions and can now both rapidly dissociate from the ribosome . The order of events is not deterministic: multiple ribosome conformations , ribosome dynamics and the lack of strong coupling between the rates of subunit rotation and the dissociation of RF1 and RF3 seem characteristic features of RF1-dependent termination . This work provides an unexpected view on the role of nucleotide exchange and GTP hydrolysis by RF3 . Although RF3-GDP or apo-RF3 can bind to the ribosome carrying RF1/RF2 ( Peske et al . , 2014; Zavialov et al . , 2001 ) , this interaction does not result in the recruitment of the factor to its binding site at the vicinity of L11 . In vitro in the absence of GTP , apo-RF3 can form a relatively stable complex with PostHC–RF1 ( Pallesen et al . , 2013; Shi and Joseph , 2016 ) , but this binding does not alter the dynamics of subunit rotation and does not accelerate RF1 dissociation ( this paper and [Sternberg et al . , 2009] ) . Rather , the GTP-bound form of RF3 is required to stimulate ribosome dynamics and RF1 dissociation from PostHC . Given the moderate difference in the affinities of RF3 for GTP and GDP , at cellular concentrations a large fraction of RF3 is in the GTP form . Furthermore , given the high GTP association rate , apo-RF3 will be immediately converted into the functionally active GTP form ( Peske et al . , 2014 ) ; thus , the apo-RF3–ribosome complex can only be a transient intermediate . The present experiments , most of which are performed in the presence of a GTP regeneration system , which does not allow for accumulation of the GDP- or apo-form of RF3 , show efficient factor binding , peptide release and factor recycling . We thus have no indication for an active role of nucleotide exchange or for an essential role of the GDP- or the apo-form of RF3 in termination at cellular conditions and we consider the respective models unlikely . Unexpectedly , our data suggest that GTP hydrolysis or an authentic GTP-bound form of RF3 are required to release RF3 that is arrested on the ribosome in the absence of RF1 . At the first glance , the low dissociation rate of RF3–GDPNP from the ribosome appears to contradict the results of the experiments with RF3–GTP , which show that factor dissociation is not coupled to GTP hydrolysis ( Figure 3C , F ) . We hypothesize that upon binding to the ribosome , RF3 can either form an initial binding complex from which the factor can dissociate rapidly , or enter an engaged complex , from which RF3 can only dissociate after GTP hydrolysis ( Figure 6E ) . In principle , this should result in biphasic dissociation time courses of RF3-GTP with a second slow phase corresponding to the rate of GTP hydrolysis , which we did not observe . However , in the presence of GTP the fraction of RF3 molecules that enter the engaged state may be too small to capture . As RF3–GDPNP appears to have a lower affinity to the ribosome than RF3–GTP , the transient initial RF3–ribosome complex might be too short-lived to be detected and only the stable engaged complexes are captured . Alternatively , GDPNP , as well as GTPγS or the RF3 ( H92A ) mutant may induce a conformation that hinders RF3 from dissociation but is hardly populated in the presence of GTP; in this case , the effects are purely conformational and not due to GTP hydrolysis as such . Available structures of ribosome-bound RF3 suggest that RF3 is arrested on ribosomes in the R state ( Gao et al . , 2007; Jin et al . , 2011; Zhou et al . , 2012 ) . This could explain why PostHC , with its higher propensity to be in the R state than the PreHC , is more efficient in stimulating GTP hydrolysis by RF3 ( Zavialov et al . , 2002 ) . In this respect , RF3 appears to be an unusual GTPase that differs from other translational GTPases , such as EF-G , EF-Tu and IF2 , where GTP hydrolysis is coupled to key steps on the reaction pathway of the factors and is required on all ribosome complexes . Rather , the internal clock of the RF3 GTPase ( Peske et al . , 2014 ) acts as a rescue mechanism to release RF3 recruited to complexes that do not contain RF1 . This scenario is realistic at the concentrations of factors in the cell where RF3 is much more abundant than RF1 ( Schmidt et al . , 2016 ) . The smFRET data presented here for a simple model system present a starting point to study dynamics of more natural termination complexes containing long peptide nascent chains . While model termination systems are fully functional in all steps of termination and the rate of GTP hydrolysis by RF3 is similar with the fM-Stop and fMFTI-Stop termination contexts ( Zavialov and Ehrenberg , 2003 ) , the length of the nascent peptide and the nature of the P-site tRNA may attenuate the ribosome dynamics . While currently such complexes are biochemically too heterogeneous to study , further development of smFRET techniques toward multicolor detection and better time resolution may provide a tool to decipher the dynamics of these heterogeneous assemblies .
All smFRET experiments were performed in imaging buffer ( 50 mM Tris-HCl pH 7 . 5 , 70 mM NH4Cl , 30 mM KCl , 15 mM MgCl2 , 1 mM spermidine , 8 mM putrescine , 2 . 5 mM protocatechuic acid , 50 nM protocatechuate-3 , 4-dioxygenase ( from Pseudomonas ) , 1 mM Trolox ( 6-hydroxy-2 , 5 , 7 , 8-tetramethylchromane-2-carboxylic acid ) , and 1 mM methylviologen ) . Peptide hydrolysis experiments were performed in TAKM7 buffer ( 50 mM Tris-HCl pH 7 . 5 , 70 mM NH4Cl , 30 mM KCl , 7 mM MgCl2 ) . The preparation and functional characterization of ribosomes labeled with Cy3 at protein L11 and double-labeled at S6-Cy5 and L9-Cy3 was carried out as described ( Adio et al . , 2015; Sharma et al . , 2016 ) . E . coli strain lacking L1 were obtained from the Keio collection ( CGSC#: 12041 ) and ΔL1 ribosomes purified according to the protocol used for native ribosomes ( Rodnina and Wintermeyer , 1995 ) . A single cysteine was introduced at position T202 of L1 and the protein purified as described in Fei et al . ( 2008 ) . L1 ( T202C ) was fluorescence labeled with Cy5-maleimide ( GE Healthcare ) and purified using a 5 ml HiTrap SP HP cation exchange chromatography column ( GE Healthcare ) . ΔL1 ribosomes were reconstituted by incubation with a 5-fold molar excess of L1-Cy5 for 30 min at 37°C . Excess protein was removed by centrifugation through a 30% sucrose cushion in 50 mM Tris-HCl pH 7 . 5 , 70 mM NH4Cl , 30 mM KCl , 15 mM MgCl2 , pH 7 . 5 . The RF2 construct was cloned from the E . coli K12 strain and contains the natural T246A replacement ( Wilson et al . , 2000 ) . Catalytically impaired RF1 ( G234A ) ( RF1 ( GAQ ) ) and RF2 ( G251A ) ( RF2 ( GAQ ) ) , and the respective single-cysteine variant RF1 ( S167C ) ( Sternberg et al . , 2009; Wilson et al . , 2000 ) , RF2 ( C273 ) and RF3 ( L233C ) were generated by Quickchange mutagenesis according to the standard protocol . Native cysteines were replaced by serine or alanine based on the sequence conservation analysis performed using the Consurf database . RF1 and RF2 were purified and in vitro methylated as described ( Kuhlenkoetter et al . , 2011 ) . RF3 was purified by affinity chromatography on a Ni-IDA column ( Macherey-Nagel ) followed by ion exchange chromatography on a HiTrapQ column ( Peske et al . , 2014 ) . Prior to labeling , methylated RF1 and RF2 were incubated for 30 min with a 10-fold molar excess of TCEP ( Sigma ) at room temperature ( RT ) . Cy5 maleimide ( GE Healthcare ) was dissolved in DMSO and added to the proteins ( 5- to 10-fold molar excess ) . Labeling was performed for 2 hr at RT and quenched by addition of a 10-fold molar excess of 2-mercaptoethanol over dye . Excess dye was removed by gel filtration on a PD-10 column ( GE Healthcare ) . tRNAfMet was labeled at position s4U8 with Cy3-maleimide ( Fei et al . , 2010 ) and aminoacylated and purified as described ( Milon et al . , 2007 ) . All mRNAs used in the smFRET experiments are labeled with biotin at the 5´end and were purchased from IBA ( Göttingen , Germany ) . The following sequences were used: mMetStop 5′-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGUAAACGAUU-3′ mMetPheStop 5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGUUUUAAACGAUU-3‘ mMetLysStop 5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGAAGUAAACGAUU-3‘ mMetValStop 5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGGUUUAAACGAUU-3 ‘ For the peptide hydrolysis experiments , ribosome complexes were assembled on the synthetic model mRNA , 5’-GGCAAGGAGGUAAAUAAUGUAAACGAUU-3’ ( IBA ) with a start codon followed by a stop codon . Initiation complex formation was carried out by incubating ribosomes ( 100 nM ) with a three-fold excess of IF1 , 2 and 3 , fMet-tRNAfMet , mRNA and 1 mM GTP in TAKM7 for 30 min at 37°C . To form initiation complexes with fMet-tRNAfMet-Cy3 , equal amounts of ribosomes and tRNA were used . In case of the mRNA coding for fMetStop , the initiation complex was used as PreHC . To generate PreHC on other mRNAs , an equal volume of ternary complex was added containing EF-Tu ( 1 µM ) incubated with GTP ( 1 mM ) , phosphoenolpyruvate ( 3 mM ) and pyruvate kinase ( 0 . 1 mg/ml ) in TAKM7 for 15 min at 37°C , followed by addition of Phe-tRNAPhe , Lys-tRNALys or Val-tRNAVal ( 500 nM ) . Addition of EF-G ( 100 nM ) and GTP ( 1 mM ) induced tRNA translocation to form PreHC that contains peptidyl tRNA in the P site and displays the UAA stop codon in the A site . Complexes were diluted to 1 nM with smFRET buffer ( 50 mM Tris-HCl , 70 mM NH4Cl , 30 mM KCl , 15 mM MgCl2 , 1 mM spermidine and 8 mM putrescine ) . Biotin/PEG functionalized cover slips were incubated for 5 min at room temperature with the same buffer containing additionally BSA ( 10 mg/ml ) and neutravidin ( 1 µM ) ( Thermo Scientific ) . Excess neutravidin was removed by washing the cover slip with buffer containing BSA ( 1 mg/ml ) . Ribosome complexes were applied to the surface and immobilized through the mRNA-biotin:neutravidin interaction . Images were recorded at a rate of 30 frames/s after exchanging the buffer with imaging buffer at room temperature ( 22°C ) ( Adio et al . , 2015 ) . To monitor subunit rotation of L9/S6-labeled ribosomes in the presence of release factors at steady-state conditions , imaging buffer was supplemented with RF1 , RF2 and/or RF3 ( 1 µM each ) . In experiments with RF1 ( GAQ ) or RF2 ( GAQ ) , the observation time was limited to <10 min in order to minimize peptide hydrolysis due to residual factor activity . In experiments monitoring subunit rotation by RF3 in the GTP form or in complex with non-hydrolysable GTP analogs , imaging buffer was additionally supplemented with the energy recycling system ( 1 mM GTP or 1 mM GDPNP or 1 mM GTPγS , 3 mM phosphoenolpyruvate and 0 . 1 mg/ml pyruvate kinase ) . FRET signals reporting on the time course of subunit rotation during termination were obtained by injecting RF1 or RF2 ( 100 nM ) in imaging buffer to immobilized PreHC or PostHC . To measure FRET signals reporting on the residence time of labeled release factors on PreHC or PostHC labeled at protein L11 with Cy3 , the complexes were immobilized on the cover slip . Movies were recorded upon addition of Cy5-labeled RF1 , RF2 or RF3 to a final concentration of 10 nM in imaging buffer . To study the residence time of Cy5-labeled RF3 or to study the residence time of Cy5-labeled RF1 or RF1 ( GAQ ) on ribosomes in the presence of unlabeled RF3 , imaging buffer was supplemented with unlabeled RF3 ( 1 µM ) , GTP ( 1 mM ) , phosphoenolpyruvate ( 3 mM ) and pyruvate kinase ( 0 . 1 mg/ml ) . To study the residence time of Cy5-labeled RF3 in the presence of RF1 , imaging buffer was supplemented with unlabeled RF1 ( 1 µM ) , GTP ( 1 mM ) , phosphoenolpyruvate ( 3 mM ) and pyruvate kinase ( 0 . 1 mg/ml ) . To monitor FRET signals reporting on the conformation of the P-site tRNA PreHC or PostHC labeled on protein L1 ( C202-Cy5 ) and on fMet-tRNAfMet ( thioU8-Cy3 ) or tRNAfMet ( U8-Cy3 ) were immobilized on the coverslip . Movies were recorded upon addition of imaging buffer or imaging buffer containing RF3 ( 1 µM ) . In experiments with RF3 imaging , buffer was additionally supplemented with the energy recycling system ( 1 mM GTP or 1 mM GDPNP or 1 mM GTPγS , 3 mM phosphoenolpyruvate and 0 . 1 mg/ml pyruvate kinase ) ( Sternberg et al . , 2009 ) . Fluorescence time courses for donor ( Cy3 ) and acceptor ( Cy5 ) were extracted as described ( Adio et al . , 2015; Fei et al . , 2008; Roy et al . , 2008 ) . A semi-automated algorithm ( Matlab ) was used to select anti-correlated fluorescence traces ( correlation coefficient <0 . 1 ) exhibiting characteristic single fluorophore fluorescence intensities ( Adio et al . , 2015 ) . Time traces for further analysis were selected from the dataset by choosing only those traces that contained single photobleaching steps for Cy3 and Cy5 ( as recommended in [Fei et al . , 2008] ) . The bleed-through of the Cy3 signal into the Cy5 channel was corrected using an experimentally determined coefficient ( ~0 . 13 in our experimental system [Adio et al . , 2015] ) . All trajectories were smoothed over three data points . FRET efficiency was defined as the ratio of the measured emission intensities , Cy5/ ( Cy3 +Cy5 ) ( Roy et al . , 2008 ) . FRET-histograms were fitted to Gaussian distributions using Matlab code ( Adio et al . , 2015 ) . Mean FRET values ( mean ±sd ) and population distribution ( p=area under the curve ± sd ) were calculated from three independent datasets and are summarized in Supplementary file 1 . The vbFRET software package ( http://vbfret . sourceforge . net/ ) ( Bronson et al . , 2009 ) was used for hidden Markov model ( HMM ) analysis of the FRET data . Time trajectories with only one transition per trace and with the FRET changes of less than 0 . 1 were excluded from further kinetic analysis ( Fei et al . , 2008; Sternberg et al . , 2009 ) . Individual time-resolved FRET traces were compiled into FRET probability density plots ( contour plots ) ( Blanchard et al . , 2004; Munro et al . , 2007 ) . For the experiments measuring subunit rotation of PostHC upon binding of RF1 in real time , FRET traces were synchronized at the transition to the stable N state . For the experiments measuring subunit rotation of PreHC upon binding of RF2 in real time , FRET traces were synchronized to the first N to R transition . In experiments measuring the residence time of labeled release factors , FRET traces are synchronized to the beginning of the FRET event reporting on the binding of the factor to the ribosome . One-dimensional histograms at the right side of the contour plots summarize FRET values of the first 10–30 time frames ( 0 . 3–1 . 0 s ) of the FRET signals . The photobleaching rates of the S6/L9-FRET pair were estimated as described ( Adio et al . , 2015 ) from the non-fluctuating 0 . 7 FRET trajectories obtained with PreHC , PreHC-RF1 ( GAQ ) and PostHC-RF1 , as well as from the non-fluctuating 0 . 5 FRET trajectories of PostHC*-RF3 ( GTP ) ; the photobleaching rates were in the range of 0 . 07–0 . 19 s−1 , comparable to 0 . 05–0 . 3 s−1 in ( Sternberg et al . , 2009 ) . Dwell times of individual FRET states in traces with multiple FRET states were calculated from idealized traces ( Bronson et al . , 2009 ) . Dwell time histograms were fitted to either one- or two-exponential function . Rates ( k ) were calculated by taking the inverse of dwell times . Rate constants ± standard deviations were determined from three independent datasets as described in Fei et al . ( 2011 ) ; Sternberg et al . ( 2009 ) ; Wasserman et al . ( 2016 ) and summarized in Supplementary file 1 . PreHC was prepared as described ( Peske et al . , 2014 ) and purified through sucrose cushion centrifugation . After centrifugation , ribosome pellets were resuspended in TAKM7 , frozen in liquid nitrogen and stored at −80°C . The extent of initiation was better than 95% as determined by nitrocellulose filtration and radioactive counting . PreHC ( 100 nM ) was incubated with RF3 at the indicated concentration and nucleotide ( 1 mM ) for 15 min at 37°C . Pyruvate kinase ( 0 . 1 mg/ml ) and phosphoenol pyruvate ( 3 mM ) were added in all experiments performed in the presence of GTP . Time courses were started by addition of RF1 or RF2 ( 10 nM ) . Samples were quenched with a solution containing TCA ( 10% ) and ethanol ( 50% ) . After centrifugation ( 30 min , 16 , 000 g ) , the amount of released f[3H]Met in the supernatant was quantified by radioactive counting . | Inside cells , molecular machines called ribosomes make proteins using messenger RNA as a template . However , the template contains more than just the information needed to create the protein . A ‘stop codon’ in the mRNA marks where the ribosome should stop . When this is reached a group of proteins called release factors removes the newly made protein from the ribosome . Bacteria typically have three types of release factors . RF1 and RF2 recognize the stop codon , and RF3 helps to release RF1 or RF2 from the ribosome so that it can be recycled to produce another protein . It was not fully understood how the release factors interact with the ribosome and how this terminates protein synthesis . Adio et al . used TIRF microscopy to study individual ribosomes from the commonly studied bacteria species Escherichia coli . This technique allows researchers to monitor movements of the ribosome and record how release factors bind to it . The results of the experiments performed by Adio et al . show that although RF1 and RF2 are very similar to each other , they interact with the ribosome in different ways . In addition , only RF1 relies upon RF3 to release it from the ribosome; RF2 can release itself . RF3 releases RF1 by forcing the ribosome to change shape . RF3 then uses energy produced by the breakdown of a molecule called GTP to help release itself from the ribosome . Most importantly , the findings presented by Adio et al . highlight that the movements of ribosomes and release factors during termination are only loosely coupled rather than occur in a set order . Other molecular machines are likely to work in a similar way . The results could also help us to understand the molecular basis of several human diseases , such as Duchenne muscular dystrophy and cystic fibrosis , that result from ribosomes not recognizing stop codons in the mRNA . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Dynamics of ribosomes and release factors during translation termination in E. coli |
Pseudogenes are thought to be inactive gene sequences , but recent evidence of extensive pseudogene transcription raised the question of potential function . Here we discover and characterize the sets of mouse lncRNAs induced by inflammatory signaling via TNFα . TNFα regulates hundreds of lncRNAs , including 54 pseudogene lncRNAs , several of which show exquisitely selective expression in response to specific cytokines and microbial components in a NF-κB-dependent manner . Lethe , a pseudogene lncRNA , is selectively induced by proinflammatory cytokines via NF-κB or glucocorticoid receptor agonist , and functions in negative feedback signaling to NF-κB . Lethe interacts with NF-κB subunit RelA to inhibit RelA DNA binding and target gene activation . Lethe level decreases with organismal age , a physiological state associated with increased NF-κB activity . These findings suggest that expression of pseudogenes lncRNAs are actively regulated and constitute functional regulators of inflammatory signaling .
Large scale transcriptome analyses has revealed that three quarters of the human genome may be expressed ( Djebali et al . , 2012 ) , much of it as noncoding RNA ( ncRNA ) . Over the past several years , the literature describing the functions of long noncoding RNA ( lncRNA ) has exploded with detailed reports demonstrating that lncRNA can regulate neural development ( Feng et al . , 2006; Bond et al . , 2009; Rapicavoli et al . , 2010; Rapicavoli et al . , 2011 ) , differentiation ( Dinger et al . , 2008; Guttman et al . , 2009; Loewer et al . , 2010; Guttman et al . , 2011 ) , epigenetic marks on chromatin ( Rinn et al . , 2007; Tsai et al . , 2010; Wang et al . , 2011 ) and transcription factor signaling ( Willingham et al . , 2005; Kino et al . , 2010; Gomez et al . , 2013 ) . In addition , the human genome was found to contain over 11 , 000 pseudogenes , of which 833 were expressed and associated with active chromatin ( The ENCODE Project Consortium , 2012 ) . Pseudogenes have traditionally been defined as ancestral copies of protein coding genes that arise from a gene duplication event or by a retrotransposition event that is followed by subsequent accumulation of mutations that render the pseudogene transcriptionally inactive . More recent studies have revealed that many pseudogenes are expressed as lncRNAs and can have a role in gene silencing ( Duret et al . , 2006 ) , and cancer ( Poliseno et al . , 2010; Kalyana-Sundaram et al . , 2012 ) . The pro-inflammatory cytokine TNFα acts through the transcription factor NF-κB to play a key role in innate and adaptive immunity , inflammation , apoptosis , and aging . Dysregulation of NF-κB plays an important role in the pathophysiology of inflammatory disease , when proinflammatory cytokines drive NF-κB activation , which in turn drives production of proinflammatory cytokines . The NF-κB family is composed five proteins in mammals: RelA ( p65 ) , RelB , c-Rel , p50 ( NF-κB1 ) , and p52 ( NF-κB2 ) . NF-κB family members form dimers , either as heterodimers or homodimers , which can act to positively or negatively regulate target gene expression . Under normal conditions NF-κB dimers are sequestered in the cytoplasm by binding to IκB proteins . Activation by binding of ligand to a receptor on the cell surface leads to a signaling cascade which leads to phosphorylation , rapid ubiquitination , and degradation of the IκB proteins . This reveals a nuclear localization sequence on NF-κB . NF-κB is then translocated to the nucleus where it binds DNA to activate or repress transcription . Importantly , only Rel family members ( RelA , RelB , and c-Rel ) can activate transcription . p50 and p52 can form heterodimers with RelA family members to activate transcription , or alternatively form homodimers to compete for NF-κB binding sites reviewed in Hayden and Ghosh ( 2012 ) . Here , we use paired-end directional sequencing to identify the effects of TNFα stimulation on the entire transcriptome of mouse embryonic fibroblast ( MEF ) cells . TNFα regulates the transcription of 3596 protein coding genes , 48 annotated lncRNAs , 54 pseudogene lncRNAs , and 64 de novo lncRNAs . We validate a subset of these lncRNAs , and classify them by response to various microbial components and proinflammatory cytokines , dependence on RelA and subcellular localization . We identify an lncRNA pseudogene , Lethe ( named after the mythological river of forgetfulness , for its role in negative feedback ) , which is expressed in response to proinflammatory cytokines TNFα and IL-1β , and the anti-inflammatory agent , dexamthasone , but is not responsive to microbial components , and is primarily found on the chromatin . Lethe is regulated by RelA , independent of pseudogene family members and proximal genes . Additionally , Lethe is dramatically downregulated in aged spleen . Finally , Lethe binds directly to RelA to inhibit NF-κB DNA binding activity . These findings suggest that Lethe may function as a novel negative regulator of NF-κB , to help fine tune the inflammatory response .
We hypothesized that NF-κB regulates the expression of lncRNAs just as it regulates the expression of coding genes and microRNAs ( Boldin and Baltimore , 2012 ) . To determine whether NF-κB regulates the expression of lncRNAs , we performed paired-end directional RNA-seq on wildtype ( WT ) MEFs before treatment and after treatment for 1 . 5 , 6 and 24 hr with 20 ng/ml of TNFα . On average , more than 20 million reads were mapped to the mouse genome ( mm9 assembly ) for each treatment condition ( Supplementary file 1A ) . First , reads were mapped to the mouse mm9 reference genome using TopHat ( Trapnell et al . , 2009 ) . Using an in-house generated script , RefSeq and Ensemble annotated transcripts’ expression in the form of RPKM ( reads per kilobase of exon model per million ) were obtained and those transcripts with at least a twofold change in expression and an average RPKM > 1 , were defined as significant . Reference based de novo transcriptome assembly of mapped reads was performed using two methods . Raw reads were mapped using TopHat and de novo transcriptome assembly of mapped reads was performed using , Cufflinks ( Trapnell et al . , 2010 ) and Scripture ( Guttman et al . , 2010 ) in parallel . RefSeq and Ensemble annotated transcripts were downloaded from UCSC table browser , and these annotated transcripts were filtered out from Scripture and Cufflinks-assembled transcriptomes to yield about 1500 novel de novo isoforms that are expressed at an RPKM > 1 . Because many isoforms mapped to a single locus , we further filtered the list of novel transcripts by applying promoter regions as defined by H3K4me3 via chromatin immunoprecipitation sequencing ( ChIP-Seq ) , which yielded 184 novel loci . To further refine the candidate transcripts , we extracted the raw reads that mapped to those 184 loci , processed a de novo transcript assembly through Trinity ( Grabherr et al . , 2011 ) and determined the Coding Potential Calculator ( CPC ) score of each transcript ( Kong et al . , 2007 ) to identify 64 novel de novo lncRNAs ( Figure 1A , Supplementary file 1B ) . 10 . 7554/eLife . 00762 . 003Figure 1 . TNFα regulates the transcription of many coding and noncoding genes . ( A ) Workflow for strategy for discovery of NF-κB regulated lncRNAs . ( B ) 3596 RefSeq protein coding genes are regulated by TNFα . Values are normalized to the 0 hr time point . ( C ) 244 RefSeq ncRNAs are regulated by TNFα . ( D ) 64 de novo lncRNAs are regulated by TNFα . ( E ) The fraction of all RefSeq ncRNAs for each class of transcript . ( F ) 54 pseudogene lncRNAs are regulated by TNFα . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 00310 . 7554/eLife . 00762 . 004Figure 1—Figure supplement 1 . Heatmap of RefSeq genes . Mean centered heatmap of RefSeq protein coding genes and RefSeq lncRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 004 In this way , 3596 protein coding transcripts , 244 ncRNAs and 64 de novo lncRNAs were detected in these experiments . Many RefSeq protein coding genes that had been shown to be regulated by NF-κB were induced by TNFα including Gadd45b , Sod2 , Nfkbia , Relb , Cdkn2a , and Il6 . Additionally , the RNA-seq data showed the expected oscillatory gene expression pattern of NF-κB dependent gene expression , and notably the dynamic range by RNA-seq is greater than previously observed with microarrays ( Kawahara et al . , 2011 ) . Interestingly , the 244 RefSeq ncRNAs showed a similar expression pattern as the protein coding genes , where peak expression or repression levels were observed at 1 . 5 and 24 hr post stimulation . A subset also showed maximal repression at 6 hr in both classes . 84 ncRNAs were at least twofold upregulated when compare to untreated at a least one time point . In contrast , the vast majority ( 59 out of 64 ) of de novo lncRNAs were primarily repressed upon TNFα treatment ( Figure 1B–D , Supplementary file 1B–D ) . A similar result in which most de novo lncRNAs were down regulated after treatment was seen in response to estrogen in breast cancer cells ( Hah et al . , 2011 ) . Next , we divided the RefSeq ncRNAs by their RefSeq annotation into four classes , pri-miRNAs ( 40% ) , RNaseP , SnoRNA , ScaRNA ( 19% ) , pseudogene lncRNA ( 22% ) and annotated lncRNAs ( 19% ) ( Figure 1E ) . Interestingly , only 11 of 96 pri-miRNAs were upregulated with TNFα treatment . In contrast about 23 of 45 housekeeping RNAs ( RNaseP , scaRNA , snoRNA ) and 37 of 54 pseudogene lncRNAs were upregulated . Finally , 12 out of 48 annotated lncRNA were upregulated , mirroring what we see in the de novo lncRNAs . To further examine the pseudogene component of the lncRNAs , we created a heatmap of pseudogene lncRNA , and observed the same oscillatory gene expression pattern that was observed in the protein coding genes ( Figure 1F and Supplementary file 1E ) . We determined that the pseudogene Rps15a-ps4 ( herein named Lethe ) , had the highest expression changes of any pseudogene with an RPKM > 1 . Additionally , we observed that Gapdh had seven pseudogenes that were identified as induced by TNFα , but we were unable to validate this result with qRT-PCR . We selected Refseq genes with significant differential expression over the time course ( FDR < 0 . 05 , SAMseq ) and varied by at least twofold , yielding 3690 significant transcripts ( Supplementary file 1F ) . We organized their patterns of temporal expression by mean-centered hierarchical clustering ( Figure 1—figure supplement 1 ) , and determined which lncRNAs clustered with known NF-κB regulated genes . From our list , we chose to validate and further characterize Cox2 Divergent , Gp96 Convergent , H2-T23/24AS , HoxA11AS , Lethe , Pbrm1 Convergent , Scripture 16 , 612 and Scripture 60 , 588 . Our directional paired-end RNA-Seq data revealed TNFα regulation of many lncRNA transcripts which include divergent , antisense , convergent , and intergenic transcripts . Since the functional relationship between genomic organization and expression is unknown , we chose to validate and further characterize the lncRNAs expression alongside the closest protein coding gene under a variety of different stimuli by qRT-PCR . In our subset of lncRNAs , we found that most lncRNAs are co-regulated with their protein-coding gene . This is not surprising since we chose to validate lncRNAs that were close to genes that were regulated by TNFα . One notable exception was Lethe . Although the Lethe is expressed from the same strand as Gmeb1 and close to the 3′ terminus of Gmeb1 , qRT-PCR showed that Lethe was specifically induced by TNFα and IL-1β , whereas Gmeb1 expression did not significantly change , confirming that the Lethe is not an extension of the 3′ UTR of Gmeb1 ( Figure 2A ) . 10 . 7554/eLife . 00762 . 005Figure 2 . LncRNAs distinguish between different stimuli and are regulated by NF-κB . ( A ) Validation of lncRNAs expression alongside the closest protein coding gene under a variety of different stimuli by qRT-PCR . Genomic organization is shown below . MEFs were treated with 20 ng/ml TNFα for 0 and 6 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is normalized to Actin levels ( mean ± SD ) . ( B ) LncRNAs are regulated by RelA . qRT-PCR in WT and RelA−/− littermate cells under a variety of different stimuli . MEFs were treated with 20 ng/ml TNFα for 0 and 6 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is normalized to Actin levels ( mean ± SD ) . ( C ) Endogenous RelA is recruited to the promoters of lncRNAs . MEFs were treated with 20 ng/ml TNFα for 0 and 15 min . ChIP with α-RelA antibodies was performed and RelA percent enrichment relative to input is shown ( mean ± SD , Nkfbia , p<0 . 0518; Gp96 Convergent , p<0 . 007; Lethe , p<0 . 002 ) . ( D ) LncRNAs are found throughout the cell . Cellular fractionation was performed and fraction found in the chromatin , nucleus and cytoplasm is shown . MEFs were treated with 20 ng/ml TNFα for 6 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown ( mean ± SD is shown ) . ( E ) LncRNAs are found on the chromatin . MEFs were treated with 20 ng/ml TNFα for 6 hr . RNA-IP with α-H3 antibodies was performed . RNA was isolated and Quantitative Taqman real time RT-PCR of the indicated RNAs is shown ( mean ± SD , Lethe p<0 . 004 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 00510 . 7554/eLife . 00762 . 006Figure 2—figure supplement 1 . ( A ) Western analysis of RelA protein levels . ( B ) PolyA+ Lethe is found on the chromatin . Immunoblot of wildtype and RelA−/− MEFs . Cellular fractionation was performed , total RNA was purified and polyA+ selection was performed . The fraction polyA+ RNA found in the chromatin , nucleus and cytoplasm is shown . MEFs were treated with 20 ng/ml TNFα for 6 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown ( mean ± SD is shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 006 NF-κB signaling can be initiated in response to many different stimuli including in response to proinflammatory cytokines like TNFα and IL-1β as well as in response to microbial components via the TLR family ( Hayden and Ghosh , 2012 ) . Therefore , we validated our lncRNA candidates in response to TNFα , IL-1β , and agonists of Toll like receptors ( TLR ) 1 , 2 , 3 , 4 , or 7 ( Figure 2A–B ) . TLRs are pattern-recognition receptors for pathogen components from bacteria , fungi , or viruses , and play key roles in controlling the innate and adaptive immunity ( Kawai and Akira , 2010 ) . Indeed , we found that many of the lncRNAs are upregulated in response to distinct stimuli . Most notable , Cox2 Divergent is upregulated in response to proinflammatory cytokines and TLR1-4 agonists . In contrast , Gp96 Convergent is only expressed in response to TNFα . H2-T32/24AS is responsive to only TNFα and TLR3 agonists , whereas HoxA11AS is expressed in response to TLR3 agonists and actually down regulated by TNFα stimulation . Lethe is upregulated in response to the proinflammatory cytokines TNFα and IL-1β , but not TLR agonists , indicating it may have a function in inflammation , but not in native immunity . Pbrm1 Convergent is highly upregulated in response to IL-1β , and to a lesser extent TNFα , and TLR4 and 7 agonists . These results demonstrate that lncRNAs are dynamically and specifically regulated in response to different stimuli , suggesting that the pattern of lncRNAs can serve as an internal representation of a cell’s exposure to distinct inflammatory and pathogenic signals . Next , we wanted to determine if the lncRNAs were directly regulated by NF-κB . To address this question , we used two different methods . First , we performed qRT-PCR in RelA−/− littermate cells alongside WT cells ( Figure 2—figure supplement 1 ) . Cox2 Divergent is dramatically upregulated in response to proinflammatory cytokines and TLR1-4 agonists in WT and to an even larger extent , in RelA−/− cells indicating that it is not directly regulated by NF-κB component RelA . Additionally , HoxA11AS , H2-T23/24AS , Gp96 Convergent and Scripture 16 , 612 all show some induction in RelA−/− cells . In contrast , induction of Lethe , Pbrm1 Convergent and Scripture 60 , 588 is largely abrogated in RelA−/− cells , indicating that RelA is required to induce these lncRNAs ( Figure 2B ) . Second , we performed RelA chromatin immunoprecipitation ( ChIP ) in WT MEFs . Upon TNFα signaling , RelA was found to bind to the promoters of Nfkbia , Gp96 Convergent and Lethe , but was not detected on the promoters of Cox2 Divergent or Pbrm1 Convergent , or Dll1 , a negative control ( Figure 2C ) . These results indicate that Lethe and Gp96 Convergent are directly transcriptional targets of RelA , with Lethe being particularly dependent on RelA for induction . To determine where our candidate genes are located within the cell we performed subcellular fractionation on cells stimulated with TNFα for 6 hr . We found that the subcellular distribution of TNFα−induced lncRNAs vary in a transcript-specific manner . Gapdh was tested as a control and found to be evenly distributed between the nucleus and cytoplasm with little transcript found on the chromatin . Likewise , Cox2 Divergent , Gp96 Convergent , and H2-T23/24AS were evenly distributed between nucleus and cytoplasm . HoxA11AS was mostly nuclear with some transcript found in the cytoplasm , but not on the chromatin . Interestingly , Lethe , Pbrm1 Convergent , Scripture 16 , 612 and Scripture 60 , 588 were found mostly on the chromatin , with a smaller fraction in the nucleus . These results indicate that Lethe , Pbrm1 , Scripture 16 , 612 and Scripture 60 , 588 may be directly involved in gene regulation by interacting with the chromatin ( Figure 2D ) . To further determine if it is the nascent transcript or processed transcript that is found on the chromatin , we performed polyA selection on our subcellular fractions and analyzed the results by qRT-PCR . Interestingly , polyadenylated Lethe RNA is still preferentially associated with chromatin compared to two control mRNAs , Actin and Nfkbia ( Figure 2—figure supplement 1 ) , indicating that full length Lethe is associated with chromatin . Finally , we performed H3-RNA immunoprecipitation ( RIP ) to directly test if lncRNAs are found on the chromatin after cells were stimulated with TNFα for 6 hr . We found that Lethe and Pbrm1 Convergent are both found on the chromatin , while Cox2 Divergent and Gp96 Convergent were not detected , confirming our fractionation results ( Figure 2E ) . Results from Figure 2 are summarized in Table 1 . 10 . 7554/eLife . 00762 . 007Table 1 . Classification of TNF regulated lncRNAsDOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 007TNFαIL-1βTLR1TLR2TLR3TLR4TLR5TLR7RelA-depH3-RIPCox2 Divergent++++++n . d . n . d . −n . d . Gp96 Convergent+n . d . n . d . n . d . n . d . n . d . n . d . n . d . +n . d . H2-T23/24AS+n . d . n . d . n . d . +n . d . n . d . n . d . −−HoxA11AS−−−−+−−−−−Lethe++−−−−−−++Pbrm1 Convergent++−−−−−+−/++Scripture 16 , 612−+++−−−−n . d . n . t . Scripture 60 , 588−−−−−−−−n . d . n . t . These results demonstrate that lncRNAs are regulated by diverse and specific stimuli . Additionally , lncRNAs are directly regulated by NF-κB . Finallyp , the subcellular distribution of TNFα−induced lncRNAs varies by transcript . n . d . , not detectable . n . t . , not tested . Since Lethe is an Rps15a pseudogene , we wanted to determine if there are other pseudogenes that are directly regulated by TNFα . We obtained Taqman probes against non-repetitive sequences unique to each pseudogene member , and examined Rps15a pseudogene family members as well as Rps15a to determine if they are regulated by TNFα ( Figure 3A , Figure 3—figure supplement 1 ) . qRT-PCR analysis showed that Nfkbia ( a positive control ) is dynamically regulated in response to TNFα , as is Lethe . Rps15a and Rps15a-ps6 , another Rps15a pseudogene lncRNA , are both transcribed but are not regulated by TNFα ( Figure 3B ) . 10 . 7554/eLife . 00762 . 008Figure 3 . Lethe is a pseudogene lncRNAs that is regulated by NF-κB , Glucocorticoid Receptor and in aging . ( A ) Gene structure , homology and Taqman probe design of Rps15a and pseudogene family members . ( B ) Lethe is induced by TNFα , but other family members are not . MEFs were treated with 20 ng/ml TNFα for 0 and 1 . 5 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown normalized to Actin levels ( mean ± SD , p<0 . 012 ) . ( C ) Genomic organization of Lethe with RNA-Seq data at time 0 , 1 . 5 , 6 and 24 hr post TNFα treatment . Lethe is located on mouse chromosome 4 between Gmeb1 and Ythd2 . Gmeb1 and Ythd2 are not induced by TNFα stimulation . ( D ) Lethe is induced by dexamethasone treatment , but not other nuclear hormone receptor agonists . MEFs were treated with either 10 nM vitamin D , 100 nM methyltrienolone , 100 nM estradiol , or 1 μM dexamethasone for 0 and 6 hr . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown normalized to Actin levels ( mean ± SD , p<0 . 003 ) . ( E ) Lethe is down-regulated in aged mice . Lethe is expressed in young spleen from male and female mice . Five mice were used for each sex and time point . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown normalized to Actin levels ( mean ± SD , Lethe p<0 . 001 , Gmeb1 p<0 . 003 ) . ANOVA analysis was performed to determine significance . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 00810 . 7554/eLife . 00762 . 009Figure 3—figure supplement 1 . Alignment of Lethe with Rps15a-ps6 . ClustalW2 alignment was performed on Lethe ( Rps15a-ps4 ) and Rps15a-ps6 . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 009 Lethe is 697 bp long unspliced lncRNA , and its locus on chromosome four lays approximately 500 bp downstream of Gmeb1 and 8 kb upstream of Ythdf2 on the minus strand ( Figure 3C ) . Lethe is dramatically induced upon TNFα stimulation at 1 . 5 hr and 24 hr and repressed at 6 hr , in an expression pattern that is characteristic of other NF-κB regulated transcripts . Importantly , expression of its two neighbor mRNA genes was not changed by TNFα stimulation ( Figure 3C ) , indicating that Lethe is independently regulated . It is known that glucocorticoid receptor ( GR ) and NF-κB share many target gene sites ( Rao et al . , 2011 ) . Therefore we tested whether Lethe could be induced upon stimulation with a number of nuclear hormone agonists including the GR agonist , dexamethasone . We found that Lethe is upregulated in response to anti-inflammatory agent , dexamethasone , but not in response to other nuclear hormone receptor agonists examined , including Vitamin D ( Vitamin D Receptor ) , methyltrienolone ( Androgen Receptor ) and estradiol ( Estrogen Receptor ) ( Figure 3D ) . Thus , Lethe is a pseudogene lncRNA that is induced by both inflammatory stimuli and an anti-inflammatory therapeutic . Recent work has shown that the transcription factor binding motif most strongly associated with aging is NF-κB ( Adler et al . , 2007 ) . To determine if Lethe is expressed in old tissue as a result of constant NF-κB signaling , we tested a panel of tissues , including liver , lung , kidney , skin , spleen , cortex ( brain ) , and skeletal muscle . Lethe is expressed in male and female spleen , but not detectable in other tissues . Interestingly , Lethe is downregulated with age: 20-fold and 160-fold in males and females respectively ( Figure 3E ) . Neither Nfkbia nor Gmeb1 expression changes with age or sex . We hypothesized that Lethe acts in trans to regulate NF-κB function . Therefore we performed loss-of-function experiments and measuring expression of canonical NF-κB members . We used chemically modified chimeric antisense oligonucleotides ( ASO ) which have been shown to be effective at knocking-down expression of nuclear ncRNAs ( Ideue et al . , 2009 ) . ASO blockade inhibited the TNFα induction of Lethe , and we monitored the induction of two NF-κB target genes by qRT-PCR . Nfkbia level was significantly higher than TNFα stimulated ASO control in one of the two ASOs tested , while Nfkb2 was induced significantly for both ASOs tested ( Figure 4A ) . This result indicates that Lethe may act as a repressor of NF-κB activity . 10 . 7554/eLife . 00762 . 010Figure 4 . Lethe Binds to RelA and inhibits RelA occupancy of DNA . ( A ) Increased expression of NF-κB regulated genes in Lethe knockdown cells . Quantitative Taqman real time RT-PCR of the indicated RNAs is shown normalized to Actin levels ( mean ± SD , p<0 . 05 is shown ) ( B ) Lethe inhibits TNFα induced reporter gene expression . RLU of 3x-κB reporter activity and mutant reporter activity ( mean ± SD , p<0 . 05 is shown ) in CMV_Lethe transfected 293T cells . Reporter constructs are diagrammed above . ( C ) Endogenous RelA recruitment to the promoters of target genes is reduced in the presence of Lethe . 293T expressing CMV_GFP or CMV_Lethe were treated with 20 ng/ml TNFα for 15 min . ChIP with α-RelA antibodies was performed and RelA percent enrichment relative to input is shown ( mean ± SD; Il6 , p<0 . 033; Sod2 , p<0 . 001; Il8 , p<0 . 003; Nfkbia , p<0 . 015 ) . ( D ) Lethe binds to RelA . MEFs were treated with 20 ng/ml TNFα for 6 hr . RNA-IP with α-RelA antibodies was performed . RNA was isolated and Quantitative Taqman real time RT-PCR of the indicated RNAs is shown ( mean ± SD , p<0 . 020 ) . ( E ) Lethe expression blocks DNA binding of the RelA homodimer to its target . NF-κB EMSA of GFP or Lethe transfected 293T nuclear extracts treated with 20 ng/ml TNFα for 15 min . Extracts were pretreated with unlabeled NF-κB ( specific ) or CREB ( nonspecific competitor ) , or α-RelA antibodies for 15 min prior to incubation with probe . ( F ) Model for Lethe regulation of gene expression . Upon addition of TNFα or dexamthasone , Lethe is transcribed . Lethe can then bind to RelA–RelA homodimers and block binding to other NF-κB response elements , inhibiting NF-κB . DOI: http://dx . doi . org/10 . 7554/eLife . 00762 . 010 Conversely , we overexpressed Lethe or GFP in the presence of an NF-κB luciferase reporter gene after TNFα stimulation . Lethe expression , but not GFP expression , repressed NF-κB reporter gene activity . Additionally , Lethe can increase the repression NF-κB luciferase reporter gene expression in a dose dependent manner . To determine if Lethe’s effect on reporter gene activity was specific to NF-κB mediated reporter gene expression , we mutated the κB binding sites out of the reporter plasmid . As expected , the repression was no longer observed , indicating that Lethe requires NF-κB to repress reporter gene expression ( Figure 4B ) . The TNFα inducible repression of NF-κB luciferase reporter gene expression indicates that Lethe may affect the ability of RelA to bind to target promoters . To test this possibility , 293T cells were transfected with Lethe and ChIP was performed with RelA antibodies or IgG . In response to TNFα , Lethe expression significantly decreased RelA occupancy of several NF-κB target genes including Il6 , Sod2 , Il8 , and Nfkbia ( Figure 4C ) . Immunoblot analysis confirmed that that Lethe does not lower RelA protein level ( Figure 4C ) . These results indicate that Lethe acts to inhibit NF-κB binding to the chromatin . We reasoned that Lethe may bind RelA directly . RelA-RIP retrieved Lethe , but not Gapdh in MEFs stimulated with TNFα for 6 hr . Interestingly , other lncRNAs , Cox2 Divergent and Gp96 Convergent did not bind to RelA ( Figure 4D ) . To further explore the relationship between Lethe and RelA , NF-κB DNA binding was assessed by electro mobility shift assays ( EMSA ) . 293T cells were transfected with either CMV_GFP or CMV_Lethe . 48 hours post transfection cells were stimulated with TNFα for 15 min before nuclear lysates were prepared . As expected , TNF-stimulated extracts contained NF-κB activity , which are shifted by radiolabelled NF-κB probes , specifically competed away by cold NF-κB probes , and supershifted by anti-RelA antibody . Notably Lethe expression blocks DNA binding of the RelA homodimer , but not other isoforms ( Figure 4E ) . These results indicate that Lethe may act as an inhibitor of NF-κB by binding directly to the RelA homodimer , and blocking RelA’s ability to bind DNA ( Figure 4F ) .
Recent large scale RNA-Seq experiments have revealed that lncRNAs are dynamically expressed in normal tissues through development , differentiation , in response to different stimuli and as an organism ages ( Guttman et al . , 2009; Hah et al . , 2011; Guttman et al . , 2011; Chang et al . , 2013 ) . However , most large scale sequencing experiments discard the pseudogene transcriptional contribution . In depth analyses have revealed that pseudogenes are important drivers and suppressors of human cancers ( Poliseno et al . , 2010; Tay et al . , 2011; Kalyana-Sundaram et al . , 2012 ) . In this study we performed paired-end directional sequencing to identify novel transcripts that are regulated by TNFα signaling . We identified 48 annotated lncRNAs , 64 de novo lncRNAs and 54 pseudogene lncRNAs that are differentially regulated with TNFαstimulation . While prior studies have reported lncRNA induction by endotoxin ( Guttman et al . , 2009; De Santa et al . , 2010 ) , the specificity of the response and requirement of NF-κB were not known . Here , we have characterized a number of transcripts that are induced by a panel of microbial components and inflammatory cytokines . Notably , many lncRNAs are uniquely regulated by specific stimuli in a RelA-dependent fashion . Hence , the repertoire of specific transcriptional programs downstream of inflammatory and innate immunity signaling is expanded by the recognition of lncRNAs reported here . Lethe is a pseudogene lncRNA that comprises an unexpected a regulator of the inflammatory response . Lethe is uniquely induced by the inflammatory cytokines TNF-α and IL-1β , and Lethe inhibits NF-κB by physical interaction to inhibit RelA binding to DNA . Lethe is thus a negative feedback inhibitor of NF-κB signaling , and its mode of action is that of a decoy lncRNA ( Wang and Chang , 2011 ) . Lethe’s mechanism of action is reminiscent of similar to Gas5 or PANDA lncRNAs that titrate glucocorticoid receptor or NF-YA transcription factors away from their cognate binding sites , respectively ( Kino et al . , 2010; Hung et al . , 2011 ) . Endogenous polyadenylated Lethe RNA is highly associated with the chromatin fractions . Nonetheless , the current data do not distinguish whether the functional pool of Lethe is on chromatin , in nucleoplasm , or both . When Lethe is fused to a SV40 polyadenylation signal ( which will efficiently cause primary transcript processing and polyadenylation ) , Lethe expression inhibited NF-κB -dependent gene expression in a dose-dependent manner ( Figure 4B ) , suggesting that chromatin tethering as a primary transcript is not strictly required . Aging is associated with and requires activation of NF-κB-mediated gene expression programs ( Adler et al . , 2007; Southworth et al . , 2009 ) ; the age-associated loss of Lethe expression we observed may be one of the causes for increased NF-κB activity in aging . Intriguingly , Lethe is also selectively induced by dexamethasone , an anti-inflammatory glucocorticoid agonist , but not by other nuclear hormone receptor agonists tested . These results raise the tantalizing concept that an anti-inflammatory therapeutic acts in part by directly activating the negative feedback system of pro-inflammatory signaling ( Figure 4F ) . Lethe’s age dependent down regulation is especially interesting because inflammatory diseases such as lupus , rheumatoid arthritis , and ulcerative colitis have a higher incidence in females indicating that Lethe may have a protector role in the inflammation response that is lost with age . LncRNAs may be particularly suited to play such a balancing role in cellular signaling because its regulatory elements can receive and potentially integrate multiple input signals . The fact that lncRNA expression tends to be more tissue- and state-specific than mRNAs suggests lncRNAs are well positioned to adjudicate and diversify signaling networks in a context-specific manner ( Ravasi et al . , 2006; Cabili et al . , 2011; Djebali et al . , 2012 ) . Our work adds to the concept that some pseudogenes may have function as lncRNAs . Current literature suggests that pseudogenes are under very little selective pressure and therefore can rapidly evolve . While many pseudogenes are likely to be genetic fossils that do not have any function , perhaps the best example of pseudogene functionalization as lncRNA comes from Xist . Xist evolved from the pseudogene degeneration of a protein coding gene in the placental mammalian lineage ( Duret et al . , 2006 ) , and is now essential for dosage-compensation and X chromosome inactivation in female mammals . Moreover , some pseudogenes may act as endogenous inhibitors of microRNAs in vivo ( Poliseno et al . , 2010; Salmena et al . , 2011; Ebert and Sharp , 2010 ) . However , there is often no correlation between pseudogene and cognate wild-type genes across many tissues ( Kalyana-Sundaram et al . , 2012 ) , suggesting that the primary role of pseudogenes is not to act as an endogenous inhibitor of microRNAs . The specificity of Lethe expression highlights the need to accurately annotate pseudogenes in high throughput analysis and the need to further explore the roles of pseudogenes found throughout the genome .
All animal experiments were approved by the Stanford University Institutional Animal Care and Use Committee and the University of Michigan . Young and old mouse strain and husbandry conditions as described in ( Miller et al . , 2011 ) . Young and old mice were 4-month and 22-month-old respectively . Primary WT and RelA−/− MEFs were harvested from littermate 13 . 5-day-old embryos using standard methods and propagated in DMEM ( Invitrogen , Carlsbad , CA ) plus 10% FBS . MEFs were passaged a total of four times before all experiments . All experiments were performed in a minimum of two independently derived MEF lines of the same genotype . 293T cells were grown in DMEM plus 10% FBS . Cells were transfected with Fugene6 ( Promega , E2691 ) per the manufacturer’s instructions , and were harvested two days post-transfection . TNFα ( 210-TA-050 ) and IL-1β ( 201-LB-005 ) were ordered from R&D Systems , LPS from Escherichia coli 0111:B4 ( L5293 ) was ordered from Sigma-Aldrich , and the Human TLR1-9 Agonist kit ( tlrl-kit1hw ) was purchased from InvivoGen . Dexamethasone was obtained from Sigma ( D4902 ) . vitamin D , methyltrienolone , and estradiol were a gift from Brian Feldman . Antibodies specific for RelA ( p65 ) ( ab7970; Abcam ) , Histone H3 ( ab1791; Abcam ) , Histone H3 ( tri methyl K4 ) antibody ( ab8580; Abcam ) , Rabbit Control IgG ( ab46540; Abcam ) , and Actin ( A5316; Sigma ) are from the indicated sources . 3xκB Luciferase reporter , pTK-Renilla , and pCMV_GFP were obtained from lab stocks , were sequenced and compared to NCBI for confirmation . pCMV_Lethe was cloned from genomic DNA using primers listed in Supplementary file 1G . Total RNA was isolated from MEFs treated with 20 ng/ml TNFαfor the indicated times before RNA extraction with Trizol ( 10 , 296-010; Invitrogen ) , followed by RNAeasy kit ( 74 , 104; Qiagen ) and treated with Turbo DNAse Free Kit ( AM1907; Ambion ) . Poly-A RNA was selected for using the MicroPoly ( A ) Purist kit ( AM1919; Ambion ) . 200 ng polyA RNA was used for each library . Paired End Directional library construction was performed for dUTP libraries as described ( Levin et al . , 2010 ) except libraries were size selected by gel purification after ligation and after PCR amplification . Libraries were sequenced with an Illumina Genome Analyzer II by the Stanford Functional Genomics Facility . Sequencing reads were mapped to the mouse genome ( mm9 assembly ) using TopHat ( version 1 . 1 . 3 ) ( Trapnell et al . , 2009 ) . Each sample generated 23–32 million mapped sequences . Reference-based de novo transcritome assembly was performed using Cufflinks ( version 0 . 9 . 3 ) ( Trapnell et al . , 2010 ) and Scripture ( Guttman et al . , 2010 ) . RefSeq and Ensemble annotated transcripts were filtered out from Scripture and Cufflinks assembled transcriptomes . Transcripts with less than RPKM > 1were also removed . De novo transcript assembly was processed through Trinity ( Grabherr et al . , 2011 ) and CPC scores were determined ( Kong et al . , 2007 ) . To determine the number of statistically significant differentially expressed genes for hierarchical clustering , we performed SAMseq , a nonparametric method for estimating significance in RNA-seq data ( Li and Tibshirani , 2011 ) and discovered 3690 significant transcripts with FDR < 0 . 05 . DNA was cross-linked for 10 min with 1% formaldehyde and stopped in 0 . 125 M glycine . Purified chromatin was sonicated to ∼250 bp using the Bioruptor ( Diagenode , Inc . , Delville , NJ ) and incubated with the IgG or Histone H3 ( tri methyl K4 ) as previously described in http://farnham . genomecenter . ucdavis . edu/pdf/FarnhamLabChIP%20Protocol . pdf . ChIP-seq libraries were made and sequenced as above after second strand synthesis . Size-selected libraries of 200–300 bp length were used for Illumina deep-sequencing . Raw reads from ChIP-Seq were mapped to mouse genome ( mm9 assembly ) using Bowtie ( version 0 . 12 . 6 ) ( Langmead et al . , 2009 ) and H3K4me3 peaks were called out using MACS ( Zhang et al . , 2008 ) . MEFs were treated with TNF-α ( 20 ng/ml ) , 10 ng/ml human IL-1β , 100 ng/ml LPS ( E . coli 055:B5 ) , 100 ng/ml Pam3CSK4 , 108 cells/ml HMLK , 25 μg/ml poly ( I:C ) , 25 μg/ml poly LMW ( I:C ) , 10 μg/ml LPS ( from E . coli strain K12 ) , 100 ng/ml recombinant flagellin ( Salmonella Typhimurium ) , 5 μg/ml imiquimod-R837 , 10 nM vitamin D , 100 nM methyltrienolone , 100 nM estradiol , or 1 μM dexamethasone for 6 hr . Total RNA was prepared as described above . RNA was analyzed on a LightCycler 480 by RT-qPCR using total RNA ( 100 ng ) , Taqman One Step RT-PCR master mix ( 4309169; Life Technologies ) . Assays are listed in Supplementary file 1G . Reactions were in triplicate for each sample and performed a minimum of three times . Data were normalized to Actin levels . Young and old mice tissue was mixed with QIAzol ( Qiagen ) in a 2-ml tube containing a 5-mm stainless steel bead ( Qiagen ) and was then disrupted on a tissue lyser . CHCl3 was mixed to the homogenate and after centrifugation the aqueous solution was apply to a RNeasy column ( Qiagen ) . The RNA purification was then finished on an automated QIAcube system ( Qiagen ) and included a DNAse treatment . MEFs were treated with TNF-α ( 20 ng/ml ) for 15 min . ChIP was performed as above for ChIP-Seq . Chromatin was sonicated to 500 bp . 293T cells were treated with TNF-α ( 20 ng/ml ) for 10 min . DNA was cross-linked for 10 min with 1% formaldehyde and stopped in 0 . 125 M glycine . Purified chromatin was sonicated to ∼500 bp using the Bioruptor ( Diagenode , Inc ) and incubated with 2 μg RelA antibodies or IgG at 4°C overnight . Immunoprecipitation was performed with the Rainin Purespeed tips ( PT-2-A5 ) per manufacturer’s instructions . DNA was analyzed on a LightCycler 480 ( Roche ) using LightCycler 480 SYBR Green I Master Mix ( 4707516001; Roche ) per manufacturer’s instructions . Primers are listed in Supplementary file 1G . MEFs were treated with TNF-α ( 20 ng/ml ) for 6 hr . Cells were fractioned as previously described ( Méndez and Stillman , 2000 ) . RNA was extracted and analyzed as above . MEFs were treated with TNF-α ( 20 ng/ml ) for 6 hr . IP was performed as described above for ChIP except all buffers were pH 7 . 0 and cells were cross-linked with 1% glutaraldehyde . RNA was extracted and analyzed as above . MEF cells were nucleofected using the Nucleofector for mouse embryonic fibroblasts per manufacturer’s instructions ( VPL-1004 , Lonza ) except 500 , 000 cells were nucleofected per condition in 1 μM ASO and plated in one well of a 6-well plate . The ASOs ( IDT ) were designed as described in ( Ideue et al . , 2009 ) to increase stability . 48 hr post nucleofection , cells were treated with 20 ng/ml TNF-α for 6 hr prior to RNA extraction . 293T cells were transfected with Fugene6 per the manufacturer’s instructions 24 hr post plating with 1 μg of 3xκB Luciferase reporter construct , 50 ng of pTK-Renilla and a total of 150 ng of expression plasmids for GFP and Lethe in a 12-well plate . Each condition was performed in triplicate . After 48 hr , cells treated with 20 ng/ml TNF-α for 6 hr . Cells were harvested 2 days post-transfection and luciferase was measured per manufacturer’s instructions with the Dual-Luciferase Reporter System ( E1910; Promega ) . Luciferase values were normalized to Renilla to control for transfection efficiency . Experiments were repeated three independent times . 293T cells were transfected with Fugene6 per the manufacturer’s instructions 24 hr post plating with 10 μg pCMV GFP or pCMV Lethe . After 48 hr , cells treated with 20 ng/ml TNF-α for 10 min , washed two times with PBS and lysed . Nuclear lysates were prepared as previously described ( Schmitt et al . , 2011 ) . NF-κB DNA binding activity was measured using the Gel Shift Assay System ( E3300; Promega ) per manufacturer’s instructions . Briefly , 8 μg of nuclear lysate were incubated with 10 pg of 32P-labeled DNA probe for 15 min . For supershift , lysate was preincubated with 1 μg RelA antibody for 10 min . For competitive and non-competitive experiments , 100-fold molar excess unlabeled NF-κB or CREB probe were preincubated with lysate for 10 min . Complexes were separated by electrophoresis on 6% TBE gels ( EC6265; Invitrogen ) and assayed by PhosophorImager analysis . Assays were repeated three independent times . Deep sequencing data in this study are available for download from Gene Expression Omnibus ( http://www . ncbi . nlm . nih . gov/geo ) ( accession ID: GSE47494 ) . | The simplest account of gene expression is that DNA is transcribed into messenger RNA , which is then translated into a protein . However , not all genes encode proteins; for some it is the RNA molecule itself that is the end product . Many of these ‘non-coding RNAs’ are thought to be involved in regulating the expression of other genes , but their exact functions are unknown . Pseudogenes are genes that have lost their protein-coding abilities as a result of mutations they have accumulated mutations over the course of evolution . They were previously referred to as ‘junk DNA’ or ‘dead genes’ because they were thought to be completely non-functional , lacking even the ability to encode RNA . However , recent work has shown that pseudogenes are in fact transcribed into long non-coding RNAs , and these are now the focus of much research . Here , Rapicavoli et al . report that certain pseudogenes and long non-coding RNAs are involved in regulating the immune response . Specific and distinct pseudogene-derived long RNAs are made when cells are exposed to different kinds of infections . Immune cells such as macrophages and lymphocytes produce a protein called tumor necrosis factor alpha ( TNFα ) , which is involved in triggering fever and inflammation . TNFα exerts these effects by binding to and activating a transcription factor called NF-κB , which then moves to the nucleus and binds to DNA , regulating the expression of genes that encode immune proteins . Rapicavoli et al . found that the production of a long non-coding RNA called Lethe ( after the ‘river of forgetfulness’ in Greek mythology ) increases when TNFα activates NF-κB . Surprisingly , however , Lethe then binds to NF-κB and prevents it from interacting with DNA , thereby reducing the production of various inflammatory proteins . This is the first time that a pseudogene has been shown to have an active role in regulating signaling pathways involved in inflammation , and raises the possibility that other pseudogenes may also influence distinct feedback loops and signaling networks . It suggests that many novel functions for pseudogenes and long non-coding RNAs remain to be discovered . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2013 | A mammalian pseudogene lncRNA at the interface of inflammation and anti-inflammatory therapeutics |
Interprotein electron transfer underpins the essential processes of life and relies on the formation of specific , yet transient protein-protein interactions . In biological systems , the detoxification of sulfite is catalyzed by the sulfite-oxidizing enzymes ( SOEs ) , which interact with an electron acceptor for catalytic turnover . Here , we report the structural and functional analyses of the SOE SorT from Sinorhizobium meliloti and its cognate electron acceptor SorU . Kinetic and thermodynamic analyses of the SorT/SorU interaction show the complex is dynamic in solution , and that the proteins interact with Kd = 13 . 5 ± 0 . 8 μM . The crystal structures of the oxidized SorT and SorU , both in isolation and in complex , reveal the interface to be remarkably electrostatic , with an unusually large number of direct hydrogen bonding interactions . The assembly of the complex is accompanied by an adjustment in the structure of SorU , and conformational sampling provides a mechanism for dissociation of the SorT/SorU assembly .
Although electron transfer reactions are key biochemical events , which underpin fundamental processes , such as respiration and photosynthesis , the study of the molecular details of the interprotein interactions at their core can be largely intractable . In particular , atomic resolution crystal structures of electron transfer complexes are rare , due to their fundamentally transient nature ( Antonyuk et al . , 2013 ) . Electron transfer pathways are made up of chains of redox proteins , which provide a path for the controlled flow of electrons and rely on efficient docking of protein redox partners through noncovalent , dynamic protein-protein interfaces ( Moser et al . , 1992 ) . Complementary electrostatic surfaces , hydrophobic interactions and dynamics at the protein-protein interface have all been proposed to contribute to efficient interprotein electron transfer ( Leys and Scrutton , 2004 ) , with a strong correlation between the driving force for the reaction , the distance between redox centers and the rate of electron transfer ( Moser et al . , 1992; Marcus and Sutin , 1985 ) . Interprotein electron transfer processes are central to the redox conversions of cellular sulfur compounds , which are an evolutionarily ancient type of metabolism that has existed as long as cellular life ( Schidlowski , 1979; Kappler et al . , 2008 ) . Sulfur-containing compounds mediate many crucial reactions in the cell ( for example , in coenzyme A , sulfur containing amino acids or glutathione ) , but their reactivity also makes them potentially toxic ( Kappler , 2011 ) . Sulfite in particular , is a highly reactive sulfur compound that can cause damage to proteins , DNA and lipids , resulting in oxidative stress and irreversible cellular damage ( Feng et al . , 2007 ) . In most cells , the detoxification of sulfite by oxidation to sulfate ( Equation 1 ) is catalyzed by sulfite oxidizing enzymes ( SOEs ) ( Kisker et al . , 1997 ) . ( 1 ) SO32- + H2O → SO42- + 2H+ + 2e- SOEs from plants , higher animals and bacteria have been characterized and they all catalyze the same fundamental reaction . However , their cellular functions , catalytic properties ( Kappler , 2011; Feng et al . , 2007; Kappler and Wilson , 2009; Hille , 2002; Hänsch et al . , 2007 ) and the identities of their natural electron acceptors vary significantly . Some SOEs transfer electrons to oxygen ( Schrader et al . , 2003 ) , while others interact with redox proteins such as cytochrome c ( Kisker et al . , 1997; Kappler and Wilson , 2009; Low et al . , 2011; Bailey et al . , 2009; Cohen and Fridovich , 1971; Cohen and Fridovich , 1971; Cohen et al . , 1971 ) or as yet unknown cellular components ( Kappler , 2011; Low et al . , 2011; Bailey et al . , 2009; Wilson and Kappler , 2009 ) . To date , three unique crystal structures of SOEs have been reported: from chicken , plant and bacteria , which differ significantly in their domain architectures and redox cofactor compositions . However , none of these studies show details of a SOE in complex with its external electron acceptor ( Kisker et al . , 1997; Schrader et al . , 2003; Kappler and Bailey , 2005 ) . At present , no structural information on the molecular interactions of any of these enzymes with their respective electron acceptors is available and the determinants that dictate the type of electron acceptor individual SOEs employ , while maintaining the efficiency of the basic enzyme reaction are open questions . ( Kappler , 2011; Kappler , 2008 ) Here , we have investigated an electron transfer complex involving the periplasmic SorT sulfite dehydrogenase from the α-Proteobacterium Sinorhizobium meliloti , which represents a structurally uncharacterized type of SOE , and its electron acceptor , the c-type cytochrome SorU ( Low et al . , 2011; Wilson and Kappler , 2009 ) . In S . meliloti the SorT sulfite dehydrogenase is part of a sulfite detoxification system that is induced in response to the degradation of sulfur containing substrates such as the organosulfonate taurine ( Wilson and Kappler , 2009 ) . Electrons derived from sulfite oxidation are passed on to the SorU cytochrome , and likely then to cytochrome oxidase , as S . meliloti is capable of sulfite respiration ( Low et al . , 2011 ) . Here , we report the crystal structures of both the isolated SorT and SorU proteins and the biochemical and structural analyses of the SorT/SorU electron transfer complex . This is the first time that a crystal structure of a molybdenum enzyme in complex with its external electron acceptor has been solved .
Sulfite-oxidizing enzymes , particularly those from bacteria , are known to be highly efficient catalysts ( Kappler , 2011; Kappler and Enemark , 2015 ) . Previous work has established that SorT is able to transfer electrons to the SorU cytochrome that is encoded on the same operon; however , no kinetic details of the interaction were reported ( Low et al . , 2011 ) . With the artificial electron acceptor ferricyanide , SorT was shown to have a turnover number of 338 ± 3 s-1 ( Low et al . , 2011; Wilson and Kappler , 2009 ) . Employing SorU as the substrate , we analyzed the kinetics of the interaction between SorT and SorU and found the interaction to be fast and highly specific , with a KM ( SorU ) of 32 ± 5 μM and a kcat of 140 ± 11 s-1 , confirming that SorU is the natural electron acceptor of SorT . Measurement of the thermodynamics of the SorT/SorU interaction by isothermal titration calorimetry ( ITC ) revealed a dissociation constant of Kd = 13 . 5 ± 0 . 8 μM with a determined stoichiometry of 0 . 8 ± 0 . 2 . These values are in the range observed for other electron transfer complexes ( Dell'acqua et al . , 2008; Pettigrew et al . , 2003 ) and match a model where SorT sequentially transfers two electrons , derived from sulfite oxidation , to two SorU molecules . In other words , the SorT/SorU complex must form twice ( with two different ferric SorU protein molecules ) to complete the oxidative half reaction of SorT . The KM of SorT for SorU is very close to its affinity for sulfite ( KM = 15 . 5 ± 1 . 9 μM [Wilson and Kappler , 2009] ) , and the turnover number in the SorU-based assay is ~40% of that seen with ferricyanide as the electron acceptor ( Wilson and Kappler , 2009 ) , where no significant reorientation and docking of the electron acceptor is required . These data reveal interesting details about the formation of the SorT/SorU complex , which appears to form with similar affinities between the two proteins when both are oxidized as in the ITC experiments and in a system where SorT constantly undergoes oxidation and reduction ( SorU-based enzyme assay ) . In addition , the similarity between the determined values of Kd and KM indicates that the affinity of SorT for SorU is unaffected by the presence of substrate or product . However , the kinetic parameters for this interaction are clearly distinct from those for other sulfite-oxidizing enzymes , such as the bacterial SorAB enzyme or chicken sulfite oxidase ( CSO ) , both of which have much higher affinities for their respective electron accepting cytochromes c ( both with KM ( Cyt c ) ca . 2 μM ) ( Kappler et al . , 2006 ) . The catalytic turnover of CSO is relatively slow ( kcat = 47 . 5 ± 1 . 9 s-1 ) , due to a mechanism requiring internal rearrangement , while turnover of SorAB with its natural electron acceptor ( 334 ± 11 s-1 ) is significantly faster than that for SorT ( Kappler and Enemark , 2015; Kappler et al . , 2006 ) . In order to investigate whether there are any structural reasons for these differences , we solved the crystal structure of SorT by molecular replacement and refined it to 2 . 4 Å resolution . The structure shows two homodimeric assemblies per asymmetric unit ( Table 1 ) , which is in agreement with the quaternary structure as determined by MALLS ( Low et al . , 2011; Wilson and Kappler , 2009 ) . Unexpectedly however , within the SorT homodimer the protomers are oriented in a head-to-tail orientation ( Figure 1A ) , a subunit arrangement that has not been previously observed in structures of sulfite-oxidizing enzymes . In keeping with the nomenclature applied to other structurally-characterized SOEs , a ‘dimerization’ domain typically defines the interface between the two monomers ( Kisker et al . , 1997; Schrader et al . , 2003 ) , but the structure of the SorT dimer does not follow this paradigm . 10 . 7554/eLife . 09066 . 003Figure 1 . The crystal structures of the SorT and SorU proteins in isolation . ( A ) The structure of the SorT homodimer . Molecule A in blue/yellow; molecule B in gray ( with transparent surface ) . For molecule A , the ‘SUOX-fold domain’ and ‘dimerization domain’ are represented in blue and yellow , respectively . The molybdopterin cofactor is shown as sticks within the SUOX-fold domain . The corresponding domains of the opposing protomer ( shown in molecular surface representation ) , which constitute the dimer interface are colored to highlight the ‘head-to-tail’ dimer arrangement . INSET: a closer view of the molybdenum binding-site: the molybdenum atom ( green sphere ) is coordinated by two dithioline ligands from the molybdopterin ( yellow spheres ) , residue Cys 127 , an axial oxo ligand and an equatorial hydroxo or water ligand ( red spheres ) . ( B ) The structure of SorU . The main three helices are labeled and the heme cofactor is shown in red . INSET: the heme binding site with the heme cofactor , coordinating residues , covalent links to Cys 50 and 57 and hydrophobic residues lining binding site: Phe 39 , Val 62 , Val 76 , Val 80 , Val 101 highlighted . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 00310 . 7554/eLife . 09066 . 004Table 1 . Data collection and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 004SorTSorUSorT/SorU complexData collection Space GroupP21F222P21212Cell dimensionsa , b , c ( Å ) 96 . 0 , 92 . 2 , 109 . 470 . 9 , 129 . 2 , 197 . 0109 . 6 , 95 . 8 , 49 . 9α , β , γ ( ° ) 90 , 89 . 7 , 9090 , 90 , 9090 , 90 , 90X-ray sourceAUS MX2AUS MX2AUS MX2λ ( Å ) 0 . 9500 . 9540 . 954DetectorADSC Quantum 315rADSC Quantum 315rADSC Quantum 315rResolution range ( Å ) 50-2 . 4 ( 2 . 43-2 . 35 ) a50-2 . 2 ( 2 . 28-2 . 20 ) 50-2 . 5 ( 2 . 50-2 . 59 ) Observed reflections2405219634064273Unique reflections779712345318883Completeness ( % ) 98 . 4 ( 99 . 4 ) 99 . 9 ( 100 ) 99 . 3 ( 99 . 6 ) Multiplicity3 . 1 ( 3 . 1 ) 4 . 1 ( 4 . 1 ) 3 . 4 ( 3 . 4 ) <I/σ ( I ) >6 . 7 ( 2 . 1 ) 8 . 9 ( 1 . 6 ) 8 . 8 ( 1 . 7 ) Rmerge ( % ) b15 . 9 ( 66 . 7 ) 13 . 3 ( 76 . 6 ) 13 . 9 ( 76 . 6 ) Refinement Reflections in working set740242211317781Reflections in test set39271201960Protomers per ASU441Total atoms ( non-H ) 1137829413422Protein atoms1089025423290Metal atoms442Water atoms37622763Other atoms10816867Rwork ( % ) c20 . 8 ( 31 . 7 ) 19 . 2 ( 30 . 8 ) 21 . 1 ( 30 . 2 ) Rfree ( % ) d23 . 9 ( 34 . 7 ) 24 . 0 ( 34 . 3 ) 26 . 0 ( 36 . 5 ) Rmsd bond lengths ( Å ) 0 . 0080 . 0060 . 012Rmsd bond angles ( deg ) 1 . 080 . 911 . 41<B> ( Å2 ) e32 . 520 . 638 . 0Cruickshank's DPI0 . 070 . 230 . 49PDB ID4PW34PWA4PW9aValues in parenthesis are for highest-resolution shellb Rmerge = ∑hkl ∑i | Ii ( hkl ) - <I ( hkl ) > |/∑hkl ∑i Ii ( hkl ) c Rwork = ∑h | Fobs – Fcalc |/∑hFobsd Calculated as for Rwork using 10% of the diffraction data that had been excluded from the refinementeAs calculated by BAVERAGE ( Winn et al . , 2011 ) Nevertheless , the fold of the SorT monomers is similar to those of other SOEs ( Kisker et al . , 1997; Schrader et al . , 2003; Kappler and Bailey , 2005 ) , comprising a central ‘SUOX-fold’ domain ( Workun et al . , 2008 ) that harbors the Mo active site and a ‘dimerization’ domain ( Figure 1A ) . The SorT active site has a square-pyramidal geometry seen in all other SOE structures with a five coordinate molybdenum atom and a single tricyclic pyranopterin cofactor ( Figure 1A , Table 2 ) ( George and Pickering , 1999 ) . 10 . 7554/eLife . 09066 . 005Table 2 . Mo coordination geometry in the active site of SorT . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 005BondDistance ( Å ) Mo-S1 ( pterin ) 2 . 4Mo-S2 ( pterin ) 2 . 4Mo-S ( Cys 127 ) 2 . 3Mo=O1 . 7Mo-OH/H2O1 . 9 Single electron reduction of SorT to its EPR active MoV form was achieved using a combination of Ti ( III ) citrate and a suite of organic redox mediators . The MoV EPR spectrum is similar to the so-called ‘high pH’ EPR signature of SOEs ( Figure 2A , Table 3 , Appendix 1 ) . An additional feature is the presence of superhyperfine coupling between the unpaired electron on the Mo ion and two nearby I = ½ nuclei . This implies that the equatorially coordinated O-donor is an aqua ligand at pH 8 ( Figure 2B ( ii ) ) or that a hydroxido ligand is hydrogen bonded with a water molecule whose proximal H-atom is coupled with the electron spin on Mo ( Figure 2B ( i ) , Appendix 1 ) . 10 . 7554/eLife . 09066 . 006Figure 2 . EPR analysis of the SorT protein . ( A ) X-band EPR spectra of the Mo ( V ) center in SorT . ( a ) First and ( b ) second derivative EPR spectra of SorT at 0 mV vs NHE in tricine pH 8 . 0 , υ= 9 . 43462 GHz , T = 136 . 3 K . ( c ) Computer simulation of the second derivative spectrum with the spin Hamiltonian parameters listed in Table 3; ( d , e ) Expansion of spectra ( b ) and ( c ) , respectively . ( B ) Schematic structures of the ( i ) high and ( ii ) low pH forms of sulfite oxidase . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 00610 . 7554/eLife . 09066 . 007Table 3 . Spin Hamiltonian parameters for the Mo ( V ) center of SorT and various low and high pH forms of human , avian , plant and bacterial sulfite oxidases . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 007SpeciesParameterXYZβoRefSorTg1 . 949301 . 959971 . 98632-A ( 95Mo ) b20 . 536 . 053 . 926A ( 1H ) b , c3 . 54 . 04 . 80SorAcg1 . 95411 . 96611 . 9914-dKlein , et al . , 2013 Human SOg Low pH1 . 96461 . 97232 . 0023-Enemark , et al . , 2010 A ( 1H ) b11 . 477 . 107 . 71-Enemark , et al . , 2010 Chicken SOg ( Low pH ) 1 . 96581 . 97202 . 0037-Drew and Hanson , 2009 A ( 1H ) b11 . 937 . 377 . 95-Drew and Hanson , 2009 g ( High pH ) 1 . 95311 . 96411 . 9872Drew and Hanson , 2009 A . Thaliana SOg ( Low pH ) 1 . 9631 . 9742 . 005-Enemark , et al . , 2006 A ( 1H ) b11 . 99 . 210 . 3--Enemark , et al . , 2006 g ( High pH ) 1 . 9561 . 9641 . 989-Enemark , et al . , 2006 aNon-coincident angle between g and A ( rotation about x axis ) . bUnits 10-4 cm-1 . c Two magnetically equivalent protons ( I=1/2 ) were included in the computer simulated spectra . c95Mo hyperfine couplings were unresolved and the shoulders on gz were incorrectly attributed to 95Mo hyperfine resonances . dEuler angles were not determined . Within the SorT dimer , the subunit interface involves both the 'SUOX-fold' and the ‘dimerization’ domains ( Figure 1A ) , resulting in a buried surface area of ~1280 Å2 per monomer , which is approximately 9% of the solvent-accessible surface of each monomer . The functional consequences and structural origins of these significant differences among the quaternary structures of SOEs are at this point unknown , as all of these enzymes are highly catalytically active , have similar active site structures and no known kinetic cooperativity that would imply a functional role for the different oligomeric assemblies ( Kappler and Wilson , 2009; Hänsch et al . , 2007; Wilson and Rajagopalan , 2004 ) . Despite the dynamic nature of the SorT/SorU interaction , it was possible to co-crystallize SorT with SorU , resulting in the crystal structure of the SorT/SorU complex , where a single SorT/SorU entity is present per asymmetric unit , and the application of crystallographic 2-fold symmetry reveals a SorU/SorT2/SorU assembly ( Figure 3A , Table 1 ) . Small-angle X-ray scattering ( SAXS ) with a sample prepared as a stoichiometric mixture of SorT and SorU ( Figure 3B , Table 4 , Appendix 2-Figure 1 ) confirmed that the structure observed in the crystal is preserved in solution . 10 . 7554/eLife . 09066 . 008Figure 3 . The structure of the SorT/SorU electron transfer complex . ( A ) The asymmetric unit from the crystal structure of the SorT/SorU complex contains the functional electron transfer complex . The SorU/SorT2/SorU complex is revealed by the application of crystallographic symmetry operators . The positions of the redox active molybdenum ( SorT ) and heme c ( SorU ) cofactors are indicated . ( B ) Two views of an overlay of the SorU/SorT2/SorU crystal structure with the averaged and filtered dummy atom model from 10 ab initio reconstructions as revealed by SAXS analyses . ( C ) ‘Open-book unfolding’ of SorT/SorU complex ( SorT is shown in blue , SorU in red ) indicating the ‘footprint’ of interfacing residues from each protein . ( D ) The same view as Panel C , showing the charge complementarity of the SorT/SorU interface ( areas of positive charge in blue , negative charge in red and neutral in white ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 00810 . 7554/eLife . 09066 . 009Table 4 . Data collection and processing parameters for analysis of the SorT/SorU complex in solution by Small Angle X-ray Scattering ( SAXS ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 009Data collection parameters InstrumentSAXSess ( Anton Paar ) Beam geometry10 mm slitAH , LH ( Å-1 ) , GNOM beam geometry definition0 . 28 , 0 . 12q-range measured ( Å-1 ) 0 . 01-0 . 400Exposure time ( min ) 60 ( 4 x 15 ) SorT2SorU2 concentration range ( mg mL-1 ) 2 . 75-5 . 5Temperature ( ºC ) 10Structural parameters* Rg ( Å ) , I ( 0 ) ( cm-1 ) from Guinier ( desmeared data ) q*Rg < 1 . 330 . 8 ± 0 . 4 , 0 . 223 ± 0 . 002Rg ( Å ) , I ( 0 ) ( cm-1 ) from P ( r ) ( q-range 0 . 01 – 0 . 25 Å-1 ) 32 . 0 ± 0 . 3 , 0 . 235 ± 0 . 002dmax ( Å ) from P ( r ) 110Molecular mass determination* Molecular mass Mrfrom Guinier I ( 0 ) ( ratio with expected ) 108741 ( 0 . 984 ) Molecular mass Mrfrom P ( r ) I ( 0 ) ( ratio with expected ) 114593 ( 1 . 037 ) SorT2SorU2 parameters calculated from sequence and chemical composition Molecular volume ( Å3 ) 134385Molecular weight Mr ( Da ) 110556Partial specific volume ( cm3 g-1 ) 0 . 732Contrast ( X-rays ) ( Δρ x 1010 cm-2 ) 2 . 895Modeling results and validation Crystal structure Rg , dmax ( Å ) SorT/SorU2/SorT SorT 31 . 3 , 108 27 . 9 , 99Crystal structure compare to desmeared I ( q ) ( χ-value ) SorT/SorU2/SorT ( q-range 0 . 01 – 0 . 15 Å-1 ) SorT ( q-range 0 . 01 – 0 . 15 Å-1 ) 1 . 7 2 . 3Results from 10 ab initio shape restorations . P1 symmetry: Average molecular volume ( Å3 ) Normalised spatial distribution ( NSD ) and NSD variation χ value for fit to desmeared data 140800 0 . 508 ( 0 . 008 ) 1 . 8Software employed Calculation of expected Mr , △ρ and υ valuesMULChPrimary data reduction , I ( q ) vs q SAXSQuant 1DDesmearingSAXSQuantGuinier analysisPRIMUSP ( r ) analysisGNOMModel I ( q ) from crystal coordinatesCRYSOLab initio shape restorationsDAMMIN3D graphics representationsPYMOL*Reported for 2 . 75 mg ml-1 measurement . The central assembly within the SorU/SorT2/SorU complex is the SorT homodimer , which is identical to the structure of the SorT homodimer alone ( Figures 1A and 3A , Table 1 ) . The structure of the SorU protein , both within the SorT/SorU complex and when crystallized in isolation ( Figure 1B , Table 1 ) , is predominantly α-helical with three major α-helices arranged to form a bundle that frames the heme-binding site ( Figure 1B ) ( C47XXC50H , axial ligands: His 51 and Met 87 ) . In the SorT/SorU complex , the SorU protein docks within a pocket adjacent to the SorT active site , with the heme cofactor located at the protein-protein interface ( Figures 3A and 4A ) . The shortest ‘edge-to-edge’ distance between the SorT Mo atom and the propionate group from the SorU heme c cofactor is 8 . 2 Å , which is well within the distance for fast electron transfer through the protein medium ( Page et al . , 1999 ) . PATHWAY analysis ( Onuchic et al . , 1992 ) ( Table 5 ) further indicates that the dominant electron tunneling pathway from SorT to SorU proceeds from the Mo atom , via the coordinating H2O/OH- , to the guanidinium group of SorT residue Arg 78 and across the protein-protein interface to the heme propionate group and to the pyrrole ring of heme c to the heme iron ( Figure 4B , Table 5 ) . 10 . 7554/eLife . 09066 . 010Figure 4 . Orientation of the redox cofactors in the crystal structure of the SorT/SorU electron transfer complex . ( A ) Electron density map in the region of the SorT/SorU interface . The SorT molecule is represented in blue and the SorU molecule in red . The 2Fo-Fc electron density map ( contoured at 1σ ) is shown as a blue net and the redox cofactors ( molybdenum and heme ) are colored according to the representation in Panel B . ( B ) Pathway for electron transfer ( Beratan et al . , 1992 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 01010 . 7554/eLife . 09066 . 011Table 5 . Electron transfer parameters between SorT ( Mo ) and SorU ( Fe ) as calculated by PATHWAYS ( Onuchic et al . , 1992 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 011Distance ( Mo-Fe , Å ) 16 . 5 ÅAtomic packing density ( ρ ) 0 . 97Average decay exponential ( β ) 0 . 97Electronic coupling ( HDA ) 3 . 4 x 10-4Maximum ET rate ( s-1 ) 1 . 2 x 107 Specific structural adaptations of the SorU protein take place when the SorT/SorU assembly is formed ( Figure 5B ) . A surface loop on SorU ( residues 82–93 ) moves away from the SorT/SorU complex interface , leading to a reorientation of the heme ligand residue Met 87 so that a different Met 87 rotamer coordinates the iron in the SorU structures within and outside of the complex ( Figure 5B ) . This change in the structure of SorU is required to allow a Mo-heme edge-to-edge distance of ‘closest approach’ of ca . 8 Å within the SorT/SorU assembly . Without this adjustment ( for example , if the SorU residue 82–93 loop structure remained rigid ) the closest approach for the redox cofactors would be ca . 10 Å . 10 . 7554/eLife . 09066 . 012Figure 5 . Comparisons of ( A ) the SorT/SorU and SorAB structures and ( B ) the structures of SorT and SorU within and outside of the electron transfer complex . ( A ) Structures of the SorT/SorU ( left ) and SorAB ( right ) complexes , where the Cα traces of the heme-containing protomers are colored according to temperature factor . ( B ) Superposition of the SorT and SorU structures within and outside of the electron transfer complex , highlighting conformational changes that were observed to accompany complex formation . The crystal structures of SorT and SorU within the SorT/SorU complex are shown in gray , and the superposed structures of SorT and SorU determined alone are shown in blue and red respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 012 Changes in the conformation of axial heme ligands are known to alter the orbital interactions between the iron atom and the ligand , which can change the redox properties of the heme group ( Tai et al . , 2013 ) . However , this does not appear to be the case here as the redox potential of the SorU heme was determined by optical spectroelectrochemistry to be +108 ( ±10 ) and +111 ( ±10 ) mV vs . NHE ( pH 8 . 0 ) , respectively , in the presence and absence of SorT ( Table 6 , Figure 6B ) . 10 . 7554/eLife . 09066 . 013Figure 6 . Redox analyses of the SorT protein and the SorT/SorU complex . ( A ) Plot of EPR intensity ( Ip ) at 343 mT ( from MoV form of SorT ) as a function of solution redox potential ( E mV vs NHE ) . The solid line is a fit to the equation I ( E ) = Ip1 =10 ( E-E1 ) /59 + 10 ( E2-E ) /59 using the potentials E1 = MoVI/V = +110 ( ±10 ) mV and E2 MoV/IV-18 ( ±10 ) mV vs NHE ) . ( B ) Electronic spectra of ferric and ferrous SorU obtained from spectroelectrochemistry . Inset: plot of absorbance at 550 nm ( ferrous α-band ) and 406 nm ( ferric Soret band ) as a function of applied potential . The solid lines are theoretical curves based on the equation Abs = ( εox10 ( E-E' ) /59 ) +εred ) 1 + 10 ( E-E' ) /59Ctot where the extinction coefficients refer to the oxidized and reduced forms of the protein and Abs is the absorbance at this same wavelength . ctot is the total protein concentration . The redox potential ( E' = +111 mV vs NHE ) was obtained by global analysis of all potential dependent spectra across all wavelengths with the program ReactLab Redox ( Maeder and King ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 01310 . 7554/eLife . 09066 . 014Table 6 . Redox potential values for SorT and SorUa . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 014ProteinCoupleEº ( mV vs NHE ) SorTMoVI/V+110 ( ±10 ) MoV/IV-18 ( ±10 ) SorUFeIII/II+108 ( ±10 ) SorU ( in the presence of SorT ) FeIII/II+111 ( ±10 ) aRedox potentials of SorT were determined by redox potentiometry , and SorU redox potentials by optical spectroelectrochemistry . The MoVI/V and MoV/IV redox potentials of SorT ( +110 ( ±10 ) mV and -18 ( ±10 ) mV vs . NHE ( pH 8 ) ) were determined by an EPR-monitored redox titration where the initial EPR-silent MoVI form is reduced to the EPR-active MoV state ( Figure 2A ) , which then gives way to the EPR-silent MoIV form at low potential , resulting in a bell-shaped curve ( Table 6 , Figure 6A ) . The high MoVI/V potential at pH 8 matches that of the ferris/ferrous redox couple . ( 3 , 4 ) . The structural change in SorU indicates that an ‘induced fit’ mechanism is responsible for the formation of a productive SorT/SorU electron transfer complex . This type of mechanism has in the past been used to describe electron transfer complexes ( involving electron transfer flavoproteins or ferredoxin reductases ) where the protein partners include mobile domains , and where conformational change is necessary for the creation of high affinity protein-protein interfaces ( Senda et al . , 2007; Toogood et al . , 2007 ) . However , although facilitating redox interactions , these systems are distinct from the docking mechanism seen for the SorT and SorU proteins which accompanies modifications to the structure of SorU and allows the two redox centers to attain positions of closest approach for fast electron transfer . The SorT/SorU complex interface shows significant charge complementarity , with the negative charge on SorU correlating with a concentration of positive charges at the SorU binding site on SorT ( Figures 3C , D ) . Unlike other known cytochromes c that can act as electron acceptors to SOEs ( Low et al . , 2011; Brody and Hille , 1999; Kappler et al . , 2000 ) , the electrostatic surface of SorU has an overall negative charge ( Figures 3C , D ) . The positive charge on the SorT surface therefore explains the low catalytic activity of SorT with horse heart cytochrome c ( 7 U/mg , pI ~10 ) , the natural electron acceptor for vertebrate SOEs , compared to the high activity observed with SorU ( 212 U/mg; pI ~4 ) ( Kappler , 2011; Low et al . , 2011; Wilson and Kappler , 2009 ) . In its electrostatic nature , the interaction surface between SorT and SorU is unusual . Structures of other cytochrome-containing electron transfer complexes ( Nojiri et al . , 2009; Axelrod et al . , 2002; Pelletier and Kraut , 1992; Solmaz and Hunte , 2008 ) show binding interfaces characterized by a ‘ring’ of electrostatic interactions that encompass contact surfaces that are predominantly hydrophobic . In fact , the ‘steering’ of electron transfer partners by electrostatic interactions , accompanied by ‘tuning’ via hydrophobic interactions is a dominant observation for protein-protein electron transfer complexes ( Nojiri et al . , 2009; Axelrod et al . , 2002; Pelletier and Kraut , 1992; Solmaz and Hunte , 2008 ) . In addition to the electrostatic interactions that support the formation of the SorT/SorU complex , there are six hydrogen bonds found at the SorT/SorU protein-protein interface , as well as a salt bridge between the SorT active site residue Arg 78 and a propionate group of the SorU heme moiety ( Table 7 , Figure 7 ) . This is an unusually large number in comparison with structures of other cytochrome-containing electron transfer complexes ( Nojiri et al . , 2009; Axelrod et al . , 2002; Pelletier and Kraut , 1992; Solmaz and Hunte , 2008 ) , which tend to have fewer hydrogen bonds and lack salt bridges . In fact , direct hydrogen bonds between electron transfer proteins are generally considered unfavorable for a transient interaction because of energetically disadvantageous desolvation ( Miyashita et al . , 2003 ) . Also significant is the observation that no intermolecular interactions at the interface are mediated by water molecules ( Figure 7 ) ( Nojiri et al . , 2009; Gray and Winkler , 2005 ) . In fact , very little ordered water is observed in proximity to the interfacing region of the SorT molecule ( a total of 2 water molecules only and these are hydrogen bonded to the SorT molecule rather than mediating the SorT/SorU interaction ) . However , the current analysis is limited by the moderate resolution of the current structure ( 2 . 6 Å ) , which as a consequence , includes only ca . 0 . 15 modeled water molecules per residue . 10 . 7554/eLife . 09066 . 015Table 7 . Comparison of the protein-protein interfaces in the SorT/SorU and SorAB structures . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 015ParameterSorT/SorUaSorABbSorTSorUSorASorBAverage relative B factorc ( Å2 ) 0 . 91 . 51 . 01 . 1Buried surface area ( Å2 ) d64469612541380Interfacing residuesd31214633Hydrogen-bonds630eSalt-bridges12eShape complementarity statisticf0 . 630 . 77aThis workbPDB code 2BLF ( Kappler and Bailey , 2005 ) cCalculated as the average for the protomer of interest divided by the average for the entire complex structure . d ( Krissinel and Henrick , 2007 ) eTaken from ( Kappler and Bailey , 2005 ) f ( Lawrence and Colman , 1993 ) 10 . 7554/eLife . 09066 . 016Figure 7 . The bonding network at the interface of SorT/SorU . ( A ) An open-book representation depicts residues involved in forming stable bonds at the interface between SorT and SorU as corresponding color patches mapped onto the molecular surface . ( B ) Stereoview of the interface between SorT and SorU . Bonding residues are shown as sticks with bonds shown as dashes between atoms . SorT is shown in light grey and SorU is shown in magenta . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 016 Compared with the subunit interface of the SorAB bacterial sulfite dehydrogenase , which contains 30 hydrogen-bonds and 2 salt bridges ( Kappler and Bailey , 2005 ) , supporting a permanent , heterodimeric complex of a heme c subunit ( SorB ) and a Mo cofactor containing ( SorA ) catalytic subunit ( Kappler and Bailey , 2005 ) ( Table 7 ) , the extent of the subunit interactions in the SorT/SorU structure is modest . The difference between the permanent SorAB and the transient SorT/SorU complexes is also illustrated by calculations of the shape complementarity and the buried surface areas between the protomers ( Lawrence and Colman , 1993 ) , with the latter being about twice as large for the SorAB assembly than for SorT/SorU ( Table 7 ) . Interestingly , much of the additional contact area between molecules in the SorAB structure derives from the SorB N-terminal structure ( residues B501-B518 , PDB 2BLF ) , which extends away from the core of the subunit , wraps around the SorA ‘SUOX-fold' domain and contributes one salt-bridge and 6 hydrogen bonding interactions ( Table 7 ) ( Kappler and Bailey , 2005 ) . This feature is absent from the SorU structure . Despite the intricate assembly of interactions at the protein-protein interface , but in agreement with the kinetics and thermodynamics of the SorT/SorU interaction in solution , the SorT/SorU contact is dynamic , as illustrated by a temperature factor analysis of the complex structure . Within the SorT/SorU complex , the SorU protein shows a significantly increased average atomic temperature factor ( reflecting significant flexibility ) relative to the structure of the SorT protomer ( 1 . 5 versus 0 . 9 Å2 , respectively; Table 7 ) . Furthermore , the relative temperature factors per residue for the SorU molecule increase with increasing distance from the SorT/SorU interface ( Figure 5A ) , indicating that the SorU molecule is dynamic relative to SorT within the crystalline lattice . In contrast , the ‘static’ SorAB complex shows uniform , low temperature factors for both redox subunits ( Table 7 , Figure 5A ) . This observation is an exquisite illustration of ‘conformational sampling’ within the SorT/SorU electron transfer complex , which results from the conformational flexibility of one protein redox partner relative to the other and both facilitates electron exchange by accessing the optimal orientations of each redox partner and promotes fast dissociation of the complex following transfer ( Leys and Scrutton , 2004; van Amsterdam et al . , 2002 ) .
By describing the structure of the SorT/SorU complex in this work , we report the first example of a structure of an SOE in complex with its external electron acceptor; all previous structures of SOEs being of the enzymes and their internal heme domains or subunits only ( Kisker et al . , 1997; Schrader et al . , 2003; Kappler and Bailey , 2005 ) . The structure of the SorT/SorU complex therefore allows insights into electron transfer in what is thought to be a highly prevalent type of bacterial SOE ( Low et al . , 2011 ) and into protein-protein electron transfer in general . While the complex shows dynamic adaptations similar to those demonstrated previously for electron transfer complexes and has a dissociation constant of the right order of magnitude , it also shows some features that have not been seen in electron transfer complexes , namely an interface that is stabilized by a relatively large number of hydrogen bonds and salt bridges , and an induced fit docking mechanism . The structure of the protein-protein interface in the SorT/SorU structure is particularly intriguing . The relative lack of bound water molecules , mirrors more the observations made of permanent heterodimeric complexes than transient interactions ( Gray and Winkler , 2005 ) . Previous investigations into the factors that influence protein-protein docking for electron transfer have shown that the strength of the protein-protein interaction correlates linearly with the product of the total charges on the protein partners ( Trana et al . , 2012; Xiong et al . , 2010 ) . In this way , the affinity between the SorT/SorU proteins ( Kd = 13 . 5 ± 0 . 8 μM ) correlates with the fast measured turnover rates ( kcat of 140 ± 11 s-1 ) and with the predominantly electrostatic nature of the protein-protein interface . The turnover number for SorT with SorU as the electron acceptor is also in the range of values measured for the human sulfite oxidase ( HSO ) and CSO ( 25 . 0 ± 1 . 3 s-1 and 47 . 5 ± 1 . 9 s-1 , respectively ) , but significantly slower than that observed for the permanent SorAB complex ( 345 ± 11 s-1 ) ( Kappler et al . , 2006; Wilson and Rajagopalan , 2004; Brody and Hille , 1999 ) . Importantly , in the structure of SorAB , the docking site of the heme subunit ( SorB ) with the SorA subunit is almost identically positioned to that seen for SorT/SorU , with major differences existing only in the number of hydrogen bonds and salt bridges in the protein-protein interface . For the CSO and HSO enzymes , docking of the mobile heme b domain near the Mo active site has been proposed to be similar to that seen for SorT/SorU ( Utesch and Mroginski , 2010 ) . It should be noted , however , that the docking events in CSO ( and HSO ) and between SorT/SorU serve fundamentally different purposes: for CSO and HSO , domain docking enables intramolecular electron transfer and involves a heme domain that is an intrinsic part of the enzyme . This is a step that precedes interactions with the external electron acceptor for these enzymes . In contrast , and despite the fact that it is occupying a similar docking site to that predicted for CSO and HSO ( Utesch and Mroginski , 2010 ) , SorU is the external electron acceptor for SorT and the electron transfer is intermolecular . The SorT/SorU complex described here thus represents an elegant compromise between the requirements for fast and efficient electron transfer and reaction specificity . It also illustrates new aspects for highly dynamic protein-protein interactions: ( i ) A relatively large number of hydrogen bonds and salt bridges may be required to form the initial stable protein complex , but this does not preclude a dynamic protein – protein interaction; ( ii ) relatively subtle structural adjustments in one redox partner ( SorU ) can facilitate electron transfer by ideally locating the redox active cofactors in close proximity; ( iii ) the comparatively complex binding interface in SorT/SorU can be counterbalanced by the conformational sampling of one protein relative to the other , which enables the rapid dissolution of the complex following electron exchange . It remains to be seen whether these principles apply to other SOE – external electron acceptor interactions . Future work should focus on investigating interaction interfaces in currently little studied SOEs where new types of interactions may be present as for many of these enzymes currently no external electron acceptor is known .
Recombinant SorT and SorU proteins were overproduced and purified as previously described ( Low et al . , 2011 ) , with minor modifications . SorT was crystallized by hanging drop vapor diffusion with drops consisting of equal volumes ( 2 μL ) of protein and crystallization solution ( 0 . 1 M HEPES pH 7 . 5 , 8% ethylene glycol , 0 . 1 M manganese ( II ) chloride tetrahydrate and 17 . 5% PEG 10 , 000 ) at 20°C . Crystals were cryoprotected in reservoir solution with 30% glycerol before flash-cooling in liquid nitrogen . Small ( ca . 20 × 10 × 10 μ ) crystals of SorU were grown in drops containing equal volumes ( 2 μL ) of protein and reservoir solution ( 1 . 8 M tri-sodium citrate , pH 5 . 5 , 0 . 1 M glycine ) , which were harvested and flash-cooled in liquid nitrogen without additional cryoprotection . Purified SorT ( 20 mM Tris pH 7 . 8 , 2 . 5% glycerol ) and SorU ( 20 mM Tris pH 7 . 8 , 150 mM NaCl ) were mixed and incubated on ice at a molar ratio of 2:1 ( SorU:SorT; total protein concentration 8 mgmL-1 ) before crystallization via hanging-drop vapor diffusion with a reservoir solution containing 0 . 2 M sodium formate , 0 . 1 M Bis-Tris propane pH 7 . 5 and 20% PEG 3350 . Crystals grew to a maximum size of ca . 150 x 100 x 20 μ in 4 days at 20°C and were flash-cooled in liquid nitrogen after brief soaking in mother liquor containing 30% glycerol . All diffraction data were collected on an ADSC Quantum 315r detector at the Australian Synchrotron on beamline MX2 at 100 K and were processed with HKL2000 ( Otwinowski and Minor , 1997 ) . Unit cell parameters and data collection statistics are presented in Table 1 . The crystal structure of SorT was solved by molecular replacement using PHASER ( McCoy et al . , 2007 ) with a search model generated with CHAINSAW ( Chainsaw , 2008 ) from the SorA portion of the SorAB crystal structure ( 29 . 0% sequence identity , Protein Data Bank entry 2BLF [Kappler and Bailey , 2005] ) as a template ( Larkin et al . , 2007 ) . The resulting model was refined by iterative cycles of amplitude based twin refinement ( using twin operators H , K , L and –H , -K , L with estimated twin fractions of 0 . 495 and 0 . 505 respectively ) within REFMAC ( Murshudov et al . , 2011 ) , interspersed with manual inspection and correction against calculated electron density maps using COOT ( Emsley and Coot , 2004 ) . The refinement of the model converged with residuals R = 0 . 208 and Rfree = 0 . 239 ( Table 1 ) . The structure of the SorT/SorU complex was solved by molecular replacement using PHASER ( McCoy et al . , 2007 ) , with the refined SorT structure as a search model . Initial rounds of refinement yielded a difference Fourier electron density map , which clearly showed positive difference density for the location of one molecule of SorU per asymmetric unit , which was manually built using COOT ( Emsley and Coot , 2004 ) . Refinement was carried out with REFMAC5 ( Murshudov et al . , 2011 ) and PHENIX ( Adams et al . , 2002 ) and converged with residuals R = 0 . 211 and Rfree = 0 . 260 ( Table 1 ) . The refined SorU model , from the SorT/SorU complex structure , was used as a search model to solve the SorU structure by molecular replacement using PHASER ( McCoy et al . , 2007 ) . Refinement was carried out with REFMAC5 and PHENIX ( Adams et al . , 2002 ) and converged with residuals R= 0 . 192 and Rfree = 0 . 240 . All structures were judged to have excellent geometry as determined by MOLPROBITY ( Chen et al . , 2010 ) ( Table 1 ) . SAXS analysis of the SorT/SorU complex was performed in a buffer of 20 mM Tris pH 7 . 8 , 2 . 5% v/v glycerol . Purified SorU and SorT were mixed and incubated on ice at a molar ratio of 2:1 ( SorU:SorT ) , generating two samples of total protein concentrations 2 . 75 and 6 . 25 mgmL-1 , respectively . SAXS data were measured as described previously ( Jeffries et al . , 2011 ) with the data collection parameters listed in Table 4 . Data were reduced to I ( q ) vs q ( q=4πsinθλ , where q=4sin2θ is the scattering angle ) using the program SAXSquant that includes corrections for sample absorbance , detector sensitivity , and the slit geometry of the instrument . Intensities were placed on an absolute scale using the known scattering from H2O . Protein scattering was obtained by subtraction of the scattering from the matched solvents ( 20 mM Tris pH 7 . 8 , 2 . 5% v/v glycerol obtained from the flow-through after protein concentration by centrifugal ultrafiltration ) . Molecular weight ( Mr ) estimates for the proteins were made using the equation from Orthaber ( Orthaber et al . , 2000 ) : Mr=NAI ( 0 ) C△ρM2 where NAis Avogadro’s number , C is the protein concentration and △ρM=△ρυ , where △ρ is the protein contrast and υ the partial specific volume , both of which were determined using the program MULCh ( Whitten et al . , 2008 ) . The ATSAS program package ( Volkov and Svergun , 2003 ) was used for data analysis and modeling , with the specific programs used detailed in Table 4 , along with the data ranges and results of each of the calculations . Further detail on data interpretation and analysis for these experiments is detailed in Appendix 2 . Continuous-wave X-band ( ca . 9 GHz ) ( CW ) electron paramagnetic resonance ( EPR ) spectra were recorded with a Bruker Elexsys E580 CW/pulsed EPR spectrometer fitted with a super high Q resonator; the microwave frequency and magnetic field were calibrated with a Bruker microwave frequency counter and a Bruker ER 036TM Teslameter , respectively . A microwave power of 20 mW was used and optimal spectral resolution was obtained by keeping the modulation amplitude to a 1/10 of the linewidth . A flow-through cryostat in conjunction with a Eurotherm ( B-VT-2000 ) variable temperature controller provided temperatures of 127–133 K at the sample position in the cavity . Bruker’s Xepr ( version 2 . 6b . 45 ) software was used to control the data acquisition including , spectrometer tuning , signal averaging , temperature control and visualization of the spectra . Computer simulation of the EPR spectra were performed with the following spin Hamiltonian ( Equation 2 ) ( 2 ) H = βB·g·S + S·A ( 95 , 97Mo ) · I -gnβB·I + ∑i=12 ( S·A ( 1H ) ·I - gnβnB·I ) using the XSophe-Sophe-XeprView ( version 1 . 1 . 4 ) computer simulation software suite ( Hanson et al . , 2004; Hanson et al . , 2013 ) on a personal computer , running the Mandriva Linux v2010 . 2 operating system . Further detail on data interpretation and analysis for these experiments is detailed in Appendix 1 . The MoIV/V and MoV/VI redox potentials of SorT were determined by an EPR-monitored redox titration carried out in a Belle technology anaerobic box . The protein solution ( 1 . 5 mL , 40-90 μM in Tris-HCl , pH 8 . 0 and 10% glycerol ) also contained the following redox mediators at concentrations of ~50 µM: diaminodurol ( 2 , 3 , 5 , 6-tetramethylphenylene-1 , 4-diamine , Em , 7 +276 mV ) , dichlorophenolindophenol ( Em , 7 +217 mV ) , 2 , 6-dimethylbenzoquinone ( Em , 7 +180 mV ) , phenazine methosulfate ( Em , 7 +80 mV ) , 2 , 5-dihydroxybenzoquinone ( Em , 7 –60 mV ) indigo trisulfonate ( Em , 7 -90 mV ) , 2-hydroxy-1 , 4-naphthoquinone ( Em , 7 -152 mV ) and anthraquinone 2-sulfonate ( Em , 7 -230 mV ) . The reductant was Ti ( III ) citrate and the oxidant was NaS2O8 ( both ~100 mM ) . After addition of titrant and equilibration ( 15–30 min ) , the equilibrium potential was measured with a combination Pt wire/AgAgCl redox electrode attached to a Hanna 8417 meter calibrated against the quinhydrone redox couple ( Eo′ ( pH 7 ) = +284 mV vs NHE ) . A 100 μL aliquot of protein was withdrawn and transferred to an EPR tube ( in the anaerobic box ) which was then sealed and then carefully frozen in liquid nitrogen ( outside the box ) . Potentials for all experiments were measured with a combination Pt wire-Ag/AgCl electrode attached to a Hanna 8417 meter . The intensity of the MoV signal ( I ) was recorded as a function of measured potential ( E ) , Figure 6 . Spectroelectrochemistry of SorU in isolation and the SorT:SorU complex was performed with a Bioanalytical Systems BAS100B/W potentiostat connected to a Bioanalytical Systems thin layer spectroelectrochemical cell ( 0 . 5 or 1 mm pathlength ) bearing a transparent Au mini-grid working electrode , a Pt wire counter and Ag/AgCl reference electrode . Redox mediators ( all Fe complexes ) used in the experiment were employed at concentrations of 50 μM ( Figure 8 ) . 10 . 7554/eLife . 09066 . 017Figure 8 . Redox mediators employed in optical spectroelectrochemistry experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 09066 . 017 None exhibit any significant absorption in the spectral range of interest at micromolar concentrations . The total solution volume was ca . 700 μL . The buffer was 20 mM Tris ( pH 8 ) containing 200 mM NaCl as supporting electrolyte . The SorU concentration was ca . 50 μM , while experiments on the SorU:SorT complex used approximately equal concentrations of both proteins ( 50 μM ) . Spectra were acquired within a Belle Technology anaerobic box with an Ocean Optics USB2000 fibre optic spectrophotometer . Initially , the cell potential was poised at ca . -100 mV and the system was allowed to equilibrate until no further spectral changes were apparent ( fully reduced SorU ) . The potentials were then increased in 50 mV increments and the spectrum was measured when no further changes were seen ( 5–10 min ) . When the protein was fully oxidized the potential was scanned in the reverse direction in 50 mV intervals to establish reversibility . Data were fitted using the program ReactLab Redox ( Maeder and King ) . Affinity measurements were conducted using a Microcal ITC200 system ( GE Healthcare ) at 25°C using SorT and SorU in buffer ( 25 mM HEPES pH 7 . 5 , 150 mM NaCl and 2 . 5% glycerol ) at final concentrations of 300 μM and 30 μM , respectively . SorT at a concentration of 300 μM was titrated with eighteen injections ( 2 . 0 μl each ) of SorU . All affinity measurements were performed in triplicate and fitted using a single site mode . Protein concentrations were estimated using Bicinchoninic acid ( BCA ) protein assay kit ( Thermo Fisher Scientific , Waltham , MA ) . Sulfite dehydrogenase enzyme assays were carried out as described previously ( Low et al . , 2011; Wilson and Kappler , 2009; Kappler et al . , 2000 ) . The reduced – oxidized extinction coefficient for SorU at 550 nm was 17 . 486 mM-1 cm-1 as determined by spectroelectrochemistry . Data fitting was carried out using Sigmaplot 12 ( Systat ) . | A key feature of many important chemical reactions in cells is the transfer of particles called electrons from one molecule to another . The sulfite oxidizing enzymes ( or SOEs ) are a group of enzymes that are found in many organisms . These enzymes convert sulfite , which is a very reactive compound that can damage cells , into another compound called sulfate . As part of this process the SOE transfers electrons from sulfite to other molecules , such as oxygen or a protein called cytochrome c . In the past , researchers have described the three-dimensional structure of three SOEs using a technique called X-ray crystallography . However , it has been difficult to study how SOEs pass electrons to other molecules because of the temporary nature of the interactions . McGrath et al . studied an SOE called SorT , which is found in bacteria . The SorT enzyme passes electrons from sulfite to another protein called SorU . McGrath used X-ray crystallography to determine the three-dimensional structures of versions of these proteins from a bacterium called Sinorhizobium meliloti . This included structures of the proteins on their own , and when they were bound to each other . These structures revealed that a subtle change in the shape of SorU occurs when the proteins interact , which enables an electron to be quickly transferred . McGrath et al . also found that the interface between the two proteins showed an unexpectedly high number of contact sites . These strengthen the interaction between the two proteins , which helps to make electron transfer more efficient . However , these contact sites do not prevent the two proteins from quickly moving apart after the electrons have been transferred . The next challenge is to find out whether these observations are common to SOEs from other forms of life . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Structural basis of interprotein electron transfer in bacterial sulfite oxidation |
Previously , we reported that long-term memory ( LTM ) in Aplysia can be reinstated by truncated ( partial ) training following its disruption by reconsolidation blockade and inhibition of PKM ( Chen et al . , 2014 ) . Here , we report that LTM can be induced by partial training after disruption of original consolidation by protein synthesis inhibition ( PSI ) begun shortly after training . But when PSI occurs during training , partial training cannot subsequently establish LTM . Furthermore , we find that inhibition of DNA methyltransferase ( DNMT ) , whether during training or shortly afterwards , blocks consolidation of LTM and prevents its subsequent induction by truncated training; moreover , later inhibition of DNMT eliminates consolidated LTM . Thus , the consolidation of LTM depends on two functionally distinct phases of protein synthesis: an early phase that appears to prime LTM; and a later phase whose successful completion is necessary for the normal expression of LTM . Both the consolidation and maintenance of LTM depend on DNA methylation .
Since the pioneering work of the Flexners ( Flexner et al . , 1963 ) , Agranoff ( Agranoff and Klinger , 1964 ) and their colleagues , it has been widely accepted that memory consolidation—the process by which a labile , short-term memory trace is transformed into a stable , long-term trace ( Lechner et al . , 1999; Müller and Pilzecker , 1900 ) —requires protein synthesis ( Davis and Squire , 1984; Goelet et al . , 1986; Hernandez and Abel , 2008 ) . Moreover , the protein synthesis underlying memory consolidation has been observed to exhibit a temporal gradient: inhibition of protein synthesis during or shortly after training appears to be maximally disruptive of memory consolidation; significantly less amnesia is caused by protein synthesis inhibition ( PSI ) that commences approximately an hour or more after training ( Davis and Squire , 1984 ) . In general , there has been little evidence for a functional distinction with respect to memory consolidation between protein synthesis that occurs during training—hereafter ‘early’ protein synthesis—and protein synthesis that occurs within the first hour or so after the end of training—hereafter ‘late’ protein synthesis . Thus , studies have classically observed significant disruptive effects on the consolidation of long-term memory ( LTM ) whether a protein synthesis inhibitor is applied either immediately prior to , or immediately after , training ( e . g . , Agranoff et al . , 1966; Barondes and Cohen , 1968; Flexner and Flexner , 1968 ) . Consequently , both early and late protein synthesis are commonly regarded as participating in a more-or-less unitary consolidative process . In particular , protein synthesis is believed to mediate critical late events in memory consolidation , including late gene transcription via the synthesis of transcription factors , such as the CCAAT/enhancer-binding protein ( C/EBP ) , and the consequent synthesis of proteins involved in the construction of new synaptic connections ( Bailey et al . , 2015; Kandel et al . , 2014 ) . One mechanism increasingly implicated in the consolidation of LTM is the epigenetic process of DNA methylation ( Levenson et al . , 2006; Maddox et al . , 2014; Miller et al . , 2008; Monsey et al . , 2011; Oliveira , 2016; Rajasethupathy et al . , 2012 ) . However , the relationship between protein synthesis and DNA methylation in memory consolidation is unclear . Mechanistically , is protein synthesis upstream or downstream of DNA methylation during consolidation ? DNA methylation is usually associated with gene silencing . If DNA methylation is required for the synthesis of necessary consolidative proteins , this would imply that a prerequisite for this synthesis is the silencing of one or more repressor genes . On the other hand , it is possible that activation of DNA methyltransferase ( DNMT ) , the family of enzymes that catalyze the transfer of a methyl group to DNA , during memory consolidation itself depends on protein synthesis . Of course , these two possibilities are not mutually exclusive . Here , we have examined the potentially distinctive roles of early and late protein synthesis in the consolidation of the LTM for behavioral sensitization in Aplysia . In addition , we have tested the effect on memory consolidation of both early and late inhibition of DNA methylation . We find that LTM can be induced by partial training , which is insufficient to induce LTM in naïve ( untrained ) animals , after the disruption of LTM by late , but not early , administration of a protein synthesis inhibitor . By contrast , both early and late inhibition of DNMT block LTM consolidation as indicated by the preclusion of subsequent memory induction by partial training . These results point to a functional distinction between early and late protein synthesis in memory consolidation , and suggest a potential role for early protein synthesis in DNA methylation . Finally , we show that inhibition of DNMT disrupts not only the consolidation , but also the persistence , of LTM; thus , the maintenance of consolidated LTM requires ongoing DNA methylation .
Animals were given training that induced long-term sensitization ( LTS ) of the siphon-withdrawal reflex ( SWR ) ; this training ( hereafter 5X training ) consisted of five bouts of tail shocks spaced 20 min apart ( Cai et al . , 2011 , 2012; Chen et al . , 2014 ) ( Figure 1 ) . Control animals received no training . Then , ~15 min ( range = 10–20 min ) after training , trained animals received an intrahemocoelic injection of anisomycin , a protein synthesis inhibitor , or vehicle solution . Control animals received an injection of vehicle solution at the equivalent experimental time . At 24 h after training the duration of the SWR was tested in all the animals ( 24-h posttest ) , after which two of the groups—the control group ( Control-Veh-3XTrained ) and a group that had received the long-term training plus the posttraining injection of anisomycin ( 5XTrained-Aniso-3XTrained ) were given additional sensitization training , which consisted of three bouts of tail shocks spaced 20 min apart . Previously , we found that this training ( hereafter 3X training ) , which is insufficient to induce LTM in naïve animals , can successfully reinstate LTM following its disruption by inhibition of PKM Apl III—the Aplysia homolog of PKMζ ( Bougie et al . , 2012 , 2009 ) —or memory reconsolidation blockade ( Chen et al . , 2014 ) . All the groups , including the two that received the 5X training but not the 3X training ( 5XTrained-Veh and 5XTrained-Aniso groups ) , were given another test at 48 h after training or at the equivalent experimental time ( 48-h posttest ) . 10 . 7554/eLife . 18299 . 003Figure 1 . LTS can be established by truncated training following its disruption by posttraining PSI . ( A ) Experimental protocols . The times of the pretests , training , posttests , and drug/vehicle injections are shown relative to the end of the fifth bout of sensitization training ( time = 0 ) . The time of the intrahemocoelic injection of anisomycin or vehicle is indicated by the red arrow . After the 24-h posttest , animals in the Control-Veh-3XTrained and 5XTrained-Aniso-3XTrained groups received truncated sensitization training ( 3 bouts of tail shocks ) . ( B ) The mean duration of the SWR measured at 24 h and 48 h for the Control-Veh-3XTrained ( n = 7 ) , 5XTrained-Veh ( n = 6 ) , 5XTrained-Aniso ( n = 5 ) , and 5XTrained-Aniso-3XTrained ( n = 6 ) groups . A repeated-measures ANOVA indicated that there was a significant group x time interaction ( F[6 , 40] = 210 . 9 , p < 0 . 0001 ) . Subsequent planned comparisons indicated that the overall differences among the four groups were highly significant on all of the posttests ( 24 h , F[3 , 20] = 456 . 7 , p < 0 . 0001; and 48 h , F[3 , 20] = 250 . 6 , p < 0 . 0001 ) . SNK posthoc tests revealed that the initial sensitization training produced significant LTS , as indicated by the increased mean duration of the SWR , in the 5XTrained-Veh group ( 56 . 7 ± 2 . 2 s ) at 24 h compared with that in the Control-Veh-3XTrained ( 1 . 9 ± 0 . 9 s , p < 0 . 001 ) . The mean duration of the SWR in the 5XTrained-Veh group at 24 h was also significantly longer than that in the 5XTrained-Aniso ( 1 . 4 ± 0 . 4 s , p < 0 . 001 ) , and 5XTrained-Aniso-3XTrained groups ( 1 . 7 ± 0 . 7 s , p < 0 . 001 ) . The differences among the Control-Veh-3XTrained , 5XTrained-Aniso and 5XTrained-Aniso-3XTrained groups were not significant at 24 h . The mean duration of the SWR in the 5XTrained-Veh group ( 58 . 2 ± 1 . 8 s ) was still protracted at 48 h , and was significantly longer than that in the Control-Veh-3XTrained group ( 1 . 1 ± 0 . 2 s ) , as well as that in the 5XTrained-Aniso group ( 1 . 2 ± 0 . 2 s , p < 0 . 001 for both comparisons ) . LTS was induced in the 5XTrained-Aniso-3XTrained group by the three additional tail shocks applied after the 24-h posttest . The mean duration of the SWR in this group at 48 h was 49 . 7 ± 3 . 4 s , which was significantly longer than that for the Control-Veh-3XTrained group . In addition , the mean duration of the reflex in the 5XTrained-Aniso-3XTrained group was longer than that in 5XTrained-Aniso group , but still significantly shorter than that in the 5XTrained-Veh group at 48 h ( p < 0 . 01 ) . Asterisks , comparisons of the 5XTrained-Veh group with the Control-Veh-3XTrained group , the 5XTrained-Aniso group , and 5XTrained-Aniso-3XTrained group at 24 h; and comparisons of the 5XTrained-Veh group with the Control-Veh-3XTrained group and the 5XTrained-Aniso group at 48 h . Pound signs , comparison of the 5XTrained-Veh with the 5XTrained-Aniso-3XTrained group at 48 h . Plus signs , comparisons of the 5XTrained-Aniso-3XTrained group with the Control-Veh-3XTrained and 5XTrained-Aniso groups at 48 h . Here and in subsequent figures one symbol indicates p < 0 . 05; two symbols , p < 0 . 01; and three symbols , p < 0 . 001 . Error bars in this and subsequent figures represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 003 The group given 5X training followed by vehicle injection ( 5XTrained-Veh group ) exhibited significant sensitization at 24 h after training compared with the two groups that received the 5X training followed by anisomycin injection ( 5XTrained-Aniso and 5XTrained-Aniso-3XTrained groups ) , and with the control group ( Control-Veh-3XTrained ) . The subsequent 3X training produced LTS in the 5XTrained-Aniso-3XTrained group , as shown by the results for the 48-h posttest . ( Notice that the 3X training did not induce LTS in animals that did not receive prior 5X training . ) Thus , although posttraining PSI produced complete amnesia for sensitization at 24 h , it did not preclude the subsequent establishment of LTS by partial training . Previous work ( Montarolo et al . , 1986 ) examined the effect of PSI at various times after training on serotonin ( 5HT ) -dependent , long-term facilitation ( LTF ) of the sensorimotor synapse in dissociated cell culture , an in vitro homolog of LTS in Aplysia ( Kandel , 2001 ) . This work demonstrated that LTF , tested at 24 h after training , was disrupted by anisomycin applied during a 3-h period that extended from 1 h before the onset of spaced 5HT training through 0 . 5 h after training ( Montarolo et al . , 1986 ) . ( The 5HT training lasted 1 . 5 h . ) By contrast , a 3-h period of anisomycin treatment beginning at either 0 . 5 h or 4 h after the end of 5HT training did not block LTF . Considering our behavioral results ( above ) , we wished to determine whether exposure to a protein synthesis inhibitor immediately after 5HT training would disrupt LTF at 24 h and , if so , whether LTF could subsequently be induced by partial training . Accordingly , some sensorimotor cocultures were treated with anisomycin ( 10 µM , 2 h ) immediately after training with five 5-min pulses of 5HT ( 100 µM; 5X5HT training ) , spaced at 15-min intervals ( Cai et al . , 2008; Chen et al . , 2014 ) ( Figure 2A ) . To test whether LTF could be induced by partial training following its potential disruption by posttraining anisomycin treatment , three spaced pulses of 5HT ( 3X5HT training ) were used . The experiment included a group of cocultures ( 5X5HT group ) that received the full 5HT training , but not the posttraining anisomycin , as well as a group ( Control ) that received neither the 5HT nor the posttraining anisomycin . Finally , there was a group ( 3X5HT ) that received the three pulses of 5HT at 24 h , but not the initial 5X5HT training . ( Cocultures not treated with a drug at a particular point in the experiment were treated with standard perfusion medium instead . ) The application of anisomycin immediately after the 5X5HT training blocked LTF at 48 h ( 5X5HT-Aniso group; Figure 2B , C ) . However , three pulses of 5HT applied 24 h after the original 5HT training and subsequent PSI induced LTF at 48 h ( 5X5HT-Aniso-3X5HT group ) . Notice that the mean EPSP in the 5X5HT-Aniso-3X5HT group was not significantly different from that in the 5X5HT group , which indicates that the supplemental partial training induced normal LTF . These cellular results accord with our behavioral finding that , although immediate posttraining PSI disrupts the consolidation of LTM , later abbreviated training can result in full LTM . Montarolo et al . ( 1986 ) observed significant LTF when cocultures were treated with anisomycin for 3 h , and even for 22 h , beginning 0 . 5 h after the end of 5X5HT training; the present results , taken together with those of Montarolo et al . , indicate that protein synthesis during a period of 30 min or so immediately following the 5X5HT training is critical for the normal consolidation of LTF in Aplysia . 10 . 7554/eLife . 18299 . 004Figure 2 . Partial training induces LTF following its disruption by PSI immediately after long-term training . ( A ) Experimental protocols . The initial training consisted of five 5-min pulses of 100 µM 5HT ( 5X5HT ) spaced at 15-min intervals . Cocultures in the 5X5HT-Aniso and 5X5HT-Aniso-3X5HT groups were treated with anisomycin ( 10 µM , red bar ) for 2 h immediately after the 5X5HT training . Three 5-min pulses of 5HT ( 100 µM; 3X5HT training ) were given to cocultures in the 5X5HT-Aniso-3X5HT group at 24 h after 5X5HT training , as well as to cocultures in the 3X5HT group at the equivalent experimental time . ( B ) Sample EPSPs . Each pair of traces shows EPSPs recorded from the same coculture on the pretest and posttest . Scale bars: 10 mV , 100 ms . ( C ) Graph presenting the mean normalized EPSPs , measured at 48 h , for the five experimental groups: Control ( n = 13 ) , 3X5HT ( n = 12 ) , 5X5HT ( n = 16 ) , 5X5HT-Aniso ( n = 12 ) , and 5X5HT-Aniso-3X5HT ( n = 7 ) . A one-way ANOVA indicated that the overall differences among the five groups were highly significant ( F[4 , 55] = 7 . 9 , p < 0 . 0001 ) . SNK posthoc tests showed that the mean normalized EPSP in the 5X5HT group ( 216 . 6% ± 30 . 6% ) at 48 h was significantly larger than that in the Control ( 113 . 1% ± 11 . 1% , p < 0 . 01 ) , 3X5HT ( 90 . 8% ± 17 . 9% , p < 0 . 001 ) , and 5X5HT-Aniso ( 76 . 1% ± 16 . 4% , p < 0 . 001 ) groups . The mean normalized EPSP in the 5X5HT-Aniso-3X5HT group ( 196 . 5% ± 26 . 8% ) was also significantly larger than that in the Control ( p < 0 . 05 ) , 3X5HT ( p < 0 . 05 ) , and 5X5HT-Aniso ( p < 0 . 05 ) groups . None of the other differences among the groups was significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 004 Castellucci et al . ( 1989 ) found that PSI during the original period of sensitization training produced amnesia . We wished to determine whether LTM could be induced by partial training following its disruption by PSI during sensitization training . Accordingly , we performed an experiment using the identical protocol as that shown in Figure 1 , except that the animals received an injection of either anisomycin or vehicle ~15 min before the onset of the 5X training ( Figure 3 ) . The 5X training induced LTM at the 24-h posttest in animals given the vehicle ( Veh-5XTrained group ) , but not in animals given anisomycin ( Aniso-5XTrained and Aniso-5XTrained-3XTrained groups ) prior to training . Furthermore , 3X training did not produce LTM in animals that received the protein synthesis inhibitor prior to 5X training , as indicated by the lack of sensitization in the Aniso-5XTrained-3XTrained group at 48 h . In contrast to posttraining PSI , therefore , PSI during training prevented induction of LTM by the supplemental truncated training . 10 . 7554/eLife . 18299 . 005Figure 3 . LTS cannot be induced by partial training when PSI occurs during the original ( 5X ) sensitization training . ( A ) Experimental protocols . The times at which the pretests , training , posttests , and drug/vehicle injections occurred are shown relative to the end of the last training session . The red arrow indicates when either anisomycin or vehicle was injected into the hemocoel . ( B ) The mean duration of the SWR measured at 24 h and 48 h for the Veh-Control-3XTrained ( n = 5 ) , Veh-5XTrained ( n = 8 ) , Aniso-5XTrained ( n = 6 ) , and Aniso-5XTrained-3XTrained ( n = 6 ) groups . A repeated-measures ANOVA showed a significant group x time interaction ( F[6 , 42] = 40 . 9 , p < 0 . 0001 ) . Planned comparisons indicated that the group differences were highly significant for both 24-h ( F[3 , 21] = 43 . 9 , p < 0 . 0001 ) and 48-h ( F[3 , 21] = 45 . 4 , p < 0 . 0001 ) posttests . SNK posthoc tests revealed that the 5X training produced sensitization of the SWR in the Veh-5XTrained group ( mean duration = 50 . 1 ± 5 . 6 s ) at 24 h compared with the results for the Veh-Control-3XTrained group ( mean duration of the SWR = 1 . 4 ± 0 . 4 s , p < 0 . 001 ) . In addition , the SWR in the Veh-5XTrained group was significantly longer than that in the Aniso-5XTrained group ( 2 . 2 ± 1 . 2 s , p < 0 . 001 ) and the Aniso-5XTrained-3XTrained group ( 4 . 0 ± 2 . 3 s , p < 0 . 001 ) . The differences among the Veh-Control-3XTrained , Aniso-5XTrained , and Aniso-5XTrained-3XTrained groups were not significant at 24 h . The SWR in the Veh-5XTrained group ( mean duration = 49 . 8 ± 5 . 8 s ) remained sensitized at 48 h , as indicated by the comparison with the reflex in the Veh-Control-3XTrained group ( mean duration = 1 . 2 ± 0 . 2 s ) . The SWR was also significantly prolonged in the Veh-5XTrained group compared with that in the Aniso-5XTrained group ( mean duration = 2 . 0 ± 1 . 0 s , p < 0 . 001 for both comparisons ) . The three tail shocks applied after the 24-h posttest did not establish LTS in the Aniso-5XTrained-3XTrained group . The mean duration of the SWR in this group at 48 h was 2 . 2 ± 1 . 2 s , which was not significantly different from that in the Veh-Control-3XTrained and Aniso-5XTrained groups . The SWR of the Veh-5XTrained group at 48 h was significantly longer than that in the Aniso-5XTrained-3XTrained group ( p < 0 . 001 ) . Asterisks , comparisons of the Veh-5XTrained group with the Veh-Control-3XTrained group , the Aniso-5XTrained group , and the Aniso-5XTrained-3XTrained group . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 005 Why should PSI during training be so devastating for the consolidation of LTM ? An intriguing possibility is that PSI during training obstructs DNA methylation required for memory consolidation . To test this possibility , we performed experiments in which the DNA methyltransferase ( DNMT ) inhibitor RG108 was injected into animals just before the onset of 5X training ( Figure 4 ) . DNMT inhibition during 5X training resulted in amnesia at 24 h and 48 h posttraining ( comparison of the Veh-5XTrained group with the RG-5XTrained group ) ; furthermore , subsequent 3X training did not induce LTM , as shown by the 48-h data ( comparisons of the RG-5XTrained-3XTrained group with the Veh-5XTrained and Veh-Control-3XTrained groups ) . 10 . 7554/eLife . 18299 . 006Figure 4 . DNMT inhibition during the original ( 5X ) sensitization training precludes the ability of subsequent partial training to induce LTS . ( A ) Experimental protocol . The times of occurrence of the pretests , training , posttests , and drug/vehicle injections are shown relative to the end of the last training session . Either RG108 or vehicle was injected into the hemocoel at the time indicated by the red arrow . ( B ) The mean duration of the SWR measured at 24 h and 48 h for the Veh-Control-3XTrained ( n = 7 ) , Veh-5XTrained ( n = 7 ) , RG-5XTrained ( n = 8 ) , and RG-5XTrained-3XTrained ( n = 7 ) groups . A repeated-measures ANOVA indicated that there was a significant group x time interaction ( F[6 , 50] = 73 . 6 , p < 0 . 0001 ) . Subsequent planned comparisons showed that the overall differences among the four groups for the 24-h and 48-h posttests were highly significant ( 24 h , F[3 , 25] = 197 . 9 , p < 0 . 0001; and 48 h , F[3 , 25] = 82 . 8 , p < 0 . 0001 ) . As revealed by SNK posthoc tests , the SWR exhibited sensitization at 24 h in the Veh-5XTrained group ( mean duration = 54 . 0 ± 3 . 4 s ) compared with that in the Veh-Control-3XTrained group ( mean duration = 1 . 3 ± 0 . 3 s , p < 0 . 001 ) . The differences in duration of the SWR at 24 h among the Veh-Control-3XTrained , RG-5XTrained ( 3 . 6 ± 1 . 3 s ) , and RG-5XTrained-3XTrained ( 1 . 9 ± 0 . 6 s ) groups were not significant . Sensitization of the SWR was maintained in the Veh-5XTrained group ( mean duration of the reflex = 55 . 3 ± 3 . 6 s ) at 48 h , as shown by the comparison with the Veh-Control-3XTrained group ( mean duration of the reflex = 1 . 6 ± 0 . 6 s , p < 0 . 001 ) . There were no significant differences among the Veh-Control-3XTrained , RG-5XTrained ( mean duration of the SWR = 3 . 1 ± 1 . 2 s ) , and RG-5XTrained-3XTrained ( mean duration of the SWR = 5 . 7 ± 4 . 4 s ) groups at 48 h , indicating that the three additional bouts of tail shocks given to the latter group after the 24-h posttest failed to induce LTS . Asterisks , comparisons of the Veh-5XTrained group with the Veh-Control-3XTrained group , the RG-5XTrained group , and the RG-5XTrained-3XTrained group . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 006 Next we examined whether DNMT inhibition that commenced after long-term training caused amnesia and , if so , whether subsequent abbreviated training resulted in LTM . Accordingly , we performed an experiment like the previous one , except that RG108 was injected into some animals 10–20 min after , rather than before , 5X training ( Figure 5 ) . Posttraining DNMT inhibition , like pretraining DNMT inhibition , blocked the consolidation of LTM , as shown by the absence of LTS at 24 h and 48 h ( comparisons of the 5XTrained-Veh group with the Control-Veh-3XTrained and 5XTrained-RG groups ) . In addition , supplemental 3X training following posttraining DNMT inhibition did not induce LTM ( comparisons of the 5XTrained-RG-3XTrained group with the 5XTrained-Veh and Control-Veh-3XTrained groups ) . 10 . 7554/eLife . 18299 . 007Figure 5 . Posttraining inhibition of DNMT precludes later induction of LTS by partial training . ( A ) Experimental protocol . The times at which the pretests , training , posttests , and drug/vehicle injections occurred are shown relative to the end of the last training session . The time of the intrahemocoelic injection of either RG108 or vehicle is indicated by the red arrow . After the 24-h posttest , animals in the Control-Veh-3XTrained and 5XTrained-RG-3XTrained groups received 3X sensitization training . ( B ) The mean duration of the SWR measured at 24 h and 48 h for the Control-Veh-3XTrained ( n = 8 ) , 5XTrained-Veh ( n = 8 ) , 5XTrained-RG ( n = 7 ) , and 5XTrained-RG-3XTrained ( n = 6 ) groups . A repeated-measures ANOVA showed that the group x time interaction was significant ( F[6 , 50] = 64 . 7 , p < 0 . 0001 ) . The overall differences among the four groups for the 24-h and 48-h posttests were highly significant , as indicated by a one-way ANOVA ( 24 h , F[3 , 25] = 82 . 6 , p < 0 . 0001; and 48 h , F[3 , 25] = 69 . 2 , p < 0 . 0001 ) . SNK posthoc tests revealed significantly greater sensitization in the 5XTrained-Veh group at 24 h ( mean duration of the SWR = 53 . 1 ± 5 . 1 s ) than in the Control-Veh-3XTrained group ( mean duration of the SWR = 2 . 1 ± 0 . 9 s , p < 0 . 001 ) . The differences among the 5XTrained-RG ( mean duration of the SWR = 1 . 7 ± 0 . 7 s ) , 5XTrained-RG-3XTrained ( mean duration of the SWR = 2 . 2 ± 0 . 7 s ) , and Control-Veh-3XTrained groups at 24 h were not significant . Sensitization persisted in the 5XTrained-Veh group at 48 h ( mean duration of the SWR = 48 . 3 ± 5 . 0 s ) compared with the Control-Veh-3XTrained group ( mean duration of the SWR = 1 . 6 ± 0 . 3 s , p < 0 . 001 ) . The failure of the 3X training to induce sensitization in the 5XTrained-RG-3XTrained group was shown by the lack of significant differences between this group ( mean duration of the SWR = 2 . 8 ± 1 . 2 s ) and the Control-Veh-3XTrained group at 48 h . There was also no significant difference between the mean duration of the reflex in the 5XTrained-RG-3XTrained group and that in the 5XTrained-RG ( 2 . 3 ± 1 . 0 s ) group at 48 h . Asterisks , comparisons of the 5XTrained-Veh group with the Control-Veh-3XTrained group , the 5XTrained-RG group , and the 5XTrained-RG-3XTrained group . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 007 The failure of partial training to induce LTM following posttraining RG108 treatment , coupled with its ability to establish LTM following posttraining PSI , suggests that DNA methylation is a prerequisite for the establishment of the occult priming trace accessed by partial training after posttraining PSI . To examine this possibility , we gave animals long-term behavioral sensitization training and then administered the DNMT inhibitor 24 h after training ( Figure 6 ) . All animals that received the 5X training exhibited LTS at 24 h ( 5XTrained-Veh , 5XTrained-RG and 5XTrained-RG-3XTrained groups ) , as indicated by the comparison with the control group ( Control-Veh-3XTrained ) . LTS was also present at 48 h in animals that received the original long-term training and an injection of the vehicle after the 24-h test ( 5XTrained-Veh group ) , but not in animals that received the 5X training and an injection of RG108 at 24 h ( 5XTrained-RG and 5XTrained-RG-3XTrained groups ) . Moreover , truncated training did not restore LTM in the RG108-treated animals ( comparison between the Control-Veh-3XTrained and 5XTrained-RG-3XTrained groups ) . An additional experiment using a different DNMT inhibitor—5-azacytidine ( 5-aza ) —yielded identical results ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 18299 . 008Figure 6 . Inhibition of DNMT with RG108 eliminates established LTS in Aplysia . ( A ) Experimental protocol . The occurrences of the pretests , training , posttests , and drug/vehicle injections are shown relative to the end of the last training session . Either RG108 or vehicle was injected into the animals at the time indicated by the red arrow . After the 48-h posttest , animals in the Control-Veh-3XTrained and 5XTrained-RG-3XTrained groups received 3 bouts of sensitization training . ( B ) RG108 treatment at 24 h after training abolished LTS . There were four experimental groups: Control-Veh-3XTrained group ( n = 6 ) , 5XTrained-Veh group ( n = 6 ) , 5XTrained-RG group ( n = 6 ) , and 5XTrained-RG-3XTrained group ( n = 6 ) . A repeated-measures ANOVA disclosed a significant group x time interaction ( F[9 , 60] = 22 . 9 , p < 0 . 0001 ) . Subsequent planned comparisons showed that the overall differences among the four groups for the 24- , 48- and 72-h posttests were highly significant ( 24 h , F[3 , 20] = 13 . 8 , p < 0 . 0001; 48 h , F[3 , 20] = 28 . 6 , p < 0 . 0001; and 72 h , F[3 , 20] = 27 . 9 , p < 0 . 0001 ) . Animals in all three groups trained with five bouts of tail shocks exhibited significant sensitization at 24 h , as indicated by SNK posthoc tests . Thus , the mean SWR was longer in the 5XTrained-Veh ( 45 . 7 ± 6 . 9 s ) , 5XTrained-RG ( 42 . 2 ± 6 . 7 s ) , and 5XTrained-RG-3XTrained ( 47 . 5 ± 6 . 5 s ) groups than that in the Control-Veh-3XTrained group ( 2 . 0 ± 0 . 7 s; p < 0 . 001 for each comparison ) . However , although the 5XTrained-Veh group exhibited significant sensitization on both the 48-h ( mean SWR = 43 . 7 ± 7 . 6 s ) and 72-h ( mean SWR = 41 . 0 ± 7 . 3 s ) posttests , sensitization was absent in both groups of RG108-treated animals after 24 h . Posthoc tests revealed no significant differences for any of the comparisons between the Control-Veh-3XTrained group and the 5XTrained-RG group , or the 5XTrained-RG-3XTrained group , on the posttests after 24 h . Therefore , inhibiting DNMT with RG108 24 h after training erased established LTS . There was no evidence of spontaneous recovery of sensitization over the 48-h period after RG108 injection; furthermore , three additional bouts of training failed to reinstate LTS . Asterisks , comparisons of the 5XTrained-Veh , 5XTrained-RG , and 5XTrained-RG-3XTrained groups with the Control-Veh-3XTrained group at 24 h; and comparison of the 5XTrained-Veh group with the Control-Veh-3XTrained group at 48 h and 72 h . Plus signs , comparisons of the 5XTrained-Veh group with the 5XTrained-RG and 5XTrained-RG-3XTrained groups at 48 h and 72 h . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 00810 . 7554/eLife . 18299 . 009Figure 6—figure supplement 1 . Inhibition of DNMT with 5-azadeoxycytidine ( 5-aza ) also eliminates established LTS in Aplysia . ( A ) Experimental protocol . The occurrences of the pretests , training , posttests , and drug/vehicle injections are shown relative to the end of the last training session . The red arrow indicates the time at which either the drug or the vehicle was injected into animals . After the 48-h posttest , animals in the Control-Veh-3XTrained and 5XTrained-5aza-3XTrained groups received 3 additional bouts of sensitization training . ( B ) 5-aza treatment at 24 h after training abolished LTS . There were four experimental groups: Control-Veh-3XTrained group ( n = 11 ) , 5XTrained-Veh group ( n = 8 ) , 5XTrained-5aza group ( n = 8 ) , and 5XTrained-5aza-3XTrained group ( n = 10 ) . A repeated-measures ANOVA indicated that there was a significant group x time interaction ( F[9 , 99] = 132 . 2 , p < 0 . 0001 ) . The overall differences among the four groups at 24 h , 48 h and 72 h were highly significant ( 24 h , F[3 , 33] = 145 . 9 , p < 0 . 0001; 48 h , F[3 , 33] = 145 . 7 , p < 0 . 0001;and 72 h , F[3 , 33] = 99 . 3 , p < 0 . 0001 ) , as revealed by a one-way ANOVA . SNK posthoc tests on the 24-h data showed that the mean SWR was significantly more prolonged in the 5XTrained-Veh ( 53 . 9 ± 3 . 5 s ) , 5XTrained-5aza ( 59 . 4 ± 0 . 6 s ) , and 5XTrained-5aza-3XTrained ( 55 . 0 ± 3 . 4 s ) groups than that in the Control-Veh-3XTrained group ( 1 . 3 ± 0 . 3 s; p < 0 . 001 for each comparison ) . The 5XTrained-Veh group also exhibited significant sensitization at both 48 h ( mean duration of the SWR = 55 . 3 ± 4 . 8 s ) and 72 h ( mean duration of the SWR = 54 . 4 ± 5 . 6 s; p < 0 . 001 for both comparisons with the Control-Veh-3XTrained group ) . However , sensitization was absent in the 5XTrained-5aza and 5XTrained-5aza-3XTrained groups at both 48 h and 72 h , as indicated by comparisons with the Control-Veh-3XTrained group . There were no significant differences for any of the comparisons among the Control-Veh-3XTrained , 5XTrained-5aza , and 5XTrained-5aza-3XTrained groups for the tests after 24 h . Asterisks , comparisons of the 5XTrained-Veh , 5XTrained-5aza , and 5XTrained-5aza-3XTrained groups with the Control-Veh-3XTrained group at 24 h; and comparison of the 5XTrained-Veh group with the Control-Veh-3XTrained at 48 h and 72 h . Plus signs , comparisons of the 5XTrained-Veh group with the 5XTrained-5aza and 5XTrained-5aza-3XTrained groups at 48 h and 72 h . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 009 Because the animals were tested at 24 h , which was just before the application of the DNMT inhibitor , it might be argued that the subsequent disruption of LTM was mediated by blockade of memory reconsolidation ( Cai et al . , 2012; Nader , 2015 ) , rather than by the interruption of ongoing gene silencing per se . To examine this possibility , we performed an experiment like that shown in Figure 6 , but omitted the 24-h test of the withdrawal reflex ( Figure 7 ) . Application of the DNMT inhibitor 5-aza a little more than 24 h after the 5X training eliminated LTM , as indicated by the lack of sensitization at 48 h in the 5XTrained-5aza and 5XTrained-5aza-3XTrained groups . In addition , treatment with 5-aza precluded the subsequent establishment of LTM by partial training ( 5XTrained-5aza-3XTrained group ) . Thus , the disruption of consolidated LTM by DNMT inhibition after training cannot be attributed to reconsolidation blockade . 10 . 7554/eLife . 18299 . 010Figure 7 . Disruption of established LTS with inhibition of DNMT is not a reconsolidation-related phenomenon . ( A ) Experimental protocol . The times at which the pretests , training , posttests , and drug/vehicle injections occurred are shown relative to the end of the last training session . The intrahemocoelic injection of either drug or vehicle is indicated by the red arrow . The animals did not receive a 24 h test prior to the drug/vehicle injection . After the 48-h posttest , animals in the Control-Veh-3XTrained and 5XTrained-5aza-3XTrained groups received 3 bouts of sensitization training ( 3X training ) . ( B ) 5-aza injection abolished established LTS in the absence of a posttest at 24 h . Four experimental groups were included: Control-Veh-3XTrained group ( n = 7 ) , 5XTrained-Veh group ( n = 7 ) , 5XTrained-5aza group ( n = 7 ) , and 5XTrained-5aza-3XTrained group ( n = 8 ) . A repeated-measures ANOVA indicated that there was a significant group x time interaction ( F[6 , 50] = 105 . 8 , p < 0 . 0001 ) . Subsequent planned one-way ANOVAs showed that the overall differences among the four groups at both 48 h and 72 h were highly significant ( 48 h , F[3 , 25] = 385 . 4 , p < 0 . 0001;and 72 h , F[3 , 25] = 183 . 3 , p < 0 . 0001 ) . SNK posthoc tests revealed that the SWR in the 5XTrained-Veh group was significantly sensitized at both 48 h ( mean = 56 . 1 ± 2 . 9 s ) and 72 h ( mean = 52 . 4 ± 3 . 7 s ) compared with that in the Control-Veh-3XTrained group ( p < 0 . 001 for each comparison ) . Furthermore , the mean duration of the SWR in the 5XTrained-Veh group was significantly longer than that in the 5XTrained-5aza ( 1 . 1 ± 0 . 1 s at 48 h , p < 0 . 001; and 1 . 6 ± 0 . 6 s at 72 h , p < 0 . 001 ) and 5XTrained-5aza-3XTrained ( 1 . 1 ± 0 . 1 s at 48 h , p < 0 . 001; and 1 . 9 ± 0 . 9 s at 72 h , p < 0 . 001 ) groups . The Control-Veh-3XTrained , 5XTrained-5aza , and 5XTrained-5aza-3XTrained groups did not differ significantly at either 48 h or 72 h . Thus , the erasure of established LTS by inhibition of DNMT ( Figure 6 ) did not require elicitation of the SWR immediately preceding the drug injection . Asterisks , comparisons of the 5XTrained-Veh group with the Control-Veh-3XTrained , 5XTrained-5aza , and 5XTrained-5aza-3XTrained groups at 48 h and 72 h . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 010 A potential explanation for the devastating effect of DNMT inhibition on LTM at 24 h is that the memory for LTS is not fully consolidated by this time . To test whether inhibition of DNA methylation would eliminate LTM at a later posttraining time , we performed experiments in which RG108 was administered 5 d after training . Here , animals were given the standard 5X training , tested for LTS 5 d later , and then given an injection of either RG108 or vehicle solution ( LATE treatment condition ) ( Figure 8A ) . All animals were retested at 6 d , after which some animals were given 3X training . Finally , all animals were tested once more at 7 d . The 5X training produced LTS that persisted for 7 d after training ( 5XTrained-VehLATE group vs . Control-VehLATE-3XTrained group ) ( Figure 8B ) . DNMT inhibition at 5 d after 5X training eliminated LTS ( comparisons between the 5XTrained-RGLATE and 5XTrained-VehLATE groups , and between the 5XTrained-RGLATE and Control-VehLATE-3XTrained groups ) , and partial retraining at 6 d failed to reinstate it ( comparisons between the 5XTrained-RGLATE-3XTrained and 5XTrained-VehLATE groups , and between the 5XTrained-RGLATE-3XTrained and Control-VehLATE-3XTrained groups ) . 10 . 7554/eLife . 18299 . 011Figure 8 . RG108 treatment 5 days after training abolishes LTS in Aplysia . ( A ) Experimental protocol . The occurrences of the pretests , training , posttests , and drug/vehicle injections are shown relative to the end of the last training session . The red arrow indicates the time of the intrahemocoelic injection of RG108 or vehicle . After the day six posttest , animals in some groups received partial sensitization training ( 3 bouts of tail shocks ) . ( B ) RG108 injection at day five after training ( LATE treatment ) erased LTS . There were four experimental groups: Control-VehLATE-3XTrained group ( n = 6 ) , 5XTrained-VehLATE group ( n = 5 ) , 5XTrained-RGLATE group ( n = 6 ) , and 5XTrained-RGLATE-3XTrained group ( n = 6 ) . A repeated-measures ANOVA indicated that there was a significant group x time interaction ( F[9 , 57] = 66 . 3 , p < 0 . 0001 ) . Subsequent planned comparisons showed that the overall differences among the four groups on days 5 , 6 and 7 were highly significant ( day 5 , F[3 , 19] = 43 . 7 , p < 0 . 0001; day 6 , F[3 , 19] = 105 . 1 , p < 0 . 0001; and day 7 , F[3 , 19] = 252 . 5 , p < 0 . 0001 ) . There was significant sensitization at day five prior to RG108/vehicle injection in the 5XTrained-VehLATE ( mean duration of the SWR = 55 . 0 ± 5 . 0 s ) , 5XTrained-RGLATE ( mean duration of the SWR = 56 . 7 ± 2 . 1 s ) , and 5XTrained-RGLATE-3XTrained ( mean duration of the SWR = 49 . 3 ± 6 . 1 s ) groups compared with the Control-VehLATE-3XTrained group ( mean duration of the SWR = 1 . 7 ± 0 . 7 s ) ( p < 0 . 001 for each comparison ) . The 5XTrained-VehLATE group also exhibited robust sensitization on day 6 ( mean duration of the SWR = 52 . 4 ± 5 . 5 s ) and 7 ( mean duration of the SWR = 53 . 0 ± 3 . 3 s ) compared with the Control-VehLATE-3XTrained group . Sensitization was absent in the 5XTrained-RGLATE group on day 6 and 7; thus , there was no spontaneous recovery of LTS during the 48-h period after the application of RG108 . Three bouts of training shortly after the day six posttest did not restore LTS in the 5XTrained-RGLATE-3XTrained group the next day . In particular , the mean duration of the SWR in the 5XTrained-RGLATE-3XTrained group ( 2 . 5 ± 0 . 8 s ) on day 7 was not significantly different from that in the Control-VehLATE-3XTrained group , and was significantly shorter than that in the 5XTrained-VehLATE group ( p < 0 . 001 ) . Asterisks , comparisons of the 5XTrained-VehLATE , 5XTrained-RGLATE , and 5XTrained-RGLATE-3XTrained groups with the Control-VehLATE-3XTrained group on day 5; and comparison of the 5XTrained-VehLATE group with the Control-VehLATE-3XTrained group on days 6 and 7 . Plus signs , comparisons of the 5XTrained-VehLATE group with the 5XTrained-RGLATE and 5XTrained-RGLATE-3XTrained groups on days 6 and 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 011 Another argument against the notion that consolidated LTM can be erased by DNMT inhibition is that DNMT inhibitors may injure the animals , or degrade their health or responsiveness in other ways . As a test for nonspecific , health-related effects of RG108 , we assessed whether LTS could be reinduced in animals by 5X training following the elimination of LTS by administration of the DNMT inhibitor . We performed an experiment like that shown in Figure 6 , except that animals given RG108 at 24 h received 5 bouts of training ( that is , full LTS training ) at 48 h rather than 3 bouts of training ( partial training ) ( Figure 9A ) . As before , LTS was absent 24 h after administration of RG108 ( comparisons between the 5XTrained-Veh group and the 5XTrained-RG and 5XTrained-RG-5XTrained groups at 48 h ) ; nonetheless , 5X training after the 48-h test successfully reinduced LTS ( comparisons between the 5XTrained-RG-5XTrained group and the 5XTrained-Veh and the 5XTrained-RG groups at 72 h ) ( Figure 9B ) . Thus , the apparent elimination of LTM following treatment with RG108 cannot be ascribed to a deleterious effect of the drug on the health of the animals . Taken together , our results strongly argue that the maintenance of LTM in Aplysia requires ongoing DNA methylation . 10 . 7554/eLife . 18299 . 012Figure 9 . Animals can relearn following elimination of LTS by DNMT inhibition . ( A ) Experimental protocol . The times at which the pretests , training , posttests , and drug/vehicle injections occurred are shown relative to the end of the last training session . The time of the intrahemocoelic injection of either RG108 or vehicle is indicated by the red arrow . After the 48-h posttest , animals in the 5XTrained-RG-5XTrained group received a second round of full sensitization training ( five bouts of electrical tail shocks ) . ( B ) Sensitization retraining produced LTS in animals following erasure of LTM by RG108 . There were four experimental groups: Control-Veh group ( n = 6 ) , 5XTrained-Veh group ( n = 6 ) , 5XTrained-RG group ( n = 4 ) , and 5XTrained-RG-5XTrained group ( n = 6 ) . A repeated-measures ANOVA indicated that the group x time interaction was significant ( F[9 , 54] = 61 . 4 , p < 0 . 0001 ) . Subsequent one-way ANOVAs indicated that the overall differences among the four groups at 24 h , 48 h and 72 h were highly significant ( 24 h , F[3 , 18] = 33 . 3 , p < 0 . 0001; 48 h , F[3 , 18] = 70 . 7 , p < 0 . 0001; and 72 h , F[3 , 18] = 54 . 9 , p < 0 . 0001 ) . SNK posthoc tests performed on the 24-h data revealed that the initial training produced significant sensitization in the 5XTrained-Veh group ( mean duration of the SWR = 50 . 8 ± 6 . 1 s ) , 5XTrained-RG group ( mean duration of the SWR = 57 . 8 ± 2 . 3 s ) , and 5XTrained-RG-5XTrained group ( mean duration of the SWR = 54 . 2 ± 5 . 8 s ) compared with Control-Veh group ( mean duration of the SWR = 1 . 2 ± 0 . 2 s; p < 0 . 001 for each comparison ) . The SWR of 5XTrained-Veh group also exhibited sensitization at 48 h ( mean duration = 49 . 2 ± 5 . 2 s ) and 72 h ( mean duration = 45 . 2 ± 5 . 6 s ) compared with the Control-Veh group ( 48 h , mean duration = 1 . 2 ± 0 . 2 s , p < 0 . 001; and 72 h , mean duration = 1 . 2 ± 0 . 2 s , p < 0 . 001 ) . Sensitization memory was significantly disrupted at 48 h in both the 5XTrained-RG ( mean SWR = 1 . 3 ± 0 . 3 s ) and 5XTrained-RG-5XTrained ( mean SWR = 2 . 2 ± 0 . 7 s ) groups by the RG108 injection immediately after the 24-h posttest ( p > 0 . 05 for the comparisons with the Control-Veh group ) . Retraining after the 48-h posttest reestablished full LTM . The mean duration of the SWR in the 5XTrained-RG-5XTrained group at 72 h ( 55 . 8 ± 4 . 2 s ) was significantly greater than that for the Control-Veh group ( mean duration = 1 . 2 ± 0 . 2 s , p < 0 . 001 ) , as well as for the 5XTrained-RG group at 72 h ( 2 . 5 ± 1 . 0 s , p < 0 . 001 ) . Asterisks , comparisons of the 5XTrained-Veh , 5XTrained-RG , and 5XTrained-RG-5XTrained groups with the Control-Veh group at 24 h; comparisons of the 5XTrained-Veh group with the Control-Veh , 5XTrained-RG , and 5XTrained-RG-5XTrained groups at 48 h; and comparison of the 5XTrained-Veh group with the Control-Veh and 5XTrained-RG groups at 72 h . Plus signs , comparison of the 5XTrained-RG-5XTrained group with the 5XTrained-RG group at 72 h . DOI: http://dx . doi . org/10 . 7554/eLife . 18299 . 012
We have shown that protein synthesis during and shortly after sensitization training is essential for the normal consolidation of LTM in Aplysia . Our results therefore confirm previous results obtained in Aplysia by Montarolo et al . ( 1986 ) and Castellucci et al . ( 1989 ) , as well in vertebrates by many groups ( reviewed in Davis and Squire , 1984; Hernandez and Abel , 2008 ) . In addition , however , we have significantly extended prior findings regarding protein synthesis and memory consolidation through our demonstration that LTM can be induced by supplemental partial training following its disruption by PSI shortly after the original ( full ) LTS training , but not following PSI during the original LTS training . Thus , the present results reveal a novel functional distinction between the mnemonic role of protein synthesis during training and that of protein synthesis shortly after training . An early study in Aplysia indicated that bath-applied anisomycin ( 18 µM ) produces rapid ( ≤15 min ) , nearly complete ( 95–99% ) inhibition of protein synthesis , as measured by the incorporation of leucine into proteins in identified central neurons ( Schwartz et al . , 1971 ) . Because the pretraining injections of anisomycin in our study were made 10–20 min prior to the onset of training , and because the duration of the 5X training in our study was ~80 min , the pretraining anisomycin treatment would be expected to produce >90% disruption of protein synthesis in the animals throughout much , if not all , of the training period . The posttraining injections of anisomycin in our study were made 10–20 min after the end of 5X training; if one assumes a maximum post-injection time of 15 min for the onset of significant PSI within the central nervous system ( CNS ) of the animals ( Schwartz et al . , 1971 ) —drugs injected into the hemocoel of Aplysia have ready access to the CNS due to the open circulatory system and lack of a blood-brain barrier in gastropod mollusks ( Abbott et al . , 1986 ) —then the posttraining injections of anisomycin should have begun to inhibit protein synthesis by >90% within 30 min after the end of 5X training . Our results indicate that proteins synthesized during training ( early protein synthesis ) play a special role in the consolidation of LTM . Specifically , early protein synthesis causes the generation of a priming component that allows LTM to be later established by partial training if it is disrupted by posttraining PSI ( Figures 1 and 2 ) . Barzilai et al . ( 1989 ) reported that a 1 . 5-h treatment with 5HT produces the rapid onset ( <30 min ) of the synthesis of several ( unidentified ) proteins; in addition , the synthesis of some of these proteins depends on gene transcription , suggesting that they represent immediate-early proteins . Thus , the priming component induced by early protein synthesis in our study may be the product of immediate-early gene transcription . In support of this idea , Rajasethupathy et al . ( 2012 ) found that exposure to five spaced pulses of 5HT down regulated the expression of the transcriptional repressor CREB2 in Aplysia sensory neurons; moreover , this downregulation depended on methylation of the CREB2 gene , because it was blocked by the DNMT inhibitor RG108 . CREB2 is regarded as a memory suppressor in Aplysia ( Abel et al . , 1998 ) ( although see Hu et al . , 2015 ) ; blockade of CREB2 activity ( by means of a function-blocking antiserum ) in Aplysia sensory neurons has been shown to facilitate the induction of LTF ( Bartsch et al . , 1995 ) . The memory suppressive effect of CREB2 is due , at least in part , to the repression of CREB1 activation and the downstream expression of immediate-early genes , including the C/EBP gene ( Alberini et al . , 1994; Alberini and Kandel , 2014 ) . Thus , the expression of CREB1-dependent immediate-early genes , enabled through 5HT-induced DNA methylation of CREB2 , could form the priming memory component revealed by the present experiments . According to this idea , DNA methylation would be upstream from the early protein synthesis required for memory consolidation in Aplysia . The blockade of protein synthesis shortly after ( i . e . , starting ~30 min after ) training , albeit disruptive of LTM , as indicated by the absence of LTM at 24 h following posttraining PSI , nonetheless does not preclude the subsequent induction of LTM by partial training . ( Note that our results appear to partly contradict those of Montarolo et al . ( 1986 ) who observed that anisomycin applied to sensorimotor cocultures starting at 30 min after 5X5HT training did not block LTF; however , the onset of the anisomycin application in our experiments was ~15 min earlier than in the Montarolo et al . study , which may explain the apparent discrepancy in results . ) In support of the present findings , Shobe et al . ( 2016 ) have recently reported that consolidation of the LTM for sensitization is disrupted by posttraining application of the protein synthesis inhibitor emetine to a reduced preparation of Aplysia . Therefore , PSI after training does not disrupt the memory primer , which is induced by early protein synthesis ( Figure 3 ) . It seems likely , moreover , that it is the persistence of this primer that underlies the ability of truncated training to reinstate LTM following its disruption by reconsolidation blockade or inhibition of PKM ( Chen et al . , 2014 ) . How can truncated training establish LTM following impairment of memory consolidation by posttraining PSI , and also reinstate consolidated LTM after its disruption by reconsolidation blockade or inhibition of PKM ( Chen et al . , 2014 ) ? One possibility is that protein synthesis—in addition to signaling by one or more growth-related factors ( Hu et al . , 2004; Kopec et al . , 2015; Zhang et al . , 1997 ) —during long-term training results in persistent activation of mitogen-activated protein kinase ( MAPK ) ( Martin et al . , 1997; Sharma et al . , 2003b ) , and it is the sustained MAPK activity that enables reinstatement of LTM by partial training . In support of this idea , sensitization training that is sufficient to induce LTM has recently been found to cause prolonged ( 1 h ) posttraining activation of MAPK ( Shobe et al . , 2016 ) . But posttraining PSI disrupts this persistent MAPK activity ( Shobe et al . , 2016 ) , and we have shown that the memory priming signal is maintained despite posttraining blockade of translation ( Figures 1 and 2 ) . Although the priming signal cannot therefore be MAPK itself , it could be a signal downstream from activated MAPK . Martin et al . ( 1997 ) showed that during 5HT-induced LTF MAPK translocates to the nucleus of sensory neurons; this nuclear translocation is required for LTF . Possibly , the priming signal is a nuclear change downstream of MAPK . Epigenetic modifications are attractive candidates for the priming signal . Indeed , MAPK-dependent increases in the phosphorylation of histone H3 have been implicated in LTM in rats ( Chwang et al . , 2006 ) and snails ( Danilova et al . , 2010 ) . Furthermore , we previously showed that inhibition of histone deacetylase ( HDAC ) permitted 3X sensitization training to induce LTM in naïve , untrained animals ( Chen et al . , 2014 ) . Thus , persistent histone modifications induced by 5X training may provide the scaffolding necessary for later reconstruction of LTM by partial training during memory reinstatement . Besides MAPK-dependent epigenetic modifications , other candidates for the priming signal include small , non-coding RNAs , the expression of which is induced by 5HT training during LTF ( Rajasethupathy et al . , 2012 ) . In future research we will seek to identify the memory priming signal . Our demonstration that LTM can be fully established by abbreviated training after being disrupted by posttraining PSI echoes recent findings for contextual fear memory in mice by Tonegawa and colleagues ( Ryan et al . , 2015 ) . These investigators used optogenetic stimulation of hippocampal neurons that had been active during fear conditioning to restore LTM after the induction of retrograde amnesia by posttraining treatment with anisomycin . Similar to our finding that LTF was absent at 24 h in sensorimotor cocultures after 5X5HT training followed by exposure to anisomycin ( Figure 2 ) , Ryan et al . observed an absence of learning-induced long-term potentiation ( LTP ) ( Herring and Nicoll , 2016 ) in hippocampal slices from anisomycin-treated animals; this synaptic disruption correlated with retrograde amnesia . Despite the lack of persistent synaptic changes widely regarded as hallmarks of consolidated memory ( Bailey et al . , 2015; Dudai et al . , 2015 ) , some aspect of LTM nonetheless endured in the two studies . Ryan et al . ( 2015 ) did not test whether LTM could be induced by optogenetic stimulation subsequent to PSI during fear conditioning , nor did they propose a specific storage mechanism for LTM . Our data point to DNA methylation as being essential for the consolidation of LTM . Prior work in mammals also supports a critical role for DNA methylation in memory consolidation ( Halder et al . , 2016; Levenson et al . , 2006; Miller et al . , 2008; Miller and Sweatt , 2007; Monsey et al . , 2011; Oliveira , 2016 ) . More relevantly , Kandel and colleagues reported that RG108 blocks the establishment of LTF in Aplysia sensorimotor cocultures , and that this effect is due , at least in part , to blockade of 5HT-induced methylation of the gene for the transcriptional repressor CREB2 ( Rajasethupathy et al . , 2012 ) . The present results elaborate upon this idea; besides showing that the consolidation of LTM in Aplysia requires epigenetic suppression of one or more memory repressor processes , our results suggest the intriguing possibility that this suppression may depend on early protein synthesis . Although little is known at present regarding the potential role of protein synthesis in DNA modification , a report that activity-dependent induction of a specific gene , Gadd45b , is essential for DNA demethylation in the mammalian hippocampus ( Ma et al . , 2009 ) is consistent with this idea . Furthermore , in their study of the epigenetic regulation of memory consolidation in Aplysia , Rajasethupathy et al . ( 2012 ) found that DNA methylation of CREB2 was regulated by a neuronally expressed Piwi protein . Perhaps the early protein synthesis mediating the consolidation of LTM involves the expression of Piwi in Aplysia . In the absence of direct evidence that protein synthesis triggers DNA methylation in Aplysia , however , it is at least as plausible that the process of DNA methylation is upstream , rather than downstream , of early protein synthesis in the consolidation of LTM , as discussed above . Given , as our results indicate , that the consolidation of LTM in Aplysia depends critically on the silencing of one or more genes whose protein products act to repress memory , what are potential candidates for this memory repressive function ? An obvious candidate , of course , is CREB2 ( Bartsch et al . , 1995; Rajasethupathy et al . , 2012 ) , but there are others . For example , phosphatases have been proposed to subserve memory repression in mammals . Miller and Sweatt ( 2007 ) found that infusion of a DNMT inhibitor into the hippocampus of rats immediately after contextual fear conditioning blocked the consolidation of fear memory as assessed 24 h later , and that this effect was due , in part , to DNA methylation of the gene for protein phosphatase 1 ( PP1 ) . Another phosphatase that may subserve memory repression , and whose gene may become silenced by DNA methylation during learning , is calcineurin ( protein phosphatase 2B ) ( Baumgärtel and Mansuy , 2012 ) . Calcineurin activity suppresses the induction of hippocampal LTP ( Winder et al . , 1998; Winder and Sweatt , 2001 ) . Moreover , genetically overexpressing calcineurin in the brains of mice disrupts the consolidation of LTM ( Mansuy et al . , 1998 ) , whereas genetically inhibiting calcineurin enhances hippocampal LTP and LTM in mice ( Malleret et al . , 2001 ) . In addition , Baumgärtel et al . ( 2008 ) reported that calcineurin activity is inhibited in the amygdala during the consolidation of conditioned taste aversion ( CTA ) in mice , and that the level of calcineurin activity during learning determines the strength of the CTA memory . In Aplysia the effects of genetically inhibiting or overexpressing either PP1 or calcineurin on LTM have yet to be examined . However , both phosphatases modulate the CREB-mediated response to extracellular stimuli in Aplysia signaling pathways ( Hawkins et al . , 2006 ) . Moreover , pharmacological inhibition of calcineurin has been shown to facilitate the induction of the LTM for sensitization in Aplysia ( Sharma et al . , 2003a ) . Besides demonstrating a role for DNA methylation in memory consolidation , the present study shows that ongoing DNA methylation plays a crucial role in memory maintenance . This finding resonates with those from recent work in mammals ( Halder et al . , 2016; Miller et al . , 2010; Mizuno et al . , 2012 ) , as well as in invertebrates ( Biergans et al . , 2015; Lukowiak et al . , 2014 ) . We do not know the identity of the gene or genes whose persistent methylation underlies memory maintenance in Aplysia . The gene for CREB2 is one possibility; but a recent study using sensorimotor cocultures found that a long-lasting increase in postsynaptic CREB2 expression mediated the maintenance of LTF beginning 48 h after the induction of synaptic plasticity ( Hu et al . , 2015 ) , which is inconsistent with a role for the continuous silencing of the CREB2 gene in the maintenance of sensitization memory . Interestingly , Miller et al . ( 2010 ) observed persistent DNA methylation of the gene for calcineurin in the anterior cingulate cortex of rats trained contextual fear conditioning . The notion that ongoing suppression of calcineurin activity via gene silencing mediates LTM maintenance in Aplysia is attractive in light of previous data implicating calcineurin activity in the inhibition of LTM ( Sharma et al . , 2003a ) ; nonetheless , to date no direct evidence supports a role for calcineurin in the maintenance of memory in Aplysia . A unique and fascinating aspect of the present results is that they enable a direct comparison of the effect on the persistence of a single , specific form of memory—LTS in Aplysia—of inhibiting DNA methylation with those for two other manipulations that have been purported to eliminate consolidated LTM , inhibition of PKMζ ( Sacktor , 2011 ) and reconsolidation blockade ( Nader , 2015 ) . Previously , we showed that although inhibition of PKM Apl III , the Aplysia homolog of PKMζ ( Bougie et al . , 2012 , 2009 ) , and reconsolidation blockade both disrupt the consolidated LTM for behavioral sensitization in Aplysia ( Cai et al . , 2011 , 2012 ) , the memory can nonetheless be fully reinstated using truncated sensitization training ( Chen et al . , 2014 ) . By contrast , as shown in the present study , the LTM for sensitization cannot be reinstated after its disruption by inhibition of DNMT ( Figures 6–9 ) . This finding indicates that the ongoing DNA methylation of one or more genes is a precondition for memory maintenance , and , moreover , that some essential priming component of LTM , one whose persistence enables LTM reinstatement , must be impervious to the disruptive effects of PKM inhibition and of reconsolidation blockade , but eliminable by inhibition of DNMT . In summary , we have found that two functionally distinct periods of protein synthesis regulate the consolidation of LTM in Aplysia . The earlier period occurs during training ( and , possibly , extends into the immediate posttraining period as well ) ; it involves the production of a memory priming element . The later period starts within 30 min after training; proteins synthesized during this posttraining period are required for the normal expression of LTM . Nonetheless , inhibition of the synthesis of these late proteins , albeit disruptive of LTM , does not impair the priming element , whose occult presence permits LTM to be established by supplemental partial training following its disruption by posttraining PSI . Finally , we have shown that both the consolidation and maintenance of LTM in Aplysia depend , in part , on gene silencing via DNA methylation . Future work will be required to determine the identity of the gene/s whose DNA methylation is required for the induction and persistence of LTM , and whether this methylation is triggered by early protein synthesis .
Adult Aplysia californica ( 80–120 g ) were supplied by Alacrity Marine Biological Services ( Redondo Beach , CA , USA ) and housed in cooled ( 12–14°C ) , aerated seawater . The behavioral training , testing , and drug injection methods were described in our previous study ( Chen et al . , 2014 ) . The cell culture , electrophysiological recording , 5HT training , and anisomycin treatment methods were like those previously described ( Cai et al . , 2011; Chen et al . , 2014 ) . The protein synthesis inhibitor , anisomycin , was prepared for injection as was previously done ( Chen et al . , 2014 ) . The DNA methyltransferase ( DNMT ) inhibitors RG108 ( Sigma , St Louis , MO ) and 5-azadeoxycytidine ( 5-aza; EMD Millipore , Billerica , MA ) were dissolved in DMSO to a concentration of 25 mM to make a stock solution . To inhibit DNMT , a volume of 100 μl per 100 g of body weight of either RG108 or 5-aza was injected into the animals . The specific times at which the intrahemocoelic injections were made are indicated in the relevant figures . SPSS 22 . 0 ( IBM , Armonk , NY ) was used for the statistical analysis . For the analysis of the behavioral data , a repeated-measures two-way analysis of variance ( ANOVA ) was first performed on the overall data . Once the repeated-measures ANOVA showed a significant interaction , one-way ANOVAs were carried out on the separate test times , followed by Student-Newman-Keuls ( SNK ) posthoc tests for pairwise comparisons . For the synaptic data , the EPSP amplitude of the posttest was normalized to that of the pretest for the same coculture . The normalized data were expressed as means ± SEM . A one-way ANOVA was performed on the overall data , followed by Student-Newman-Keuls ( SNK ) posthoc tests for pairwise comparisons . All reported levels of significance represent two-tailed values unless otherwise indicated . | The formation of long-term memory depends on new proteins being made in the brain . These new proteins are used partly to build the new connections among neurons that essentially store the memory , and must be made within a critical period of time . Experiments on animals have found that new proteins must be made during or shortly after training to form a stable memory; if protein synthesis is blocked during this period , the memory will not be stabilized ( a process also known as memory consolidation ) . Changes that alter the activity of genes in neurons also play essential roles in memory consolidation . One such change involves the attachment of a methyl group – a molecule that contains one carbon atom surrounded by three hydrogen atoms – to the DNA of a gene . This process , called DNA methylation , typically inhibits the activity of the gene . Pearce et al . looked at how completely preventing protein synthesis and DNA methylation disrupted memory consolidation in a type of marine snail called Aplysia . Previously , researchers have exploited this animal’s simple nervous system and behavior to discover basic biological mechanisms of memory that are common to all animals . The snails were given training that increased the likelihood that they would reflexively withdraw part of their body ( called the siphon ) in response to touch . When Pearce et al . inhibited protein synthesis soon after training , the snails did not remember the training when tested 24 hours later , as expected . Further analysis showed , however , that a trace of the memory , referred to as the “priming trace” , remained . Snails that had this priming trace could form a long-term memory after partial training , whereas untrained snails did not form memories after such partial training . Inhibiting the synthesis of proteins during the original training blocked the priming trace , as did inhibiting DNA methylation during or after training . Moreover , inhibiting DNA methylation erased a previously established memory and prevented it from being reinstated by partial training . Overall , the findings of Pearce et al . show that proteins produced in the brain by learning have multiple roles . In addition , both the consolidation and maintenance of long-term memory depend on one or more genes that otherwise suppress memory being inhibited via DNA methylation . Future work will now aim to identify the priming trace and the genes that suppress memory . Knowledge of the priming trace could lead to new treatments for memory-related disorders such as Alzheimer’s disease . Furthermore , identifying genes that can suppress memory might allow us to reduce some of the harmful effects of traumatic experience . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Role of protein synthesis and DNA methylation in the consolidation and maintenance of long-term memory in Aplysia |
Piezo1 ion channels mediate the conversion of mechanical forces into electrical signals and are critical for responsiveness to touch in metazoans . The apparent mechanical sensitivity of Piezo1 varies substantially across cellular environments , stimulating methods and protocols , raising the fundamental questions of what precise physical stimulus activates the channel and how its stimulus sensitivity is regulated . Here , we measured Piezo1 currents evoked by membrane stretch in three patch configurations , while simultaneously visualizing and measuring membrane geometry . Building on this approach , we developed protocols to minimize resting membrane curvature and tension prior to probing Piezo1 activity . We find that Piezo1 responds to lateral membrane tension with exquisite sensitivity as compared to other mechanically activated channels and that resting tension can drive channel inactivation , thereby tuning overall mechanical sensitivity of Piezo1 . Our results explain how Piezo1 can function efficiently and with adaptable sensitivity as a sensor of mechanical stimulation in diverse cellular contexts .
Mechanosensation is essential for cells ranging from bacteria , which must regulate cell volume in response to harsh osmotic stress , to Merkel cells and sensory neurons in human fingertips , which are exquisitely sensitive to gentle touch ( Ranade et al . , 2015 ) . Mechanosensation is initiated through the opening of mechanosensitive ion channels , including the K2P family in vertebrates , NOMPC in Drosophila , and the DEG/ENaC family in Caenorhabditis elegans ( Ranade et al . , 2015 ) . Recently , Piezo proteins were identified as the pore-forming subunits of an excitatory ( non-selective cation ) mechanosensitive channel in metazoans ( Coste et al . , 2010; 2012; Faucherre et al . , 2013; Kim et al . , 2012; Schneider et al . , 2014 ) . Piezos are large proteins with >2500 residues that lack homology to any known proteins ( Coste et al . , 2015 ) . The two mammalian isoforms , Piezo1 and Piezo2 , are widely expressed and play key roles in many physiological processes , including vascular development , red blood cell volume regulation , lineage choice in neural stem cells , and touch sensation in Merkel cells and DRG neurons ( Cahalan et al . , 2015; Li et al . , 2014; Maksimovic et al . , 2014; Pathak et al . , 2014; Ranade et al . , 2014a; Woo et al . , 2014 ) . In mouse , knockout of either isoform is lethal , further emphasizing the functional importance of the protein ( Ranade et al . , 2014a ) . Despite an increased understanding of the various roles Piezos play in many biological processes , the activation mechanism , including the precise physical stimulus that initiates pore opening ( gating ) , is unknown . The recent medium-resolution cryo-electron microscopy structure of mouse Piezo1 revealed many features of the coarse channel architecture , but does not provide conclusive clues about the activation mechanism ( Ge et al . , 2015 ) . In vivo , Piezos respond to diverse forces , including laminar flow and cellular compression ( Lee et al . , 2014; Li et al . , 2014; Ranade et al . , 2014a ) . In heterologous systems , two techniques used to evoke channel activity are direct stimulation of the cell by touching with a blunt glass pipette and application of negative pressure to stretch the membrane in a patch pipette ( Coste et al . , 2010 ) . Both stimuli induce several geometric and energetic changes in the membrane , including alterations in curvature and tension , any of which could in principle activate mechanically activated ion channels ( Sukharev and Corey , 2004 ) . For example , lateral membrane tension is the stimulus for the well-characterized bacterial mechanosensitive ion channel MscL ( Moe and Blount , 2005; Sukharev et al . , 1999 ) . To date , no accessory proteins of Piezos have been identified that could tether the channel to the cellular matrix , suggesting that , as for MscL , the activating stimulus may be directly transmitted through the bilayer . A key feature of Piezos is that during a sustained stimulus , currents decay ( inactivate ) with a typical time course of tens of milliseconds , suggesting continuous modulation of channel availability ( Coste et al . , 2010 ) . Consistent with a high physiological importance for inactivation , Piezo mutations that alter inactivation kinetics are linked to several human diseases , including dehydrated hereditary stomatocytosis , xerocytosis , Marden-Walker and Gordon syndromes , and distal arthrogryposis ( Albuisson et al . , 2013; Bae et al . , 2013; Coste et al . , 2013; McMillin et al . , 2014 ) . Importantly , several of these gain-of-function mutations not only reduce the rate and/or extent of inactivation , but also apparently sensitize Piezos to pressure ( Bae et al . , 2013 ) . Additionally , the only known agonist for Piezo ion channels , Yoda1 , both sensitizes Piezo1 to pressure and slows inactivation ( Syeda et al . , 2015 ) . The pressure sensitivity of Piezos also varies with cell type , with previous reports of half-maximal pressure for activation for Piezo1 ranging from ~−15 to −40 mmHg ( Coste et al . , 2010; Li et al . , 2014; Pathak et al . , 2014 ) . Local membrane tension and curvature also vary among and even within cells , with potentially important implications for Piezo function . Piezo sensitivity is also modulated by proteins including EPAC1 and STOML3; differential expression of these and other modulators both among cells and within a single cell may also regulate the overall sensitivity ( Eijkelkamp et al . , 2013; Poole et al . , 2014 ) . A mechanistic understanding of the link between the activating stimulus , local membrane environment , inactivation , and channel sensitivity may provide specific strategies for pharmacological modulation of Piezo activity . Here , we combine high-resolution , high-contrast imaging with electrophysiology to investigate whether membrane curvature or lateral membrane tension is the physical stimulus for activation of Piezo1 and how sensitivity to this stimulus might be intrinsically regulated . We find that Piezo1 is activated by membrane tension with a high degree of sensitivity ( T50 = 1 . 4 ± 0 . 1 mN/m ) and that this sensitivity is directly influenced by resting membrane tension .
In order to assay the influence of membrane curvature and lateral membrane tension on Piezo ion channel activity , we transiently transfected HEK293t cells with mouse Piezo1-IRES-GFP and performed electrophysiological recordings during application of negative or positive pressure to the membrane patch by using a high-speed pressure clamp system , while simultaneously imaging the membrane inside the patch pipette with high resolution ( 400x ) , differential interference contrast ( DIC ) microscopy . Consistent with previous reports , we observed that in cell-attached patches , negative pressure induced rapidly-inactivating inward currents at −80 mV that are characteristic of Piezo1 ion channels ( Coste et al . , 2010; 2012 ) . Currents first became apparent at −5 mmHg , increased with the magnitude of pressure , and reached saturation at ~−30 mmHg ( Figure 1A ) . As expected , our simultaneous imaging approach revealed that negative pressure also induced a convex curvature of the membrane and that the radius of curvature decreased with increasing magnitudes of pressure ( Figure 1A; Video 1 ) . 10 . 7554/eLife . 12088 . 003Figure 1 . Electrophysiology and high-contrast imaging of Piezo-containing membranes . Pressure-step protocol , representative currents and corresponding images from individual cell-attached patches from a HEK293t cell expressing mouse Piezo1-IRES-GFP , upon negative ( A ) and positive ( B ) pressure stimulation . Pressure-step protocol , respective representative currents and corresponding images from individual inside-out patches upon negative ( C ) and positive ( D ) pressure stimulation . Pressure-step protocol , respective representative currents and images from individual outside-out patches upon negative ( E ) and positive ( F ) pressure stimulation . All patches were held at −80 mV . Scale bars are 2 µm for all images . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 00310 . 7554/eLife . 12088 . 004Video 1 . Response of cell-attached HEK293t patch to stimulation with negative pressure in −5 mmHg increments; acquired at 7 . 9 frames/s , played at 50 frames/s . Video corresponds to cell in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 004 Next , we wanted to probe how Piezo1 would respond to opposite ( concave ) membrane curvature . We therefore applied positive pressure to our membrane patches , which indeed inverted the membrane curvature . Piezo1 currents were also reliably induced by this positive pressure protocol ( Figure 1B ) . Peak current amplitudes again increased with the magnitude of pressure , albeit with a striking difference: pressure responses were right-shifted , as currents first became apparent at +15 mmHg and often did not reach saturation before rupture of the patch . Interestingly , we also often observed small currents upon release of small ( +5 mmHg ) positive pressure stimuli , discussed further below . Together , these observations suggest that in cellular membranes , Piezo1 is activated by both convex and concave membrane curvature . We next wanted to test if membrane sidedness and mechanical stability conferred by the cytoskeleton had any influence in this activation process . We therefore repeated these experiments in both inside-out and outside-out patches , both of which result in disrupted cytoskeletal structure and in the latter , inverted membrane leaflets with respect to the patch pipette . For inside-out patches , both positive and negative pressure again evoked transient inward currents through Piezo1 ion channels ( Figure 1C and D ) . In this patch configuration , the optical density of the membrane was reduced compared to cell-attached patches , consistent with the idea that less cytoskeleton is retained in a cell-detached configuration ( Suchyna et al . , 2009 ) . However , DIC imaging was still sufficiently sensitive to resolve the convex and concave membrane curvatures induced by positive and negative pressure protocols , respectively . In the outside-out configuration , we were only able to consistently visualize the membrane during application of positive pressure , but not during negative pressure , perhaps because excess membrane folded within the tip was conformationally flexible and therefore not resolved at our imaging speed ( Ruknudin et al . , 1991 ) . Still , we obtained the same result as for the other patch configurations: both positive and negative pressure robustly activated Piezo1 currents ( Figure 1E and F ) . In contrast to cell-attached and inside-out patches , we did not consistently observe decay of currents elicited in the outside-out configuration . In all of the above experiments , only negligible currents were elicited in cells transfected with empty vector ( pcDNA ) , showing that the currents induced by membrane curvature are specifically mediated by Piezo1 ( Figure 2A–C ) . While Piezo1-mediated currents were reliably evoked by bidirectional pressure in all patch configurations , individual patches did show some expected variability in their precise current amplitude levels , mostly among different patch configurations . Several variables could potentially change among configurations , including patch surface area , cytoskeletal content , and stability of the gigaseal , any or all of which could in theory contribute to the observed differences . Specifically , currents elicited by positive pressure in outside-out patches were typically larger than in the other configurations , consistent with the observation that a larger surface area of membrane is preserved within the confines of the gigaseal in this configuration ( Figure 1F ) . In contrast , currents elicited by positive pressure in inside-out patches were small , likely because patches that survived multiple positive pressure pulses were biased towards those made using smaller pipettes ( Figure 1D ) . 10 . 7554/eLife . 12088 . 005Figure 2 . Mean Piezo1 current responses for all patch configurations upon positive and negative pressure stimulation . ( A ) Pressure-evoked currents from cell-attached patches from HEK293t cells expressing empty vector ( pcDNA; open circles ) or Mouse Piezo1-IRES-GFP ( closed circles ) . Separate patches were tested for positive and negative pressure stimulation . N = 7 for pcDNA at negative pressure , N = 6 for pcDNA at positive pressure , N = 15 for Piezo1 at negative pressure and N = 12 for Piezo1 at positive pressure . ( B ) Pressure-evoked currents from inside-out patches from HEK293t cells expressing empty vector ( pcDNA; open circles ) or Mouse Piezo1-IRES-GFP ( closed circles ) . Separate patches were tested for positive and negative pressure stimulation . N = 4 for pcDNA at negative pressure , N = 3 for pcDNA at positive pressure , N = 10 for Piezo1 at negative pressure and N = 7 for Piezo1 at positive pressure . ( C ) Pressure-evoked currents from outside-out patches from HEK293t cells expressing empty vector ( pcDNA; open circles ) or Mouse Piezo1-IRES-GFP ( closed circles ) . Separate patches were tested for positive and negative pressure stimulation . N = 3 for pcDNA at negative pressure , N = 7 for pcDNA at positive pressure , N = 6 for Piezo1 at negative pressure and N = 11 for Piezo1 at positive pressure . ( D ) Normalized mean current-pressure relations for all six configurations . For each individual patch currents were normalized to the peak current for that patch . All data points are mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 005 We also observed differences in the pressure required for half-maximal activation ( P50 ) of Piezo1 among patch configurations and directions of curvature ( Figure 2D ) . Cell-attached patches stimulated with negative pressure required less pressure for activation ( P5o = −16 . 7 ± 2 . 8 mmHg; N=15 ) than inside-out patches ( P50 = −35 . 8 ± 4 . 0 mmHg; N=10; P<0 . 05 ) . In contrast , cell-attached and inside-out patches required similar pressures for activation with positive pressure ( cell-attached: P50 = +19 . 3 ± 1 . 2 mmHg; N=12; inside-out: P50 = +26 . 7 ± 4 . 2 mmHg; N=7; P>0 . 05 ) . These values for positive pressure are likely an underestimate of the true P50 values for the positive pressure in both configurations , due to premature rupture of some patches before reaching saturation . For outside-out patches , much larger pressures were required for activation than other configurations ( positive pressure: P50 = +70 . 0 ± 9 . 6 mmHg; N=11; negative pressure: P50 = −39 . 8 ± 3 . 1 mmHg; N = 6 ) . Together , these experiments provide strong evidence that Piezo1 can be activated by both convex and concave membrane curvature . Further , as cytoskeletal content varies with patch configuration , the different sensitivities we observed among configurations and with convex versus concave geometry suggests that Piezo1 sensitivity to a given stimulus ( in this case , pressure ) may vary significantly with the amount of cytoskeletal content and its sidedness , i . e . , whether the cytoskeleton is subjected into a convex vs . concave geometry ( Suchyna et al . , 2009 ) . The radius of curvature ( R ) of a surface exposed to a pressure difference ( Δp ) is directly related to lateral tension ( T ) , as described by Laplace’s law: T = R·Δp/2 . However , while curvature can be either positive or negative ( convex or concave ) , tension is a symmetrical quantity . The fact that Piezo1 ion channels respond well to both convex and concave curvature therefore suggests qualitatively that Piezo1 might be activated by lateral membrane tension . To investigate this symmetric relationship quantitatively , we next measured the radius of membrane curvature ( R ) for each patch and pressure step ( Δp ) with a custom script written in Igor-Pro ( WaveMetrics , Lake Oswego , OR ) and calculated the membrane tension using Laplace’s law ( Figure 3A ) . We focused our analysis on the cell-attached and inside-out configurations , which had the highest quality images ( Figure 1A and C ) . 10 . 7554/eLife . 12088 . 006Figure 3 . Measurement of membrane curvature and quantification of membrane tension . ( A ) Representative image of cell-attached patch and schematic showing orientation of membrane . The solid red line marks the measured position of the membrane and the dashed yellow line is a circular fit to this position . Both steps were performed using a script written in Igor Pro ( see Materials and methods ) . For this representative patch the radius R from the fit ( solid yellow line ) was 2 . 87 µm . ( B ) Current-tension histogram for Piezo1 responses to negative pressure in cell-attached patches from HEK293t cells . For each cell , current-pressure curves were fit with a sigmoid , and each response normalized to the plateau from the fit . Tension was calculated using the measured membrane curvature from the corresponding image for each response and normalized current plotted against tension ( gray circles ) . Data were binned ( bin width 1 mN/m ) and pooled ( black bars; mean ± s . e . m ) . Binned data were fit with a Boltzmann function: Imax/ ( 1+exp ( - ( T-T50] ) /k] ) ) where Imax is the maximal normalized current , T is tension , T50 is the tension of half-maximal activation , and k is the slope factor . The standard deviation of the normalized amplitude for each bin was used to weight the fit . Fit parameters Imax = 0 . 99±0 . 01 , T50 = 2 . 7±0 . 1 mN/m , k = 0 . 8±0 . 1 . N = 15 cells and 218 responses . ( C ) Current-tension histogram for Piezo1 responses to negative pressure in inside-out patches from HEK293t cells . Plot was generated as described in ( B ) . Fit parameters: Imax = 0 . 81±0 . 04 , T50 = 4 . 7±0 . 3 mN/m , k = 1 . 2±0 . 1 . N = 10 cells and 123 responses . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 006 In order to pool data from multiple cells , we normalized each individual patch to its plateau current in response to saturating stimuli ( obtained from a Boltzmann fit; see Materials and methods ) and calculated current amplitude histograms as a function of membrane tension ( Figure 3B and C ) . The binned data were then fit , using the standard deviations for each bin to weight the fit . For cell-attached patches , we found that Piezo1 channel activity is well described by a Boltzmann function with a tension of half-maximal activation T50 = 2 . 7 ± 0 . 1 mN/m and a slope factor k = 0 . 8 ± 0 . 1 ( N = 15 ) . Similarly , for inside-out patches , the distribution was also well-described by a Boltzmann function; however , patches required slightly more tension for activation ( T50 = 4 . 7 ± 0 . 3 mN/m; N = 10; P<0 . 001 vs cell-attached ) and had a shallower slope factor ( k = 1 . 2 ± 0 . 1; P<0 . 001 vs . cell-attached ) . Notably , in both configurations the likelihood of a given pulse leading to patch rupture increased sharply starting around 10 mN/m; this is consistent with the lytic tension of the gigaseal , previously reported to be ~10 mN/m ( Suchyna et al . , 2009 ) . The fact that inside-out patches require a greater increase in tension to open Piezos is also consistent with the notion that the intrinsic resting tension is different in these two configurations ( see below ) . Together , these results demonstrate quantitatively that Piezo1 activation is consistent with membrane tension as the principal stimulus . In addition , this reveals that Piezo1 is a tension sensor with higher sensitivity than previously reported for other mechanically activated ion channels such as MscS and MscL , ( T50 = ~5 mN/m and ~10 mN/m , respectively , when reconstituted in asolectin liposomes or lipid bilayers ( Moe and Blount , 2005; Nomura et al . , 2012; Sukharev , 2002; Sukharev et al . , 1999 ) . The resting tension in a cell-attached gigaseal patch has previously been estimated to be on the order of 0 . 5–4 . 0 mN/m , which is similar in magnitude to the tension of half-maximal activation ( T50 = 2 . 7 ± 0 . 1 mN/m ) we determined in cell-attached patches ( Opsahl and Webb , 1994 ) . Additionally , we observed that membrane patches are already substantially curved at rest , i . e . in the absence of any external pressure difference ( Δp = 0 mmHg ) ( Figure 1A ) . Finally , as mentioned above , we noticed in our cell-attached and inside-out recordings that stimulation with +5 mmHg was not sufficient to activate channels , but that instead a current was evoked upon pressure release ( Figures 1B and D ) . We therefore hypothesized that even at rest ( Δp = 0 mmHg ) , a substantial fraction of Piezo1 ion channels might be stimulated and subsequently inactivated . From this , we predicted that a small positive pressure stimulus , that precisely compensates the resting curvature , should effectively zero membrane tension in the patch dome and therefore , if sufficiently long , such a stimulus would allow Piezos to recover from inactivation . To test these predictions systematically , we again utilized our ability to image and thereby precisely measure membrane curvature while simultaneously measuring Piezo1 currents . Previously , pressure prepulses have been used to modulate the resting state of K2P and MscS mechanosensitive ion channels prior to assaying availability ( Akitake et al . , 2005; Honore et al . , 2006 ) . Here , we developed a novel prepulse protocol , in which we applied pressure steps of varying amplitudes for 5 s ( 0—+10 mmHg , △=1 mmHg ) , followed by a pressure release to 0 mmHg ( Figure 4A ) . Strikingly , with this first prepulse protocol we were able to elicit robust rapidly-inactivating currents in cell-attached patches not during the presence of pressure , but rather upon its release . The current amplitudes depended strongly on the prepulse amplitude in a U-shaped manner , i . e . , currents were maximal after prepulses of ~ +5–6 mmHg and decreased for smaller or larger prepulses ( Figure 4A , B ) . We never observed currents upon release of pressure in cells transfected with empty vector ( pcDNA ) , indicating that these currents were indeed Piezo1-mediated ( Figure 4B ) . Importantly , the biphasic dependence of current on prepulse pressure amplitude was tightly linked to membrane curvature . Specifically , we observed for each individual patch that peak currents occurred at or near the minimal curvature , i . e . , when the membrane was flattest ( Figure 4C ) . Averaging data from N = 14 patches further showed that currents are maximal precisely after prepulses that minimize membrane curvature ( R→∞ ) ( Figure 4D ) . 10 . 7554/eLife . 12088 . 007Figure 4 . Activation of Piezo1 currents upon release of pressure stimulation . ( A ) Left , pressure stimulus protocol and representative currents showing activation of Piezo1 ion channels in a cell-attached patch upon release of a 5 s positive pressure stimulus . Holding potential was −80 mV . Right , corresponding images for 0 , +6 , and +10 mmHg pressure steps with membrane patch radius R fit superimposed ( red dashes ) and calculated radius indicated below . ( B ) Mean peak current upon release of a 5 s positive pressure pulse ( 0 to +10 mmHg ) for cells transfected with empty vector ( pcDNA; N = 9 cells ) and with mouse Piezo1 ( N = 14 cells ) . ( C ) Current-radius relationships for six representative measurements performed as shown in ( A ) . The solid black line is showing the measurement in ( A ) . ( D ) Normalized current-radius relationship for all measurements . For each individual patch currents were normalized to the maximal response from that patch and plotted versus inverse radius . Data were binned ( bin width 0 . 05 µm-1 ) ; bars represent mean normalized amplitude ± s . e . m . for each bin . N = 14 cell-attached patches and 148 responses . ( E ) Pressure-stimulus protocol and representative currents showing the time course of current increase with longer prepulse duration in a patch expressing mouse Piezo1 . ( F ) Mean peak current as a function of prepulse duration for cells transfected with pcDNA or Piezo1 ( N = 9 and N = 11 , respectively ) . For each Piezo1 patch , the prepulse amplitude that caused maximal current for that cell ( determined with protocol in ( A ) ) was used . For our patch pipette sizes this was typically +5 or +6 mmHg; +5 mmHg was used for all pcDNA patches . ( G ) Normalized mean peak current as a function of prepulse duration for cells transfected with Piezo1 . For each individual patch , currents were normalized to maximal response from that patch . Mean data were fit with an exponential function I=Imax + A*exp ( -t-t0 ) /tau . Fit parameters Imax = 0 . 82±0 . 02 , A = 0 . 49±0 . 02 , tau = 2 . 4±0 . 3 ms . N = 11 cells . All data points are mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 007 This result raised the possibility that Piezo1 ion channels could be sensing changes in tension , rather than absolute tension . However , the fact that pressure stimuli that overcompensate resting membrane tension and induce opposite curvature lead to reductions in current amplitude upon pressure release make it implausible that Piezo channels sense changes in tension . Rather , the most direct explanation is that a transient reduction in tension by flattening the membrane patch allows for recovery of Piezo1 ion channels from inactivation . A second prediction from our hypothesis was that Piezo1 recovery from inactivation should manifest itself in a specific time course . We therefore next investigated the relationship between current amplitude and prepulse duration . First , using the above protocol , we determined for each individual patch the precise prepulse amplitude that produced the greatest current upon release of pressure . For our patch pipette sizes ( typically 2–3 MΩ in our standard solutions ) , this was typically +5 mmHg or +6 mmHg . Second , we applied exactly this optimal prepulse stimulus for varying durations from 300 ms to 10 s ( △=0 . 75x ) , followed by a return to 0 mmHg ( Figure 4E ) . Using this second prepulse protocol , we found that the increase in current amplitude with prepulse duration followed an exponential time course with τ = 2 . 4 ± 0 . 3 s ( Figure 4F , G ) . This time constant is comparable to the time-dependence of recovery of Piezo2 ion channels from inactivation by whole-cell poking assays , providing further evidence that the process we are observing reflects recovery from inactivation ( Coste et al . , 2013 ) . Importantly , this finding demonstrates that under standard recording conditions , a substantial fraction of Piezo ion channels is inactivated prior to pressure stimulation . Thus far , our data demonstrate that membrane tension is a potent modulator of Piezo1 responsiveness to a single subsequent stimulus . With this information in hand , we next asked whether the overall sensitivity of Piezo1 could be altered by resting membrane tension . In addition to our own data , two previously reported observations suggested this might be possible . First , human gain-of-function mutations in Piezo1 that reduce channel inactivation also apparently sensitize Piezo1 to pressure stimulation ( Bae et al . , 2013 ) . Second , the only known chemical Piezo1 agonist , Yoda1 , both antagonizes inactivation and shifts the P50 curve towards smaller values ( Syeda et al . , 2015 ) . To test how removing membrane tension would affect overall Piezo sensitivity , we developed a third prepulse protocol . We applied alternating 5 s prepulses of 0 , +5 , or +10 mmHg; each prepulse was followed by a 300 ms test pulse of varying amplitude ( 0 to −50 mmHg; △=5 mmHg ) ( Figure 5A ) . We chose these precise prepulse amplitudes because we had previously observed that +5 mmHg prepulses nearly flattened the membrane , while +10 mmHg prepulses induced opposite ( concave ) curvature that was roughly equivalent in magnitude to the resting ( convex ) curvature ( Figure 4A ) . We chose a 5 s prepulse duration because this was sufficient for nearly complete recovery from inactivation while minimizing premature rupture of patches during long pulses to positive pressures ( Figure 4E ) . As before , this experiment was performed with simultaneous imaging of cell-attached membrane patches . 10 . 7554/eLife . 12088 . 008Figure 5 . Overall Piezo1 sensitivity is regulated by resting membrane tension . ( A ) Stimulus protocol and representative currents from a cell-attached HEK293t cell patch expressing mouse Piezo1-IRES-GFP . The test pulse for these currents was −10 mmHg ( thick purple line ) ; holding potential was −80 mV . Inset shows test currents at magnified scale . ( B ) Peak current-pressure relationships for test pulses ( 0 to −50mmHg , Δ5 mmHg ) following 5 s 0 mmHg , +5 mmHg and +10 mmHg prepulses . . All data points are mean ± s . e . m . N = 8 cell-attached patches ( pcDNA ) and 11 cell-attached patches ( Piezo1 ) . ( C ) Mean patch curvature as a function of time during protocol performed shown in ( A ) . Representative images of one individual patch are shown above . Each patch was tested with no prepulse ( 0 mmHg ) , a +5 mmHg prepulse , and a +10 mmHg prepulse at each test pressure before advancing to the next test pressure . Grey markers show inverse radius during rest periods ( 0 mmHg , between stimuli ) , purple markers show inverse radius during 300 ms test pulses ( 0 to −50 mmHg , Δ5 mmHg ) , orange markers show inverse radius during +5 mmHg or +10 mmHg prepulse . All data points are mean ± s . e . m . N = 11 for cell-attached patches . ( D–F ) Normalized current-tension relationships obtained from protocol shown in ( A ) using no prepulse ( 0 mmHg ) ( D ) , +5 mmHg prepulse ( E ) and +10 mmHg prepulse ( F ) . Currents from individual patches are normalized to the maximal response for each patch . Data were pooled and binned ( bin width 1 mN/m ) ; bars represent mean ± s . e . m . N = 11 patches . Binned data were fit with a Boltzmann function I = Imax/ ( 1+exp ( - ( T-T50/k ) ) where I is normalized current , Imax is the plateau , T is tension , T50 is the tension of half-maximal activation , and k is the slope factor . The standard deviation of the normalized amplitude for each bin was used to weight the fit . Fit parameters for no prepulse ( 0 mmHg ) : Imax = 0 . 84±0 . 02 , T50 = 2 . 2±0 . 1 mN/m , k = 0 . 8±0 . 1 . For +5 mmHg prepulse: Imax = 0 . 85±0 . 01 , T50 = 1 . 4±0 . 1 mN/m , k = 0 . 7±0 . 1 . For +10 mmHg prepulse: Imax = 0 . 70±0 . 04 , T50 = 1 . 8±0 . 2 mN/m , k = 1 . 1±0 . 2 . ( G ) Fits from D-F overlayed ( solid line ) with 95% confidence intervals ( dashed lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12088 . 008 The effect of the +5 mmHg prepulse on the overall current-pressure relationship was striking: When preceded by a +5 mmHg prepulse , currents elicited by a subsequent test pulse were greatly increased in amplitude . Additionally , the pressure of half-maximal activation was shifted by ~9 mmHg towards lower pressures ( P50 = −16 . 8 ± 2 . 8 mmHg with a 0 mmHg prepulse and P50 = −7 . 7 ± 1 . 1 mmHg with a +5 mmHg prepulse; N=11; P<0 . 05 ) ( Figure 5A and B ) . Importantly , prepulses of +10 mmHg did not affect the overall pressure sensitivity as compared to 0 mmHg prepulses ( P50 = −16 . 8 ± 2 . 8 mmHg with a 0 mmHg prepulse and P50 = −13 . 5 ± 2 . 9 mmHg with a +10 mmHg prepulse; N=11; P>0 . 05 ) . As before ( Figure 2A ) , only negligible currents were elicited in patches transfected with empty vector ( pcDNA ) , even during test pulses preceded by a +5 mmHg prepulse , indicating the increase in current at lower pressures did not result from novel recruitment of endogenous mechanosensitive channels in HEK293t cells ( Figure 5B ) . These results suggest that pressure prepulses that minimize membrane tension shift overall Piezo sensitivity maximally leftwards . Importantly , we found that during the long duration of the prepulse experiment , membrane geometry was sufficiently stable to reversibly and reliably alternate between different membrane curvatures ( Figure 5C ) . While there are slight variations in resting radius throughout the duration of the recording , likely due to slight creep of the patch , these are minor compared to the large changes in radius induced by the pre- and test pulses . To establish the effect of prepulses on overall Piezo sensitivity quantitatively , we calculated membrane tension during the test pulse from the corresponding images . We found that the tension of half-maximal activation was indeed leftward shifted , from T50 = 2 . 2 ± 0 . 1 mN/m without a prepulse ( 0 mmHg ) to T50 = 1 . 4 ± 0 . 1 mN/m with a +5 mmHg prepulse ( P<0 . 001; N=11; Figure 5D , E , and G ) . Upon a +10 mmHg prepulse , the T50 was also leftward shifted to that without a prepulse , but to a lesser extent than with a +5 mmHg prepulse ( T50 = 1 . 8 ± 0 . 2 mN/m; P=0 . 03; Figure 5F and G ) . The slope factor was unaltered by prepulse amplitude ( 0 mmHg prepulse , k = 0 . 8 ± 0 . 1; +5 mmHg prepulse , k = 0 . 7 ± 0 . 1; P=0 . 53 vs 0 mmHg; +10 mmHg prepulse , k = 1 . 1 ± 0 . 2; P=0 . 07 vs . 0 mmHg ) . Altogether , our approach unmasks the inherent tension sensitivity of Piezo1 and demonstrates that it can be substantially modulated by resting membrane tension . Our results imply that in a native and undisturbed cell , the same mechanism might affect overall sensitivity of Piezo1 .
We originally set out to identify the physical stimulus that activates Piezo1 ion channels . We chose to examine Piezo sensitivity in membrane patches from living cells , which differs from the bottom-up approaches of artificial lipid bilayers or micelle systems , in that the lipid composition is heterogeneous , the channel protein is not purified from possible interaction partners , and cellular mechanical stability is maintained . We found that similar to the vertebrate mechanosensitive K+ channels TREK-1 and TRAAK , Piezo1 responds robustly to both positive and negative pressure ( Brohawn et al . , 2014b ) . There are several possible mechanisms for how mechanosensitive ion channels may convert physical force into pore opening ( Ranade et al . , 2015; Sukharev and Corey , 2004 ) . In a tethered mechanism , force could be transmitted to the channel through auxiliary proteins , whereas in a bilayer mechanism , force could be transmitted directly through the lipid bilayer . In the latter case , the channel must sense either membrane curvature or lateral membrane tension . Several well-studied mechanosensitive ion channels have previously been demonstrated to sense lateral membrane tension , including prokaryotic channels MscS and MscL , as well as TREK-1 and TRAAK ( Brohawn et al . , 2014b; Moe and Blount , 2005; Sokabe et al . , 1991; Sukharev , 1999; 2002 ) . However , Piezos are distinct from each of these previously identified tension sensors in that they are much larger , with many more predicted transmembrane domains , and in that they share no homology on a primary sequence or overall architectural level ( Brohawn et al . , 2014a; Coste et al . , 2015; Ge et al . , 2015; Kamajaya et al . , 2014 ) . Our simultaneous imaging with electrophysiology revealed that both convex and concave macroscopic curvature in the membrane patch induce channel opening . While it is theoretically possible that Piezo1 senses convex and concave membrane curvature with equal sensitivity , this mechanism would require curvature sensing structures that are symmetrical . However , Piezo proteins do not contain any amino acid sequences with similarity to any known curvature-sensing proteins , and the recently obtained cryo-electron microscopy structure shows no symmetrical features with respect to the plane of the bilayer , making this possibility unlikely ( Antonny , 2011; Ge et al . , 2015; McMahon and Boucrot , 2015 ) . It is also possible that local membrane curvature and tension are not strictly coupled to macroscopic membrane curvature . Previous reports indicate that alterations in local curvature induced by asymmetric incorporation of lipids can change the response of tension-gated channels to pressure ( Perozo et al . , 2002 ) ; this will also have to be tested for Piezo1 . Consistent with the idea that Piezo1 senses lateral membrane tension , we were able to combine electrophysiology and imaging to quantify the tension required for activation of Piezo1 in cell-attached and inside-out patches . Our data suggest that Piezo is a unique mechanosensor , as Piezo1 probed in a cellular environment is much more sensitive to tension ( T50 = 1 . 4 ± 0 . 1 mN/m ) than are MscL and MscS in azolectin liposomes ( T50 ~5–10 mN/m ) , although we cannot rule out differences arising from the different local lipid environment in eukaryotic cells versus reconstituted systems ( Moe and Blount , 2005; Nomura et al . , 2012; Sukharev , 2002; Sukharev et al . , 1999 ) . Together , our data support the evolutionary need for this novel mechanosensor; unlike the prokaryotic ‘release valves’ MscL and MscS , Piezo is able to detect very small changes in membrane tension , an ability that is essential for its role in intricate physiological processes ranging from sensing renal and blood flow to detection of light touch ( Cahalan et al . , 2015; Maksimovic et al . , 2014; Ranade et al . , 2014a; Ranade et al . , 2014b; Woo et al . , 2014 ) . Observing Piezo1 activity in multiple patch configurations also allowed us to make several observations . First , sensitivity of Piezo1 to tension differed in cell-attached versus inside-out patches . One potential explanation for this difference is the varying amount of cytoskeleton retained in the two configurations , which contributes to membrane properties , including tension ( Gauthier et al . , 2012 ) . Consistent with this , we observed lower optical density in inside-out patches , which are known to contain less cytoskeleton ( Suchyna et al . , 2009 ) . Together , this predicts that the cytoskeleton is an important regulator of Piezo1 sensitivity , a finding that is reconciled with previous reports that inhibition of actin polymerization with cytochalsin-D inhibits whole-cell Piezo1 currents evoked by direct stimulus with a glass pipette , but increases opening in cell-attached pressure-evoked currents ( Gnanasambandam et al . , 2015; Gottlieb et al . , 2012 ) . In fact , we cannot rule out a role of the cytoskeleton in directly contributing or even being essential to Piezo activation . We also observed differences in both rate and extent of decay of Piezo1-mediated currents among patch configurations . While the mechanism of Piezo1 inactivation is unknown , a previous study reports that inactivation is irreversibly lost with repeated stimulation , which may explain why we observed the least inactivation in outside-out patches , in which the membrane undergoes the most manipulation prior to assaying activity ( Gottlieb et al . , 2012 ) . Second , the reported sensitivity of Piezo1 to pressure shows remarkable variation among cell types and stimulation protocols . For example , while the P50 is typically reported to be ~ −30 mmHg for heterologously expressed Piezo1 in HEK293t cells , in neural stem cells , the P50 was ~−13 mmHg , whereas the P50 was ~−40 mmHg in the breast cancer cell line MCF-7 ( Coste et al . , 2010; Li et al . , 2014; Pathak et al . , 2014 ) . One explanation for this is that Piezo1 sensitivity is modulated by interaction with other cellular components , which may be differentially expressed: For example , Piezo1 mechanosensitivity requires the presence of phosphoinositides , which are depleted over time in excised patches ( Borbiro et al . , 2015 ) ; the integral membrane protein STOML3 also greatly increases Piezo1 sensitivity ( Poole et al . , 2014 ) . Here , we found that an additional modulator of Piezo sensitivity is resting membrane tension . Importantly , cellular membrane tension varies greatly with cell type; even within one cell local tension depends on factors including lipid composition , cytoskeletal contacts , the extracellular matrix , and others ( Blumenthal et al . , 2014; Hoffman , 2014; Vasquez et al . , 2014 ) . As Piezos are thus under differing tension depending on their cellular expression and localization , the ability of tension to modulate sensitivity gives Piezo1 a broad tuning curve that primes it to respond to physiologically relevant changes in tension at that location . Importantly , this makes Piezos robust sensors of membrane tension in the remarkably wide variety of cell types in which they are expressed . Third , our results also identify inactivation as an important physiological modulator of overall Piezo1 sensitivity . Interestingly , the currents we observed after Piezo channels recovered from inactivation during a transient period of zero tension ( i . e . , +5 mmHg prepulses ) may have important physiological relevance: a stimulus that results in a local reduction of membrane tension may lead to increased Piezo activity in this region upon release of this stimulus , perhaps providing a mechanism by which cells can sense not only the onset , but also the offset of a stimulus . The interplay between activation and inactivation may also make Piezo1 most sensitive to rapidly applied mechanical stimuli , similar to previous reports for MscS , as slowly applied stimuli will lead to gradual accumulation of channels in inactivated states ( Akitake et al . , 2005 ) . The advent of structural information about the various domains of Piezo , as well as its overall architecture will be extremely useful in identifying the structural correlates of both the mechanosensor and the inactivation mechanism , as activation and inactivation combine to dictate overall sensitivity ( Ge et al . , 2015; Kamajaya et al . , 2014 ) . Finally , we anticipate that our simple prepulse protocol will provide a useful tool for measuring the inherent mechanosensitivity in different cells and irrespective of inactivation kinetics by manipulating the curvature of the membrane to minimize tension prior to testing sensitivity . The prepulse amplitude required for flattening of the membrane and removal of resting tension will vary with the particulars of a given system , including cell type and pipette size , but can be measured even in the absence of imaging data with the protocol in Figure 4B , by using the pressure at which the maximal offset current amplitude is evoked .
Human embryonic kidney HEK293t cells ( ATCC # 3579061 ) were provided and authenticated ( STR authenticated and verified mycoplasma-free ) by the Duke Cell Culture Facility . Cells were grown in DMEM ( Life Technologies ) with 10% heat-inactivated fetal bovine serum ( Clontech Laboratories , Mountain View , CA ) , 50 units/ml penicillin , and 50 mg/ml streptomycin ( Life Technologies , Carlsbad , CA ) . Cells were transiently transfected in 6 well plates in the presence of 10 μM ruthenium red with Mouse Piezo1-IRES-GFP ( 3 μg ) or empty vector ( pcDNA3 . 1 ( - ) and GFP ) using Fugene ( Promega , Madison , WI ) ~48 hr before recording . Transfected cells were reseeded at low density the day before recording in 50 mm glass-bottomed dishes ( P50G-0-30-F; MatTek Corporation , Ashland , MA ) coated with Poly-L-lysine and laminin . Patch-clamp recordings were performed at room temperature using an EPC10 amplifier and Patchmaster software ( HEKA Elektronik , Lambrecht , Germany ) . Data were sampled at 5 kHz and filtered at 2 . 9 kHz . Borosilicate glass pipettes ( 1 . 5 OD , 0 . 85 ID; Sutter Instrument Company , Novato , CA ) had a resistance of 1 . 5–4 MΩ when filled with pipette buffer solution ( in mM: 130 NaCl , 5 KCl , 10 HEPES , 10 TEACl , 1 CaCl2 , 1 MgCl2 , pH = 7 . 3 with NaOH ) . The standard bath solution was ( in mM ) : 140 KCl , 10 HEPES , 1 MgCl2 , 10 glucose , pH = 7 . 3 with KOH . Pipettes were angled at ~15° with respect to the glass cover slip to optimize image contrast . Pressure was controlled with a high-speed pressure clamp system ( HSPC-1; ALA Scientific Instruments , Farmingdale , NY ) . Patches were held at −80 mV and stimulated with pressure-step protocols described in the manuscript . Unless stated otherwise , sweeps were separated by 10 s to allow for recovery from inactivation . Electrophysiological recordings were only analyzed for patches with a seal resistance of at least 1GΩ and maximal pressure-induced currents of at least 50 pA for cell-attached and inside-out patches with negative pressure and 20 pA for all other configurations . Only one patch was excised from each cell; with the exception of Figure 4F , which required prior determination of the appropriate prepulse for each patch using the protocol in 4B , only one protocol was performed on each patch . Analysis was performed with Igor Pro 6 . 22A ( WaveMetrics , Lake Oswego , OR ) . Baseline currents before pressure stimulation were subtracted off-line and peak currents measured at each pressure . Student’s t-test or ANOVA followed by Tukey-Kramer comparison of pairs of means were used to assess statistical significance . Images of the cell membrane inside the patch pipette were captured at a rate of ~7 . 5 frames per second ( 125 ms exposure ) at a resolution of 61 . 5 pixels/μm using a Plan Apo ( 100x ) DIC oil objective coupled with a Coolsnap ES camera and 4x relay lens ( Nikon Instruments Inc , Melville , NY ) . During imaging , the focal plane was continuously adjusted to center on the contact points of the membrane with pipette walls , indicated by the ‘crossing over’ of the lines corresponding to the pipette ( see Video 1 ) . Short ( 300 ms ) pressure pulses were used to minimize membrane ‘creep’ and excessive movement of the membrane out of the focal plane . Images were extracted from videos in NIS-Elements ( Nikon Instruments Inc , Melville , NY ) and imported into Igor Pro ( WaveMetrics , Lake Oswego , OR ) . To identify the membrane geometry , a line scan parallel to the pipette walls was performed to localize the minimum pixel intensity for each line over a rolling average of 9 pixels; the script , custom-written in IgorPro , is available in our Github Repository ( github . com/GrandlLab ) . These positions were then fit with a circle to obtain the radius ( R ) . Tension ( T ) was calculated for every pressure step Δp using Laplace’s law: T= R·Δp/2 . For pooling and binning data , current-pressure responses for individual cells were fit with a Boltzmann function ( Imax/ ( 1+exp ( - ( P-P50 ) /k ) ) ) and individual current amplitudes were re-normalized to the plateau response to saturating stimuli from the fit ( IMax ) . To calculate the tension required for activation , binned data were fit with a Boltzmann function ( I = Imax/ ( 1+exp ( - ( T-T50 ) /k ) ) ) ; the fit was weighted with the standard deviations from each bin . | Piezo ion channels are proteins that are embedded in the cell membranes of many types of tissue , including the heart , lung , skin and kidney . These proteins are essential for many biological processes , including sensing gentle touches and ensuring that blood vessels develop properly . When stimulated by mechanical forces , a central pore in the Piezo channel opens to allow positively charged ions to flow into the cell , which triggers electrical and chemical signaling processes inside the cell . However , it was not known exactly what type of mechanical stimulus is sensed by Piezo ion channels . Lewis and Grandl expressed Piezo ion channels in cultured human kidney cells , and opened them by applying pressure to parts of the cell membrane inside a glass pipette . This causes a number of changes to the membrane , including to its curvature and tension , either of which could potentially open the Piezo channels . However , Lewis and Grandl were able to calculate from images of the cell membrane inside the pipette that tension is the activating stimulus . Further experiments unexpectedly revealed that the tension that is usually present in the cell membrane is sufficient to inactivate Piezo channels and prevent them from responding to an additional mechanical stimulus . This suggests that Piezo ion channels are inherently more sensitive to tension than previously realized , which could explain why different cell types appear to have different sensitivities to pressure . Although Lewis and Grandl have now shown that Piezo channels are activated by tension , more work is needed to investigate how the Piezo ion channel senses this force , and how this leads to the channel pore opening . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"physics",
"of",
"living",
"systems",
"neuroscience"
] | 2015 | Mechanical sensitivity of Piezo1 ion channels can be tuned by cellular membrane tension |
In a dynamic environment an organism has to constantly adjust ongoing behavior to adapt to a given context . This process requires continuous monitoring of ongoing behavior to provide its meaningful interpretation . The caudate nucleus is known to have a role in behavioral monitoring , but the nature of these signals during dynamic behavior is still unclear . We recorded neuronal activity in the caudate nucleus in monkeys during categorization behavior that changed rapidly across contexts . We found that neuronal activity maintained representation of the identity and context of a recently categorized stimulus , as well as interpreted the behavioral meaningfulness of the maintained trace . The accuracy of this cognitive monitoring signal was highest for behavior for which subjects were prone to make errors . Thus , the caudate nucleus provides interpretive monitoring of ongoing behavior , which is necessary for contextually specific decisions to adapt to rapidly changing conditions .
We adapt to different situations by sorting information into behaviorally meaningful categories that can change rapidly across contexts . This categorization process allows us to interact with such environments . Eager to get to an important meeting , we drive fast , but the actual speed depends on the speed limits along the way . In this situation , categories are delineated by a category boundary ( i . e . , speed limit ) that varies with different environments . Flexible categorization requires continuous monitoring of the changes in the environment to ensure that classification reflects current task demands , even after categories are well learned . How the brain monitors such complex cognitive behavior is not well understood . The basal ganglia , particularly the caudate nucleus , are known to be essential for flexible behavior ( Wise et al . , 1996; Barnes et al . , 2005 ) and together with the interconnected prefrontal and parietal cortices play an important role in categorization ( Poldrack et al . , 1999; Poldrack et al . , 2001; Seger and Cincotta , 2005; Freedman and Assad , 2006; Seger , 2008; Antzoulatos and Miller , 2011; Ashby and Maddox , 2011; Mendez et al . , 2011; Merchant et al . , 2011 ) . Many psychiatric and neurological disorders that compromise the caudate nucleus are characterized by impairment in cognitive flexibility ( Knowlton et al . , 1996; Shohamy et al . , 2004; Montoya et al . , 2006 ) . Human and animal studies have provided extensive evidence for the critical role of this structure in category learning ( Seger and Miller , 2010; Antzoulatos and Miller , 2011 ) . Neurobiological models of categorization have suggested that the caudate nucleus , together with interconnected cortical and subcortical structures , contributes to the maintenance and switching of rules that guide categorization ( Maddox and Ashby , 2004; Ashby and Ennis , 2006 ) . Despite the established role of the caudate nucleus in flexible behavior by linking actions and outcomes ( Hikosaka et al . , 2000; Yin et al . , 2005; Graybiel , 2008; Balleine and O'Doherty , 2010 ) , its role in the monitoring of such behavior is less clear . Recent studies have highlighted the strong contribution of the caudate to post-action evaluation by monitoring behavioral performance based on reward information ( Lau and Glimcher , 2007; Ding and Gold , 2010; Thorn et al . , 2010; Kim et al . , 2013 ) . The general idea is that the caudate detects a mismatch between expected and actual outcomes and these prediction error signals tend to alert an organism about the overall level of behavioral performance . Some studies have shown that post-action neuronal activity maintained memory traces of specific actions , possibly linking them to outcomes ( Lau and Glimcher , 2007; Kim et al . , 2013 ) . Others have found that post-action activity represented outcomes independent of specific actions ( Ding and Gold , 2010 ) . Monitoring signals are sensitive to behavioral context change ( Hikosaka and Isoda , 2010 ) and can also incorporate uncertainty estimates ( Badre , 2012; Kepecs and Mainen , 2012 ) known to modulate caudate activity ( Ding and Gold , 2012; Yanike and Ferrera , 2014 ) . Thus , a systematic evaluation of caudate monitoring signals is important for understanding its role in behavioral flexibility , when changes in context and outcome are frequent . To study what aspects of cognitive flexible behavior are monitored in the caudate nucleus we recorded neuronal activity during a categorization task in which decision criteria changed rapidly across trials . We evaluated post-decision signals both at the level of individual neurons and their population activity . We found that while individual neurons were highly context specific , their population activity across ensembles of neurons provided an accurate and separable read-out of sensory and cognitive aspects of ongoing behavior .
We analyzed a total of 155 presumed projection neurons from the associative caudate nucleus ( Figure 2A ) in two monkeys performing the speed categorization task ( Monkey C: n = 91; Monkey F: n = 64 ) . For each neuron only trials in the neuron's response field were included . The data from the two monkeys were combined as they were qualitatively similar . We found that the majority of caudate neurons were responsive after , and not prior to , the animal making a decision . Therefore we focused on two post-decision periods of the task , one right after the decision ( ‘post-saccade’ , 0–400 ms after saccade onset , median saccade onset 309 ms ) and another when the correctness of the decision was revealed at 800 ms after the decision onset ( ‘reward’ , 0–600 ms after reward onset ) . We found that many neurons ( 71 out of 155 , 46% , Figure 2B , top ) showed significantly different activity between the fixation period and at least one of the post-decision periods of the task ( bootstrap test , p < 0 . 01 ) . Next , we asked whether caudate neurons represented category-related information regardless of the properties of the visual stimuli or their stimulus selectivity . We identified 38 out of 71 neurons that were sensitive to the category context ( i . e . , boundary position ) during either one or both post-decision periods ( Figure 2B , bottom ) . The firing rates of these neurons were significantly different between the two boundary positions for at least one speed either during the post-saccade ( ‘psacc’ , n = 31 ) and/or reward ( ‘rwd’ , n = 23 ) periods of the task ( Figure 3 , boostrap test , p < 0 . 05 Bonferroni corrected ) . Figure 4 shows example neurons with activity during the post-saccade ( A ) and reward ( B ) periods with a significantly different average response to one out of 8 stimuli , corresponding to speed 12 ( see Figure 4—figure supplement 1 for other examples ) , for which they responded with either significantly higher ( Figure 4A ) or lower ( Figure 4B ) firing rate on trials with the fast compared to slow boundary positions ( bootstrap test , p < 0 . 0001 ) . These neurons responded similarly to most other stimuli , irrespective of the boundary position ( Figure 4A , B , bottom ) . The majority of these post-decision neurons ( psacc: 19/31 , 61%; rwd: 15/23 , 65% ) discriminated significantly only one out of 8 stimuli and many of them ( psacc: 22/31 , 71%; rwd: 11/23 , 48% ) had significantly different neuronal activity for stimuli near the category boundaries , speeds 6 and 12 ( Figure 4E , F , bottom ) . 10 . 7554/eLife . 03727 . 004Figure 2 . Location of recording sites . ( A ) Recording chamber showing access to the associative striatum ( indicated in red square ) in Monkey F . ( B ) ( top ) Distribution of recording sites for post-saccade ( gray ) and reward ( black ) neurons across two animals . AP = 0 corresponds to the anterior commissure . ( bottom ) Location of category-related neurons with selective response to slow ( orange ) and fast ( blue ) boundaries superimposed on the location of responsive neurons ( gray ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 00410 . 7554/eLife . 03727 . 005Figure 3 . Sensitivity to the boundary position in caudate neurons . ( A and B ) Distribution of significant p-values indicating difference in spike count between slow and fast boundary positions for neurons with activity during the post-saccade ( A ) and reward ( B ) periods of the task . Only values of less than 0 . 05 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 00510 . 7554/eLife . 03727 . 006Figure 4 . Representation of category signals . ( A ) Example neurons with activity during post-saccade ( A ) and reward ( B ) periods . ( top ) Spike raster plots , each row is one trial and each dot is detected spike , and spike density functions ( mean ± SEM ) in response to speeds 6 ( left ) and 12 ( right ) . Black lines show periods of the task . Dashed black lines indicate average reaction time to saccade . ( bottom ) Average neuronal activity across stimuli sorted by the boundary position . Orange/blue dashed lines , actual boundary positions . Bars , SEM . ( C and D ) Scatter plots of CIs for speed 6 vs speed 12 across neurons with activity during post-saccade ( C , n = 31 ) and reward ( D , n = 23 ) periods . Example neurons in ( A and B ) , star and square in ( C and D ) , respectively . ( E and F ) ( top ) Average CI across speeds for neurons with activity during post-saccade ( E ) and reward ( F ) periods . ( bottom ) Proportion of neurons with a significant difference between spike counts across boundaries for each speed ( bootstrap test , p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 00610 . 7554/eLife . 03727 . 007Figure 4—figure supplement 1 . Examples of two additional caudate neurons . On average , these neurons responded differently to the same stimulus ( speed 6 ) on trials with slow ( orange ) and fast ( blue ) boundary positions during the post-saccade ( A ) and reward ( B ) periods of the task . Same notation as in Figure 4A , B . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 007 To quantify the sensitivity to the boundary position , we calculated a category index for each neuron ( CI; see ‘Materials and methods’ ) separately for each period of the task . The CI measures how well neuronal responses discriminate the same stimuli when presented with different boundaries . The index ranges between 0 and 1 , with values close to 0 indicating weak discrimination and values close to 1 indicating strong differentiation between the two categories . Similarly to the example neurons , the CIs between the stimuli close to the boundaries ( speeds 6 and 12 ) were different for most individual neurons in both periods of the task ( Figure 4C , D ) . When averaged across neurons , the CIs for the two most difficult stimuli were the highest during both periods of the task ( Figure 4E , F , top ) . The ability of individual neurons to multiplex their function according to the relevant boundary position is inconsistent with the known sensitivity of caudate neurons to the rate of reward coding , which varied similarly with two boundaries ( Cromwell and Schultz , 2003; Hassani et al . , 2001 ) . These results show that individual neurons were limited in their ability to monitor behavior across multiple boundaries . However , as a population , the neuronal activity of these neurons could differentiate stimuli near both category boundaries . Next we tested whether combined signals across pools of individual neurons could provide a reliable population code ( Pouget et al . , 2003 ) to monitor behaviors across multiple boundaries . We predicted on each trial the stimulus speed and category boundary position ( for a total of 16 conditions: 8 × 2 ) based on the neuronal activity of populations of independently recorded neurons ( see ‘Materials and methods’ ) . To evaluate the prediction accuracy , we used the proportion of correct estimates for either speed or boundary position . We found that on each trial , the population activity of ensembles of caudate neurons provided a reliable read-out of the two signals: the identity ( speed ) of the previously categorized stimulus and the context ( boundary position ) in which the stimulus was categorized with above chance accuracy during the post-saccade ( Figure 5A , B; n = 31 ) and reward ( Figure 5C , D; n = 23 ) periods of the task . To account for variable levels of the subjects' performance , we obtained the prediction accuracy for the speed and boundary position separately for correct only trials and for all trials ( correct and incorrect ) separately . On average , the prediction accuracies for the two signals were significantly better on correct trials compared to all trials for the post-saccade ( Figure 5A , B; 1-way ANOVA ( trial type ) , p < 0 . 0001 ) and reward ( Figure 5C , D; p < 0 . 0001 ) populations . We also evaluated decoding accuracy as a function of the number of neurons ( Figure 5—figure supplement 1 ) . The decoding performance converged towards an asymptote faster for the boundary position ( ∼5 neurons ) compared to the speed ( ∼10–15 neurons ) in each neuronal population . This suggests a more redundant neuronal code for the boundary position . 10 . 7554/eLife . 03727 . 008Figure 5 . Prediction accuracy for speed and category boundary . The average prediction accuracy for speed ( A , C ) , across two boundary position , and boundary position ( B , D ) , across all speeds , for correct only trials ( black ) and all trials ( correct and incorrect , gray ) separately for the post-saccade ( A , B ) and reward ( C , D ) neuronal populations . Corresponding chance levels are shown in dashed line . Orange/blue dashed lines , actual slow/fast boundary positions . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 00810 . 7554/eLife . 03727 . 009Figure 5—figure supplement 1 . Prediction accuracy and population size . Prediction accuracy for stimulus speed ( A ) and boundary position ( B ) as a function of neuronal population size for neurons with activity during post-saccade ( black ) and reward ( gray ) periods of the task . The decoding performance plateaued faster for the boundary position ( ∼5 neurons ) compared to the speed ( ∼10–15 neurons ) in each neuronal population suggesting more redundant coding for the boundary position . Chance levels are shown in dashed line . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 009 To further quantify the relationship between the accuracy of the population read-out and subjects' behavioral performance for each stimulus , we determined the prediction accuracy of the stimulus speed by splitting trials into slow and fast boundary positions . We found that the accuracy of the population read-out for the identity of the previously categorized stimuli varied with the categorization difficulty at each boundary . This was shown by a significant interaction between the speed and boundary position ( Figure 6A , C; 2-way ANOVA ( speed , boundary ) , p < 0 . 001 ) in each neuronal population . To correlate subjects' categorization performance and the reliability of the population representation , we averaged the prediction accuracy between the slow and fast boundary positions across all speeds and plotted it as a function of the distance to the boundary for post-saccade ( Figure 6B ) and reward ( Figure 6D ) populations . The average prediction accuracy was the highest for the near boundary stimuli on the right , coinciding with the highest behavioral error rate ( Figure 1B , D ) , then decreased with the distance to the boundary . The population read out of the previous stimulus speed was strongly correlated with subjects' behavioral performance , expressed as an error rate , averaged across two boundaries for each speed ( Figure 1B ) , similarly in both neuronal populations ( Figure 6B , D; psacc: R2 = 0 . 85; rwd: R2 = 0 . 76 ) . Thus , the population was best at maintaining the sensory representation of stimuli , for which the subjects were prone to make categorization errors . 10 . 7554/eLife . 03727 . 010Figure 6 . Population read out of speed and boundary position . ( A and C ) Proportion of correct estimates for each stimulus speed separately for slow ( upper bars ) and fast ( lower bars ) boundary positions for post-saccade ( A ) and reward ( C ) populations for correct only ( gray ) and incorrect ( open ) trials . ( B and D ) Average prediction accuracy for speed ( black circles ) and behavioral error rate ( mean ± SEM , dashed and solid lines ) as a function of stimulus' position to the boundary ( dashed line ) for each neuronal population . The prediction accuracy for the most extreme stimuli ( speeds 2 , 4 , 14 , and 16 ) deviated from the behavioral error rate function , possibly due to a greater perceptual uncertainty , about the identity of the stimulus or boundary cue , compared to the intermediate stimuli . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 010 We then compared the predication accuracy between trials when subjects categorized the same stimuli correctly and incorrectly for a subset of stimuli , speeds 6 and 12 , with sufficient number of incorrect trials ( see ‘Materials and methods’ ) . In both neuronal populations , the incorrect categorical judgments were followed by the failure to represent the identity of the preceding stimulus ( Figure 6A , C ) . In contrast , correct categorization of the same stimuli , either near or far from the boundary , were followed by a reliable but different read-out of stimulus speed based on the population activity ( 2-way ANOVA [trial type , boundary] , p < 0 . 001 ) . These results suggest that outcome uncertainty near the boundaries can act as a gating mechanism to enhance representation of the boundary stimuli to potentially overcome decision uncertainty . We reasoned that if this neuronal code is cognitive and reflects categorization process , instead of reward processing , then even irrelevant information which could be pertinent for categorization should be processed in a task meaningful way . We asked if caudate neurons encoded category-irrelevant features of otherwise relevant stimuli and whether the post-decision evaluation of the irrelevant information reflected categorization difficulty . We tested this hypothesis by computing how well we could predict discrimination accuracy between two opposite directions of dot motions for each stimulus separately for the slow and fast boundaries based on neuronal activity ( see ‘Materials and methods’ ) . We found that even those features of stimuli ( i . e . , dots direction ) that were not linked to category identity were selectively suppressed when stimuli were near the boundary compared to when they were far from the boundary in both neuronal populations ( see ‘Materials and methods’; Figure 7A , B; 3-way ANOVA ( direction , speed , boundary ) , p < 0 . 01 ) . Similarly to single unit studies ( Hussar and Pasternak , 2009 ) , the post-decision population read-out in the caudate strongly reflected behavioral relevance of stimuli during categorization . 10 . 7554/eLife . 03727 . 011Figure 7 . Contextual modulation during categorization . ( A and B ) Prediction accuracy for two opposite directions of dots motion ( up or down ) for each stimulus for post-saccade ( A ) and reward ( B ) neuronal populations . ( C and D ) Discrimination accuracy across pairs of neighboring speeds ( seven pairs , 2–4 , 4–6 , … , 14–16 ) separately for trials with each boundary position for post-saccade ( C ) and reward ( D ) neuronal populations . The thickness of each bar corresponds to the average discrimination accuracy with ± SEM . Same notation as in above . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 01110 . 7554/eLife . 03727 . 012Figure 7—figure supplement 1 . Effect of categorical perception on discrimination of near boundary stimuli in individual neurons . Ratio Difference indexes for each neuron with activity during post-saccade ( A ) and reward ( B ) periods showing difference in neuronal activity between speed pairs ( 4/6 or 12/14 ) when categorized as same ( Ratio Diff: intracategory ) or different ( Ratio Diff: intercategory ) categories . For each neuron , a pair ( 4/6 or 12/14 ) with the maximum Ratio Diff is plotted for each period of the task . Orange/blue indicates the maximum Ratio Diff for pairs 4/6 or 12/14 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03727 . 012 The bias towards processing near boundary stimuli in the caudate population code suggests that it could represent the position of each boundary , either actual or inferred . The actual boundary positions ( speeds 5/13 for slow/fast boundaries respectively ) were never explicitly given and the subjects had to infer the positions of each boundary through training . We tested whether caudate population represented the position of each boundary by comparing the discrimination accuracy for the seven pairs of neighboring speeds ( 2–4 , … , 14–16 ) separately for each boundary and neuronal population ( see ‘Materials and methods’ ) . We reasoned that the greater discrimination accuracy between pairs of neighboring speeds would correspond to the more accurate neuronal representation of the internal estimates of the boundary position . We found that immediately after the decision , the post-saccade population activity provided signals specific to all behaviorally relevant boundary positions . Figure 7C shows that high discrimination accuracy occurred for stimuli across the actual boundary positions ( speed pairs 4/6 and 12/14 ) , and across the average internal estimates of the boundary positions ( 6/8 and 10/12 , respectively ) . Figure 7D shows a different pattern during reward delivery . The strength of the discrimination accuracy peaked for the speeds across the actual boundary positions ( 4/6 and 12/14 ) , but remained elevated for intermediate speeds regardless of the distance to the boundary . These differences in population read-out codes possibly reflect different putative synaptic pooling mechanisms , with the pooling during the reward period better suited to linking stimuli with dual class membership to the received outcome . We confirmed that this signal reflects categorical perception as opposed to representation of stimulus identity . We reasoned that any two stimuli should be discriminated better when they are judged to be in a different class as opposed to being in the same class . This is often referred to as the boundary accentuation effect ( Goldstone , 1994 ) . The results in Figure 7C , D confirmed this prediction , the average discrimination accuracy between the near boundary speeds ( 4/6 and 12/14 ) being significantly greater when the stimuli were classified in two different ( slow vs fast ) compared to the same ( slow or fast ) categories in both populations ( psacc: 0 . 86 ± 0 . 02 vs 0 . 63 ± 0 . 02; rwd: 0 . 9 ± 0 . 01 vs 0 . 82 ± 0 . 01 , permutation test , p < 0 . 001 ) . To understand the representation at the level of individual neurons , for each neuron we calculated a ratio difference index ( Ratio Diff ) measuring the difference in firing for two pairs of speeds ( 4/6 and 12/14 ) separately for slow and fast boundary trials . Consistent with the population findings , individual neurons had significantly larger differences in firing rate to the near boundary stimuli when in different categories compared to the same category both during the post-saccade ( boostrap test , p < 0 . 04 ) and reward ( p < 0 . 01 ) periods of the task ( Figure 7—figure supplement 1 ) . These results suggest that the caudate neuronal code parses information into behaviorally relevant categories and that the code is most informative for the stimuli with the most erroneous performance .
We found that caudate neurons monitored flexible categorization behavior by providing distinct representation of decision-pertinent variables , from stimulus specific features to more cognitive signals reflecting internal estimates of decision variables . The monitoring signal in the caudate was not only accurate at representing ongoing behavior , but it also interpreted behavioral consequences in light of changing demands of the task . The accuracy of the monitoring signal was inversely related to behavioral proficiency , suggesting that the caudate nucleus allocates more efficient coding for changing and/or uncertain information critical for behavioral flexibility . These findings provide direct evidence for the role of the caudate nucleus in online monitoring of complex cognitive behavior , possibly allowing for contextually specific decisions to adapt to rapidly changing context ( Daw et al . , 2006; Hikosaka and Isoda , 2010; Pearson and Platt , 2013 ) . It is well-established that the caudate nucleus supports cognitive function , including categorization , by utilizing reward or motivation information . However , how cognitive and reward signals interact to support complex behavior is still unclear . In this report , we showed that the monitoring signals in the caudate nucleus reflecting cognitive variables were shaped by reward modulation , as more accurate representation correlated with the most variable reward outcome . We found that individual neurons were highly selective at linking sensory and reward information . In other words , the neurons did not generalize across sensory inputs to code reward-related information . However , their population response provided a reliable and separable representation of both cognitive and motivational signals on a trial-by-trial basis . The reported monitoring signals in the caudate can reflect the outcome prediction and/or the rate of reward coding ( Hollerman et al . , 1998; Hassani et al . , 2001; Cromwell and Schultz , 2003 ) . The motivational signals potentially shape categorical representation in the caudate to monitor context specific decisions to adapt to rapidly changing conditions . Our findings are consistent with the growing evidence for the strong contribution of the caudate nucleus to post-decision monitoring and evaluation ( Lau and Glimcher , 2007; Ding and Gold , 2010; Thorn et al . , 2010 ) . Traditional views postulate that the striatum , which includes the caudate nucleus , contributes to behavioral evaluation through updates based on differences between the observed and expected outcomes through reinforcement learning ( Daw et al . , 2006; Williams and Eskandar , 2006 ) . Recent views suggest that such updates can also incorporate uncertainty estimates , similar to the prefrontal cortex ( Badre , 2012; Kepecs et al . , 2008; Kepecs and Mainen , 2012 ) . Caudate neuronal activity is sensitive to both stimulus ( Ding and Gold , 2012 ) and outcome uncertainty ( Yanike and Ferrera , 2014 ) , which can potentially shape more efficient coding of changing or uncertain information . Our findings that the caudate nucleus provides online monitoring of flexible behavior are consistent with a recent study showing that individual caudate neurons coded values of visual objects in a flexible manner ( Kim and Hikosaka , 2013 ) . Specifically , neurons in the head of the caudate nucleus represented changes in recent value of objects while monkeys made saccades to visual objects with different values and inactivation of this structure disrupted their behavioral preference for high value objects . While we also recorded in the anterior part of the caudate nucleus ( see ‘Materials and methods’ ) , our experimental paradigm was different . During the categorization task , monkeys had to maintained representation of both stable ( i . e . , always far from boundaries ) and changing ( i . e . , near the boundaries ) values of dot stimuli to perform successfully , well after the initial learning had occurred . Therefore , our findings are complementary to the aforementioned study because we showed that the caudate nucleus provides abstract cognitive monitoring signals beyond flexible reward associations . Taken together , these results suggest that the anterior caudate plays an important role in the change detection network ( Isoda and Hikosaka , 2011; Pearson and Platt , 2013 ) contributing to a rapid adjustment of behavior . Our findings that the representation of category-relevant information was enhanced while category-irrelevant information was suppressed near the boundaries are consistent with the possible role of the caudate nucleus in the working memory updating . Computational models ( O'Reilly and Frank , 2006 ) have suggested that the striatum contributes to a selective gating of information flow into the working memory in the prefrontal cortex . The present findings suggest that such selective gating in the caudate nucleus can occur via a neuronal population code , which allows for independent read-out of all task variables by downstream structures . In our experimental paradigm the direction of the dots motion was irrelevant for the speed categorization and the animals were never explicitly required to utilize that information . Yet , the accuracy of the population read-out for the dots motion direction in the caudate was suppressed near the boundary compared to that far from the boundary . These results suggest that the caudate nucleus sorts information by its relevance to the task at hand , possibly automatically , to potentially affect ongoing behavior ( Badre , 2012 ) . Finally , our findings suggest that unrevealing the nature of neuronal population code can greatly benefit our understanding of the neural basis of complex cognitive behavior . We found that individual neurons provided a context-specific code , while their population read-out covered all aspects of behavior and multiple separable signals could be reliably extracted . Previous studies on visual categorization have shown that representation of multiple categories in the prefrontal cortex at the level of single neurons was either distributed or sparse depending on how much the visual categories overlapped ( Cromer et al . , 2010; Roy et al . , 2010 ) . How these differences translate into population codes remains to be studied . The results of our experiment show that the caudate neuronal population code can reliably represent all aspects of cognitive complexity during flexible behavior .
Two adult male rhesus monkeys ( Macaca mulatta , Monkey C: 8 . 2 kg and Monkey F: 11 . 5 kg ) were used in the experiments . All ‘Materials and methods’ and treatments were in accordance with NIH guidelines and approved by the Institutional Animal Care and Use Committee at Columbia University and the New York State Psychiatric Institute . Prior to the experiments each animal was implanted with a scleral search coil , head post and recording chamber under aseptic conditions using isoflurane anesthesia . The animals received postoperative analgesics during postsurgical recovery . The positions of the recording chambers were guided by monkeys' individual MRI atlases . The recording chamber ( 20 mm in diameter ) for Monkey C was placed on the scull over the arcuate sulcus positioned at stereotaxic coordinates 20 mm anterior and 15 mm lateral allowing access to the anterior caudate nucleus via the frontal eye fields ( FEF ) . Monkey F was sequentially implanted with two different recording chambers . The first recording chamber ( 20 mm in diameter ) was placed at 25 mm anterior and 18 mm lateral positioned over the acruate sulcus . The second recording chamber ( 20 × 30 mm oval ) was centered at 15 mm anterior and 12 mm lateral . We used tungsten or glass coated electrodes with impedance ranging from 1–3 . 5 MΩ . The signals were amplified , filtered and passed to a real-time action potential detection . Action potentials were converted to TTL pulses that were stored together with the behavioral data . We also stored individual waveforms on each channel for further offline analysis . To identify the anterior caudate nucleus we used a number of criteria . We used depth measurements and identified the position of the caudate relative to the FEF . Also the dorsal edge of the caudate was identified by the presence of injury potentials . We identified the phasically active neurons by their low baseline activity ( 1–3 Hz ) . On each recording track we made sure to identify a tonically active neuron ( 4–8 Hz ) , in fact during some sessions we simultaneously recorded a pair of tonically and phasically active neurons . Post-saccade and reward neurons tended to be distributed without any distinct spatial organization in the associative caudate ( Figure 2B ) . Monkeys were trained to sit in a primate chair for the duration of the experiments with their heads restrained . They performed behavioral experiments and received liquid reward for correctly executing the behavioral task . We trained two adult monkeys to categorize the speed of moving random dot patterns depending on the position of a category boundary ( Ferrera et al . , 2009 ) . The stimulus set consisted of random-dot patterns moving at 8 different speeds ( 2 , 4 , 6 , … , 16 deg/s ) with coherence equal to 1 . The direction of random dot motion also varied randomly . On each trial , dot direction was selected from a set of two opposite directions . Generally , the directions were ‘up’ and ‘down’ although other axes of motion were also tried . Animals were never asked to judge the direction of dot motion , thus we consider this stimulus dimension as category-irrelevant . On a given trial , monkeys judged the speed of motion as ‘slow’ or ‘fast’ depending on one of two reference speeds . Each trial started with a fixation cue presented at the center of the computer screen . After animals fixated for 400 ms ( baseline period ) , one of two boundary cues ( blue or orange squares ) indicating the reference speed was presented for 800 ms ( cue period ) , followed by the random-dot stimulus together with two spatially located targets ( decision period ) . Monkeys were trained to associate speeds faster than the reference with the green target and slower than the reference with the red target . They indicated their judgment by making a saccade within 800 ms to one of the targets , the positions of which were randomized across trials . Feedback was provided at the end of the trial . Correctly categorized stimuli were followed by two drops of water and a high tone , while incorrectly categorized stimuli were followed by a low tone and no reward . The trials were separated by a 2000 ms inter-trial interval . The task had a block-randomized design; each trial type was presented randomly from a block and animals had to complete each trial-type to progress to the next block . The full design comprised 64 trial types: 8 speeds × 2 directions × 2 boundaries × 2 target locations . On a small fraction of trials ( ∼0 . 13 ) animals broke fixation during the decision period of the task without making a choice ( fixation break trials ) . The fixation break trials were reshuffled with the different trials in a given block and were not immediately repeated . The average reaction time to abort on the fixation break trials was 348 ± 12 ms and was similar between the two animals ( monkey: F , 328 ± 13 ms , and monkey C: 367 ± 18 ms; 1 tail t test p = 0 . 08 ) . Trials on which animals broke fixation during the baseline or cue periods of the task were excluded from the analyses . We used a memory guided saccade task with 8 spatial targets at 45° intervals to identify task-related neurons . We identified each neuron's response field by finding the spatial location which evoked the maximum firing rate during one of the task periods in the memory guided saccade task . In the speed categorization task , we placed one of the spatial targets in the response field and one in the location opposite to the response field of a neuron . Some of the cells that were responsive during the memory guided saccade task were not responsive during the categorization task ( 21/176 cells , 12% ) . These cells were excluded from the present analyses . We analyzed a total of 155 cells with the average of 494 ± 94 trials per session ( range [285–1393] ) , which was similar between the two animals ( Monkey F: 551 ± 92; Monkey C: 444 ± 105; 1 tail t test , p = 0 . 36 ) . The monkeys performed on average similar number of trials with slow and fast boundary positions across all speeds ( mean slow boundary: 226 ± 16 trials; mean fast boundary: 219 ± 15; paired t test p = 0 . 74 ) and on average 27 ± 2 trials for each speed . For each neuron , we only included trials in a neuron's response field during the categorization task . We identified all task related neurons by a bootstrap test ( p < 0 . 05 ) comparing baseline firing rate during the initial fixation period ( 400 ms ) with the average spike counts during each of four different task periods: cue ( 0–700 ms ) , decision ( 0 ≤ 800 ms ) , post-saccade ( 0–400 ms ) , and reward ( 0–600 ms ) . We focused out analyses on the post-decision activity and only included neurons with significant task-related activity either during 400 ms after saccade ( ‘post-saccade’ activity aligned to saccade onset ) and 600 ms during reward ( ‘reward’ activity aligned to the reward onset ) periods of the task . For each neuron , only trials in a neuron's response field were included . The data from the two monkeys were combined as they were qualitatively similar . The majority of these post-decision neurons ( psacc: 19/31 , 61%; rwd: 15/23 , 65% ) discriminated significantly only one out of eight stimuli and many of them had significantly different neuronal activity for stimuli on the inside near the category boundaries , speeds 6 and 12 ( psacc: 22/31 , 71%; rwd: 11/23 , 48% ) . | The ability to adapt behavior in a changing environment is a hallmark of intelligent systems . From adjusting our driving speed to match road conditions to responding to a last-minute change of plans , mental flexibility underpins much of our day-to-day functioning . To perform optimally , an animal must continuously monitor its own behavior and adjust it according to circumstances . A region of the brain called the caudate nucleus is thought to contribute to this process by keeping track of the relation between an action and its outcomes , but it is not clear how it monitors cognitive aspects of ongoing behavior . Yanike and Ferrera have clarified this process by recording electrical activity from the caudate nucleus in two monkeys as they categorized visual stimuli . The monkeys viewed a moving stimulus and classified it as ‘fast’ or ‘slow’ relative to a reference speed that varied from trial to trial . The monkeys were trained to use two different references speeds and were told which reference speed to use at the start of each trial . They used an eye movement to indicate their decision . Most neurons within the caudate nucleus responded after the monkey had made a decision , suggesting that these neurons might be involved in evaluating the decision that had just been made . The response of the neurons depended on the stimulus speed , and also on the category ( fast or slow ) in which the stimulus belonged . This observation indicates that the caudate nucleus tracked the context ( reference speed ) as well as the stimulus speed . Yanike and Ferrera also showed that the response of the entire population of caudate neurons could be decoded to reveal both the speed of the stimulus and whether the monkey had categorized it as fast or slow . This shows that after a decision has been made , neurons continue to signal both the stimulus and the context in which that stimulus was presented . Such ‘post-decision’ monitoring is important for anticipating the outcome of the decision . Overall the results suggest that the caudate nucleus helps animals to adapt their behavior to rapidly changing circumstances by supporting decision-making that takes context into account . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Interpretive monitoring in the caudate nucleus |
The pioneer factor hypothesis ( PFH ) states that pioneer factors ( PFs ) are a subclass of transcription factors ( TFs ) that bind to and open inaccessible sites and then recruit non-pioneer factors ( non-PFs ) that activate batteries of silent genes . The PFH predicts that ectopic gene activation requires the sequential activity of qualitatively different TFs . We tested the PFH by expressing the endodermal PF FOXA1 and non-PF HNF4A in K562 lymphoblast cells . While co-expression of FOXA1 and HNF4A activated a burst of endoderm-specific gene expression , we found no evidence for a functional distinction between these two TFs . When expressed independently , both TFs bound and opened inaccessible sites , activated endodermal genes , and ‘pioneered’ for each other , although FOXA1 required fewer copies of its motif for binding . A subset of targets required both TFs , but the predominant mode of action at these targets did not conform to the sequential activity predicted by the PFH . From these results , we hypothesize an alternative to the PFH where ‘pioneer activity’ depends not on categorically different TFs but rather on the affinity of interaction between TF and DNA .
Transcription factors ( TFs ) face steric hindrance when instances of their motifs are occluded by nucleosomes ( Kornberg , 1974; Kaplan et al . , 2009 ) . This barrier prevents spurious transcription but must be overcome during development when TFs activate batteries of silent genes . The pioneer factor hypothesis ( PFH ) describes how TFs recognize and activate nucleosome-occluded targets . According to the PFH , categorically different TFs cooperate sequentially to activate their targets . Pioneer factors ( PFs ) bind to and open inaccessible sites and then recruit non-pioneer factors ( non-PFs ) that are responsible for recruiting additional factors to initiate gene expression ( McPherson et al . , 1993; Shim et al . , 1998; Cirillo et al . , 1998; Cirillo et al . , 2002 ) . PFs also play a primary role in cellular reprogramming by first engaging silent regulatory sites of ectopic lineages ( Iwafuchi-Doi and Zaret , 2014 ) . Continuous overexpression of PFs and non-PFs can lead to a variety of lineage conversions ( Wapinski et al . , 2013; Matsuda et al . , 2019; Soufi et al . , 2015; Soufi et al . , 2012; Sekiya and Suzuki , 2011; Morris et al . , 2014 ) . The conversion from embryonic fibroblasts to induced endoderm progenitors offers one clear example ( Sekiya and Suzuki , 2011; Morris et al . , 2014 ) . This reprogramming cocktail combines the canonical PF FOXA1 ( Cirillo et al . , 2002 ) and non-PF HNF4A ( Karagianni et al . , 2020 ) and is suggested to rely upon the sequential activity of FOXA1 followed by HNF4A ( Horisawa et al . , 2020 ) . The PFH makes strong predictions about the activities of ectopically expressed PFs and non-PFs . Because PFs are defined by their ability to bind nucleosome-occluded instances of their motifs , the PFH predicts that PFs should bind to a large fraction of their motifs . However , similar to other TFs , PFs only bind a limited subset of their inaccessible motifs ( Barozzi et al . , 2014; Mayran et al . , 2018; Donaghey et al . , 2018; Manandhar et al . , 2017 ) . There are chromatin states that are prohibitive to PF binding ( Mayran et al . , 2018; Zaret and Mango , 2016 ) , and , in at least two cases , FOXA1 requires help from other TFs to bind at its sites ( Donaghey et al . , 2018; Swinstead et al . , 2016 ) . These examples suggest that PFs are not always sufficient to open inaccessible chromatin . The PFH also predicts that non-PFs should only bind at accessible sites , yet the bacterial protein LexA can pioneer inaccessible sites in mammalian cells ( Miller and Widom , 2003 ) . These observations , and the absence of direct genome-wide interrogations of the PFH , prompted us to design experiments to test major predictions made by the PFH using FOXA1 and HNF4A as a model PF and non-PF . To test these predictions , we expressed FOXA1 and HNF4A separately and together in K562 lymphoblast cells and then measured their effects on DNA-binding , chromatin accessibility , and gene activation . In contrast to the predictions of the PFH , we found that both FOXA1 and HNF4A could independently bind to inaccessible instances of their motifs , induce chromatin accessibility , and activate endoderm-specific gene expression . The only notable distinction between the two factors was that HNF4A required more copies of its motif to bind . When expressed together , co-binding could only be explained in a minority of cases by sequential FOXA1 and HNF4A activity . Instead , most co-bound sites required concurrent co-expression of both factors , which suggests cooperativity between these TFs at certain repressive genomic locations . We suggest that our findings present an alternative to the PFH that eliminates the categorical distinction between PFs and non-PFs and instead posits that the energy required to pioneer occluded sites ( ‘pioneer activity’ ) comes from the affinity of interaction between TFs and DNA .
We tested predictions of the PFH using FOXA1 as a model endoderm PF and HNF4A as a model non-PF . Because PFs are defined by their behavior in ectopic settings , we expressed FOXA1 and HNF4A in mesoderm-derived K562 lymphoblast cells . These cells express neither FOXA1 nor HNF4A and present a different complement of chromatin and cofactors . Thus , any ectopic signature that we observe is due primarily to the TFs themselves . We focused only on the initial response to TF expression to capture primary mechanisms of TF behavior and not the secondary effects that can lead to cellular conversion and that may confound our analyses . To perform these experiments , we created lentiviruses that inducibly express either FOXA1 or HNF4A ( Figure 1A ) . We created cassettes in which a doxycycline-inducible promoter drives either FOXA1 or HNF4A and cloned these cassettes separately into a lentiviral vector ( Meerbrey et al . , 2011 ) that constitutively expresses green fluorescent protein ( GFP ) . Although PFs are typically expressed at supraphysiological levels ( Ng et al . , 2021; Davis et al . , 1987 ) , we infected K562 cells with each vector at a multiplicity of infection ( MOI ) of 1 to limit the degree of nonspecific effects . We then used flow cytometry to sort single cells and selected FOXA1 and HNF4A clones that had similar GFP levels to ensure that our clones carried a similar transgene load . Finally , we performed both doxycycline titration induction and time-course experiments to identify the minimum doxycycline concentration and treatment time for robust TF activity . We observed that 0 . 5 µg/ml doxycycline for 24 hr was the minimal treatment condition that allowed FOXA1 and HNF4A , and their respective target genes ALB and APOB , to reach a plateau of expression ( Figure 1—figure supplement 1 ) . At this concentration , both FOXA1 and HNF4A were induced approximately 1000-fold ( Figure 1—figure supplement 1 ) . We used these conditions in our subsequent experiments . The first prediction of the PFH is that co-expression of FOXA1 and HNF4A should be sufficient to induce ectopic tissue-specific gene expression . We tested this prediction by infecting our FOXA1 clonal line with HNF4A-expressing lentivirus to generate a double expression clonal line , hereafter referred to as FOXA1-HNF4A . Upon co-induction in K562 cells , we observed strong enrichment for both liver- and intestine-specific gene activation; FOXA1-HNF4A activated 91 liver-specific genes ( 18 expected , p<10–38 , cumulative hypergeometric ) and 38 intestinal genes ( 9 expected by chance , p<10–13 , cumulative hypergeometric ) ( Figure 1B ) . The dual liver and intestine enrichment that we observed is consistent with the finding that intestinal gene regulatory networks appear during reprogramming experiments that aim to use FOXA1-HNF4A to convert embryonic fibroblasts to the liver lineage ( Morris et al . , 2014 ) . We conclude that FOXA1 and HNF4A are sufficient to activate endoderm-specific gene expression in the ectopic K562 line . Where ectopic genes are activated in K562 cells , the PFH predicts co-binding of FOXA1 and HNF4A at inaccessible sites and induction of chromatin accessibility . Alternatively , FOXA1 and HNF4A may not be able to overcome the K562 chromatin environment and instead activate gene expression by binding exclusively to accessible K562 sites . To distinguish between these possibilities , we measured FOXA1 and HNF4A binding by CUT&Tag ( Kaya-Okur et al . , 2019 ) after induction , and chromatin accessibility by ATAC-seq ( Buenrostro et al . , 2015 ) both before and after doxycycline induction . At the liver-specific locus ALB , FOXA1 and HNF4A co-bound at inaccessible sites and increased accessibility ( Figure 1C ) . This pattern was consistent surrounding FOXA1-HNF4A-activated liver genes: 43 of the 53 co-bound sites within 50 kb of a FOXA1-HNF4A-activated gene were inaccessible prior to induction , and the accessibility signal at these co-bound sites increased substantially upon induction ( Figure 1D and E ) . Although we focused on functional binding surrounding activated liver genes , these patterns were consistent across the genome . The vast majority of both FOXA1 and HNF4A binding sites fell within sites that were inaccessible prior to induction ( -dox ) ( Figure 1—figure supplement 2 ) , and both FOXA1 and HNF4A opened the majority of the inaccessible sites to which they bound ( Figure 1—figure supplement 2 ) . These results show that despite an entirely ectopic complement of chromatin and cofactors within mesoderm-derived K562 cells , the endodermal TFs FOXA1 and HNF4A can find and activate the correct genes . Most individual binding by FOXA1 and HNF4A near their co-activated genes occurred at the same sites bound in HepG2 liver cells ( Partridge et al . , 2020; Figure 1—figure supplement 2 ) . Altogether we conclude that when co-expressed , FOXA1 and HNF4A conform to the predictions of the PFH and that cis-regulatory sequences are sufficient to guide their activity within an ectopic cell type . We next sought to test whether ectopic tissue-specific gene expression in K562 cells results from the sequential activity of FOXA1 and HNF4A , as predicted by the PFH . The sequential activity model predicts that HNF4A will not bind to its sites without FOXA1 , and that FOXA1 will not activate expression without HNF4A , such that neither FOXA1 nor HNF4A should activate tissue-specific gene expression when expressed alone . To test this prediction , we used the single-expression K562 lines to induce either FOXA1 or HNF4A alone and measured mRNA expression by RNA-seq . FOXA1 induction resulted in strong liver-specific enrichment ( p<10–4 , cumulative hypergeometric ) and weak intestinal-specific enrichment ( not significant ) ( Figure 2A ) , while HNF4A induction resulted in both strong liver-specific enrichment ( p<10–8 , cumulative hypergeometric ) and strong intestinal-specific enrichment ( p<10–15 , cumulative hypergeometric ) ( Figure 2B ) . Importantly , neither FOXA1 nor HNF4A are expressed within K562 cells nor did they induce expression of the other TF , suggesting that the expression changes we observed were due to the independent effects of either FOXA1 or HNF4A . When expressed individually , FOXA1 and HNF4A activated largely independent sets of liver genes ( Figure 2C ) and intestinal genes ( Figure 2D ) . FOXA1 activates liver genes enriched for fibrinolysis and complement activation ( Supplementary file 1 ) , whereas HNF4A activates liver genes enriched for cholesterol import and lipoprotein remodeling ( Supplementary file 2 ) . Thus , in contrast to the predictions of the PFH , FOXA1 and HNF4A are each sufficient to induce separate and specific endodermal responses when expressed alone in K562 cells . Our results raised the possibility that both FOXA1 and HNF4A can bind and open inaccessible instances of their motifs . To test this , we induced FOXA1 and HNF4A expression individually and then measured each factor’s binding profile and their accessibility profiles before and after induction . FOXA1 induction resulted in FOXA1 binding and induced accessibility adjacent to ARG1 , a liver-specific gene that is silent in K562 cells ( Figure 3A ) , while HNF4A alone bound and induced accessibility at sites nearby the liver-specific gene APOC3 ( Figure 3B ) . This pattern was consistent across liver-specific loci . 34 of the 59 FOXA1 binding sites within 50 kb of a FOXA1-activated liver gene were inaccessible and opened upon induction ( Figure 3C and E ) as was the case for 39 of the 76 HNF4A binding sites ( Figure 3D and F ) . We observed similar patterns genome-wide . FOXA1 and HNF4A bound primarily to sites that were inaccessible prior to induction ( -dox ) ( Figure 3—figure supplement 1 ) , opened them ( Figure 3—figure supplement 1 ) , and in regions surrounding activated genes , most binding occurred at the same sites bound in HepG2 liver cells ( Figure 3—figure supplement 1 ) . We conclude that FOXA1 and HNF4A have roughly equivalent abilities to bind and open inaccessible sites . We sought to reconcile these findings with what the PFH had predicted . We first considered whether , in the absence of FOXA1 , native K562 TFs were ‘pioneering’ for HNF4A . A de novo motif discovery analysis of the 500 bp centered on inaccessible FOXA1 or HNF4A binding sites revealed strong enrichment for each TF’s motif , but no other strong signals . Similarly , we found no evidence for enrichment of predicted K562 PFs AP1 ( FOS/JUN; MA0099 . 2; Biddie et al . , 2011 ) , GATA1 ( MA0035 . 4; Iwafuchi-Doi and Zaret , 2014 ) , MYB ( MA0100 . 1; Lemma et al . , 2021 ) , or SPI1 ( PU . 1; MA0080 . 1; Iwafuchi-Doi and Zaret , 2014 ) either in inaccessible binding sites over randomly chosen sites or in HNF4A over FOXA1 binding sites ( Figure 3—figure supplement 2 ) . Thus , the similar activities of FOXA1 and HNF4A are not explained by pioneering activity provided by endogenous K562 TFs . We next considered whether differences in FOXA1 and HNF4A motif content could explain our results . We focused on binding sites surrounding activated liver genes and used FOXA1 and HNF4A position weight matrices ( Figure 3G ) to count occurrences in the 500 bp of sequence surrounding these sites . Sites independently pioneered by FOXA1 contained between 2–4 motifs , while sites pioneered by HNF4A contained 3–6 motifs ( Figure 3H ) . This is despite the fact that the FOXA1 motif occurs more frequently across the genome than the HNF4A motif ( Figure 3—figure supplement 3 ) . This observation is consistent with data showing that FOXA1 has higher affinity for its binding site than HNF4A ( Fernandez Garcia et al . , 2019; Rufibach et al . , 2006; Jiang et al . , 1997 ) and suggests that there may not be anything categorically different about FOXA1 and HNF4A , but rather that ‘pioneer activity’ may depend on the affinity of interaction between TF and DNA . Another possible explanation for our results could be that at the concentrations TFs are expressed in cellular reprogramming , the differences between PFs and non-PFs are no longer apparent . We took advantage of our doxycycline-inducible system to test this hypothesis by lowering the doxycycline concentration from 0 . 5 µg/ml to 0 . 05 µg/ml , thus dropping the TF concentration significantly ( Figure 1—figure supplement 1 ) . We then remeasured binding and expression . We found that lower induction resulted in far fewer FOXA1 and HNF4A genome-wide binding events ( Figure 3—figure supplement 4 ) . This effect was even more pronounced when we subset the binding events into sites that were either accessible or inaccessible prior to induction . Both FOXA1 and HNF4A shifted from binding predominantly inaccessible sites to binding predominantly accessible sites ( Figure 3—figure supplement 4 ) . Thus , binding of both factors depends on a balance of TF concentration and accessibility state , and the results from expression profiling in the lower induction regime are consistent with this idea . Whereas FOXA1 and HNF4A previously activated 33 and 47 liver genes , at the lower induction rate they activated 8 and 30 , respectively ( Figure 3—figure supplement 4 ) . Thus , lowering the induction levels had strong effects on the activities of both FOXA1 and HNF4A , but did not reveal qualitative differences between the two TFs . These results suggest that the induction conditions in cellular reprogramming do not mask differences between the TFs , a result consistent with the fact that the PFH was developed to explain the properties of cellular reprogramming cocktails . In addition to those genes independently activated by FOXA1 and HNF4A , there is an additional set of 31 liver genes that are not activated until both FOXA1 and HNF4A are present ( Figure 4A ) . We therefore asked whether these 31 liver genes are activated sequentially , as predicted by the PFH . If these genes conform to the PFH , then we would expect that at every gene there are nearby sites where FOXA1 binds individually and where FOXA1 and HNF4A co-bind when expressed together . This would be evidence for FOXA1 ‘pioneering’ sites for later HNF4A binding and so we have called these sites ‘FOXA1 pioneered’ ( FP ) . Sites are ‘HNF4A pioneered’ ( HP ) if HNF4A binds individually and FOXA1 and HNF4A co-bind when expressed together and sites are ‘cooperatively bound’ ( CB ) if neither TF binds individually but both do when expressed together . When there is sequential binding of the two TFs it is apparent in comparisons of the single versus double expression clones , whereas obligate cooperativity between the TFs results in binding that is observed only in the double expression clone . There are examples of each modality surrounding AMDHD1 , a liver-specific gene co-activated by FOXA1 and HNF4A ( Figure 4B ) . When we examine all of the liver genes only activated by FOXA1-HNF4A co-expression , we find that in contradiction with the PFH , there are roughly equal numbers of FP , HP , and CB sites ( Figure 4C ) . Therefore , in most cases , genes that require joint FOXA1-HNF4A activity do not rely on sequential FOXA1-then-HNF4A behavior . The patterns of genome-wide co-binding and accessibility of FOXA1 and HNF4A follow similar trends . Of the 11 , 402 co-bound sites , 2023 were FP , 3398 were HP , and 2192 were CB ( Figure 4D ) and FOXA1-induced differentially accessible peaks explain a minority of the FOXA1-HNF4A differentially accessible peaks ( Figure 4—figure supplement 1 ) . Cooperative binding may be more important in less accessible parts of the region as there are more CB sites in ChromHMM-labeled ( Ernst and Kellis , 2012 ) heterochromatic and repressed regions , and there are more FP and HP sites in promoter and enhancer regions ( Figure 4E ) . The correlation between TF binding and factors such as TF binding strength , motif content , TF concentration , and accessibility state has so far suggested that an affinity model may explain ectopic FOXA1 and HNF4A behavior . Thus , we predicted that motif counts would explain genome-wide binding patterns . Because it requires more energy to bind at inaccessible sites than accessible sites , we predicted that there would be more motifs at inaccessible binding sites than at accessible sites , and that this motif distribution would be higher than that found in random genomic sequences . When we examined the 500 bp of sequence centered upon genome-wide TF binding sites , we found that for both FOXA1 and HNF4A , inaccessible binding sites had higher motif content than accessible binding sites and that these inaccessible binding sites had higher motif content than random inaccessible regions ( Figure 5A and B ) . A simple motif threshold could predict binding , though only when predicting inaccessible sites ( Figure 5C ) . We also predicted that if FOXA1 and HNF4A are not categorically different , then we would find similar trends between the motifs for the two TFs . We predicted that total FOXA1 and HNF4A motif count at inaccessible sites would be higher than at random sites , and that FP or HP sites would have more FOXA1 or HNF4A sites , respectively , than CB sites . When we examined the 500 bp of sequence centered upon genome-wide co-bound sites , we found that there was higher total motif content at inaccessible binding sites as compared to random ( Figure 5D ) and that FOXA1 and HNF4A motif content was higher at FP or HP sites , respectively , than CB sites ( Figure 5E ) . And like individually bound sites , a motif threshold could only predict inaccessible binding behavior ( Figure 5F , top panels ) . The motif threshold was somewhat effective at differentiating between FP or HP versus CB sites ( Figure 5F , lower panel ) . Altogether , these results further support our hypothesis that affinity better explains ectopic FOXA1 and HNF4a ‘pioneer activity’ than the current formulation of the PFH .
In contrast to the predictions of the PFH , we found that both the canonical PF FOXA1 and non-PF HNF4A can independently bind inaccessible sites , increase accessibility , and activate nearby endodermal genes in a mesodermal cell line . Some endodermal genes require the joint activity of both TFs , but the predominant mode of action at these targets does not conform to the predicted sequential activity of FOXA1 followed by HNF4A . These observations suggest that we do not need to invoke the PFH to explain FOXA1 and HNF4A’s behavior in ectopic K562 cells and that instead we may use the affinity of interaction between each TF and its target sites to explain its behavior . An affinity model assumes that there is nothing categorically different between FOXA1 and HNF4A . We hypothesize that differences still exist between TFs’ abilities to bind at nucleosome-occluded sites but that ‘pioneer activity’ is a spectrum not a binary classifier . The probability of a binding event depends on the intrinsic binding ability of the TF and the motif count at a potential binding site . Previous measures of intrinsic binding strength that show FOXA1 binds more tightly than HNF4A ( Fernandez Garcia et al . , 2019; Rufibach et al . , 2006; Jiang et al . , 1997 ) may explain why in our assays FOXA1 requires fewer copies of its motif to bind . In fact , FOXA1 has a three-dimensional , histone-like structure that may explain its superior binding strength ( Clark et al . , 1993 ) . However , given the right sequence context , HNF4A also displays pioneer activity . We hypothesize that HNF4A was misclassified because of both developmental timing and indirect assays of pioneer activity . FOXA1 precedes HNF4A during hepatic development ( Lau et al . , 2018 ) and studies have traditionally established PF status by using endogenous binding or genome-wide chromatin marks . Perhaps sequential activity of FOXA1 and HNF4A is necessary during hepatic development , but our data show that both TFs are sufficient to independently activate silent genes . We further hypothesize that our findings may extend to other reprogramming cocktails that combine PFs and non-PFs . While our study is limited to two TFs at two concentrations in one cell line , other data support our hypothesis . Early reprogramming of fibroblasts to myoblasts relied solely upon the ectopic overexpression of MyoD , without an accompanying non-PF ( Davis et al . , 1987; Choi et al . , 1990 ) and new reprogramming cocktails have been tested and validated in a large-scale screen for single , cell-autonomous reprogramming TFs ( Ng et al . , 2021 ) . Increasing the efficiency of reprogramming cocktails that depend on multiple TFs will require distinguishing between the independent and cooperative effects of TFs . For example , our finding that HNF4A independently activates more intestine-specific genes than FOXA1 raises the possibility that titrating down HNF4A activity during reprogramming could result in a more liver-specific profile . Such fine-tuning of TF activities has been suggested as an option to improve the success of other reprogramming cocktails ( Ma et al . , 2015; Wang et al . , 2015; Vaseghi et al . , 2016 ) . Although we found clear instances of sites independently pioneered by either FOXA1 or HNF4A , not all sites containing multiple motifs were pioneered in K562 cells , which comports with studies showing that the sequence context in which motifs occur also plays an important role in determining whether sites will be pioneered or not . GAL4’s ability to bind nucleosomal DNA templates depends both on the number of copies of its motif ( Taylor et al . , 1991 ) and the positioning of the motif in the nucleosome ( Vettese-Dadey et al . , 1994 ) . Precise nucleosome positioning also dictates TP53 and OCT4 pioneering behavior ( Yu and Buck , 2019; Huertas et al . , 2020 ) . A TF’s motif affinity , motif count , and the presence of cofactor motifs are all strong predictors of pioneer activity ( Yan et al . , 2018; Manandhar et al . , 2017; Donaghey et al . , 2018; Heinz et al . , 2010; Boyes and Felsenfeld , 1996; Minderjahn et al . , 2020; Meers et al . , 2019 ) and certain types of heterochromatic patterning have been labeled ‘pioneer resistant’ ( Mayran et al . , 2018 ) . Thus , we hypothesize that general pioneer activity may best be summarized by the free energy balance between TFs , nucleosomes , and DNA ( Polach and Widom , 1996; Mirny , 2010 ) rather than as a property of specific classes of TFs .
We grew K562 cells ( ATCC CCL-243 , Manassas , VA ) in Iscove’s Modified Dulbecco Serum supplemented with 10% fetal bovine serum , 1% penicillin-streptomycin , and 1% nonessential amino acids . We used these cells to generate our clonal lines ( FOXA1 , HNF4A , and FOXA1-HNF4A ) , and we thank Washington University in St . Louis Genome Engineering and the iPSC Center for their help confirming K562 identity with STR profiling and testing for mycoplasma contamination . When it was time to conduct one of our functional assays , we split FOXA1- , HNF4A- , or FOXA1-HNF4A-expressing cells into replicate flasks and then treated with either ±0 . 5 µg/ml or 0 . 05 µg/ml doxycycline ( Sigma #D9891-1G ) for 24 hr . We used PCR to add V5 epitope tags to the 3′ end of FOXA1 ( Addgene #120438 , Watertown , MA ) and HNF4A ( Addgene #120450 ) constructs and then used HiFi DNA Assembly ( NEB #E2621L , Ipswich , MA ) to clone each construct into a pINDUCER21 doxycycline-inducible lentiviral vector ( Addgene #46948 ) . All primers are listed in Supplementary file 3 . The Hope Center Viral Vector Core at Washington University in St . Louis then generated and titered high-concentration virus . We infected human K562 cells at a MOI of 1 by spinoculation at 800G for 30 min in the presence of 10 µg/ml polybrene ( Sigma #TR1003G , St . Louis , MO ) , passaged the cells for 3 days , and then selected for positively infected cells by single-cell sorting on GFP+ into 96-well plates . Finally , we used qPCR to select for clones that had high inducibility of TF and target gene expression ( Figure 1—figure supplement 1 ) . We extracted RNA from 1e6 cells/sample with the PureLink RNA Mini ( Invitrogen #12183020 , Waltham , MA ) column extraction kit and completed on-column DNA digestion with PureLink DNase ( Invitrogen #12185010 ) . We quantified and assessed the quality of the RNA with an Agilent 2200 Tapestation instrument and then either froze down pure RNA for later RNA-sequencing library preparation or used ReadyScript cDNA Synthesis Mix ( Sigma #RDRT-100RXN ) to produce cDNA for qPCR . We performed qPCR with SYBR Green PCR Master Mix ( Applied Biosystems #4301955 , Waltham , MA ) and gene-specific and housekeeping primers ( Supplementary file 3 ) . We generated three replicates of ±doxycycline-treated RNA-sequencing libraries with the NEBNext Ultra II Directional RNA Library Prep Kit ( NEB #E7765S ) . We quantified and assessed the quality of the libraries with an Agilent 2200 Tapestation instrument , size selected with AMPure XP beads ( Beckman Coulter #A63880 , Brea , CA ) , and then sequenced the libraries with 75 bp paired-end reads on an Illumina NextSeq 500 instrument . We quantified transcripts with Salmon ( Patro et al . , 2017 ) , filtered out any with fewer than 10 reads , and then called differentially expressed transcripts with DESeq2 ( Love et al . , 2014 ) . A gene was called differentially upregulated if it had a log2fold change of at least 1 and was called ‘activated’ if it had fewer than 50 normalized reads in the uninduced control . A gene was called ‘tissue-specific’ according to the Human Protein Atlas definition of tissue enrichment ( Uhlén et al . , 2015 ) , which is if a gene is at least fourfold higher expressed in the tissue of interest than in any other tissue as measured by deep sequencing of RNA from the tissue of interest . We followed the Omni-ATAC protocol ( Corces et al . , 2017 ) to generate two replicates of ±doxycycline-treated low-background ATAC-sequencing libraries . We isolated 2e5 cells/sample and then extracted 5e4 nuclei/sample for tagmentation and library preparation . We quantified and assessed the quality of the libraries with an Agilent 2200 Tapestation instrument , size selected with AMPure XP beads , and then sequenced the libraries with 75 bp paired-end reads on an Illumina NextSeq 500 instrument . We aligned transcripts with bowtie2 ( Langmead and Salzberg , 2012 ) with the parameters: --local -X2000 , generated RPKM normalized BigWig files for visualization with deepTools bamCoverage ( Ramírez et al . , 2016 ) , and then called peaks at low stringency with MACS2 ( p=0 . 01 ) ( Zhang et al . , 2008 ) . With these peaks , we either called reproducible peaks with IDR ( FDR of 0 . 05 ) ( Li et al . , 2011 ) or used DiffBind ( Stark and Brown , 2011 ) to call differential peaks . We calculated the Fraction of Reads in Peaks ( FRiP ) with the Subread featureCounts tool ( Liao et al . , 2014 ) , counting reads for each replicate in the IDR-merged peak list ( Supplementary file 4 ) . We followed the CUTANA Direct-to-PCR CUT&Tag protocol ( EpiCypher , Chapel Hill , NC ) to generate two replicates of low-background CUT&Tag libraries . We isolated 1e5 cells/sample , extracted nuclei with Concanavalin A paramagnetic beads ( EpiCypher #21-1401 ) , and then either used rabbit anti-human FOXA1 monoclonal antibody ( Cell Signaling #53528 , Danvers , MA ) , mouse anti-human HNF4A monoclonal antibody ( Invitrogen #MA1-199 ) , or rabbit anti-human histone H3K4me3 polyclonal antibody ( EpiCypher #13-0041 ) as a positive control . We amplified this signal with either goat anti-rabbit ( EpiCypher #13-0047 ) or goat anti-mouse ( EpiCypher #13-0048 ) polyclonal secondary antibodies . For a negative control , we omitted the primary antibody and checked for any nonspecific pull-down . Finally , we used CUTANA pAG-Tn5 ( EpiCypher #15-1017 ) to tagment the genomic regions surrounding each bound antibody complex . We quantified and assessed the quality of the libraries with an Agilent 2200 Tapestation instrument , size selected with AMPure XP beads , and then sequenced the libraries with 150 bp paired-end reads on an Illumina NextSeq 500 instrument . When we assessed our libraries with the Agilent Tapestation instrument , we found that our negative controls had minimal signal . This is expected in the protocol , and as such sequencing the sample is recommended as optional ( Kaya-Okur et al . , 2020 ) . For this reason , we sequenced only our positive samples . We aligned our samples with bowtie2 ( Langmead and Salzberg , 2012 ) using recommended parameters ( Kaya-Okur et al . , 2020 ) : --very-sensitive --end-to-end --no-mixed --no-discordant -I 10X700 , created RPKM normalized BigWig files with deepTools bamCoverage ( Ramírez et al . , 2016 ) , and called peaks with MACS2 ( p=1e-5 ) ( Zhang et al . , 2008 ) with recommended parameters ( Kaya-Okur et al . , 2019 ) . We calculated the FRiP with Subread featureCounts tool ( Liao et al . , 2014; Supplementary file 5 ) . We then combined overlapping peaks from replicate samples using BEDTools intersect ( Quinlan and Hall , 2010 ) . We attributed binding sites to genes if they were within 50 kb ( 25 kb up- and 25 kb downstream ) of the gene’s TSS . Because co-binding occurred less frequently , we attributed co-binding sites to genes if they were within 100 kb of the gene’s TSS . ‘FOXA1 pioneered’ sites were those where we identified overlapping FOXA1 and HNF4A binding peaks within 100 kb of a gene that was only activated by FOXA1 and HNF4A and where there was also an overlapping FOXA1 binding peak , when FOXA1 was expressed alone . ‘HNF4A pioneered’ sites were those where we identified overlapping FOXA1 and HNF4A binding peaks within 100 kb of a gene that was only activated by FOXA1 and HNF4A and where was also an overlapping HNF4A binding peak , when HNF4A was expressed alone . And ‘cooperatively bound’ sites were those where we identified overlapping FOXA1 and HNF4A binding peaks within 100 kb of a gene that was only activated by FOXA1 and HNF4A and where there was neither a FOXA1 nor HNF4A binding peak . We generated lists of tissue-specific genes for each tissue by extracting ‘enriched genes’ from the Human Protein Atlas , as detailed above . We then computed hypergeometric assays to determine if our activated genes were enriched in any tissue-specific gene set . Finally , we used Panther gene ontology analysis to identify enriched biological processes . We visualized the signal from our functional assays by loading each file into the Integrated Genome Viewer ( Robinson et al . , 2011 ) using hg19 as reference . We then used the computeMatrix function in reference point mode and plotProfile function , both with default parameters , in the deepTools suite ( Ramírez et al . , 2016 ) to display aggregated CUT&Tag and ATAC-sequencing signals across indicated genomic regions . Before running motif scans , we extracted 500 bp of sequence centered on the binding sites of interest . Then , we used STREME ( Bailey , 2021 ) for de novo motif discovery and FIMO ( Grant et al . , 2011 ) for specific motif occurrence counting . We used 1e-3 as a p-value threshold and JASPAR ( Fornes et al . , 2020 ) PWMs for FOXA1 ( MA0148 . 1 ) and HNF4A ( MA0114 . 2 ) . To use motif content to predict binding , we lowered the p-value threshold to 0 to allow for weak motif contributions and then summed the motif content for each sequence . A simple threshold on this aggregate score was used as a classifier , with the receiver operating characteristic ( ROC ) curves generated by sweeping this threshold and plotting the resulting true-positive rates against false-positive rates . We used ChromHMM annotations ( Ernst and Kellis , 2012 ) to characterize the epigenetic profile of FOXA1 and HNF4A binding sites . | Cells only use a fraction of their genetic information to make the proteins they need . The rest is carefully packaged away and tightly bundled in structures called nucleosomes . This physically shields the DNA from being accessed by transcription factors – the molecular actors that can read genes and kickstart the protein production process . Effectively , the genetic sequences inside nucleosomes are being silenced . However , during development , transcription factors must overcome this nucleosome barrier and activate silent genes to program cells . The pioneer factor hypothesis describes how this may be possible: first , ‘pioneer’ transcription factors can bind to and ‘open up’ nucleosomes to make target genes accessible . Then , non-pioneer factors can access the genetic sequence and recruit cofactors that begin copying the now-exposed genetic information . The widely accepted theory is based on studies of two proteins – FOXA1 , an archetypal pioneer factor , and HNF4A , a non-pioneer factor – but the predictions of the pioneer factor hypothesis have yet to be explicitly tested . To do so , Hansen et al . expressed FOXA1 and HNF4A , separately and together , in cells which do not usually make these proteins . They then assessed how the proteins could bind to DNA and impact gene accessibility and transcription . The experiments demonstrate that FOXA1 and HNF4A do not necessarily follow the two-step activation predicted by the pioneer factor hypothesis . When expressed independently , both transcription factors bound and opened inaccessible sites , activated target genes , and ‘pioneered’ for each other . Similar patterns were observed across the genome . The only notable distinction between the two factors was that FOXA1 , the archetypal pioneering factor , required fewer copies of its target sequence to bind DNA than HNF4A . These findings led Hansen et al . to propose an alternative theory to the pioneer factor hypothesis which eliminates the categorical distinction between pioneer and non-pioneer factors . Overall , this work has implications for how biologists understand the way that transcription factors activate silent genes during development . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2022 | A test of the pioneer factor hypothesis using ectopic liver gene activation |
Many cell-intrinsic mechanisms have been shown to regulate neuronal subtype specification in the mammalian neocortex . However , how much cell environment is crucial for subtype determination still remained unclear . Here , we show that knockdown of Protocadherin20 ( Pcdh20 ) , which is expressed in post-migratory neurons of layer 4 ( L4 ) lineage , caused the cells to localize in L2/3 . The ectopically positioned “future L4 neurons” lost their L4 characteristics but acquired L2/3 characteristics . Knockdown of a cytoskeletal protein in the future L4 neurons , which caused random disruption of positioning , also showed that those accidentally located in L4 acquired the L4 characteristics . Moreover , restoration of positioning of the Pcdh20-knockdown neurons into L4 rescued the specification failure . We further suggest that the thalamocortical axons provide a positional cue to specify L4 identity . These results suggest that the L4 identity is not completely determined at the time of birth but ensured by the surrounding environment after appropriate positioning .
The mammalian neocortex consists of six layers , each containing one or more distinct subtype of neurons that share cell morphology , birth dates and connections with other regions of the central nervous system ( CNS ) ( Greig et al . , 2013; Leone et al . , 2008; Molyneaux et al . , 2007 ) . The neurons are born in the ventricular zone ( VZ ) and subventricular zone ( SVZ ) , migrate radially toward the pial surface , to eventually come to reside beneath the marginal zone ( MZ ) . Later-born neurons migrate past earlier-born neurons to more superficial layers , which results in the inside-out patterning of the cortical plate ( CP ) ( Angevine and Sidman , 1961; Berry and Rogers , 1965; Molyneaux et al . , 2007; Rakic , 1974 ) . The processes involved in the differentiation and specification of distinct neuronal subtypes during development of the neocortex have been under investigation for a long time . Several determinants of the identities of neuronal subtypes have been identified , including Tbr1 ( Hevner et al . , 2001 ) , Fezf2 ( Chen et al . , 2005a; 2005b; Molyneaux et al . , 2005 ) , Ctip2 ( Arlotta et al . , 2005 ) , Satb2 ( Alcamo et al . , 2008; Britanova et al . , 2008; Leone et al . , 2014 ) , Otx1 ( Weimann et al . , 1999 ) , Brn1/2 ( Dominguez et al . , 2013; McEvilly et al . , 2002; Sugitani et al . , 2002 ) , Sox5 ( Kwan et al . , 2008; Lai et al . , 2008 ) , and CoupTFI ( Tomassy et al . , 2010 ) , which are transcription factors expressed typically in a neuronal subtype-specific manner . However , the mechanisms by which neurons obtain such subtype-specific characteristics during development still remain largely unclear . Given that each layer of the neocortex is occupied by neurons born at around the same time ( Angevine and Sidman , 1961; Rakic , 1974 ) , the hypothesis of temporal regulation of the progenitor cells , according to which the subtype of neurons is specified depending on their birth date , has been widely accepted ( Dehay and Kennedy , 2007; Molyneaux et al . , 2007 ) . Consistent with this hypothesis , we observed that the cortical neurons acquire a birth-date-dependent segregation mechanism before their somas reach the MZ ( Ajioka and Nakajima , 2005 ) . Previous transplantation studies of ferret cerebral cortical neurons suggest that the ultimate laminar fate ( whether they eventually come to reside in the superficial layers or in the deep layers ) is determined , at least to some extent , in the progenitor cells ( McConnell and Kaznowski , 1991 ) . Recent in vitro culture studies also suggested a cell-intrinsic mechanism of subtype specification of cortical neurons ( Eiraku et al . , 2008; Gaspard et al . , 2008; Shen et al . , 2006 ) ; however , only a limited number of subtype specific markers were applied in these studies , still leaving it an open question whether all laminar fates are intrinsically determined . Although this contention had attracted little attention until recently , it is also conceivable that cortical lamination or appropriate cell positioning in the CP is required for full differentiation of neurons . The neocortex of mutant mice , such as homozygous mutant mice for Relnrl ( also known as reeler ) , which contains an almost normal set of cortical neurons ( Dekimoto et al . , 2010; Hevner et al . , 2003 ) , but severely disorganized cortical lamination , shows some changes in subtype specification ( Polleux et al . , 2001 ) . In addition , several subtype-specific characteristics of cortical neurons appear in the post-migratory phase , when the cortical layers begin to develop , implying that additional events of subtype specification may also take place . For example , pyramidal neurons begin to develop their apical dendrites only after the neurons settle beneath the MZ ( Bayer and Altman , 1991; Marin-Padilla , 1984; Tabata and Nakajima , 2001 ) . Furthermore , a number of subtype-specific molecular markers can be detected only in post-migratory neurons ( Alcamo et al . , 2008; Arlotta et al . , 2005; Britanova et al . , 2008; Kwan et al . , 2008; Lai et al . , 2008 ) . Importantly , there is a high degree of plasticity in the identity of postmitotic cortical neurons , which was revealed by ectopic expression of a subtype-specific transcription factor ( De la Rossa et al . , 2013 ) . We previously reported a set of genes that are preferentially expressed in the superficial region of the developing CP , including the primitive cortical zone ( Sekine et al . , 2011 ) , of the mouse , and proposed that several events critical for proper neuronal maturation and layer formation must take place beneath the MZ ( Sekine et al . , 2012; Tachikawa et al . , 2008 ) . However , it is still not established whether the neuronal subtype might also be affected by the microenvironment in the cortical layers after radial migration . L4 of the neocortex is composed of several types of spiny neurons , including the spiny stellate cells ( Staiger et al . , 2004 ) , that integrate thalamic inputs into cortical networks ( Lopez-Bendito and Molnar , 2003; Petersen , 2007 ) . In contrast , most of the other neocortical neurons transmit output signals to subcortical regions or other cortical regions such as the contralateral cortex ( Arlotta et al . , 2005; Thomson and Bannister , 2003 ) . Thalamocortical axons terminate primarily in L4 , and the signals received in L4 are transduced mainly to L2/3 neurons ( Lopez-Bendito and Molnar , 2003; Thomson and Bannister , 2003 ) . Reflecting their specific functions , L4 neurons show a round shape and granular morphology with multiple short dendrites , which is unique , and distinct from other neocortical “pyramidal” neurons . In addition to these unique morphological characteristics , several genes are also expressed preferentially in L4 neurons , such as Rorb ( also known as RORβ ) , Unc5d and Coup-tf1 ( Nakagawa and O'Leary , 2003; Pouchelon et al . , 2014; Zhong et al . , 2004; Zhou et al . , 1999 ) , suggesting underlying some mechanisms that might regulate the specification of L4 neurons at a molecular level . In this study , we newly identified a non-clustered protocadherin , Protocadherin20 ( Pcdh20 ) , which encodes a putative transmembrane protein , as a gene that begins to be expressed in immature postmitotic neurons , and is expressed in L4 neurons in the postnatal stages . The protocadherin family of proteins , whose adhesion strength is usually very weak or undetectable , belong to the cadherin superfamily of calcium-dependent cell-cell adhesion molecules , and are widely expressed throughout the CNS ( Kim et al . , 2007; Morishita and Yagi , 2007; Suzuki , 2000; Takeichi , 2007 ) . Although several members of this protein family , including Pcdh20 , show layer-specific expression in the CP ( Kim et al . , 2007 ) , it still remains largely unclear whether these genes may play a role in neocortical layer formation and subtype specification of neocortical neurons . Here , we investigated the role of Pcdh20 during the development of L4 . We found that Pcdh20 regulated the positioning of “future L4 neurons” into L4 via RhoA signaling , and that its knockdown not only caused malpositioning of these neurons into L2/3 , but also caused them to acquire L2/3 characteristics . Recovery of positioning of the Pcdh20-knockdown neurons to L4 also rescued the defect of their specification . Furthermore , thalamocortical axons appeared to provide a positional cue to immature L4 neurons . Our data indicate that Pcdh20 plays an essential role in the specification of L4 neurons through regulating positioning of the cells after radial migration to beneath the MZ .
To identify the genes responsible for the post-migratory events during layer formation , we explored genes that are preferentially expressed in the superficial part of the CP , where the radial migration of neurons eventually ends , on embryonic day 16 . 5 ( E16 . 5 ) and E18 . 5 ( Tachikawa et al . , 2008 ) , and are expressed in a layer-specific manner in the neocortex on postnatal day 7 ( P7 ) . As a result , we found that Pcdh20 is preferentially expressed in L4 on P7 ( Figure 1A–A” ) . No layer-specific signals were detected with the sense probe in the P7 neocortex ( Figure 1A’” ) . On E18 . 5 , Pcdh20 was expressed beneath the MZ ( Figure 1B , B’ ) in the somatosensory cortex , where a large fraction of the “future L4 neurons” resides after radial neuronal migration ( Ajioka and Nakajima , 2005 ) ( see also Figure 2H ) . We only found weak expression of Pcdh20 in the E14 . 0 and E16 . 5 neocortex ( Figure 1C , C’ , D , D’ ) , where “future L4 neurons” were being produced and were migrating ( Ajioka and Nakajima , 2005 ) . The expression levels of Pcdh20 were also analyzed by quantitative RT-PCR , and it was confirmed that the expression levels of Pcdh20 mRNA in the early stages were much lower than those in the postnatal stages ( Figure 1E ) . These results suggest that Pcdh20 begins to be expressed strongly only at a relatively late stage of radial migration toward the MZ . 10 . 7554/eLife . 10907 . 003Figure 1 . Expression of Pcdh20 mRNA in the developing neocortex . ( A–D ) In situ hybridization for Pcdh20 was performed in the E14 . 0 , E16 . 5 , E18 . 5 and P7 neocortex . The boxed regions in A–D are shown at higher magnification in A’–D’ . Nuclear staining with DAPI of the section adjacent to A’ shows the laminar structure of the neocortex ( A” ) . No layer-specific signals were detected with the sense probe in the P7 neocortex ( A”’ ) . Expression of Pcdh20 was weak in the E14 . 0 and E16 . 5 neocortex , but was clearly evident in the E18 . 5 neocortex; strong expression was observed in the P7 brain . ( E ) Quantitative RT-PCR analysis was performed at the indicated stages using Pcdh20-specific primers . Values are means ± SEM of three biological replicates . Scale bars , 1 mm ( A ) ; 500 µm ( B–D ) ; 200 µm ( A’ , A” ) ; 100 µm ( B’–D’ ) . CP , cortical plate; IZ , intermediate zone; MZ , marginal zone; SVZ , subventricular zone; VZ , ventricular zone . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 00310 . 7554/eLife . 10907 . 004Figure 2 . Pcdh20 is required for correct positioning of “future L4 neurons” . ( A ) 293T cells were transfected with a control vector ( CONsh ) , an shRNA vector targeting Pcdh20 mRNA ( PC20sh ) , or PC20sh_mut ( which harbours point mutations in PC20sh ) together with an HA-tagged Pcdh20 expression vector and a GFP expression vector . The cells were subjected to immunoblotting with antibodies to HA and GFP . ( B ) CONsh or PC20sh vector together with GFP vector was introduced on E14 . 0 cortices by in utero electroporation . Two days later , the cortices were removed , dissociated and cultured for 4 days in vitro . The GFP-positive cells were FACS sorted , and the amounts of Pcdh20 mRNA were then analyzed by RT-qPCR . The Pcdh20 levels were normalized by the expression of β-actin . Values are means ± SEM of three biological replicates . ( C ) CONsh or PC20sh vector together with GFP vector was introduced in dissociated E15 . 5 cortical cells by electroporation . Four days later , the cells were subjected to immunoblotting with antibodies to Pcdh20 and GFP . ( D , E ) CONsh , PC20sh , PC20sh_m , or PC20UTRsh vector together with a GFP vector was electroporated into the lateral ventricle of E14 . 0 embryos , then , the P7 brains were fixed and analyzed . The sections were counterstained with propidium iodide ( PI , magenta ) . Most GFP-positive cells in the control experiment were located in L4 , while the cells carrying the PC20sh or PC20UTRsh vector were located mainly in L2/3 . Results of quantitative analyses of D are presented in E ( n = 6 CONsh , n = 6 PC20sh , n = 5 PC20sh_mut , n = 5 PC20UTRsh ) . Details are described in Materials and Methods . ( F ) 293T cells were transfected with CONsh or PC20sh together with wild-type Pcdh20 ( wtPcdh20 ) or a resistant form of Pcdh20 harboring mutations in the PC20sh-targeting site ( resPcdh20 ) . The cells were subjected to immunoblotting with antibodies to Pcdh20 and GFP . ( G ) PC20sh vector together with resPcdh20 was injected , and the brains were analyzed as in E . Results of quantitative analyses of Figure 2—figure supplement 1A are presented in G ( n = 4 PC20sh+pCAGGS , n = 5 PC20sh+resPcdh20 ) . ( H–K ) CONsh ( H , J ) or PC20sh ( I , K ) vector together with a GFP vector was electroporated into E14 . 0 brains , then , E18 . 0 ( H , I ) and P2 ( J , K ) brains were analyzed . The sections were counterstained with PI ( magenta ) . The boxed regions are shown at higher magnification in the insets ( J , K ) . Note that most GFP-positive cells with the PC20sh vector migrated normally , but were malpositioned in the P2 brains . ( L , M ) Quantitative data from E18 . 0 ( H , I ) ( n = 4 CONsh , n = 4 PC20sh ) and P2 ( J , K ) ( n = 4 CONsh , n = 4 PC20sh ) brains are presented . Scale bars , 200 µm ( D , H , J ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 00410 . 7554/eLife . 10907 . 005Figure 2—figure supplement 1 . Pcdh20 is required for correct positioning of “future L4 neurons” . ( A , B ) PC20sh vector together with resPcdh20 was injected , and the brains were analyzed as in Figure 2D . ( B , C ) Experiments were done as in Figure 2D and E , except that electroporation was performed into E12 . 5 brains . ( D ) Sequential electroporation of GFP , followed 30 min later by that of PC20sh and mCherry , was performed . Control cells are magenta in color in the merged figure ( indicating that they were positive for only mCherry ) , while the PC20sh-transfected cells are green . ( E , F ) Experiments were done as in Figure 2D and E , except that electroporation was performed into E15 . 0 brains . Scale bars , 200 µm ( A , B , E ) ; 100 µm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 005 To investigate a possible role of Pcdh20 in the L4 formation , we utilized a vector-based RNA interference ( RNAi ) technique with short hairpin RNAs ( shRNAs ) to reduce the expression level of Pcdh20 during cortical development . First , we examined the knockdown efficiency of the shRNA vectors on ectopically expressed Pcdh20 . We found that expression of an shRNA vector targeting Pcdh20 ( hereinafter referred to as PC20sh ) was associated with a markedly reduced protein expression level of Pcdh20 as compared with that of a control shRNA ( CONsh ) ( Figure 2A ) . On the other hand , expression of a mutant shRNA vector harboring three point mutations in PC20sh ( PC20sh_mut ) did not significantly affect the expression level of Pcdh20 ( Figure 2A ) . Furthermore , this knockdown vector was found to markedly decrease the endogenous expression levels of Pcdh20 mRNA ( Figure 2B ) as well as protein ( Figure 2C ) in primary cortical cultures . To examine the in vivo role of Pcdh20 during cortical development , we transferred RNAi vectors into living embryos by in utero electroporation ( Tabata and Nakajima , 2001; Tabata and Nakajima , 2003 ) . Various RNAi vectors together with a green fluorescence protein ( GFP ) -expressing vector were injected into the lateral ventricles of the mouse embryos on E14 . 0 and introduced into cortical cells by electroporation . First , the pups were sacrificed on P7 , by which time , the basic structure of the neocortex was already expected to have formed . In the controls , most of the GFP-positive cells with CONsh or PC20sh_mut in the somatosensory cortex were located in L4 ( Figure 2D , E ) . On the other hand , electroporation of PC20sh changed the laminar location of the GFP-positive cells to more superficial layers ( Figure 2D , E ) . In addition , another shRNA vector targeting the 3’UTR of the Pcdh20 gene also disrupted the laminar positioning of the electroporated cells ( Figure 2D , E ) . The specificity of PC20sh for Pcdh20 was further confirmed by an experiment in which co-introduction of an RNAi-resistant Pcdh20-expressing vector ( resPcdh20 ) with PC20sh recovered the defect of neuronal positioning of the PC20sh-expressing cells ( Figure 2F , G; Figure 2—figure supplement 1A ) . We also analyzed the effects of Pcdh20 knockdown on deep layer neurons by transfecting the shRNA vectors on E12 . 5 , when L5 and L6 neurons were expected to be produced . We found that Pcdh20 knockdown in the deeper layer neurons hardly affected the cell positioning ( Figure 2—figure supplement 1B , C ) , suggesting the specific function of Pcdh20 in L4 neurons . These results together suggest the requirement of Pcdh20 for correct positioning of the cells in L4 . This function of Pcdh20 appeared to be cell-autonomous , since sequential electroporation of mCherry fluorescent protein , followed by a mixture of GFP and PC20sh , showed that the Pcdh20 knockdown did not change the positioning of the mCherry-single positive control cells ( Figure 2—figure supplement 1D ) . Given that Pcdh20 is required for correct positioning of future L4 neurons , we next examined whether ectopic expression of Pcdh20 in other layers could change their positioning to L4 . The results revealed that ectopic expression of Pcdh20 did not cause repositioning of L2/3 neurons to L4 ( Figure 2—figure supplement 1E , F ) , but caused the same cells to be located more broadly in L2/3 as compared with the control cells ( Figure 2—figure supplement 1E , F ) , suggesting that Pcdh20 expression by itself is not sufficient to change the location and morphology of L2/3 neurons to L4 , and that some other factor ( s ) are also involved in the Pcdh20-dependent positioning of the L4 neurons . The malpositioning of the neurons by Pcdh20 knockdown could be caused by impaired neuronal migration , despite the very low expression levels of Pcdh20 during the radial migration of the neurons . To determine whether the Pcdh20 knockdown affected the neuronal positioning by inhibiting radial migration of the neurons toward the MZ or by a mechanism operative after the radial migration , a series of time-course experiments were performed . Brains electroporated on E14 . 0 were analyzed on E16 . 5 , E18 . 0 and P2 , the time-points roughly corresponding to mid-migration , end of migration and post-migration , respectively ( Figure 2H–K and data not shown ) . On E16 . 5 and E18 . 0 , few differences were observed between the control and Pcdh20-knockdown brains ( Figure 2H , I , L and data not shown ) . In contrast , analysis of the P2 brains , in which the “future L4 neurons” had started to form a layer structure beneath the primitive L2/3 , revealed that the Pcdh20-knockdown cells were located in a slightly more superficial position than that in the controls ( Figure 2J , K , M ) . These data suggest that Pcdh20 may play an important role in the post-migratory stage , when the neurons are already present beneath the MZ and start to express Pcdh20 strongly . We next characterized these ectopic “future L4 neurons” malpositioned in L2/3 using several subtype-specific molecular markers . First , we examined the expression of Rorb , a well-known L4 marker in the mature neocortex ( Nakagawa and O'Leary , 2003 ) . Immunohistochemical analysis of the P7 brains showed that while most of the GFP-positive cells in the control samples expressed Rorb , the percentage of Rorb-positive cells among the GFP-positive cells was markedly decreased by knockdown of Pcdh20 ( Figure 3A–C; Figure 3—figure supplement 1A , B ) . Expression of KCNH5 ( a potassium voltage-gated channel , subfamily H ) , another marker of L4 neurons , was also not observed in the ectopically located GFP-positive cells by Pcdh20 knockdown ( Figure 3D , E; Figure 3—figure supplement 1C , D ) . Moreover , immunostaining for NetrinG1 , a marker of thalamocortical axons ( TCAs ) , showed that the malpositioned neurons by Pcdh20 knockdown did not grossly change the growth of TCAs ( Figure 3—figure supplement 1K , L ) . These results suggest that the malpositioned Pcdh20-knockdown neurons lost the characteristics of L4 neurons . On the other hand , while the expression of Lhx2 , a L2/3 marker ( Nakagawa et al . , 1999 ) , was detected in only a small fraction of the GFP-positive neurons in the controls , Pcdh20 knockdown was associated with a markedly increased percentage of Lhx2-expressing cells ( Figure 3F–H; Figure 3—figure supplement 1E , F ) . We also found that Pcdh20 knockdown was associated with an increase in the percentage of neurons positive for Satb2 , a callosal neuron maker ( Alcamo et al . , 2008; Britanova et al . , 2008 ) ( Figure 3I–K; Figure 3—figure supplement 1G , H ) , and Tbr1 , a L2/3 and L6 marker ( Hevner et al . , 2001 ) ( Figure 3L–N; Figure 3—figure supplement 1I , J ) . These results suggest that Pcdh20 knockdown caused the neurons destined for L4 to acquire L2/3 characteristics , at least at a molecular marker level . 10 . 7554/eLife . 10907 . 006Figure 3 . Acquisition of L2/3 molecular features of “future L4 neurons” induced by Pcdh20 knockdown . ( A–N ) CONsh or PC20sh vector together with a GFP vector was electroporated into E14 . 0 brains , then , P7 brains were fixed and analyzed . The sections were immunostained for Rorb ( A , B ) , KCNH5 ( D , E ) , Lhx2 ( F , G ) , Satb2 ( I , J ) , and Tbr1 ( L , M ) . The images at low magnification are shown in Figure 3—figure supplement 1 . The results of quantitative analyses for Rorb ( C ) , Lhx2 ( H ) , Satb2 ( K ) , and Tbr1 ( N ) are presented ( n = 4 CONsh , n = 4 PC20sh ) . Pcdh20-knockdown cells showed diminished expressions of Rorb and KCNH5 , but acquired expressions of Lhx2 , Satb2 and Tbr1 . *p < 0 . 0001 , **p < 0 . 005 . Scale bar , 50 µm ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 00610 . 7554/eLife . 10907 . 007Figure 3—figure supplement 1 . Pcdh20 knockdown resulted in “future L4 neurons” acquiring L2/3 characteristics . ( A–J ) Images shown in Figure 3A–M are presented at low magnification . ( K , L ) CONsh or PC20sh vector together with a GFP vector was electroporated into E14 . 0 brains , then , P7 brains were fixed and analyzed . The sections were counterstained with DAPI ( blue ) and immunostained for NetrinG1 ( red ) . ( M–T ) CONsh or PC20sh vector together with a GFP vector was electroporated into E14 . 0 brains , then , E18 . 0 ( M–P ) and P2 ( Q–T ) brains were analyzed . The sections were immunostained for Rorb and Lhx2 . Scale bars , 100 µm ( A , K ) ; 50 µm ( M , Q ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 00710 . 7554/eLife . 10907 . 008Figure 3—figure supplement 2 . Postmitotic function of Pcdh20 regulates the development of L4 neurons . ( A ) The experimental procedure to identify postmitotic cells from VZ cells . ( B ) The brains treated as in A were immunostained for BrdU , GFP and Rorb . Typical images of this experiment are shown . ( C ) The percentages of Rorb-positive cells among the GFP-positive and BrdU-negative cells were analyzed . The Pcdh20-knockdown cells showed loss of expression of Rorb even in a postmitotic population . ( D ) The percentages of BrdU-negative cells among GFP-positive cells are presented . The percentages of the BrdU-negative cells were not affected by the knockdown of Pcdh20 . ( E , F ) PC20sh vector together with the empty vector ( pTα1 ) or Tα1 promoter-driven resPcdh20 expression vector ( pTα1-resPcdh20 ) was electroporated into E14 . 0 brains , and P7 brains were analyzed . Results of quantitative analyses of E are presented in F . *p = 0 . 0056; NS , no significance ( p = 0 . 62 ) . Scale bars , 20 µm ( B ) ; 200 µm ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 008 We further investigated the stage at which misspecification of the laminar characteristics became evident during cortical development , by analyzing the time-course of changes in the expression of the molecular markers . We first analyzed the expression of Rorb and Lhx2 in the E18 . 0 neocortex . However , since Rorb expression was almost exclusively detected in L5 , and Lhx2 expression was uniform in all the superficial layer neurons ( both future L2/3 and L4 ) on E18 . 0 ( Figure 3—figure supplement 1M–P ) , we could not assess the specification/differentiation of L2/3 and L4 at this stage . On P2 , Rorb expression was detected in control L4 neurons , but not in Pcdh20-knockdown cells ( Figure 3—figure supplement 1Q , R ) , similar to the finding in the P7 neocortex , although Lhx2 expression was still relatively uniform in all superficial layer neurons ( Figure 3—figure supplement 1S , T ) , suggesting that at least some of the laminar characteristics had begun to be regulated by Pcdh20 by this stage . We next analyzed whether the axonal projection patterns were affected by Pcdh20 knockdown . Normally , many L4 neurons , which can be labeled by electroporation on E14 . 0 , extend axons locally within the ipsilateral cortex , while most of the L2/3 neurons , which can be labeled by electroporation on E15 . 0 , extend axons to the contralateral cortex ( Molyneaux et al . , 2007 ) . We injected Fluoro-Gold , a fluorescent retrograde axonal tracer , into the hemisphere contralateral to the electroporated side to label commissural neurons , and analyzed whether the electroporated neurons incorporated this fluorescent dye . We first confirmed that the percentage of retrogradely labeled neurons in L2/3 was much higher than that in L4 ( Figure 4A , C , D; Figure 4—figure supplement 1A , C ) . Interestingly , Pcdh20 knockdown was associated with a substantial increase in the percentage of the retrogradely labeled neurons in the electroporated cells ( Figure 4B , D; Figure 4—figure supplement 1B ) , suggesting that these axons projected to the contralateral cortex . These results suggest that the neurons labeled on E14 . 0 , which normally populate L4 , require Pcdh20 for proper L4 specification , and instead acquire L2/3 characteristics in the face of Pcdh20 knockdown . 10 . 7554/eLife . 10907 . 009Figure 4 . Alterations of axon projections and electrophysiological properties by Pcdh20 knockdown . ( A–D ) CONsh ( A , C ) or PC20sh ( B ) vector together with a GFP vector was electroporated into E14 . 0 ( A , B ) or E15 . 0 ( C ) brains . Fluoro-Gold was injected at P7 into the cortex contralateral to the side of electroporation , then , the P10 brains were fixed and analyzed . The images at low magnification are shown in Figure 4—figure supplement 1 . Results of quantitative analyses of A–C are presented in D ( n = 4 each group ) . ( E–G ) CONsh ( E , G ) or PC20sh ( F ) vector together with a GFP vector was electroporated into E14 . 0 ( E , F ) or E15 . 0 ( G ) brains . The cell morphologies were then analyzed in P14–19 brains by injection of biocytin and sequential visualization with avidin-Cy3 . The remaining cells are presented in Figure 4—figure supplement 2A–C . Pcdh20-knockdown neurons possessed thick apical dendrites , a characteristic morphologic feature of L2/3 neurons , but not of L4 neurons . ( H ) Effects on electrophysiological properties by Pcdh20 knockdown . Experiments were performed as described in E–G , and the electrophysiological properties were analyzed at P14-19 . *p < 0 . 05 ( compared with the Pcdh20-knockdown brains electroporated on E14 . 0 ) , **p < 0 . 005 , ***p = 0 . 0204 ( compared with the control brains electroporated on E14 . 0 ) . Scale bars , 50 µm ( A , E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 00910 . 7554/eLife . 10907 . 010Figure 4—figure supplement 1 . sEPSCs in Pcdh20-knockdown cells and control L2/3 and L4 cells . ( A–C ) Images shown in Figure 4A–C are presented at low magnification . ( D–F ) Examples of sEPSCs described in Figure 4H are presented . Scale bar , 100 µm ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 01010 . 7554/eLife . 10907 . 011Figure 4—figure supplement 2 . Morphological changes of “future L4 neurons” induced by knockdown of Pcdh20 . ( A–C ) The remaining cells in Figure 4E–G are presented . ( D–F ) CONsh ( D , F ) or PC20sh ( E ) vector together with a GFP vector was electroporated into E14 . 0 ( D , E ) or E15 . 0 ( F ) brains , then , P7 brains were fixed and analyzed . The sections were counterstained with PI ( magenta ) . High-magnification images are presented in the insets . The Pcdh20-knockdown cells possessed apical dendrites and axons , which are characteristics of pyramidal neurons in L2/3 of the neocortex . Scale bars , 50 µm ( A ) ; 100 µm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 011 We next compared the synaptic responses and membrane properties of the ectopic Pcdh20-knockdown neurons with those of the control L2/3 and L4 cells . Electrophysiological properties of control or Pcdh20-knockdown cells were examined in coronal slices of the somatosensory cortex prepared from the P14-19 brains that had been electroporated on E14 . 0 or E15 . 0 . Spontaneous excitatory postsynaptic currents ( sEPSCs ) were clearly detected in the Pcdh20-knockdown cells ( Figure 4H; Figure 4—figure supplement 1D–F ) , indicating that the malpositioned cells also received synaptic inputs . We also found that the Pcdh20-knockdown cells differed in some , if not in all , characteristics from the control L4 cells , while being similar in some characteristics to the L2/3 cells ( Figure 4H ) . These characteristics that conferred resemblance to the L2/3 neurons included the frequency of sEPSCs , half width of the action potential , and the input resistance ( Figure 4H ) . These data suggest that the electrophysiological properties of the malpositioned cells by Pcdh20 knockdown came to resemble those of the surrounding control L2/3 cells . Histochemical visualization with biocytin loaded through the recording patch pipette revealed that almost all control L4 neurons showed the morphology of stellate neurons , which extend their dendrites radially in many directions ( Figure 4E; Figure 4—figure supplement 2A ) . On the other hand , the malpositioned Pcdh20-knockdown neurons possessed a thick apical dendrite with terminal arborization in the L1 and a thin axon extending toward the ventricular side ( Figure 4F; Figure 4—figure supplement 2B ) , similar to the control pyramidal neurons in L2/3 ( Figure 4G; Figure 4—figure supplement 2C ) . These morphological changes were also observed in the magnified images of GFP fluorescence obtained on P7 ( Figure 4—figure supplement 2D–F ) . These changes in the electrophysiological properties and cell morphology , together with those in the expressions of molecular markers , strongly suggest that knockdown of Pcdh20 expression cause the neurons destined for L4 to acquire L2/3 neuronal characteristics . However , this conversion was incomplete , because the membrane capacitance of the Pcdh20-knockdown cells was almost the same as that of the control L4 cells and less than a half of that of the control L2/3 cells ( Figure 4H ) ; thus , the Pcdh20-knockdown cells still remained small in size , like the L4 neurons . We also noticed that Pcdh20-knockdown neurons tended to possess less well-developed basal dendrites as compared to the control L2/3 neurons ( Figure 4F , G; Figure 4—figure supplement 2 ) . The timing of terminal mitosis has been supposed to be critical for determination of the ultimate laminar fates ( Dehay and Kennedy , 2007; Molyneaux et al . , 2007 ) . It is thus possible that misspecification of the neuronal laminar identity after Pcdh20 knockdown was caused as a result of additional or delayed cell divisions . To address this question and determine whether the laminar specification was indeed changed by Pcdh20 knockdown in the postmitotic cells , we performed electroporation of the shRNA and GFP vectors on E14 . 0 and subsequently labeled the entire population of mitotic cells by serial injections of BrdU every 5 hr for 20 hr immediately after the electroporation ( Figure 3—figure supplement 2A ) . In this experiment , the population of GFP-positive and BrdU-negative cells ( referred to as GFP+/BrdU– cells ) was thought to represent postmitotic neurons or cells in the very late stages of the cell cycle in terminal mitosis in the VZ at the time of the electroporation , based on the following reasons . First , plasmid vectors are incorporated primarily into VZ cells by in utero electroporation . Second , since the duration of the S phase at this stage is about 4 hr , and since the nuclei in the S phase can be labeled for at least 2 hr after a single injection of BrdU ( Takahashi et al . , 1992 ) , injections of BrdU at intervals of up to 6 hr are sufficient for labeling all the mitotic VZ cells . Finally , the total length of the cell cycle is about 15 hr , which is shorter than the total period of BrdU application ( see ref . ( Tabata et al . , 2009 ) for more details ) . These GFP+/BrdU– cells are included in the “slowly exiting population” , a population that becomes postmitotic in the VZ and migrates out slowly into the multipolar cell accumulation zone above the VZ ( Tabata et al . , 2009 ) . Triple immunostaining of P7 brain sections for GFP , BrdU and Rorb revealed that the percentage of Rorb-positive cells in the GFP+/BrdU– population was markedly decreased by knockdown of Pcdh20 ( Figure 3—figure supplement 2B , C ) , suggesting that Pcdh20 knockdown perturbed L4 specification of the postmitotic cells . The extent of decrease in the percentage of Rorb-positive cells was comparable in all cell populations ( Figure 3C ) , suggesting that the effect of Pcdh20 knockdown was not influenced by whether the knockdown vector was introduced into a postmitotic population or a mitotic population . Importantly , the proportions of BrdU-negative cells in the entire population of GFP-positive cells were almost the same between the control and knockdown cells ( Figure 3—figure supplement 2D ) , suggesting that Pcdh20 knockdown did not alter cell cycle progression in the VZ/SVZ cells . We further investigated whether the postmitotic expression of Pcdh20 may be essential for its functions , by conducting a rescue experiment using a neuron-specific Tα1 ( α1-tubulin ) promoter-driven expression vector to express resPcdh20 ( Gloster et al . , 1994 ) . We found that a Tα1-driven resPcdh20 expression vector rescued the phenotype of malpositioning of the cells caused by Pcdh20 knockdown ( Figure 3—figure supplement 2E , F ) , further suggesting the important function of Pcdh20 in postmitotic neurons . We next sought to elucidate the molecular mechanisms involved downstream of Pcdh20 . Given that other protocadherins have been reported to regulate small GTPase RhoA ( Chung et al . , 2007; Unterseher et al . , 2004 ) and that Pcdh20 was demonstrated to exert an initial effect on neuronal morphogenesis , including dendritic morphology ( Figure 2J , K ) , which could be regulated by RhoA ( Redmond and Ghosh , 2001 ) , we focused on RhoA signaling . We first examined whether RhoA signaling was required for correct cell positioning of the L4 neurons . We inhibited RhoA signaling in the future L4 neurons by introducing a dominant-negative form of RhoA ( T19N , DN-RhoA ) . The DN-RhoA-electroporated neurons were located more superficially as compared with the control neurons ( Figure 5A , B ) , similar to the effect of Pcdh20 knockdown , suggesting that these two genes may function in the same signaling pathway to ensure correct positioning of L4 neurons . 10 . 7554/eLife . 10907 . 012Figure 5 . Involvement of RhoA in the Pcdh20-dependent development of L4 neurons . ( A , B ) The empty vector ( pCAGGS ) or dominant-negative RhoA expression vector ( DN-RhoA ) was electroporated into E14 . 0 brains , and P7 brains were analyzed . Results of quantitative analyses of A are presented in B ( n = 4 pCAGGS , n = 4 DN-RhoA ) . Many of the DN-RhoA-expressing cells were located in L2/3 . ( C , D ) The indicated vectors were electroporated into E14 . 0 brains , and P7 brains were analyzed . Results of quantitative analyses of c are presented in D ( n = 4 each group ) . Constitutively active RhoA ( CA-RhoA ) rescued the malpositioning caused by the knockdown of Pcdh20 . ( E–H ) The sections prepared from the brains in C were immunostained for Rorb and GFP . Results of quantitative analyses of E–G are presented in H ( n = 4 PC20sh+pEF , n = 4 PC20sh+CA-RhoA , n = 5 CONsh+CA-RhoA ) . Active RhoA restored the specification failure induced by Pcdh20 knockdown . ***p < 0 . 005 . Scale bars , 200 µm ( A , C ) ; 100 µm ( E on the left ) ; 50 µm ( E on the right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 01210 . 7554/eLife . 10907 . 013Figure 5—figure supplement 1 . DN-RhoA did not change the cell positioning in early stages . ( A , B ) The empty vector ( pCAGGS ) or dominant-negative RhoA expression vector ( DN-RhoA ) was electroporated into E14 . 0 brains , and E18 . 0 brains were analyzed . Results of quantitative analyses of A are presented in B ( n = 4 pCAGGS , n = 4 DN-RhoA ) . Scale bars , 200 µm ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 013 Modification of RhoA signaling might affect radial neuronal migration toward the brain surface , although its role in radial migration is controversial; inhibition of RhoA interfered with neuronal migration in some studies ( Pacary et al . , 2011; Xu et al . , 2015 ) , but it promoted migration in other cases ( Cappello et al . , 2012; Ge et al . , 2006; Nguyen et al . , 2006 ) . These previous discrepancies might have been caused by the differences in their experimental conditions . We examined the effect of DN-RhoA at E18 . 0 , when future L4 neurons just finished migration and started to mature . We found that there was no significant difference in cell positioning between control and DN-RhoA expressing cells at least at the concentration/condition we used ( Figure 5—figure supplement 1 ) , suggesting that the DN-RhoA-expressing neurons settle in the ridge of the CP similarly to normal cells after radial migration , although they could have migrated differently . Therefore , the phenotype observed at P7 ( Figure 5A ) was thought to be caused mainly by the event after radial migration toward the brain surface is completed . However , we do not exclude the possibility that there might be an additional effect of migration difference on subtype specification , especially in the case of enhanced migration , because there was a tendency ( not statistically significant ) for the DN-RhoA-expressing cells to position in deeper layers at P7 ( Figure 5A , B ( bins 4–6 ) ) . One might suspect that accelerated neuronal migration could cause premature differentiation , which would result in more cells in deeper layers , although this effect of DN-RhoA , if any , was thought to be minor . If RhoA is an intracellular effector downstream of Pcdh20 , the phenotypes observed by Pcdh20 knockdown should be rescued by activation of RhoA . To test this , we examined whether a constitutively active form of RhoA ( CA-RhoA ) could restore the phenotypes resulting from Pcdh20 knockdown . The results revealed that co-expression of CA-RhoA with PC20sh indeed rescued the malpositioning of future L4 neurons caused by Pcdh20 knockdown ( Figure 5C , D ) . Moreover , we examined the misspecification phenotype associated with Pcdh20 knockdown and found that the loss of Rorb expression by Pcdh20 knockdown was also restored by co-expression of CA-RhoA with PC20sh ( Figure 5E–H ) . These results suggest that RhoA functions downstream of Pcdh20 . The phenotypic consequences of Pcdh20 knockdown , disrupted neuronal positioning and change of layer identity , raise at least three possibilities . One is that the two phenotypes might arise independently , and the other two are that the disrupted neuronal positioning may affect the neuronal specification , or vice versa . To distinguish among these possibilities , we sought to disrupt the neuronal positioning of “future L4 neurons” using another method . We took advantage of knockdown of Doublecortin ( Dcx ) , since knockdown of this gene in the mouse neocortex has been reported to cause random disruption of neuronal positioning ( Baek et al . , 2014; Ramos et al . , 2006 ) . We focused on the cells that came to reside in the superficial layers ( L2–4 ) for the purpose of this study . The proportions of Rorb- as well as KCNH5-positive cells among the GFP-positive cells located in L2–4 were dramatically decreased by Dcx knockdown ( Figure 6A–E; Figure 6—figure supplement 1A–D ) . Correspondingly , the proportions of Lhx2- as well as Satb2-positive cells were markedly increased by Dcx knockdown ( Figure 6F–K; Figure 6—figure supplement 1E–H ) , implying the importance of cell positioning . On the other hand , no change in the proportion of Tbr1-positive cells was noted in association with Dcx knockdown ( Figure 6L–N; Figure 6—figure supplement 1I , J ) , suggesting that Tbr1 expression may not be regulated simply by cell positioning . To investigate the relationship between subtype specification and the ultimate laminar positioning , we further examined the proportions of Rorb-positive cells among Dcx-knockdown neurons residing in L2/3 or L4 . The majority of the neurons positioned in L4 expressed Rorb ( 72% ) , whereas none of the neurons in L2/3 expressed this marker ( Figure 6B ) . These results suggest that the ultimate cell positioning determines most of the neuronal characteristics of the neurons labeled on E14 . 0 , that is , into L2/3 or L4 neurons . 10 . 7554/eLife . 10907 . 014Figure 6 . Failure of subtype specification of the cells positioned ectopically in L2/3 by knockdown of Dcx . ( A–N ) Control ( Dcx_sh ( mut ) , harboring point mutations in Dcx_sh ) or Dcx_sh vector together with a GFP vector was electroporated into E14 . 0 brains , then , P7 brains were fixed and analyzed . The sections were immunostained for Rorb ( A , B ) , KCNH5 ( D , E ) , Lhx2 ( F , G ) , Satb2 ( I , J ) and Tbr1 ( L , M ) . The images at low magnification are shown in Figure 6—figure supplement 1 . The results of quantitative analyses for Rorb ( C ) , Lhx2 ( H ) , Satb2 ( K ) and Tbr1 ( N ) are presented ( n = 4 Dcx_sh ( mut ) , n = 4 Dcx_sh ) . Dcx-knockdown cells exhibited some of the phenotypes observed in the Pcdh20 knockdown experiment . ( O–Q ) Dcx_sh ( mut ) or Dcx_sh vector together with the PC20sh vector was electroporated into E14 . 0 brains , then , P7 brains were fixed and analyzed . The sections were immunostained for Rorb and GFP ( O , P ) . The results of quantitative analyses are presented in Q ( n = 4 Dcx_sh ( mut ) +PC20sh , n = 4 Dcx_sh+PC20sh ) . The cells showing recovery of positioning in L4 following introduction of the PC20sh and Dcx_sh vectors also showed recovery of Rorb expression . ( R–U ) CONsh or PC20sh vector together with a GFP vector was electroporated into E14 . 0 brains , then , E18 . 0 brains were dissociated . The neurons were cultured for 4 days ( R ) or 2 days ( S–U ) and immunostained for GFP and Rorb ( R ) or GFP ( S–U ) . Results of quantitative analyses are presented in R ( n =3 CONsh , n =3 PC20sh ) and U ( n = 22 CONsh , n = 21 PC20sh ) . *p < 0 . 0001 , **p < 0 . 02 , ***p < 0 . 005 . Scale bars , 50 µm ( A , O ) ; 20 µm ( S ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 01410 . 7554/eLife . 10907 . 015Figure 6—figure supplement 1 . Failure of subtype specification of the cells positioned ectopically in L2/3 by knockdown of Dcx . ( A–J ) Images shown in Figure 6A–M are presented at low magnification . ( K , L ) Images shown in Figure 6O and P are presented at low magnification . ( M–O ) Expression of PC20sh and Dcx_sh resulted in both phenotypes . Dcx_sh vector together with PC20sh and a GFP vector was electroporated into E14 . 0 brains , then , P7 brains were fixed and analyzed . The sections were counterstained with PI ( magenta ) . The boxed region in N is shown at a higher magnification in O . The neurons were positioned in a random manner throughout the neocortex . In addition , some of them showed well-developed apical dendrites , a feature of Pcdh20-knockdown cells , suggesting that both RNAi vectors were effective . Scale bars , 100 µm ( A , K ) ; 200 µm ( M ) ; 50 µm ( O ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 015 We also examined whether the neuronal malpositioning might be the major mechanism underlying the misspecification of the laminar identity of the Pcdh20-knockdown neurons . If neuronal malpositioning itself is critical for the misspecification of the Pcdh20-knockdown neurons , then this misspecification should be rescuable by recovering positioning of the knockdown neurons into L4 . We therefore electroporated both Pcdh20- and Dcx-knockdown vectors to induce random localization of the electroporated neurons in the CP and analyzed the subtypes of the neurons accidentally positioned in L4 in the absence of Pcdh20 . We found that the neurons were indeed positioned in a random manner throughout the neocortex ( Figure 6—figure supplement 1N–P ) , and that many of the neurons that were accidentally positioned in L4 expressed Rorb even in the absence of Pcdh20 ( Figure 6O–Q ) , suggesting that Pcdh20 controls L4 specification indirectly . To further test if Pcdh20 has a direct effect on cell fate or not , we examined subtype specification in an in vitro primary cortical culture , in which effects from the in vivo cell environment are thought to be excluded . We cultured future L4 neurons from E18 . 0 cortices , at which time almost no obvious changes associated with Pcdh20 knockdown are observed ( Figure 2H , I; Figure 3—figure supplement 1M–P ) , and looked at the subtype specification after 4 days in vitro , when these cells become Rorb-positive in vivo . We found that Pcdh20 knockdown did not change the percentage of Rorb-positive cells ( Figure 6R ) , suggesting that Pcdh20 requires the in vivo cell environment to exert its function on the L4 specification . These results altogether strongly suggest that Pcdh20 specifies L4 identity through controlling the cell positioning of future L4 neurons . In addition , in a similar culture condition , Pcdh20-knockdown neurons were found to possess a higher number of neurites than the control neurons ( Figure 6S–U ) , suggesting that Pcdh20 might , at least to some extent , directly regulate the neuronal morphology . We finally explored the mechanisms underlying the establishment of L4 specification through cell positioning . We focused on TCAs , because the timing of invasion of these axons into the neocortex and the immature L4 ( perinatal stages ) is well correlated with that of maturation of the L4 neurons ( Lopez-Bendito and Molnar , 2003 ) . It is thus possible that TCAs invading upward from the deeper part of the CP may provide a positional cue to the cortical neurons . We therefore analyzed Pcdh10 ( also known as OL-protocadherin or OL-pc ) -knockout mice , in which the TCAs are severely disturbed ( Uemura et al . , 2007 ) . First , we examined the extent to which the TCAs are lost in the neocortex of this mutant line . Immunohistochemistry for NetrinG1 , a TCA-specific marker , revealed a marked decrease of the TCAs in the brains of the Pcdh10 ( OL-pc ) knockout mice ( Figure 7A , B ) . We next examined the expressions of several subtype-specific markers to investigate layer identities on P3 and P7 . We found that expression of Rorb was markedly decreased in Pcdh10 ( OL-pc ) knockout mice ( Figure 7C–E ) . In parallel with the decrease in the expression of the L4 marker , expressions of Lhx2 and Tbr1 , both of which are not expressed in L4 in the control neocortex , were increased in the mutant neocortex ( Figure 7F–K ) . These results suggest that the neurons that are destined to become L4 neurons lose their L4 identity and instead acquire some of the L2/3 characteristics , presumably because of the absence of the TCAs . Moreover , these results are similar to the misspecification phenotype of the malpositioned neurons by Pcdh20 or Dcx knockdown , suggesting that TCAs might provide a positional cue to immature L4 neurons . 10 . 7554/eLife . 10907 . 016Figure 7 . Requirement of TCAs for correct differentiation of L4 neurons . ( A–K ) The sections prepared from Pcdh10 ( OL-pc ) knockout mice were immunostained for NetrinG1 ( A , B ) , Rorb ( C , D ) , Lhx2 ( F , G ) and Tbr1 ( I , J ) . The boxed regions are shown at higher magnification in the columns on the right . Typical images of the P7 cortices were shown . Quantitative analyses were performed on the P3 cortices ( E , H , K ) ( n = 6 Control , n = 5 Pcdh10 ( OL-pc ) KO for Rorb; n = 4 Control , n = 4 Pcdh10 ( OL-pc ) KO for Lhx2; n = 4 Control , n = 5 Pcdh10 ( OL-pc ) KO for Tbr1 ) . ( L ) A model for L4 specification suggested by our study is presented . We propose that the identities of the superficial layer neurons are not completely specified before the neurons eventually come to reside beneath the marginal zone ( left ) . After radial migration , the future L4 neurons become positioned in the lower part of the superficial cortical plate in a Pcdh20-dependent manner ( upper middle ) . This positioning enables the future L4 neurons to come in contact with TCAs , which appears to be required for further specification and maturation of the L4 neurons ( upper right ) . In the absence of Pcdh20 , the neurons reside in the upper part and are unable to receive signals from TCAs ( lower row ) . Scale bars , 200 µm ( A , C ) ; 100 µm ( C’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10907 . 016
This study provides the first evidence to show that the eventual positioning of the neurons in the neocortex contributes to the acquisition of the layer-specific characteristics of the cells . We propose that the laminar identities of neurons within the superficial layers ( L2–4 ) of the CP are not completely specified until the neurons arrive beneath the MZ; the ultimate fates of the L2/3 and L4 neurons are specified in the post-migratory phase according to the positioning of the cells . In this model , “future L4 neurons” are positioned in the deeper part of the superficial layers in a Pcdh20-RhoA-dependent manner after radial migration to the top of the CP , and then differentiate into mature L4 neurons , presumably by receiving signals from TCAs ( Figure 7L ) . Although post-migratory neurons were generally thought to just pile up on the earlier-settled neurons after radial migration , our results imply that the processes that take place after radial migration are also precisely regulated and are critical for correct laminar positioning of at least the “future L4 neurons” . Sufficient attention has not been paid in previous studies to the relationship between cell positioning and laminar identity specification ( Dehay and Kennedy , 2007; Ramos et al . , 2006 ) . In previous studies , the Relnrl/Relnrlneocortex , in which the cortical laminar structure is largely inverted and considerably disorganized , was shown to contain an almost normal set of neuronal subtypes ( Dehay and Kennedy , 2007; Hevner et al . , 2003; Polleux et al . , 1998 ) . Another relevant observation is that when future L2/3 neurons were induced to localize in L5 , they did not acquire the characteristics of L5 neurons ( Ramos et al . , 2006 ) . Here , we showed regulation by Pcdh20 of the cell positioning of “future L4 neurons” and also specification of their laminar identity , and concluded the causal effect of cell positioning regulated by Pcdh20 on L4 specification , which was supported by the results of the following experiments . First , malpositioning of the neurons by another method ( i . e . , knockdown of Dcx ) reproduced the failure of the laminar identity specification of the L4 neurons . Second , loss of Pcdh20 caused misspecification of L4 neurons only in vivo but not in an in vitro primary culture , where positional effects did not exist . Third , recovery of cell positioning into L4 even in the absence of Pcdh20 rescued the L4 characteristics . One might suspect that this model is controversial to the observation in the Relnrl/Relnrlcortex , where L4 neurons are actually generated . In this mutant , a substantial amount of TCAs was reported to grow into the mutant CP , although they pass a bizarre course ( Caviness and Frost , 1983 ) , which could be enough to induce L4 specification . Although it has long been supposed that cell fate decision of cortical neurons takes place mainly in the progenitor cells ( Dehay and Kennedy , 2007; Molyneaux et al . , 2007 ) , our data now suggest that L4 specification may also occur in postmitotic neurons , which is supported by several lines of evidence . First , the expression of Pcdh20 mRNA was undetectable or very low , if any , in progenitor cells , but became clearer in the postmitotic neurons beneath the MZ , as evaluated by in situ hybridization and higher by quantitative RT-PCR . Consistent with this finding , the distribution of the Pcdh20-knockdown neurons remained normal before and during radial neuronal migration , but became disrupted after the neurons settled beneath the MZ . Moreover , introduction of the Pcdh20-knockdown vector even in a postmitotic population caused misspecification of the L4 neurons . Finally , expression of Pcdh20 in the postmitotic neurons was able to rescue the phenotypes induced by Pcdh20 knockdown . This idea , of postmitotic regulation of subtype specification , is also supported by the fact that many subtype-specific genes play roles in postmitotic neurons ( Alcamo et al . , 2008; Arlotta et al . , 2005; Britanova et al . , 2008; Kwan et al . , 2008; Lai et al . , 2008; Leone et al . , 2014 ) and that there is a high level of plasticity in the identity of postmitotic cortical neurons ( De la Rossa et al . , 2013 ) . Surprisingly , as the result of the misspecification induced by Pcdh20 knockdown , which caused malpositioning of the “future L4 neurons” into L2/3 and loss of both the characteristic morphology of L4 neurons and of the expression of the L4 markers , the Pcdh20-knockdown neurons exhibited the morphological characteristics of pyramidal neurons and the functional characteristics of L2/3 neurons , including the characteristic marker expressions and axonal projections . These observations suggest that “future L4 neurons” have the capacity to differentiate into L2/3 neurons even in the postmitotic stage . Our results , however , also suggest regulation , at least in part , of the laminar fates in the progenitor cells , because the Pcdh20-knockdown “future L4 neurons” had “imperfect” characteristics as compared to the normal L2/3 neurons , such as poorly developed basal dendrites and relatively mild conversion of the axonal projection patterns in the knockdown neurons located in L2/3 . Our results also revealed the involvement of a mechanism downstream of Pcdh20 , namely , RhoA signaling , in the regulation of cell positioning of the future L4 neurons . We observed that the effect of dominant-negative RhoA mimicked the failure in neuronal positioning induced by Pcdh20 knockdown , and that CA-RhoA was able to restore the phenotypes of the Pcdh20-knockdown neurons . The relationship between protocadherin and RhoA has also been shown in other systems ( Chung et al . , 2007; Unterseher et al . , 2004 ) . Although we do not yet know the mechanism underlying the activation of RhoA by Pcdh20 , a mechanism similar to that observed in the regulation of Rho downstream of other protocadherins , such as paraxial protocadherin , may be operative ( Chung et al . , 2007 ) . How does Pcdh20-RhoA regulate the laminar positioning of the “future L4 neurons” ? We assume that the loss of Pcdh20 primarily leads to dysregulation of dendritic morphogenesis , followed by malpositioning of the L4 neurons , which is supported by several lines of evidence . First , Pcdh20 knockdown in an in vitro culture , where cell positional effects are ignorable , resulted in increased dendritogenesis without altering L4 fate specification . In vivo , Pcdh20- and Dcx double-knockdown neurons tended to exhibit pyramidal morphology even if they were in L4 , especially those located near the margin of L4 , also suggesting the primary effect , to some extent at least , of Pcdh20 on cell morphology . Regulation of dendritic morphogenesis by another protocadherin was also reported previously ( Shima et al . , 2004 ) . Second , in the time-course experiment of Pcdh20 knockdown , we noticed that loss of Pcdh20 first affected the dendrite morphology leading to extensive dendritic arborization , when the control cells already started to exhibit simplification of their complex apical primitive dendrites ( Figure 2J , K ) . Finally , active RhoA rescued the malpositioning caused by Pcdh20 knockdown with reduced arborization of the apical dendrites of the Pcdh20-knockdown neurons ( Figure 5C ) . Because the arborized dendrites were predominantly found in the MZ , it is possible that the abnormal dendrites might have caused the knockdown neurons to become ‘stuck’ to the MZ . How the Pcdh20-RhoA signaling is activated is yet to be defined . Considering that several members of the protocadherin family exhibit homophilic binding activity ( Kim et al . , 2007; Morishita and Yagi , 2007; Suzuki , 2000; Takeichi , 2007 ) , it is conceivable that homophilic interactions may regulate laminar formation through gathering neurons that share the adhesion activity . However , since the interactions through protocadherins , including Pcdh20 , are usually very weak or undetectable ( K . O . and K . N . unpublished data ) , there may be other ligands that activate Pcdh20 to regulate neuronal positioning . Then , what is the mechanism that gives positional cues to immature L4 neurons to exhibit cell-position-dependent subtype specification ? Since the acquisition of the L4-specific characteristics ( e . g . , simplification of apical dendrites ) by the “future L4 neurons” coincided with the arrival of the TCAs into the lower part of the CP ( Lopez-Bendito and Molnar , 2003 ) , we hypothesized that these axons could positively regulate the L4 identity specification . We demonstrated here that TCAs might be involved in correct subtype specification of L4 neurons by the analysis of TCA-deficient Pcdh10 ( OL-pc ) knockout mice . The results of other studies also suggested that defects in TCAs misspecify L4 identity ( Zhou et al . , 2010 ) and that TCAs instruct the development of modality-specific properties of cortical neurons ( Pouchelon et al . , 2014 ) . Here , we showed that TCA absence resulted in loss of L4 identity and , surprisingly , instead acquisition of some L2/3 characteristics , which recapitulated the phenotypes observed in Pcdh20 or Dcx knockdown , suggesting that TCAs might provide a positional cue to immature L4 neurons . Molecular mechanisms underlying this action still remain to be determined . Although neuronal transmission from TCAs to L4 neurons is an attractive candidate , it may not be required for this action . It was reported that defects in synaptic activity from TCAs to cortical neurons did not appear to be involved in subtype specification of L4 neurons in mice ( Hannan et al . , 2001; Iwasato et al . , 2000 ) . A more recent study showed that complete blockade of thalamocortical neurotransmission resulted in perturbed development of L4 neurons . Initial specification of cortical neurons , however , appeared largely normal ( Li et al . , 2013 ) , which occurs through the first week after birth , at the stage when specification failure was already observed in our mouse model . Finally , another intriguing finding of this study was the apparent involvement of a transmembrane protein in the subtype specification in the post-migratory stage . This suggested that the environment around the immature neurons could also influence the subtype specification of the neurons . Although postmitotic expressions of fate determinants , such as subtype-specific transcription factors , have been observed in the immature CP , it is not known whether these factors are induced intrinsically downstream of the fate determinants in the progenitor cells or in response to the extracellular environment . Therefore , it will be important to examine the role of the environment in the subtype specification of the postmitotic/post-migratory neurons in other contexts .
Pregnant ICR mice were purchased from Japan SLC ( Hamamatsu , Japan ) . The Pcdh10 ( OL-pc ) knockout mouse line generated by Lexicon Pharmaceuticals ( The Woodlands , TX ) has been described previously ( Uemura et al . , 2007 ) . The morning of vaginal plug detection was designated as E0 . 5 . All animal experiments were performed under the control of the Keio University Institutional Animal Care and Use Committee in accordance with Institutional Guidelines on Animal Experimentation at Keio University . The pSilencer 3 . 0-H1 plasmid ( Ambion , Austin , TX ) containing the H1 RNA promoter for the expression of short hairpin RNA ( shRNA ) was used for construction of the shRNA-encoding plasmids . The inserted sequences were 5’-TATTTCATAGAAGGACTGCACttgatatccgGTGCAGTCCTTCTATGAAATA-3’ ( the lower-case letters indicate the linker sequence ) for PC20sh , 5’-TATTTGATAGTAGGAGTGCACttgatatccgGTGCACTCCTACTATCAAATA-3’ ( the bold letters indicate mutated nucleotides ) for PC20sh_mut , and 5’-TTAGAGCTGCCAGTTATATCCttgatatccgGGATATAACTGGCAGCTCTAA-3’ for PC20UTRsh . The resultant plasmids targeted nt 1620–1640 or nt 3342–3362 of the Pcdh20 transcript ( GenBank Accession No . AK083114 ) . For expression of GFP or mCherry , the pCAGGS vector ( kind gift from J . Miyazaki , Osaka University ) carrying the enhanced GFP cDNA ( Clontech , Palo Alto , CA ) or mCherry ( Clontech ) was used . For expression of Pcdh20 , the gene encoding full-length mouse Pcdh20 , obtained from the FANTOM RIKEN full-length cDNA clones ( AK083114 ) ( Kawai et al . , 2001; Okazaki et al . , 2002 ) , was cloned into the plasmid vector , pCAGGS . In the case of expression of Pcdh20 in postmitotic neurons , the pTα1 vector harboring a neuron-specific Tα1 promoter vector was utilized ( kind gift from F . D . Miller ) ( Gloster et al . , 1994 ) . For expression of hemagglutinin ( HA ) -tagged Pcdh20 , an HA tag was inserted after the signal sequence of Pcdh20 . For the rescue experiments , a rescue vector harboring four mutations in the PC20sh targeting site was used . All constructs were verified by nucleotide sequencing . pCAGGS-DN-RhoA and pEF-CA-RhoA were kindly provided by T . Kawauchi ( Keio University ) and M . Hoshino ( National Center of Neurology and Psychiatry ) , and K . Kaibuchi ( Nagoya University ) , respectively . Dcx_sh targeting the 3’UTR of the Dcx gene and Dcx_sh_m harboring point mutations in Dcx_sh are described elsewhere ( Bai et al . , 2003 ) ( gift from J . LoTurco , University of Connecticut ) . Pregnant mice were deeply anesthetized , and in utero electroporation was carried out as described previously ( Tabata and Nakajima , 2001 ) . In brief , various shRNA expression vectors ( 4 mg/ml , unless otherwise mentioned ) together with the pCAGGS vector carrying the enhanced GFP cDNA ( 1 mg/ml ) were injected into the lateral ventricle of the intrauterine embryos , and electric pulses ( 33 V , 50 ms , 4 times ) were then applied using an electroporator ( CUY-21; NEPA GENE , Chiba , Japan ) with a forceps-type electrode ( CUY650P3; NEPA GENE ) . For injection of BrdU , the pregnant mice were injected intraperitoneally with a BrdU solution ( 50 mg per kg body weight; Sigma-Aldrich , St . Louis , MO ) . In the experiments of electroporation and serial injections of BrdU , we judged GFP+/BrdU– cells as postmitotic cells when electroporation was performed . One concern of this experiment is that few more cell divisions could produce GFP+/BrdU– cells through dilution of BrdU . However , if the cells divided for a few more cycles , GFP plasmids would simultaneously be diluted . It is thus supposed that the GFP+/BrdU– cells resulted from dilution of BrdU are , if any , very rare . Embryos and pups were fixed for 5 to 8 hr in phosphate-buffered saline ( PBS ) containing 4% PFA ( wt/vol ) , incubated overnight at 4°C with 20% sucrose in PBS ( wt/vol ) , embedded in OCT compound ( Sakura Finetek , Tokyo , Japan ) , and sectioned with a cryostat to obtain 14-µm-thick coronal sections . For detection of Rorb and BrdU , antigen retrieval was performed by autoclave treatment of the sections for 5 min at 105°C in 0 . 01 M sodium citrate buffer ( pH 6 . 0 ) . As primary antibodies , we used mouse antibody to BrdU ( BD Biosciences , San Diego , CA ) , rabbit antibody to GFP ( MBL , Nagoya , Japan ) , chicken antibody to GFP ( Abcam , Cambridge , UK ) , rabbit antibody to Rorb ( Diagenode , Leige , Belgium ) , mouse antibody to Rorb ( Perseus Proteomics , Tokyo , Japan ) , goat antibody to Lhx2 ( Santa Cruz , Santa Cruz , CA ) , mouse antibody to Satb2 ( Abcam ) , goat antibody to KCNH5 ( Santa Cruz ) , goat antibody to NetrinG1 ( R&D systems , Minneapolis , MN ) and rabbit antibody to Tbr1 ( kind gift from R . Hevner , University of Washington ) . Immune complexes were detected with FITC– , TRITC– or Cy5–conjugated secondary antibodies ( Jackson ImmunoResearch Laboratories , West Grove , PA ) . For nuclear staining , we used 1 µg/ml propidium iodide ( PI; Molecular Probes , Eugene , OR ) and 2 µg/ml DAPI ( Molecular Probes ) . Images were acquired using confocal microscopes ( FV300 and FV1000; Olympus , Tokyo , Japan ) . In situ hybridization of frozen sections was performed as described previously ( Tachikawa et al . , 2008 ) . Probes for Pcdh20 were prepared from the FANTOM clone set ( Kawai et al . , 2001; Okazaki et al . , 2002 ) . Total RNA was extracted from the neocortical tissues at various developmental stages using Trizol reagent ( Invitrogen , Carlsbad , CA ) , followed by DNase treatment , and subjected to RT with an oligo ( dT ) 12–18 primer ( Invitrogen ) . The resulting cDNA was subjected to real-time PCR in ABI PRISM 7500 ( Applied Biosystems , Foster City , CA ) using the Universal ProbeLibrary system ( Roche , Basel , Switzerland ) . The expression level of Pcdh20 mRNA was normalized relative to that of β-actin mRNA . The sense primer , antisense primer and , Universal ProbeLibrary number , respectively , were as follows: Pcdh20 , 5’-CTGAAGAGTGCGATGTTTCG-3’ , 5’-AGTGCAGGAGGAAAGCAAAC-3’ and probe #15; β-actin , 5’-CTAAGGCCAACCGTGAAAAG-3’ , 5’-ACCAGAGGCATACAGGGACA-3’ and probe #64 . 293T cells were grown in Dulbecco's modified Eagle's medium ( DMEM ) containing penicillin-streptomycin and 10% fetal bovine serum ( vol/vol ) . The cells were transfected with various plasmids using GeneJuice ( Novagen , Darmstadt , Germany ) for 24 hr , and lysed as described previously ( Oishi et al . , 2009 ) . For transfection into primary cortical cells , E15 . 5 cortices were dissociated with trypsin , and plasmids were introduced into the dissociated cells using the Amaxa Nucleofector system ( Lonza , Basel , Switzerland ) . The cells were cultured in Neurobasal medium containing 2% B27 and 0 . 4 mM L-glutamine . The lysates were then subjected to immunoblot analysis . The blots were probed with antibodies to HA ( Y11 , Santa Cruz ) , Pcdh20 ( Abnova , Taipei , Taiwan or Abgent , San Diego , CA ) and GFP ( MBL ) . Pups were anesthetized , and Fluoro-Gold ( 40 mg/ml; Fluorochrome , LLC ) was injected into the neocortex contralateral to the side of electroporation . Three or four injections per pup were administered , using about 200 nl of Fluoro-Gold solution per injection site . Three days later , the pups were fixed and analyzed . To quantify the pattern of migration , the position of each GFP-positive cell relative to the total distance from the bottom of the CP decided by the subplate to the outer edge of the CP ( pial surface ) was measured using the Image J software ( National Institutes of Health shareware program ) , followed by sorting into 10 bins . To quantify cell differentiation and Fluoro-Gold assays , the number of total GFP-positive cells and that of a given maker- or Fluoro-Gold-positive cells among GFP-positive cells were counted , and the percentages were calculated . For the analysis of Pcdh10 ( OL-pc ) knockout mice , the upper CP was first decided from DAPI staining , and the number of a given marker-positive cells was counted in 600-µm width of cortical columns . For all assays , more than 200 cells were counted for each sample . We repeated all experiments using at least two different litters , and at least four sections from at least three independent brains were analyzed for quantification . We analyzed the somatosensory area of the neocortex . Coronal slices of the somatosensory cortex ( 300 µm thick ) were prepared from P14-19 mice under deep anesthesia with isoflurane , and placed in artificial cerebrospinal fluid ( ACSF ) containing ( in mM ) : 126 NaCl , 3 KCl , 1 . 3 MgSO4 , 2 . 4 CaCl2 , 1 . 2 NaH2PO4 , 26 NaHCO3 , and 10 glucose , as described previously ( Yoshimura et al . , 2003 ) . For whole-cell recordings , GFP-positive neurons were detected by fluorescence observations and recorded under infrared differential interference contrast ( IR-DIC ) optics ( BX51WI , Olympus ) , with patch pipettes ( 4–6 MΩ ) filled with a solution containing ( mM ) 130 K-gluconate , 8 KCl , 1 MgCl2 , 0 . 6 EGTA , 10 HEPES , 10 Na-phosphocreatine , 3 MgATP and 0 . 5 Na2GTP ( pH 7 . 4 with KOH ) . Biocytin ( 0 . 2% ) was included in the solution to label the recorded cells intracellularly . We selected cells with a high seal resistance ( >1 GΩ ) and a series resistance ( <30 MΩ ) for the analysis . Membrane properties were assessed by injecting currents in current-clamp recordings or by applying step pulses in voltage-clamp recordings . Spontaneous EPSCs were recorded at -65 mV . After the recordings , the slices were fixed and resectioned . The sections were processed by a method using Cy3-conjugated streptavidin , and labeled neurons were imaged with a Zeiss LSM510 ( Oberkochen , Germany ) confocal microscope . The number of recorded cells in each condition was described in the figure . Primary neocortical neuronal cultures were prepared as described ( Oishi et al . , 2009 ) . Briefly , E18 . 0 neocortices that had been electroporated with shRNA vectors on E14 . 0 were dissociated with 0 . 125% trypsin and plated on to poly-D-lysine-coated dishes . The cells were cultured for the indicated period , fixed , and subjected to immunocytochemistry for GFP ( MBL ) or Rorb ( Perseus Proteomics ) . Data were represented as means ± SEM . Statistical analyses were performed using the two-tailed Welch's t-test . Differences between groups were considered to be significant at p<0 . 05 . Each p value was stated in figures or figure legends . | The outer surface of the brain ( the neocortex ) in mammals is formed out of neurons arranged into layers . These layers are laid down during embryonic development , and each layer has characteristic mix of distinct subtypes of neurons that have different forms and sizes . Previous evidence suggests that neurons that are born at around the same time end up in the same layer of the cortex , and tend to have the same form . It is also possible that the environment a newborn neuron finds itself in influences its eventual form . However , many previous studies have investigated the function of molecules that work within the neurons , and the effect that a neuron’s surroundings have on its development still remains largely unknown . Neurons that contain a protein called protocadherin20 normally end up in layer 4 of the neocortex . Oishi et al . genetically engineered mouse embryos so that the production of protocadherin20 was reduced in these neurons whilst the neocortex formed . These neurons were also tagged with a fluorescent marker , so that their eventual position and shape in the brain could be tracked . Examining the brains of the mice after they had been born showed that the tagged neurons ended up not in layer 4 , but in layers 2 and 3 of the neocortex . What is more , these neurons now looked similar to other neurons in layer 2 and 3 , as well as producing proteins and establishing connections consistent with their new location . However , further experiments that placed neurons with reduced levels of the protocadherin20 protein into layer 4 showed that these neurons acquire the characteristics of other layer 4 neurons , despite lacking a key layer 4 protein . These results therefore suggest that the eventual form of a neuron is not determined just by when it is born , but also by the environment that it finds itself in . In future studies , it will be important to clarify the molecular mechanism that provides the appropriate environment and so regulates the identity of the neurons in the developing cortex . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Identity of neocortical layer 4 neurons is specified through correct positioning into the cortex |
RNA-catalyzed RNA replication is widely believed to have supported a primordial biology . However , RNA catalysis is dependent upon RNA folding , and this yields structures that can block replication of such RNAs . To address this apparent paradox , we have re-examined the building blocks used for RNA replication . We report RNA-catalysed RNA synthesis on structured templates when using trinucleotide triphosphates ( triplets ) as substrates , catalysed by a general and accurate triplet polymerase ribozyme that emerged from in vitro evolution as a mutualistic RNA heterodimer . The triplets cooperatively invaded and unraveled even highly stable RNA secondary structures , and support non-canonical primer-free and bidirectional modes of RNA synthesis and replication . Triplet substrates thus resolve a central incongruity of RNA replication , and here allow the ribozyme to synthesise its own catalytic subunit ‘+’ and ‘–’ strands in segments and assemble them into a new active ribozyme .
The premise that some RNA sequences can catalyse and template their own replication - reciprocally synthesizing their own ‘+’ and ‘–’ strands - underpins current thinking about early genetic systems ( Crick , 1968; Orgel , 1968; Szostak et al . , 2001 ) . Any ancient ribozyme with such RNA replicase capability seems to be lost , but efforts are ongoing to recreate RNA self-replication in the laboratory ( Martin et al . , 2015 ) as a critical test of the ‘RNA world’ hypothesis ( Gilbert , 1986 ) . Early on , derivatives of naturally occurring self-splicing introns ( Doudna et al . , 1991; Green and Szostak , 1992; Hayden and Lehman , 2006 ) as well as later in vitro evolved ligase ribozymes ( Lincoln and Joyce , 2009; Sczepanski and Joyce , 2014 ) were shown to be able to assemble one of their own strands from cognate constituent RNA segments . However , a critical drawback of such systems is their need for specific preformed building blocks of at least eight nucleotides ( nt ) average length , limiting their potential for open-ended evolution , and precluding their replication from pools of random-sequence oligonucleotide substrates ( Green and Szostak , 1992; Doudna et al . , 1993 ) . In a contrasting approach , RNA polymerase ribozymes ( RPRs ) have been developed that can use general monomer building blocks ( ribonucleoside 5’ triphosphates ( NTPs ) ) in RNA-templated RNA synthesis ( Johnston et al . , 2001; Zaher and Unrau , 2007; Wochner et al . , 2011; Attwater et al . , 2013b; Horning and Joyce , 2016 ) , akin to the activity of modern proteinaceous polymerases . However , even the most highly-evolved RPRs ( Horning and Joyce , 2016 ) are substantially impeded by template secondary structures . Such structures are ubiquitous in larger , functional RNAs ( including the RPRs themselves ) and generally indispensable for function . The strong inhibitory role of this central feature of RNA leads to an antagonism between the degree to which an RNA sequence is able to fold into a defined three-dimensional structure to encode function ( such as catalysis ) and the ease with which it can be replicated ( Boza et al . , 2014 ) . This ostensible ‘structure vs . replication’ paradox would have placed stringent probability constraints on the emergence of an RNA replicase and generally impeded the ability of RNA to function as an early genetic polymer . We wondered whether this paradox might be avoided through a re-consideration of plausible building blocks for early RNA replication . Models of non-enzymatic polymerisation of all four activated ribonucleotides – the presumed source of the first RNA sequences – yield pools of di- , tri- and tetranucleotide etc . length oligonucleotides ( in decreasing abundance ) dominating the population alongside longer products ( Monnard et al . , 2003 ) . Here , we have examined whether substrates of such lengths can support RNA-catalyzed RNA replication , by developing a ribozyme capable of iterative templated ligation of 5’-triphosphorylated RNA trinucleotides ( henceforth called triplets ) . This heterodimeric triplet polymerase ribozyme demonstrated a striking capacity to copy a wide range of RNA sequences , including highly structured , previously intractable RNA templates , as well as its own catalytic domain and encoding template in segments . Its characterization revealed emergent properties of triplet-based RNA synthesis , including cooperative invasion and unraveling of stable RNA structures by triplet substrates , bi-directional ( both 5’−3’ and 3’−5’ ) and primer-free ( triplet-initiated ) RNA synthesis , and fidelity augmented by systemic properties of the random triplet pools .
We set out to explore the potential of short RNA oligonucleotides as substrates for RNA-catalyzed RNA replication . To do this , we required a ribozyme capable of general , iterative RNA-templated oligonucleotide ligation . Previously-described RNA polymerase ribozymes such as the ‘Z’ RPR ( Wochner et al . , 2011 ) can use NTPs to iteratively extend a primer hybridized to an RNA template , but do not accommodate oligonucleotides bound downstream of the primer or accept them as substrates . However , we detected a weak templated ligation activity in a truncated version of the Z RPR comprising its catalytic core domain ( Zcore ) ( Figure 1a ) , which supported incorporation of oligonucleotide substrates as short as three nt ( Figure 1—figure supplement 1 ) when incubated in the eutectic phase of water ice ( Attwater et al . , 2010; Mutschler et al . , 2015 ) . To be able to properly examine such RNA trinucleotide triphosphates ( triplets ) as replication substrates , we first sought to convert Zcore into an effective triplet polymerase ribozyme using in vitro evolution . We devised a selection strategy that required iterative templated triplet ligation by ribozymes to achieve their covalent linkage to a tagged primer ( Figure 1—figure supplement 2 ) . This enables their recovery , amplification and mutagenesis before further rounds of selection to enrich the selection pool in improved triplet polymerase ribozyme variants . We initiated selections from a library of 1 . 5 × 1015 Zcore variants with a new random 3’ N30 region under eutectic phase conditions that increase RNA half-life and enhance ribozyme activity ( Attwater et al . , 2010;Attwater et al . , 2013b ) . After 7 rounds of in-ice evolution , one-quarter of the selection pool comprised an improved ribozyme ( type 0 ) . Its core domain ( 0core , Figure 1a ) could catalyse the iterative polymerization of multiple triplets allowing us to begin to investigate the properties of triplet-based RNA replication . Significantly , we found that 0core could catalyze triplet polymerisation on a series of structured templates , which had proven intractable to the parental Z RPR ( Figure 1b ) . Here , primer extension exhibited a steep sigmoidal dependence upon triplet concentrations ( Figure 1c ) , suggestive of a cooperative invasion and unraveling of template secondary structures by the triplet substrates themselves . Although still inefficient , the fact that the nascent activity of the 0core ribozyme could already copy templates that had confounded an established RPR encouraged us to continue to seek improved triplet polymerase ribozymes to leverage this substrate behaviour . We continued selections for a further 14 rounds . At this point , the type 0 ribozyme had gone extinct , replaced by six new types of RNA each characterised by a unique 3’ domain ( Figure 2a , Figure 2—figure supplement 1 ) . Type 1 RNAs were the most abundant , comprising ~50% of pool sequences , but mysteriously were catalytically inactive with diverse mutations in their core domains . In contrast , the type 2–6 RNAs all displayed triplet polymerase activity , but fell short of the polyclonal activity of the selection pool ( Figure 2—figure supplement 2 ) . To attempt to explain this discrepancy , we explored potential interactions among the different pool lineages , and found that addition of an equimolar amount of type 1 RNA substantially enhanced triplet polymerase activity of all the other ribozyme types 2–6 ( Figure 2b ) . Dissecting type 1 RNA function , we found that 5’ truncation of the region that previously contacted the primer/template duplex ( Shechner et al . , 2009 ) did not affect its cofactor activity ( Figure 3a , Figure 3—figure supplement 1 ) . As judged by gel mobility shift ( Figure 3b ) and activity enhancement ( Figure 3—figure supplement 1 ) , type 1 RNA appears to form a 1:1 heterodimeric complex directly with active triplet polymerase ribozymes . Our attention was drawn to their selection construct-derived 5’ hairpin elements , which differed between active triplet polymerases ( ‘cap+’ , Figure 3a , Figure 2—figure supplement 1 ) and the most common type 1 variants in the selection pool where this hairpin had acquired a mutation ( yielding ‘cap–’ , Figure 3a ) . ‘cap–’ was dispensible for type 1’s cofactor activity , but when replacing ‘cap+’ in active triplet polymerases it abolished both their activity enhancement by type 1 ( Figure 3—figure supplement 1 ) and complex formation ( Figure 3b ) . This points to the ‘cap+’ hairpin as the critical site of interaction with type 1; ‘cap–’ in type 1 presumably served to deter its homodimerisation during selection . Indeed , transplanting the ‘cap+’ element could make the parental ribozymes ( Zcore and Z RPR ) receptive to activity enhancement by type 1 RNA ( Figure 2b , Figure 3—figure supplement 2 ) . The catalytically inert type 1 RNA thus represents a general , mutualistic RNA species . This molecular symbiont appears to have emerged spontaneously during in vitro evolution by forming a heterodimeric holoenzyme with triplet polymerase ribozymes , enhancing their activity to boost the recovery prospects of both complex components . In complex with type 5 ( the fastest enriching triplet polymerase ribozyme in the final selection pool ) , type 1 boosts polymerization of triplets ( or longer oligonucleotides ) to enable synthesis of long RNAs ( Figure 3c ) . Here , it became apparent that type 1 also obviates the need for ribozyme-template tethering . Due to their poor affinity for primer/template duplex ( Lawrence and Bartel , 2003 ) , RPRs generally depend upon such tethering to template ( Attwater et al . , 2010; Wochner et al . , 2011; Horning and Joyce , 2016 ) , which enhances local ribozyme concentration and promotes formation of the RPR-primer/template holoenzyme ( Attwater et al . , 2010; Attwater et al . , 2013a ) . In contrast , the triplet polymerase heterodimer appears to have a capacity for true intermolecular , sequence-general interaction with primer-template duplexes , which enables holoenzyme formation and copying of RNA templates without requiring specific ribozyme-template hybridization sites . We performed an additional five rounds of in vitro evolution to further evolve the type 5 triplet polymerase ribozyme ( now in the presence of truncated type 1 RNA ) , diversifying the previously-fixed 3’ domain reverse transcription primer binding sequence . This reselection yielded a shorter final heterodimeric triplet polymerase holoenzyme , hereafter termed ‘t5+1’ ( Figure 4 ) . This robust triplet polymerase activity now proved suitable for exploring the scope and potential of triplet-based RNA replication . As a first examination of t5+1 activity , we revisited triplet-based RNA synthesis on structured templates . To provide a stringent test of template structure inhibition , we now examined hairpin-containing templates ( 4S , 6S , 8S ) with increasing RNA hairpin stability and estimated TMs of up to 93˚C ( 8S ) . The latter had previously strongly arrested even the most advanced mononucleotide RPRs at higher temperatures ( Horning and Joyce , 2016 ) . However , using triplets as substrates t5+1 robustly copied all of these ( Figure 5a ) , even when templates were pre-folded allowing RNA secondary structures to form prior to triplet addition ( Figure 5—figure supplement 1 ) . The triplet concentration-dependent cooperative structure invasion and unraveling ( previously observed with the simpler 0core domain and partly wobble-paired RNA template structures [Figure 1b , c] ) was recapitulated with t5+1 and the highly stable 8S hairpin template ( Figure 5—figure supplement 2 ) . In contrast , dinucleotide triphosphate substrates yielded extension only up to the structured region ( Figure 5—figure supplement 2 ) . We began to explore whether triplet-based RNA synthesis by t5+1 might exhibit the generality required not just for synthesis of arbitrary structured sequences , but for replication of functional sequences ( requiring synthesis of both ‘+’ and ‘−’ strands ) . Encouragingly , t5+1 could synthesise both a functional fluorescent ‘+’ strand of the 52 nt Broccoli RNA aptamer ( Filonov et al . , 2014 ) and its encoding ‘−’ strand template from their 13 ( + ) and 12 ( − ) different constitutive triplets ( Figure 5b ) . We next turned to the critical test of generality: could triplet substrates allow self-synthesis ? As t5+1 currently lacks the efficiency to synthesise RNAs its own length , we divided the catalytic t5 ribozyme into five segments α , β , γ , δ and ε . This segmentation strategy ( akin to that used by some RNA viruses e . g . influenza ) could reduce tertiary structures ( Doudna et al . , 1991; Mutschler et al . , 2015 ) and ease product separation during RNA replication ( Szostak , 2012 ) . Starting from ~8 nt RNA primers , t5+1 achieved synthesis of the β+ , γ+ , and δ+ segments from their constitutive triplets as well as all of the ‘−’ strand segments α- , β- , γ- , δ- and ε- , but required some triplets pre-linked ( as e . g . hexa- or nonanucleotides ) for synthesis of full-length α+ and ε+ segments ( Figure 6a ) . Operating across 70 distinct ligation junctions in these reactions including AU-rich sequences , t5+1demonstrates the sequence generality for self-synthesis using triplet substrates . Notably , the average extent of ligation per junction during synthesis of t5 ‘+’ and ‘−’ strands ( 78% ) was similar to that observed when t5+1 used an unstructured model template ( 74% , Figure 3c ) upon which the parental Z and other RPRs excel ( Attwater et al . , 2013b; Horning and Joyce , 2016 ) . At this point , we tested whether the broad oligonucleotide ligation capacity of t5+1 ( Figure 3c ) might allow assembly of synthesised ‘+’ strand segments . Indeed , t5+1 could assemble these into αβ+ and γδε+ fragments , guided only by partially overlapping ‘−’ strands ( Figure 6b , Figure 6—figure supplement 1 ) . Through non-covalent association ( Vaish et al . , 2003; Mutschler et al . , 2015 ) , the ribozyme-synthesised αβ+ and γδε+ fragments spontaneously reconstituted a new catalytically active triplet polymerase ribozyme ( with in vitro transcribed type 1 RNA ) . We found that this synthesis product could regenerate fresh δ- segment using t5+1 ribozyme-synthesised δ+ ( left over from ribozyme assembly ) as a template ( Figure 6c ) , recapitulating elements of a self-replication cycle . However , while the t5+1 ribozyme displays a nascent capacity for templated synthesis of its own catalytic domain ‘+’ strands ( and ‘−’ strands ) , efficiency of both segment synthesis and assembly will need to be increased significantly to realise a full self-replication cycle ( which would also require synthesis and replication of the type 1 subunit ) . Templated ‘+’ strand self-synthesis is a central element of ribozyme self-replication . However , a limitation of our above strategy in the context of triplet-based self-replication is the continued requirement for some pre-synthesized longer oligonucleotides to act as primers and occasional substrates ( together providing here the equivalent of ~25% of triplet junctions pre-ligated ) . In particular , some specific oligonucleotide substrates were required for efficient synthesis of α+ and ε+ segments to compete out inhibitory mutual hybridisation between ‘−’ strand template and corresponding ‘+’ strand unstructured elements in the t5 ribozyme ( Figure 6—figure supplement 2 ) . In vitro selections that stabilise the ribozyme tertiary structure ( Figure 6—figure supplement 3 ) may contribute to attenuating this requirement . Additionally , more concentrated triplet substrates can successfully compete with ribozyme unstructured elements for hybridization to ‘−’ strand templates ( Figure 6—figure supplement 2 ) . The majority of specific oligonucleotides , however , were provided as primers to initiate syntheses , as required by all RPRs akin to the activity of replicative polymerases in biology . As a consequence of this , the priming sequence would effectively be excluded from evolution during replication . Furthermore , RNA oligonucleotides able to act as specific primers are unlikely to be prevalent in prebiotic substrate pools , and their depletion during successive replication cycles could lead to loss of sequence at genome ends . This ‘primer problem’ has previously been noted in the context of nonenzymatic replication ( Szostak , 2012 ) as one of the fundamental obstacles to RNA self-replication . Unexpectedly , triplet substrates provide a route to bypass the ‘primer problem’ . We observed that t5+1 can extend primers bidirectionally , in both the canonical 5’−3’ as well as the reverse 3’−5’ directions ( Figure 7a ) . This allows not only completion of RNA synthesis from either template end but also initiation from anywhere along a template , potentially allowing non-classical hierarchical or distributive RNA replication schemes as previously proposed ( Szostak , 2011 ) . Given this flexibility , we wondered if t5+1 even had a requirement for a primer oligonucleotide . Indeed , this triplet polymerase could achieve ‘primer free’ RNA synthesis ( whereby synthesis is presumably initiated by ligation of adjacent triplets anywhere on the template ) , as exemplified here for the β+ segment ( Figure 7b ) , as well as ‘primer free’ RNA replication as shown for the ‘+’ and ‘−’ strands of the γ segment , which can be replicated using triplets alone ( Figure 7c ) . Thus , the capacity of triplet substrates to pre-organise themselves on a template not only enables replication of structured templates but also allows complete copying of some RNA sequences exclusively from triplet building blocks , suggesting an alternative to the canonical end-primed replication strategies inspired by PCR . Such a ribozyme operating in a more distributive polymerisation mode might be able to replicate RNA sequences directly from the putative pools of short random RNA oligonucleotides furnished by prebiotic chemistry . Next , we investigated the consequences of using analogues of such prebiotic pools as a source of substrates for the t5+1 triplet polymerase ribozyme . Random sequence triplet pools ( ‘pppNNN’ , comprising equimolar amounts of all 64 triplets ) could be used as substrates by t5+1 in segment syntheses in place of defined triplet sets ( Figure 6—figure supplement 4 ) . Furthermore , extension activity remained robust upon pool supplementation with noncanonical dinucleotide and mononucleotide substrates ( Figure 6—figure supplement 5 ) . However , a replicase must incorporate the correct template-complementary substrate from random sequence pools , or genetic information may become irretrievably corrupted during replication ( Eigen , 1971 ) . Sequence fidelity is therefore a critical parameter of RNA replication . The fidelity challenge is exacerbated in triplet-based RNA replication by the need to discriminate between 64 distinct substrates; indeed , a previous investigation into the incorporation of individual trinucleotides indicated that misincorporations could outstrip cognate incorporation for some triplets ( Doudna et al . , 1993 ) . In order to assess the fidelity of triplet polymerase ribozymes of widely differing activity , we identified the triplets incorporated from random pppNNN triplet pools using 12 different compositionally representative N′N′N′ triplet sequences as templates . These were examined in a consistent sequence context ( 5’-GGG-N′N′N′-GGG-3’ ) and collated , which allowed an estimation of ribozyme misincorporation tendencies . On average , the starting Zcore ribozyme exhibited ~91% fidelity per position ( Figure 8a ) , lower than that described for RPRs ( 92% – 97% [Attwater et al . , 2013b; Horning and Joyce , 2016] ) . Furthermore , its accuracy exhibited a pronounced downward gradient from the first ( 5’ ) to the third ( 3’ ) triplet position , highlighting escalating risks to fidelity of synthesis founded on longer building blocks . To investigate if ribozymes could exhibit higher triplet incorporation fidelity , we had included a persistent adaptive pressure for fidelity during in vitro evolution , spiking in an excess of mispairing 3’-deoxy ‘terminator’ triplets from round nine onwards , precluding recovery of ribozymes that incorporated these mispairs ( Figure 2—source data 1 , Figure 4—source data 1 , Figure 2—figure supplement 2 ) . This yielded reshaped and improved fidelity profiles in the ‘surviving’ type 2–6 ribozymes ( Figure 8a ) . Notably , the final t5+1 ribozyme achieves an average positional fidelity of 97 . 4% using pppNNN in this sequence context , higher than the best RPR fidelity with NTPs under comparable eutectic conditions ( Attwater et al . , 2013b ) . Deep sequencing of internal triplet positions of a defined sequence ( β+ segment ) synthesised by t5+1 using pppNNN indicated similar aggregate fidelity could be achieved during longer product synthesis excluding the final triplet ( Table 1 ) . Having established that accurate triplet-based copying is possible ( in at least some sequence contexts ) , we sought to understand how the triplet polymerase ribozyme achieves it . Investigating the fidelity contributions of different t5+1 ribozyme components , we found that the type 1 RNA cofactor did not contribute; rather , fidelity gains appeared to be mediated by the newly-evolved t5 ‘ε’ 3’-domain , as its deletion ( yielding the truncated ‘αβγδ’ ribozyme ) reverted the fidelity profile towards that of Zcore ( Figure 8—figure supplement 1 ) . Presence of the ε domain did not uniformly increase fidelity , but selectively reduced the most acute errors at the second and third triplet positions ( with over 10-fold reductions for some errors , Figure 8b , Figure 8—figure supplement 2 ) . Overall error rates at the second and third triplet positions were reduced by 4-fold and 9-fold compared to Zcore ( Figure 8a ) , though increased ( 1 . 3-fold ) at the first triplet position due to a localised asymmetric tolerance of G:U wobble pairing ( Figure 8b ) . The ε domain fidelity function is contingent upon the presence of a downstream triplet , operating only with basal fidelity for final triplet incorporation ( Figure 8—figure supplement 3 ) . Dissecting the molecular determinants of the fidelity phenotype , we found that using triplet substrates modified at the third position with a 2-thiouracil in place of a uracil ( disrupting minor groove hydrogen bonding capabilities ) rendered the ε fidelity domain unable to discriminate mismatches ( Figure 8c , Figure 8—figure supplement 3 ) . Previously , a similar replacement of a uracil 2-keto group with a 2-thio modification had been shown to impair Z RPR activity when present upstream in the primer/template region ( Attwater et al . , 2013a ) , where Z is thought to rely upon sequence-general minor groove contacts through an ‘A-minor’ motif ( Shechner et al . , 2009 ) . Modification at the third triplet position reverts ε’s divergent effects on fidelity at the adjacent second and the distal first triplet positions ( Figure 8—figure supplement 3 ) ; disruption of this minor groove contact site thus abolishes overall ε fidelity domain operation . ε sensitivity to minor groove composition may be critical to its recognition of cognate Watson-Crick base pairs , reminiscent of Tetrahymena group I intron folding ( Battle and Doudna , 2002 ) and the decoding centre of the ribosome ( which also tolerates wobble pairing at the analogous ( 5’ ) triplet position ) ( Ogle et al . , 2001 ) . An important contribution to triplet fidelity also appears to arise from unexpected behaviours of the triplet substrates themselves . We observed that in some direct pair-mispair triplet contests , inclusion of their complementary triplets caused a striking ( ~3 fold ) drop in misincorporation errors ( Figure 9 ) . A potential explanation may arise from differential formation of triplet:anti-triplet dimers in the reaction: for example , more extensive pppGCC:pppGGC ( than pppACC:pppGGU ) dimer formation would selectively reduce the effective concentration of free pppGCC vs . pppACC upon inclusion of their complementary pppGGC and pppGGU . These pairwise reductions were recapitulated in the presence of random pppNNN substrate pools ( Figure 9 ) . Indeed , counterintuitively , raising pppNNN concentrations from 0 . 5 to 5 μM each almost halved the overall error rate ( Figure 9—figure supplement 1 ) . Although diverse effects upon individual misincorporations were observed , this fidelity enhancement was driven by pronounced reductions in errors where the mismatched triplet has a high GC content compared to the cognate triplet , including common G-U wobble mispairs ( Figure 9—figure supplement 1 ) . Dimer formation among pppNNN substrate pools would be expected to selectively buffer the free concentrations of the more strongly-pairing GC-rich triplets , which could promote both fidelity and sequence generality through normalization of triplet availability against template ( and complementary triplet ) binding strength . Indeed , more efficient , higher fidelity segment synthesis was observed when partially mimicking this outcome using an pppNNN pool formulated with a reduced G content ( Figure 6—figure supplement 4 , Table 1 ) . In a prebiotic scenario , substrate pool composition would have been determined by the abundance and nontemplated polymerization tendencies of the different nucleotides; large biases in these could skew triplet compositions or deplete a triplet ( resulting in mismatch incorporation ) . However , the potential for replication to proceed in different triplet registers may provide a degree of resilience towards such biases .
Here , we describe the discovery and characterization of a ribozyme ( t5+1 ) with a robust ability to polymerize RNA trinucleotide triphosphate ( triplet ) substrates . Unusually , this triplet polymerase ribozyme comprises a heterodimer of a catalytic triplet polymerase subunit ( t5 ) and a non-catalytic RNA cofactor ( type 1 ) , which enhances triplet polymerase activity and abrogates the need for template tethering . Such a quaternary structure - involving a heterodimer of a full-length and a truncated subunit - is reminiscent of the processivity factors of some proteinaceous polymerases such as the heterodimeric p66/p51 HIV reverse transcriptase holoenzyme ( Huang et al . , 1992 ) . There are multiple examples of dimerization in RNA evolution - such as the VS ribozyme ( Suslov et al . , 2015 ) , retroviral RNA genome dimerization ( Paillart et al . , 2004 ) , in vitro evolved heterodimeric RNA liposome binders ( Vlassov et al . , 2001 ) , and recently the homodimeric CORN fluorescent RNA aptamer ( Warner et al . , 2017 ) . However , the spontaneous emergence of a general , mutualistic RNA cofactor has not previously been observed for ribozymes and may suggest an underappreciated dimension to the evolutionary dynamics of ribozyme pools under stringent adaptive pressures . Indeed , the extinction of previously dominant species in the selection that were unable to benefit from type 1 enhancement ( e . g . type 0 , see Figure 3—figure supplement 2 ) and succession with cooperative RNA species ( Vaidya et al . , 2012 ) illustrates the potential for such symbioses to shape RNA molecular ecologies . The t5+1 ribozyme’s principal current shortcoming is its low catalytic efficiency . In the optimal context for mononucleotide polymerase ribozymes , this triplet polymerase heterodimer yields ~4 fold more unligated junctions than the RPR tC19Z ( Attwater et al . , 2013b ) , which itself is 240-fold slower than the currently most advanced RPR 24–3 ( Horning and Joyce , 2016 ) . Yet despite this modest catalytic power , t5+1 displays much enhanced generality in RNA synthesis and now achieves both copying of previously intractable structured RNA templates , and templated synthesis and assembly of an active ‘+’ strand copy of its catalytic domain , suggesting key contributions of the triplet substrates themselves . Indeed , one of the main findings of our work are the compelling advantages that triplet substrates appear to offer for sequence general RNA replication . For instance , when binding templates , triplets incur a lower entropic cost per position compared to canonical mononucleotides ( thus aiding copying of sequences rich in weakly pairing A and U bases ) , with particularly helpful stability contributions from intra-triplet base stacking ( Eigen , 1971 ) . Furthermore , energetically favourable inter-triplet stacking interactions appear to instigate cooperative binding and unfolding of even highly stable RNA template structures ( Figures 1b and 5a ) upon reaching the required substrate concentration threshold . In our work , this process is aided by the cold temperature and solute concentration effects of eutectic ice phase formation ( Attwater et al . , 2010; Mutschler et al . , 2015 ) . Counterintuitively , a general solution to the copying of structured RNAs arises not from conditions that disfavour base-pairing ( which would also hinder substrate binding ) , but rather from conditions that promote it . Together these favourable molecular traits serve to pre-organize the template towards a double-stranded RNA duplex with triplet junctions poised for ligation . A triplet/template duplex presents a more ordered , regular target for sequence-general ribozyme docking ( by e . g . the ε domain ) than a single stranded template ( variably prone to secondary structure formation or sequence-specific interactions with the ribozyme [Wochner et al . , 2011] ) . Such general duplex interactions also underlie other notable features observed in our triplet-based RNA synthesis such as in trans template binding ( Figure 3c ) as well as the capacity for bidirectional ( 5’−3’/3’−5’ ) and primer-free RNA synthesis ( Figure 7 ) . Contrary to expectations RNA-catalyzed triplet polymerisation can proceed with a fidelity matching or exceeding even the best mononucleotide RNA polymerase ribozymes ( Attwater et al . , 2013b; Horning and Joyce , 2016 ) . t5+1 ribozyme fidelity is due to both a readout of cognate minor groove interactions by the ribozyme ε domain ( Figure 8 ) and an unanticipated fidelity boost arising from systems-level properties of triplet pools , that appear to normalize the availability of free triplet ( and potentially longer oligonucleotide ) substrates against their base-pairing strength ( Figure 9 ) . Though further work will be required to characterize triplet pool properties , they likely involve formation of cognate or near-cognate triplet:anti-triplet interaction networks , as formation of tRNA dimers via cognate anticodon:anticodon interactions has been observed in a similar concentration range ( Eisinger and Gross , 1975 ) . While phylogenetically unrelated , mechanistic analogies between the triplet polymerase ribozyme and the ribosome are apparent . Both are RNA heterodimers that operate in a triplet register along a single-stranded RNA template , whilst enforcing a minor-groove mediated pattern of triplet or anticodon readout ( including tolerance of 5’ wobble pairing ) , suggestive of convergent adaptive solutions to the challenges of replication and decoding . It has long been speculated that the decoding centre of the small ribosomal subunit might have had its origins in an ancestral RNA replicase , but the implied triplet-based character of such a replicase was conspicuously discordant with modern mononucleotide-based replication ( Weiss and Cherry , 1993; Poole et al . , 1998; Noller , 2012 ) . The utility of triplets as substrates for RNA synthesis and self-synthesis described herein suggests that these early ideas deserve to be reconsidered . In the context of initial uncorrelated evolution of the small and large ribosomal subunits ( Petrov et al . , 2015 ) , it is tempting to speculate that an early reliance upon triplets in RNA replication could have inadvertently supplied a decoding center for translation . In conclusion , the unexpected emergent properties of triplets – including cooperative binding and unfolding of structured RNA templates , enhanced incorporation of AU-rich substrates , and error attenuation ( resulting from triplet pool interaction networks ) – argue that short RNA oligonucleotides may represent predisposed substrates for RNA-catalyzed RNA replication . Some of these benefits might also extend to codon/anticodon dynamics in early translation , and to the non-enzymatic replication of RNA ( Szostak , 2012 ) , where downstream trinucleotides have recently been shown to enhance incorporation of preceding activated mononucleotides both through stacking and positioning effects ( Vogel et al . , 2005; Zhang et al . , 2018 ) and the formation of a highly reactive intermediate ( Prywes et al . , 2016; O'Flaherty et al . , 2018 ) . Taken together , the interaction of triplet substrate pools with RNA templates promotes uncoupling of an RNA’s sequence ( i . e . information content , and associated folding tendencies ) from its replicability , thereby enhancing RNA’s capacity to serve as an informational polymer .
Standard ribozyme activity assays ( modified where specified ) comprise 5 pmol of each ribozyme annealed in 2 . 5 μl water ( 80˚C 2 min , 17˚C 10 min ) , with 2 μl of 1 M MgCl2 and 0 . 5 μl of 1 M tris•HCl pH 8 . 3 ( at 25˚C , pH raised to 9 . 2 at −7˚C ) then added on ice , and left for >5 min to ensure folding . This was added to 5 pmol each of primer and template and 50 pmol of each triplet pre-annealed in 5 μl water , then frozen on dry ice ( 10 min ) and incubated at −7˚C in a R4 series TC120 refrigerated cooling bath ( Grant ( Shepreth , UK ) ) to allow eutectic phase formation and reaction . Final pre-freezing concentrations of components are displayed throughout ( in this example , yielding 0 . 5 μM ribozyme/primer/template , 5 μM each triplet , 200 mM MgCl2 , 50 mM tris•HCl pH 8 . 3 ) . Supercooled reactions ( Figure 1c ) remained liquid by omitting the dry-ice freezing step , maintaining these concentrations . Ice crystal formation upon eutectic phase equilibration , however , concentrates all solutes ~4–5 fold ( Attwater et al . , 2010 ) to their final operational levels and cooling elevates tris-buffered pH to ~9 . 2 . Some substrate mixes ( e . g . pppNNN ) led to a higher final reaction volume , but eutectic phase equilibration restored standard operational concentrations , also applicable to the four-fold-diluted extensions with the fragmented ribozyme ( Figure 6c ) . These used 2 pmol each ribozyme/fragment annealed in 3 . 25 μl 62 mM MgCl2 , 15 mM tris•HCl pH 8 . 3 ( 37˚C 5 min , ramped to 4˚C at 0 . 1˚C/s , 4˚C 10 min ) , with pre-annealed primer/template/substrates ( 0 . 5/0 . 5/5 pmol ) added in 0 . 75 μl water . These reactions , and preparative syntheses ( Figure 6—figure supplement 1 , Figure 7c ) , were supercooled at −7˚C followed by ice crystal addition for quick freezing and optimal activity . Figure 1c extensions were set up by adding buffer , then RNAs ( preannealed together , 0 . 1 μM final concentrations ) to triplets . RNAs for ε+ syntheses were chilled on ice instead of annealing , with ribozyme/MgCl2/tris•HCl pH 8 . 3 mixed with the other RNAs at −7˚C . Oligonucleotide substrates were added equimolar to template binding sites in the primer/template/substrate anneal . NTPs , on the other hand , were added with the MgCl2/tris•HCl pH 8 . 3 to the ribozyme polymerase . At the end of standard incubations , reactions were thawed and 2 μl aliquots added to stop buffer ( 1 μl 0 . 44 M EDTA ( pH 7 . 4 ) , with urea to a 6 M final concentration and a 10–20 fold molar excess over template of complementary competing oligonucleotide ( see Supplementary file 3 ) to prevent long product/template reannealing ) . Samples were denatured ( 94˚C 5 min ) and RNAs separated by 8 M urea 1 × TBE denaturing PAGE . To avoid using potentially confounding competing oligonucleotide when purifying extension products , reactions with a biotinylated primer or template ( stopped as above ) could be purified by bead capture using MyOne C1 ( Invitrogen ) streptavidin-coated paramagnetic microbeads ( using 5 μg pre-washed beads per pmol biotinylated RNA ) in 0 . 5 × − 0 . 8 × bead buffer ( BB: 200 mM NaCl , 10 mM tris•HCl pH 7 . 4 ( at 25˚C ) , 1 mM EDTA , 0 . 1% Tween-20 ) . After washing twice in BB to remove unbound components , beads were incubated ( 1 min ) in 25 mM NaOH , 1 mM EDTA , 0 . 05% Tween-20 to denature the duplexes ( Horning and Joyce , 2016 ) . To recover biotinylated extension products ( e . g . Figure 5b left panels , Figure 6—figure supplement 1 αβ+/γδε+ , Figure 7c left panel ) the supernatant was discarded , and beads were washed first in BB with 200 mM tris•HCl pH 7 . 4 , then in BB , then heated ( 94˚C 4 min ) in 95% formamide , 10 mM EDTA to release primers for urea-PAGE . To recover extension products bound to biotinylated templates ( e . g . Figure 6—figure supplement 1 β+ , δ+ , ε+ , Figure 7c right panel ) the supernatant was removed , neutralized with 500 mM tris•HCl pH 7 . 4 , spin-concentrated using Ultracel 3K filters ( MerckMillipore , UK ) , recovered and denatured in 6M urea/10 mM EDTA before urea-PAGE . β+ synthesised in Figure 6—figure supplement 1 was not spin-concentrated , leading to a lower recovery yield; δε+ synthesis was denatured directly from the ligation reaction in 60% formamide with excess EDTA . For gel mobility shift assays ( Figure 3b ) , ribozymes were mixed at 0 . 5 μM , pre-annealed and buffer added on ice as for extension reactions , then mixed with 5 × loading buffer ( 50% glycerol , 250 mM tris•HCl pH 8 . 3 , 125 mM MgCl2 ) for separation by native PAGE ( 0 . 5 × TB , 8% 59:1 acrylamide:bisacrylamide , 25 mM MgCl2 , run in a Hoefer SE600 Chroma ( ThermoFisher , Waltham , USA ) ( upper chamber: 0 . 5 × TB 50 mM NaOAc , lower chamber: 0 . 5 × TB 25 mM Mg ( OAc ) 2 ) kept at 4˚C in a circulator bath for 6–8 hr at 10 W ) , then SYBR Gold stained as below . Fluorescent primer extension products were detected using the appropriate laser wavelength on a Typhoon Trio scanner ( GE Healthcare ( GE ) ( Chicago , USA ) ) ; gel densitometry allowed quantification of RNA synthesis efficiency . Gel contrasts in figures were linearly adjusted to optimize display of bands of differing intensities . The gel in Figure 5b ( middle panel ) was washed thrice ( 5 min ) in water , incubated with 10 μM DFHBI-1T ligand in buffer for 20 min to fold full-length broccoli aptamer ( as in [Filonov et al . , 2015] ) and scanned . The ligand was then eluted in three 1 × TBE washes ( leaving negligible background fluoresence ) , and stained in 1 × TBE with SYBR Gold ( 1:10000 ) , washed again , and re-scanned to detect all RNA products ( left panel ) ; scans were aligned via an adjacent Cy5-labelled primer extension lane ( not shown ) . Full-length product yields in the Figure 6—figure supplement 1 plus-strand syntheses were calculated by running samples of bead-eluted products ( or raw reaction for δε+ ) alongside known amounts of the positive controls indicated , followed by SYBR-Gold staining . To purify , bead-eluted products were run similarly , and excised using UV shadowing . Products were then eluted from the gel fragments in 10 mM tris•HCl pH 7 . 4 , and Spin-X column filtrate ( Costar ( Sigma-Aldrich , UK ) ) precipitated in 75% ethanol with 1 μl 1% glycogen carrier ( omitted for β+ ) . Recovered full-length product yields were calculated similarly to reaction yields for αβ+/δε+/γδε+ , or using A260s for β+ , δ+ , ε+ . Purified ribozyme- and TGK-synthesized αβ+/γδε+ fragments were sequenced by first ligating a 3’ adaptor ( 10 U/μl T4 RNA Ligase 2 truncated KQ in 1 × RNA ligase buffer ( New England Biolabs ( NEB ) , ( Ipswich , USA ) ) with 15% PEG-8000 and 2 μM AdeHDVLig at 10˚C overnight ) . These reactions were bound to MyOne C1 microbeads ( ThermoFisher ( Invitrogen ) ) , washed with BB to remove unligated adaptor , and reverse transcribed ( 50˚C 30 min ) with 1 μM HDVrec primer using Superscript III ( Invitrogen ) . Beads were washed again then PCR amplified ( five cycles with a 40˚C annealing step , then 20 cycles with a 50˚C annealing step ) using GoTaq HotStart master mix ( Promega ( Madison , USA ) ) and 0 . 8 μM each of primers P3HDV , and P5Xα8 or P5Xγ7 , for high-throughput sequencing ( Illumina ( San Diego , USA ) MiSeq or HiSeq ) after PCR product agarose gel purification . β+ syntheses’ cDNAs were amplified with P3HDV and P5Xβ6 . To estimate RNA synthesis fidelity , ribozymes extended primers using pppNNN on templates encoding CCC-XXX-CCC , where XXX were 12 different triplet sequences evenly exploring base composition and distribution ( see Supplementary file 3; for XXX = ACC , template encodes CCC-ACC-UCC to avoid a terminal run of Gs ) . Each primer/template pair ( 0 . 45/0 . 525 pmol per reaction ) was annealed in 4 mM MgCl2 , 1 mM tris•HCl pH 7 . 4 ( 80˚C 2 min , ramped to 4˚C at 0 . 1˚C/s , then kept on ice ) . The 12 pairs were combined in 0 . 27M MgCl2/67 mM tris•HCl pH 8 . 3 on ice to discourage primer-template assortment ( of which sequencing later revealed negligible levels ) . 36 pmol of each triplet in pppNNN ( equivalent to 5 μM final concentration after considering eutectic phase equilibration effects upon this more dilute reaction ) were added to a reaction vessel in 10 . 8 μl water , to which 5 . 4 μl of the primer/template/buffer mix was added followed by 7 . 2 pmol of ribozyme pre-annealed ( 80˚C 2 min 17˚C 10 min , ice >5 min ) in 1 . 8 μl water ( f . c . equivalent 1 μM , in excess over the 0 . 875 μM template to which some ribozymes could tether to enhance extension ) . Reactions were frozen and incubated ( 7 days at −7˚C ) as described above . Reactions were stopped with 3 . 6 μl 0 . 44 M EDTA and 10 . 5 pmol of each template's competing oligonucleotide ( migrating above product , with marker mutations to ensure exclusion ) , denatured with 6 M urea , and urea-PAGE separated . After alignment with a fluorescence scan of the gel , a region of the sample lane corresponding to primers extended by +4 to +14 nt was excised ( encompassing 2–4 triplet additions ) , and extension products were eluted , precipitated in 77% ethanol with 1 μl 1% glycogen carrier , washed in 85% ethanol and resuspended in water . These extension products were 3’ adaptor ligated as for fragment sequencing . Products were reverse transcribed ( 0 . 2 × adaptor ligation reaction , 1 μM HDVrec primer in Superscript III reaction , 50˚C 30 min ) and then PCR-amplified ( 1/30th reverse transcription mix , 0 . 8 μM each of primers P3HDV and P5GGGX ) for sequencing as above ( yielding 2 × 105 – 4 × 106 sequences per ribozyme assay ) . After processing and 3’ adaptor trimming , sequences corresponding to primer extended by CCC +1–3 additional triplets were collated for analysis . Variations in upstream primer sequences ( see Supplementary file 3 ) allowed the partner template to be identified for each sequenced product; the triplet incorporated after the first CCC was counted . Separately , 10 μl extensions by t5+1 of each primer/template alone with its encoded triplet and pppCCC ( and pppUCC for the ACC pair ) were combined for purification and sequencing as above , to allow isolation of the ribozyme-mediated errors resulting from inclusion of the other 62 ( 61 for ACC ) triplets in the reaction ( versus errors from sequencing , recombination etc . ) . The counts of cognate triplet ( C ) and each error triplet ( E ) in the positive control ( p ) reduced error counts in the experimental samples ( x ) to yield ribozyme-mediated error counts ( Er ) thusly: Er = Ex - Ep* ( Cx/Cp ) ( not reducing Ex below 0 , and reallocating all reductions to Cr; pppCCC counts ( and pppUCC for the ACC template ) remained uncorrected ) . For each template , counts were then collated at the first/second/third positions to yield base-specific mutation rates for each position ( Figure 8—figure supplement 2 , Figure 8—source data 1 ) . Across the 12 triplets , A , C , G , and U were encoded at each position three times; linear averages were calculated to map the position’s error profile ( Figure 8b ) and geometric means of the four nucleobases yielded the position’s overall fidelity ( Figure 8a , Figure 8—figure supplement 1 ) . Triplets ( and some other short oligonucleotides ) were prepared from NTPs by T7 RNA polymerase run-off transcription of a 5’ single-stranded DNA overhang downstream of a DNA duplex T7 promoter sequence . In most cases , the 5’ overhang encoded ( was the reverse complement of ) the desired oligonucleotide . These oligonucleotides were short enough to synthesise during the abortive initiation stage of transcription , attenuating sequence constraints on the first bases of the transcript . However , T7 RNA polymerase exhibited tendencies to skip the first ( or even second ) base ( most severe for U > C > A > G before second position purines: encoding CGU yielded some pppGU , encoding UAC yielded just pppAC ) or use oligonucleotides generated during transcription to re-initiate ( e . g . encoding GAG yielded pppGAGAG , encoding AAA yielded pppA6-9 , encoding UCC yielded pppCCC , encoding CGC yielded some pppGCGC; this tendency was most severe when the oligonucleotide could be accommodated opposite the final template bases of the promoter ) . These tendencies could be subverted by encoding additional first bases ( usually without providing the corresponding NTP ) . This initiated the oligonucleotide at the second position where skipping tendencies were lower ( e . g . encoding CUAG without CTP yielded pppUAG , encoding UUAC yielded some pppUAC ) , and reduced recruitment as initiators of products with bases not complementary to the introduced first position template base ( e . g . encoding CGAG without CTP yielded pppGAG , encoding CAA without CTP yielded pppAA and pppAAA , encoding AUCC without ATP yielded pppUCC , encoding UCGC without UTP yielded pppCGC ) . Each 30 μl transcription reaction contained 72 nmol of each desired product base as an NTP ( Roche ) ( e . g . for pppUCC , 72 nmol UTP , 144 nmol CTP ) in 1 × MegaShortScript kit buffer with 1 . 5 μl MegaShortScript T7 enzyme ( ThermoFisher ) . Also present were 15 pmol of each DNA oligonucleotide forming the transcription duplex target ( see Supplementary file 2 ) . The reactions were incubated overnight at 37˚C , stopped with 3 μl 0 . 44 M EDTA and 17 μl 10 M urea , and separated by electrophoresis ( 35 W , 4 . 5 hr ) on a 35 × 18 × 0 . 15 cm 30% 19:1 acrylamide:bis-acrylamide 3 M urea tris-borate gel . Products were identified through their relative migrations ( reflecting overall composition , fastest to slowest: C > U≈A > G ) by UV shadowing . Triplet bands were excised and eluted overnight in 10 mM tris•HCl pH 7 . 4 , and filtrate ( Spin-X ) precipitated with 0 . 3 M sodium acetate pH 5 . 5 in 85% ethanol . Pellets were washed in 85% ethanol , resuspended in water , and UV absorbances measured with a Nanodrop ND-1000 spectrophotometer ( ThermoFisher ) . Oligocalc ( Kibbe , 2007 ) was used to calculate sequence-specific concentrations and yields . pppNNN was generated by combination of equal amounts of each of the 64 triplet stocks in a lo-bind microcentrifuge tube ( Eppendorf ( Hamburg , Germany ) ) . 3’-deoxy triphosphorylated ‘terminator’ triplets were transcribed as above but using a 3’ deoxynucleoside 5’ triphosphate ( Trilink biotechnologies ) for the last position , migrating faster during PAGE than the equivalent all-RNA triplet . Triplets with 2-thiouridine residues were transcribed as for their corresponding U , replacing UTP with U2STP ( Jena Bioscience ( Jena , Germany ) ) ; incorporation and migration were similar between the two , and their concentrations were calculated from A260nm by comparison to the A260nm of mixtures of the component ribonucleotides with UTP vs . U2STP . Triplets with 2’-fluoro , 2’-deoxy positions could also be transcribed , with lower efficiency , by substituting the corresponding triphosphate ( Trilink Biotechnologies ) . The biotinylated pppGAU–Bio triplet used in γ segment synthesis ( Figure 7c ) was transcribed as for pppGAU , replacing UTP with biotin-16-aminoallyluridine-5’-triphosphate ( Trilink Biotechnologies ( San Diego , USA ) ) , quantified via by comparison to the A290nm of mixtures of the component ribonucleotides . Longer triphosphorylated oligonucleotides used in ribozyme self-synthesis were generated similarly , but using ~200 ng of fully double stranded DNA as a template . Candidate product bands were purified and the desired oligonucleotide identified by ribozyme-catalysed in-frame incorporation and , for some , fragment sequencing . Transcriptions were performed on ~15 ng/μl dsDNA using MegaShortScript enzyme and buffer ( ThermoFisher ) with 7 . 8 mM of each NTP , or , to yield a 5’ monophosphate on the product to avoid aberrant ligation , 10 mM GMP ( guanosine monophosphate ) and 2 mM of each NTP ( ‘GMP transcription’ ) . dsDNA templates for some of these ( in Supplementary file 3 ) were generated ( ‘fill-in’ ) using three cycles of mutual extension ( GoTaq HotStart , Promega ) between the associated DNA oligonucleotide and 5T7 ( or , where indicated , HDVrt for defined 3’ terminus formation [Schürer et al . , 2002] ) followed by column purification ( QiaQuick , Qiagen ) . Some 5’ biotinylated RNAs were synthesized using the TGK polymerase ( Cozens et al . , 2012 ) ( 56 μg/ml , in 1 × Thermopol buffer ( NEB ) supplemented with 3 mM MgCl2 ) to extend 5’ biotinylated RNA primers ( 0 . 75 μM ) on DNA templates ( 1 μM ) using 2 . 5 mM of each NTP ( 94˚C 30 s , 45˚C 2 min , 65˚C 30 min , 45˚C 2 min , 65˚C 30 min , then all repeated ) . Biotinylated products were bead-purified as above . 3’ biotinylation of RNAs was achieved in two stages: 3’ azidylation ( at 2 μM with 25 U/μl yeast poly-A polymerase ( ThermoFisher ) and 0 . 5 mM 2’-azido-2’-deoxycytidine triphosphate ( Trilink Biotechnologies ) for 1 hr at 37˚C ) with subsequent acidic phenol/chloroform extraction and 75% ethanol precipitation , then copper-catalysed biotin- ( PEG ) 4-alkyne ( ThermoFisher ) cycloaddition ( Winz et al . , 2012 ) with subsequent 75% ethanol precipitation followed by resuspension and buffer exchange in Ultracel 3K filters ( Amicon ) to remove residual biotin-alkyne . Round one libraries were synthesised by mutual extension of 4 nmol of oligonucleotides 1baN30 and 1GMPfo or 1GTPfo at 1 μM each in 1 × isothermal amplification buffer ( NEB ) with 250 μM each dNTP , annealed ( 80˚C 3 min , 65˚C 5 min ) before addition of 0 . 4 U/μl Bst 2 . 0 ( NEB ) and 30 min 65˚C incubation . After purification , 375 μg of each DNA ( ~1 . 5 × 1015 molecules ) were transcribed in 5 ml transcription reactions ( 36 mM tris•HCl pH 7 . 9 ( at 25˚C ) , 1 . 8 mM spermidine , 9 mM DTT , 10 . 8 mM MgCl2 , 2 mM each NTP , 1% 10 × MegaShortScript buffer , 2% 1:9 MegaShortScript:NEB T7 RNA polymerase , 37˚C overnight ) . These were treated with DNase , acid phenol/chloroform extracted and 73% ethanol precipitated prior to urea-PAGE purification , elution , filtering ( Spin-X ) and re-precipitation , yielding the 1GTP Zcore selection construct ( Supplementary file 1 ) . 10 mM GMP was present in transcriptions of the 1GMP construct , and for future transcriptions of the GMP construct selection branch and rounds 8–18; round 19–21 and reselection libraries were transcribed without GMP . Most subsequent selection rounds were transcribed in 1/10th scale transcriptions with 15 μg of DNA ( ~6 × 1013 molecules ) derived from amplification of recovered PCR products ( see later ) . For round 8 , 700 pmol DNA was formed , with Tri3CUUQ amplifying round seven merged output ( 50 pmol ) , round seven merged output recombined by StEP ( Zhao and Zha , 2006 ) ( 200 pmol ) , and 0core ribozyme with the starting 3’ N30 library domain added ( 450 pmol , but extinct at the end of selection ) . DNA encoding type 5s amplified with AACAt5s was used to generate reselection libraries by PCR amplification using primers TriGAA7GAAM and T5ba13N/T5ba20N/T5ba28N; 5 pmol of the three dsDNA products were transcribed to generate reselection constructs . An outline of the selection strategy is shown in Figure 1—figure supplement 2 , with detailed lists of selection oligonucleotides and extension parameters in Figure 1—source data 1 , Figure 2—source data 1 , Figure 4—source data 1 and Supplementary file 3 . First , selection construct was annealed with equimolar dual-5’ biotinylated primer in water ( 80˚C 2–4 min , 17˚C 10 min ) , then chilled extension buffer and triplets were added before freezing and −7˚C incubation . At the end of incubation the reaction was thawed on ice . To link the primer 3’ hydroxyl to the 5’ monophosphate of GMP constructs , selection constructs were buffer-exchanged directly after thawing using a PD-10 column ( GE ) in a cold room , into 3 ml ligation mix ( optimised to prevent ligation over gaps ) ( 2 mM MgCl2 , 50 mM tris•HCl pH 7 . 4 , 0 . 1 mM ATP , 1 mM DTT , 2 μM HOGCG ( Rounds 1–7 ) or 2 μM HOCUG ( Rounds 8–18 ) with 30 U/ml T4 RNA Ligase 2 ( NEB ) ) . After incubation at 4˚C for 1 hr , these were stopped with 2 . 2 mM EDTA and acid phenol/chloroform treated . Constructs were then precipitated with glycogen carrier and 0 . 3 M sodium acetate in isopropanol ( 55% ) before resuspension and denaturation ( 94˚C 4 min , in 6M urea 10 mM EDTA with a 3 × excess of competing oligonucleotide against the primer ) . For the reselection rounds , constructs were then treated with polynucleotide kinase ( NEB ) before denaturation to resolve the HDV-derived 2’ , 3’-cyclic phosphates and allow later adaptor ligation . Constructs were urea-PAGE separated alongside FITC-labelled RNA markers equivalent to successfully ligated constructs . The marker-adjacent gel region in the construct lane was excised , excluding the bulk unreacted construct ( judged by UV shadowing ) . Biotinylated ( primer-linked ) constructs were eluted overnight into BB with 100 μg MyOne C1 beads . After 30 μm filtering ( Partec Celltrics ( Wolflabs ( York , UK ) ) ) of the supernatant to remove gel fragments , the beads were washed in BB then denaturing buffer ( 8 M urea , 50 mM tris•HCl pH 7 . 4 , 1 mM EDTA , 0 . 1% Tween-20 , 10 μM competing oligonucleotide , 60˚C 2 min ) to confirm covalent linkage of construct to primer , before further BB washing and transfer ( to a fresh microcentrifuge tube to minimize downstream contamination ) . At this stage in the reselection , 3’ adaptors were then ligated to bead-bound constructs as above for 2 hr ( with buffer/enzyme added after bead resuspension in other reaction components including 0 . 04% Tween-20 ) , and beads BB washed and transferred again . Bead-bound constructs were now reverse-transcribed using 1 μM RTri ( or HDVRec for the reselection ) by resuspension in a Superscript III reaction with added 0 . 02% Tween-20 ( 50˚C 30 min ) . Beads were BB washed and the RNA-bound cDNA 3’ end blocked by incubation with terminal deoxynucleotidyl transferase ( ThermoFisher ) and 0 . 2 mM dideoxy-ATP ( TriLink ) with 0 . 02% added Tween-20 ( 37˚C 30 min ) , and beads were BB washed and transferred again . cDNAs were eluted ( 10 μl 0 . 1 M NaOH 0 . 1% Tween-20 20 min ) , neutralized and plus strands regenerated with 0 . 2 μM rescue oligonucleotide in an IsoAmp II universal tHDA kit ( NEB ) reaction ( 65˚C 60 min ) to read through the structured product region . Whilst at this temperature , reactions were stopped with 5 mM EDTA and one volume of BB with 50–100 μg beads to bind the nascent biotinylated plus strands at room temperature . These beads were then BB washed , NaOH washed again to discard cDNAs ( and recover only correctly-primed plus strands ) , and washed and transferred again . Each 50 μg of beads were then subjected to plus strand recovery PCR in a 100 μl GoTaq HotStart reaction with 0 . 5 μM each RTri ( or HDVrec for reselection ) and RecInt ( rounds 1–5 ) /RecIntQ ( rounds 6–9 ) /RecIntL ( rounds 10–14 , 19–21 and reselection ) /RecIntQL ( rounds 15–18 ) . The product was agarose size-purified , A260nm quantified and added to construct synthesis PCR in 3 × molar amount of the anticipated recovered RNA ( judged by test extensions ) that yielded it . This final PCR for construct transcription in the subsequent selection round used 1 μM of the indicated construct synthesis primer , plus 1 μM RTri ( or HDVrt for the reselection ) , in GoTaq HotStart reactions or ( where indicated in source data ) the GeneMorph II kit for mutagenesis ( Agilent ( Santa Clara , USA ) ) . Conditions for selections are included as source data 1 for Figures 1 , 2 and 4 . Numerical data for Figures 1c and 3b are included in Figure 1—source data 2 and Figure 3—source data 1 respectively . Numerical data and calculations for Figure 8 and its Figure 8—figure supplement 1 and 2 are supplied as Figure 8—source data 1 . Numerical data and calculations for Figure 8—figure supplement 3 are supplied as Figure 8—source data 2 . A more extensive selection of substrate concentration-dependent error rates is supplied in Figure 9—source data 1 . Sequences of ribozymes , triplet synthesis templates , and oligonucleotides used in this study are supplied in Supplementary files 1 , 2 and 3 respectively . | Life as we know it relies on three types of molecules: DNA , which stores genetic information; proteins that carry out the chemical reactions necessary for life; and RNA , which relays information between the two . However , some scientists think that before life adopted DNA and proteins , it relied primarily on RNA . Like DNA , strands of RNA contain genetic data . Yet , some RNA strands can also fold to form ribozymes , 3D structures that could have guided life’s chemical processes the way proteins do now . For early life to be built on RNA , though , this molecule must have had the ability to make copies of itself . This duplication is a chemical reaction that could be driven by an ‘RNA replicase’ ribozyme . RNA strands are made of four different letters attached to each other in a specific order . When RNA is copied , one strand acts as a template , and a replicase ribozyme would accurately guide which letters are added to the strand under construction . However , no replicase ribozyme has been observed in existing life forms; this has led scientists to try to artificially create RNA replicase ribozymes that could copy themselves . Until now , the best approaches have assumed that a replicase would add building blocks formed of a single letter one by one to grow a new strand . Yet , although ribozymes can be made to copy straight RNA templates this way , folded RNA templates – including the replicase ribozyme itself – impede copying . In this apparent paradox , a ribozyme needs to fold to copy RNA , but when folded , is itself copied poorly . Here , Attwater et al . wondered if choosing different building blocks might overcome this contradiction . Biochemical techniques were used to engineer a ribozyme that copies RNA strands by adding letters not one-by-one , but three-by-three . Using three-letter ‘triplet’ building blocks , this new ribozyme can copy various folded RNA strands , including the active part of its own sequence . This is because triplet building blocks have different , and sometimes unexpected , chemical properties compared to single-letter blocks . For example , these triplets work together to bind tightly to RNA strands and unravel structures that block RNA copying . All life on Earth today uses a triplet RNA code to make proteins from DNA , and these experiments showed how RNA triplets might have helped RNA sustain early life forms . Further work is now needed to improve the ribozyme designed by Attwater et al . for efficient self-copying . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | Ribozyme-catalysed RNA synthesis using triplet building blocks |
FcγRIIB binding to its ligand suppresses immune cell activation . A single-nucleotide polymorphic ( SNP ) change , I232T , in the transmembrane ( TM ) domain of FcγRIIB loses its suppressive function , which is clinically associated with systemic lupus erythematosus ( SLE ) . Previously , we reported that I232T tilts FcγRIIB’s TM domain . In this study , combining with molecular dynamics simulations and single-cell FRET assay , we further reveal that such tilting by I232T unexpectedly bends the FcγRIIB’s ectodomain toward plasma membrane to allosterically impede FcγRIIB’s ligand association . I232T substitution reduces in situ two-dimensional binding affinities and association rates of FcγRIIB to interact with its ligands , IgG1 , IgG2 and IgG3 by three to four folds . This allosteric regulation by an SNP provides an intrinsic molecular mechanism for the functional loss of FcγRIIB-I232T in SLE patients .
Disorders of immune components could lead to autoimmune diseases . Malfunction of an immune receptor , FcγRIIB , is generally destructive for immune system ( Niederer et al . , 2010; Pincetic et al . , 2014; Smith and Clatworthy , 2010 ) . FcγRIIB is widely expressed on most types of immune cells including B cells , plasma cells , monocytes , dendritic cells , macrophages , neutrophils , basophils , mast cells and even memory CD8+ T cells ( Starbeck-Miller et al . , 2014 ) . Among all the immune-receptors for Fc portion of IgG molecules ( FcγRs ) , FcγRIIB is unique due to its suppressive function against immune cell activation . It has been shown that single-nucleotide polymorphisms ( SNPs ) of the human FcγRIIB gene extensively influence the susceptibility toward autoimmune disorders ( Kyogoku et al . , 2002; Niederer et al . , 2010; Smith and Clatworthy , 2010 ) . A T-to-C variant in exon 5 ( rs1050501 ) of FcγRIIB causes the I232T substitution ( FcγRIIB-I232T ) within the transmembrane ( TM ) domain and is positively associated with systemic lupus erythematosus ( SLE ) in the homozygous FcγRIIB-I232T populations as reported in epidemiological studies ( Chu et al . , 2004; Clatworthy et al . , 2007; Kyogoku et al . , 2002; Niederer et al . , 2010; Siriboonrit et al . , 2003; Willcocks et al . , 2010 ) . Although a statistical linkage of the homozygous FcγRIIB-I232T polymorphism with SLE is established , comprehensive assessments and mechanistic investigations towards the inter-linkage of FcγRIIB-I232T regarding to the age of syndrome onset , progress , and clinical manifestation of SLE are still lacking . We first address this question in this report . Previous biochemical studies revealed that monocytes harboring FcγRIIB-232T ( 232T ) are hyper-activated with augmented FcγRI-triggered phospholipase D activation and calcium signaling ( Floto et al . , 2005 ) . B lymphocytes expressing 232T are of hyperactivity with abnormal elevation of PLCγ2 activation , proliferation and calcium mobilization ( Kono et al . , 2005 ) . 232T-expressing B cells lose the ability to inhibit the oligomerization of B cell receptors ( BCRs ) upon co-ligation between BCR and FcγRIIB ( Liu et al . , 2010 ) . Recent live-cell imaging studies showed that B cells expressing 232T fail to inhibit the spatial-temporal co-localization of BCR and CD19 within the B cell immunological synapses ( Xu et al . , 2014 ) . Human primary B cells from SLE patients with homozygous FcγRIIB-I232T reveal hyper-activation of PI3K ( Xu et al . , 2014 ) . Thus , it is very likely that FcγRIIB-I232T is the first example that a naturally occurring diseases-associated SNP within the TM domain of a single-pass transmembrane receptor can allosterically suppress the receptor’s ligand recognition and signaling functions . FcγRIIB’s suppressive function is triggered by its ligand engagement , while this function is disrupted by a single amino acid change in the 232th residue from Ile to Thr in FcγRIIB’s TM domain . Two early biochemical studies proposed a model of reduced affinity between 232T and lipid rafts to explain the functional relevance and effect of this natural mutation ( Floto et al . , 2005; Kono et al . , 2005 ) . A recent study also proposed a different model that I232T mutation enforces the inclination of the TM domain inside the membrane , thereby reducing the lateral mobility and inhibitory functions of FcγRIIB ( Xu et al . , 2016 ) . However , both models assumed that 232T and FcγRIIB-WT ( 232I ) have an equal capability to perceive and bind to their ligands , the IgG’s Fc portion within the antibody-antigen immune complexes . This important but experimentally un-proved pre-requisition in both models is based on the argument that 232T and 232I are identical in the amino acid sequence of their extracellular domains and thus the potential structure of ligand binding site for recognizing the ligands , that is , the IgG’s Fc portions ( Dal Porto et al . , 2004; D'Ambrosio et al . , 1996 ) . However , to date , there is no direct experimental evidence to validate this pre-requisite assumption . We also address this question in this report .
In this report , we firstly performed systemic examination over the association of FcγRIIB-I232T with clinical manifestations of SLE . We enrolled 711 unrelated Chinese SLE patients with complete clinical documents into this study ( Supplementary file 1 ) . 688 unrelated healthy Chinese volunteers with matched gender and age were also enrolled as controls ( Supplementary file 1 ) . We confirm the presence of a strong positive association of the homozygous FcγRIIB-I232T polymorphism with SLE ( χ2 = 27 . 224 , p=0 . 008 , odds ratio with 95% confidence interval ( CI ) = 1 . 927 ) ( Supplementary file 1 ) , consistent with the published epidemiological data ( Chu et al . , 2004; Clatworthy et al . , 2007; Kyogoku et al . , 2002; Niederer et al . , 2010; Siriboonrit et al . , 2003; Willcocks et al . , 2010 ) . Next , we comprehensively analyzed the clinical data for all 711 SLE patients , including 50 FcγRIIB-I232T homozygotes , 283 FcγRIIB-I232T heterozygotes and 378 FcγRIIB-WT carriers ( Table 1 and Supplementary file 2 ) . We find that the homozygous FcγRIIB-I232T polymorphism is significantly associated with early disease onset ( age at disease onset <37 , p=0 . 002 ) ( Table 1 ) . We also observe a significant association of the homozygous FcγRIIB-I232T polymorphism with more severe SLE clinical manifestations since the corresponding SLE patients present significant elevation in the amounts of anti-dsDNA antibodies ( p=0 . 004 ) , anti-nuclear antibodies ( p=0 . 021 ) and total Immunoglobulin ( Ig ) ( p=0 . 032 ) in comparison to patients carrying heterozygous FcγRIIB-I232T polymorphism or FcγRIIB-WT ( Table 1 ) . Moreover , homozygous FcγRIIB-I232T polymorphism is also significantly associated with the higher SLE disease activity index ( SLEDAI ) score ( p=0 . 014 for SLEDAI ≥12 vs . p=0 . 861 for SLEDAI <12 ) as well as more severe clinical manifestations including arthritis ( p=0 . 008 ) , anemia ( p=0 . 006 ) , leukopenia ( p=0 . 005 ) , complement decrease ( p=0 . 006 ) , hematuria ( p=0 . 004 ) and leucocyturia ( p=0 . 010 ) ( Table 1 ) . A suggestive association is also observed between homozygous FcγRIIB-I232T polymorphism and serositis ( p=0 . 063 ) ( Table 1 ) . These clinical association analyses demonstrate that SLE patients homozygous for FcγRIIB-I232T polymorphism are prone to develop more severe clinical manifestations than the patients carrying heterozygous FcγRIIB-I232T polymorphism or FcγRIIB-WT , reinforcing the importance to study the pathogenic mechanism of FcγRIIB-I232T polymorphism since this SNP occurs at a notable frequency in up to 40% ( heterozygous polymorphism ) humans ( Niederer et al . , 2010; Smith and Clatworthy , 2010 ) . Next , we examined whether I232T polymorphic substitution in the TM domain of FcγRIIB allosterically affects ligand recognition . We did this investigation as our previous observation of the inclination of the TM domain by I232T ( Xu et al . , 2016 ) led us to hypothesize that tilted TM domain of 232T may lead to ectodomain conformational changes to allosterically attenuate ligand binding . We first carried out large-scale molecular dynamics simulations ( MDS ) with modeled structures of almost full-length human FcγRIIB ( either 232I or 232T ) imbedded in the lipid bilayer ( Figure 1A and Figure 1—figure supplement 1A ) . The simulations confirm our previous results with the MDS of the TM domain of FcγRIIB only ( Xu et al . , 2016 ) , that is , I232T polymorphic substitution enforces the inclination of the TM domain ( Figure 1B , right ) . This inclination might result from the ability of H-bond formation between the side-chain Oγ atom of T232 and the backbone oxygen atom of a neighbor residue V228 ( Figure 1B , left ) . The difference of the TM domain orientation between 232I and 232T induces a distinct conformation on the ecto-membrane proximal region ( ecto-TM linker ) ( Figure 1—figure supplement 2 ) . The membrane buried non-helical region of the linker extends more in 232T than that in 232I . And the length between S218 and P221 peaks at 11 Å for 232T , about 3 Å longer than that for 232I ( Figure 1C ) . This length elongation further results in different conformation of residue P217 . The main chain dihedral angle of P217 in 232I displays two populations at 141°±23° and −50°±12° , respectively , but shifts to −40°±45° and −75°±12° in 232T ( Figure 1D and Figure 1—figure supplement 2 ) . These effects propagate and lead to a striking effect on tilting FcγRIIB’s extracellular domains toward the lipid membrane ( Figure 1E ) . Although the extracellular domains of 232I and 232T , especially their IgG binding sites , do not undergo obvious conformational change ( Figure 1—figure supplement 3 ) , their orientations toward the membrane differ significantly . The ectodomain of 232I maintains more straight-up conformation , whereas that of 232T bends down toward the lipid bilayer ( Figure 1E ) . Statistical analyses show that the ectodomain inclination angle of 232T distributes across 30 ~ 60° with a sharper single-peak at 40° ( Figure 1E ) . In contrast , the angle of 232I distributes much flatter with a favorable probability ranging from 50° to 70° ( Figure 1E ) . The distance of C1 domain to the membrane is shorter for 232T than 232I ( Figure 1E ) . To check whether these observations result from the thickness of the lipid membrane model , we carried out further simulations using lipids with shorter ( 14:0/16:1 ) or longer fatty acid tail ( 18:0/20:1 ) ( Figure 1—figure supplement 4A ) . The TM helix tilting and S218-P221 prolongation for 232T can be readily observed in these two systems ( Figure 1—figure supplement 4B–4C ) . These results suggest that I232T substitution may reduce the ligand recognition ability of FcγRIIB via two aspects . First , the tilting orientation of 232T may sterically prevent the accessibility of the IgG’s Fc portion , as significant clashes between docked Fc and the membrane are observed , although FcγRIIB’s Fc binding site is not completely buried into the membrane ( Figure 1—figure supplement 1B ) . Second , the ectodomain of 232T is less flexible ( Figure 1E ) such that the chance for FcγRIIB to associate with the ligand is greatly decreased . We next performed single-cell fluorescence resonance energy transfer ( FRET ) assay to experimentally validate whether I232T polymorphic change could allosterically bend the FcγRIIB ectodomain toward cell membrane . We fused an mTFP ( as FRET donor ) at the N-terminal of 232I or 232T ectodomain ( mTFP-232I or mTFP-232T ) and hypothesized that it should fall in the spatial proximity ( ~16 ~ 36 Å ) for FRET with plasma outer membrane labeled with octadecyl rhodamine B ( R18 , as FRET acceptor ) ( Figure 2A ) and that I232T polymorphism may exhibit an enhanced FRET efficiency . With de-quenching assay ( Chen et al . , 2015; Xu et al . , 2008 ) on A20II1 . 6 B cells expressing similar level of either mTFP-232I or mTFP-232T ( Figure 2B and C ) , we find that I232T polymorphic change indeed enhances the FRET efficiency about two folds , from ~20% in 232I to ~40% in 232T ( Figure 2C and D ) . This enhancement of FRET efficiency by I232T polymorphism indicates that 232T ectodomain prefers to a more recumbent orientation on the plasma membrane than 232I , consistent with above MDS observations . Ectodomain orientation change of a receptor can significantly affect its in situ binding affinity with its ligands ( Huang et al . , 2004 ) . We therefore predicted that titling FcγRIIB ectodomain toward plasma membrane by I232T polymorphic change may attenuate its ligand binding affinity , especially the ligand association rate . To test this hypothesis , we applied well-established single-cell biomechanical apparatus with adhesion frequency assay ( Chesla et al . , 1998; Huang et al . , 2010 ) to directly and quantitatively measure in situ two-dimensional ( 2D ) binding kinetics of either 232I or 232T binding with its ligands ( Figure 3A ) . The results show that the in situ 2D effective binding affinity of 232I with human IgG1 antibody ( anti-MERS virus S protein , or anti-S ) is about three times higher than that of 232T ( AcKa = 3 . 03 ± 0 . 15×10−7 and 0 . 80 ± 0 . 04 × 10−7 μm4 , respectively ) , whereas that with human IgG4 is hardly measured as FcγRIIB and IgG4 binding is known to be extremely weak and beyond the detection limit ( 10−8 μm4 ) of this assay ( Huang et al . , 2010 ) ( Figure 3B and F ) . Further analyses show that although the 2D off-rates of 232I and 232T from human IgG1 are similar ( 7 . 75 ± 1 . 42 and 7 . 62 ± 1 . 41 s−1 , respectively ) ( Figure 3B and H ) , the 2D effective on-rate of 232T with IgG1 is three times slower than that of 232I ( Figure 3B and G ) . These results are also confirmed by using another human IgG1 antibody ( anti-HIV1 gp120 IgG1 , or anti-gp120 ) . Both 2D effective affinity and on-rate of 232I with anti-gp120 human IgG1 are three times higher than those of 232T ( AcKa = 7 . 74 ± 0 . 24×10−7 and 2 . 43 ± 0 . 11 × 10−7 μm4 , respectively; Ackon=5 . 95±0 . 19 and 2 . 16 ± 0 . 10 × 10−7 μm4 s−1 , respectively ) , while the respective off-rates are similar ( 7 . 70 ± 0 . 83 and 8 . 90 ± 1 . 61 s−1 , respectively ) ( Figure 3C and F–H ) . I232T polymorphic change also causes the reduction of 2D affinity and on-rate of FcγRIIB’s binding with other IgG subtypes ( e . g . IgG2 and IgG3 ) ( Figure 3D–H ) . Furthermore , we aimed to rule out other factors that may potentially reduce ligand binding affinity of 232T . First , we fixed either 232I or 232T expressing B cells to exclude the effect from the reduced lateral diffusion by I232T ( Xu et al . , 2016 ) . We find that fixing B cells expressing FcγRIIB by paraformaldehyde ( PFA ) does not alter the ligand binding defects for I232T polymorphic change ( Figure 3—figure supplement 1 ) , suggesting the reduction of lateral diffusion by I232T hardly contributes to the reduction of the ligand binding affinity and on-rate . Moreover , Zhu and colleagues have also extensively discussed and experimentally proved that the lateral diffusion has negligible impact on 2D affinity of receptor-ligand binding on live cells ( Chesla et al . , 2000 ) . To further exclude potential technical artifacts in in situ single-cell adhesion frequency assay , we confirm previously known binding-enhancing FcγRIIIA-F158V polymorphism by using the 2D binding assay in this report ( Figure 3—figure supplement 2 ) . All these data strongly support that I232T polymorphic change in the TM domain of FcγRIIB allosterically tilts FcγRIIB ectodomains toward the plasma membrane , rendering steric hindrance of its ligand binding domain . As a result , 232T exhibits significantly reduced 2D affinity and association on-rate to IgG antibodies . To be noted , it is possible that similar allosteric regulation may be applied to a broad range of transmembrane receptors , for example , potentially explaining how membrane anchor pattern of CD16a influences ligand recognition ( Chesla et al . , 2000 ) . In summary , we confirm that homozygous FcγRIIB-I232T confers dramatically increased risk of developing more severe clinical manifestations in patients with SLE . The pathological relevant of I232T is caused by the inclination of the TM domain which leads to FcγRIIB ectodomain bending toward plasma membrane , significantly impairing FcγRIIB’s binding ability to IgG’s Fc portion through reducing in situ binding affinity and association rate . The hampered Fc recognition ability of FcγRIIB-I232T results in the deficiency on its inhibitory function and thus hyper-activated immune cells , potentially contributing to SLE .
The ethics committee of Peking University People's Hospital approved this study and informed consents were obtained from each patient and healthy volunteer . All the human-cell-associated experimental guidelines were approved by the Medical Ethics Committee of Peking University People’s Hospital ( approval no . 2014PHB116-01 ) and by the Medical Ethics Committee of Tsinghua University ( approval no . 20180029 ) . There were 711 patients fulfilling the 1997 revised classification criteria of the American College of Rheumatology that enrolled in this study . Healthy volunteers were recruited as controls . 4–8 ml peripheral blood was acquired from SLE patients and healthy volunteers . Genomic DNA was extracted from peripheral blood samples using the TIANamp Blood DNA Midi Kit ( Catalog#DP332-01 , TIANGEN Biotech , China ) following the manufacturer’s protocol . The TaqMan Genotyping Assays were applied for genotyping of SNP rs1050501 ( TaqMan probe C: 5’-VIC-CGCTACAGCA GTCCCAGT-NFQ-3’ , TaqMan Probe T: 5’-FAM- CGCTACAGCA ATCCCAGT-NFQ-3’ ) ( Catalog#4351376 , Life Technology ) . Amplification and genotyping analyses were performed using ABI 7300 Real-Time PCR system . Relative quantification of probes levels was calculated ( 7500 Sequence Detection System Software Version 1 . 4 , ABI ) . Few samples were genotyped by using primers ( forward: 5’-AAGGGGAGCC CTTCCCTCTGTT-3’ , reverse: 5’-CATCACCCAC CATGTCTCAC-3’ ) binding to the flanking introns of exon five as reported ( Floto et al . , 2005; Kono et al . , 2005 ) . The DNA sequencing was done by BGI ( Beijing ) . The Pearson chi-square tests were performed for the comparison of differences between cases and controls at genotype model ( recessive model CC vs . TT+TC ) . The odds ratios ( OR ) , 95% confidence intervals ( CI ) and p value for recessive model analysis were calculated using logistic regression , adjusting for age and sex . In statistical analyses , p value of less than 0 . 05 was considered statistically significant . Structure models of human FcγRIIB system ( residues A46-I310 ) were built by fusing the crystal structure of the ectodomain ( PDB code 2FCB , residues A46-Q215 ) to the transmembrane ( TM ) helix ( residues M222-R248 ) model obtained in the previous study ( Xu et al . , 2016 ) . The stalk ( residues A216-P221 ) and cytoplasmic regions ( residues K249-I310 ) were randomly placed . Different initial models were built to minimize possible artifacts in structural modeling . An asymmetric lipid bilayer with the membrane lateral area of 100 × 100 Å2 was generated with Membrane Builder in CHARMM-GUI ( Wu et al . , 2014 ) . The outer leaflet of lipid membrane contained PC , SM , and cholesterol with molar ratio 1:1:1 , and the inner leaflet of lipid membrane contained PE , PC , PS , PIP2 and cholesterol with molar ratio 4:3:2:1:5 . Different length of the lipid models were used , including the widely used PO series ( 16:0/18:1 ) which is the most common lipid within mammalian cell membranes , and other two lipid models with shorter ( 14:0/16:1 ) and longer ( 18:0/20:1 ) fatty acids . FcγRIIB models were inserted into the lipid membranes with its TM perpendicular to the bilayer surface and the ectodomain stands straight , as shown in Figure 1A . The 232I system was subsequently solvated in 100 × 100 × 203 Å3 rectagular water boxes with TIP3P water model and was neutralized by 0 . 15 M NaCl . The 232T system was obtained from the same configuration using the Mutator plugin of VMD ( Humphrey et al . , 1996 ) . The final systems contained ~0 . 20 million atoms in total . Both systems were first pre-equilibrated with the following three steps: ( 1 ) 5000 steps energy minimization with the heavy atoms of proteins and the head group of the lipids fixed , followed by two ns equilibration simulation under one fs timestep with these atoms constrained by five kcal/mol/Å2 spring; ( 2 ) 5000 steps energy minimization with the heavy atoms of protein fixed , followed by 2 ns equilibration simulation under 1 fs timestep with these atoms constrained by 1 kcal/mol/Å2 spring; ( 3 ) 4 ns equilibration simulation under 2 fs timestep with the heavy atoms of protein ecto- and TM domains constrained ( i . e . the stalk and intracellular portion is free ) by 0 . 2 kcal/mol/Å2 spring . The resulted systems were subjected to productive simulations for 200 ns with 2 fs timestep without any constrains , and the snapshots of the last 80 ns ( sampled at 10 ps intervals ) were used for detailed analyses including the probability distributions of hydrogen bonds , tilting angles of the TM helix , inclination angles of ectodomain , the distance between Ig-like C2-type one domain and lipid bilayer . The tilting angle of TM helix was defined as the angle between TM helix and membrane plane , similar as that used in previous study ( Xu et al . , 2016 ) . The inclination angle of ectodomain was defined as the angle between the membrane plane and the vector linking N-terminal of TM helix ( M222-I224 ) and linker region of Ig-like C2-type 1 and 2 domain ( S130-W132 ) . The distance between Ig-like C2-type one domain and lipid bilayer was defined as the length between center of mass ( COM ) of this domain and the heavy atoms of phospholipid head in the normal direction of bilayer . All simulations were performed with NAMD2 software ( Phillips et al . , 2005 ) using CHARMM36m force field with the CMAP correction ( MacKerell et al . , 1998 ) . The simulations were performed in NPT ensemble ( one atm , 310K ) using a Langevin thermostat and Nosé-Hoover Langevin piston method ( Feller et al . , 1995 ) , respectively . 12 Å cutoff with 10 to 12 Å smooth switching was used for the calculation of the van der Waals interactions . The electrostatic interactions were computed using the particle mesh Eward method under periodic boundary conditions . The system preparations and illustrations were conducted using VMD . 232I and 232T pHAGE plasmids were previously constructed ( Xu et al . , 2016 ) . mTFP was fused to N-terminal of either 232I or 232T in a pHAGE backbone by ClonExpress MultiS One Step Cloning Kit ( Catalog#C113 , Vazyme , China ) . Stable mTFP-232I/mTFP-232T expressing A20II1 . 6 B cell lines were acquired by lentivirus infection ( three-vector system: mTFP-232I or mTFP-232I pHAGE , psPAX2 , and pMD2 . G ) . A20II1 . 6 B cell lines expressing similar level of either mTFP-232I or mTFP-232T were obtained by multiple rounds of cell sorting ( Beckman moflo Astrios EQ ) . Either 232I or 232T expressing A20II1 . 6 B cell line was previously established ( Xu et al . , 2016 ) . FRET measurements were performed as previously described ( Chen et al . , 2015; Xu et al . , 2008 ) . Briefly , all FRET measurements were carried out on Nikon TiE C2 confocal microscope with 100x oil lens , Argon 457 nm and HeNe 561 nm laser . 1 × 106 mTFP-232I or mTFP-232T expressing A20II1 . 6 B cells were stained with 300 nM octadecyl rhodamine B ( R18 ) ( Catalog#O246 , Invitrogen ) on ice for 3 min , excited by two lasers sequentially , and imaged before and after R18 photo-bleaching . mTFP intensity was processed through Image J . FRET efficiency was then calculated by ( DQ−Q ) /DQ , where DQ and Q are de-quenched and quenched mTFP intensity , respectively . FRET efficiencies of mTFP-232I and mTFP-232T cells ( ~20 cells , respectively ) were calculated and plotted through Prism 7 ( GraphPad ) . Error bars represent SEM . Streptavidin ( SA ) coated red blood cells ( SA-RBCs ) preparation have been described previously ( Huang et al . , 2010 ) . Briefly , RBCs freshly collected from finger prick were biotinylated with biotin-PEG-SGA ( Catalog#ZZ324P050 , JenKem Technology , China ) ( Liu et al . , 2014; Wu et al . , 2019 ) , followed by incubation with streptavidin ( Catalog#C600432 , Sangon Biotech , China ) to make SA-RBCs . Human antibodies ( anti-gp120 IgG1 and IgG4 , a kind gift from Dr . Y . Shi , The Institute of Microbiology of the Chinese Academy of Sciences; anti-S IgG1 , a kind gift from Dr . L . Zhang and X . Wang , Tsinghua University; IgG2 , a kind gift from Dr . H . Wang , Hisun , China; IgG3 , Catalog#bgal-mab3 , InvivoGen ) were biotinylated by EZ-Link Sulfo-NHS-LC-Biotin kits ( Catalog# 21435 , Thermo Fisher Scientific ) . Different amount of biotinylated IgG was linked into SA-RBCs through SA-biotin interaction at RT for 30 min , respectively to produce IgG-coated RBCs . These IgG-coated RBCs were then used to measure 2D binding kinetics of FcγRIIB/IgG with adhesion frequency assay . All above experimental processes were approved by the institutional ethical review board of Zhejiang University ( approval no . 2015–006 ) . The adhesion frequency assay was applied to measure FcγRIIB/IgG in situ 2D binding kinetics . The detail experimental setup and procedure were previously described ( Huang et al . , 2010 ) . In brief , two opposing micropipettes aspirating the RBC and FcγRIIB-expressing A20II1 . 6 B cell ( either 232I or 232T ) respectively were controlled by a customized computer program to operate contact-retraction cycles . Through 50 contact-retraction cycles , the binding frequency Pa was acquired with definite contact area Ac and a series of preset contact time tc ( 0 . 1 , 0 . 2 , 0 . 5 , 1 , 2 s or longer ) . 3 ~ 4 cell pairs were tested for each contact time . And these data were then non-linearly fitted to obtain effective 2D binding affinity AcKa and off-rate koff by probabilistic kinetic model ( Chesla et al . , 1998 ) :Pa=1−exp{−mrmlAcKa ( 1−exp ( koff ) ) } , where mr and ml are receptor and ligand densities on cells , respectively . Effective 2D on-rate Ackon was then calculated as following:Ackon=AcKa×koff . In order to accurately calculate 2D binding affinity and on-rate , these two molecular densities ( mFcγRIIB and mIgG ) were determined by standard calibration beads ( Quantum Alexa Fluor647 MESF kit , Catalog#647 , Bangs Laboratories ) on flow cytometry ( Beckman CytoFLEX S ) , respectively . Binding kinetics were calculated and plotted through Prism 7 ( GraphPad ) . Error bars represent SEM . | Left unchecked the immune system can cause devastating damage to healthy tissue . To prevent this from happening , immune cells have built-in off switches that dampen their activation . One such switch is a protein called FcγRIIB that sits on the outer surface of immune cells and binds to proteins known as antibodies , which are produced as part of the immune response . Its role is to act as a brake on the immune system , and stop it from getting out of control . Overactive immune cells can lead to autoimmune diseases such as systemic lupus erythematosus , also known as SLE for short , which causes damage to the skin , joints and other organs . Previous work suggests that SLE is correlated with a specific mutation in the FcγRIIB gene , but it is unclear how the mutation and the disease are connected . Proteins are made out of building blocks called amino acids , which have different chemical properties . A swap of one amino acid for another can have big consequences for the structure of a protein . In the case of FcγRIIB , the mutation that correlates with SLE changes an amino acid called isoleucine for another called threonine . Isoleucine does not mix well with water and is commonly found buried in the middle of proteins or inside cell membranes . Threonine , on the other hand , can readily interact with the hydrogen atoms in water and other amino acids . Hu , Zhang , Sun et al . used computer simulations and imaged single human cells to find out how the isoleucine to threonine change causes immune cells to become over-activated . The experiments revealed that threonine interacts with a nearby amino acid , putting a kink in the FcγRIIB protein . This kink causes the outer part of the FcγRIIB protein to bend towards the immune cell membrane , stopping it from binding to antibodies , and putting a break on immune cells that have become hyper-activated . There is currently no cure for SLE , but understanding its causes could take us a step closer to better management of the disease . Small molecule drug treatments often target the three-dimensional shape of certain proteins , so understanding the effect of mutations at the molecular level could help with the design of new treatments in the future . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"structural",
"biology",
"and",
"molecular",
"biophysics",
"immunology",
"and",
"inflammation"
] | 2019 | FcγRIIB-I232T polymorphic change allosterically suppresses ligand binding |
Changes in glutamatergic synaptic strength in brain are dependent on AMPA-type glutamate receptor ( AMPAR ) recycling , which is assumed to occur through a single local pathway . In this study , we present evidence that AMPAR recycling occurs through different pathways regulated by synaptic activity . Without synaptic stimulation , most AMPARs recycled in dynamin-independent endosomes containing the GTPase , Arf6 . Few AMPARs recycled in dynamin-dependent endosomes labeled by transferrin receptors ( TfRs ) . AMPAR recycling was blocked by alterations in the GTPase , TC10 , which co-localized with Arf6 endosomes . TC10 mutants that reduced AMPAR recycling had no effect on increased AMPAR levels with long-term potentiation ( LTP ) and little effect on decreased AMPAR levels with long-term depression . However , internalized AMPAR levels in TfR-containing recycling endosomes increased after LTP , indicating increased AMPAR recycling through the dynamin-dependent pathway with synaptic plasticity . LTP-induced AMPAR endocytosis is inconsistent with local recycling as a source of increased surface receptors , suggesting AMPARs are trafficked from other sites .
NMDA- and AMPA-type glutamate receptors ( NMDARs/AMPARs ) are the major excitatory synaptic receptors in brain . They are held at post-synaptic densities ( PSDs ) by scaffold proteins aligning the receptors with the presynaptic glutamate release sites . Changes in synaptic strength , such as long-term potentiation ( LTP ) , long-term depression ( LTD ) ( Malinow and Malenka , 2002 ) , and homeostatic plasticity ( Pérez-Otaño and Ehlers , 2005 ) , largely reflect the number of functional synaptic AMPARs . AMPAR internalization and recycling regulates AMPAR levels at synapses . Other processes , including diffusion of extrasynaptic AMPARs outside PSDs , association and dissociation of AMPARs with PSDs , and the number of ‘slots’ that AMPAR can occupy in PSDs ( Malinow and Malenka , 2002; Ehlers , 2000; Lin et al . , 2000; Petrini et al . , 2009; Ehlers et al . , 2007 ) , also contribute to setting AMPAR levels at PSDs ( Opazo et al . , 2012 ) . During synaptic stimulation , ‘constitutive’ AMPAR recycling is increased several fold to become ‘activity-dependent’ AMPAR recycling ( Ehlers , 2000 ) . AMPARs undergo endocytosis through clathrin-coated pits during activity-dependent recycling ( Carroll et al . , 1999; Lüscher et al . , 1999 ) and before exocytosis , traffic through the recycling endosomes ( REs ) identified by co-localization with transferrin receptors ( TfRs ) ( Ehlers , 2000 ) and Rab11 ( Park et al . , 2004 ) . During LTP , REs move from the dendritic shaft into synaptic spines ( Park et al . , 2006 ) from which regulated exocytosis of AMPARs appears to occur . It is uncertain whether AMPARs are exocytosed outside of the spines and traffic to PSDs via lateral diffusion ( Ashby et al . , 2006; Yudowski et al . , 2007; Tao-Cheng et al . , 2011; Malinow and Malenka , 2002 ) or are exocytosed at specific sites near PSDs ( Mohanasundaram and Shanmugam , 2010 ) . A minimal model of AMPAR constitutive and activity-dependent recycling has emerged from these and other studies . First , a single recycling pathway is assumed that starts at clathrin-coated pits ( Blanpied et al . , 2002 ) , moves through REs , and ends with exocytosis back at the plasma membrane . Second , it is assumed that AMPAR recycling occurs locally , that is , AMPAR endocytosis and exocytosis occur at sites within the same synaptic domain . The model predicts that during LTD-AMPAR levels in recycling , endosomes and/or lysosomes at PSDs increase because endocytosis increases without increasing exocytosis from REs ( Ehlers , 2000; Fernández-Monreal et al . , 2012 ) . During LTP , endocytosis is unchanged and AMPAR levels at PSDs are predicted to increase because of their exocytosis from REs causing decreased levels of AMPARs in REs ( Park et al . , 2006; Groc and Choquet , 2006 ) . Many factors specifically affect AMPAR activity-dependent recycling without affecting constitutive recycling [e . g . , AP2 ( Lee et al . , 2002 ) , Brag2 ( Scholz et al . , 2010 ) ] , and vice versa [e . g . , NSF ( Lee et al . , 2002 ) , PIP3 ( Arendt et al . , 2010 ) ] . These studies suggest that AMPAR activity-dependent recycling is uncoupled from constitutive recycling and that separate processes underlie the two types of AMPAR recycling . Transmembrane AMPA receptor regulatory proteins ( TARPs ) , such as stargazin , interact with recycling AMPARs at synapses ( Tomita et al . , 2004; Neudauer et al . , 1998 ) . TARPs interact with neuronal isoform of PDZ-protein interacting specifically with TC10 ( nPIST ) ( Cuadra et al . , 2004 ) , suggesting that the Rho small G protein , TC10 , might be a regulator of AMPAR recycling . Here , we describe how TC10 knockdown and TC10 functional mutants reduce AMPAR surface levels and synaptic currents by altering AMPAR recycling . TC10 mutants do not alter the increases in AMPAR surface levels that occur during LTP and only partially alter decreases in AMPAR surface levels and synaptic currents that occur during LTD . Overall , our findings indicate that TC10 mutants have differential effects on constitutive and activity-dependent AMPAR recycling because AMPARs traffic through different endocytosis pathways , and activity-dependent events alter the endocytosis pathway taken by AMPARs . Furthermore , our findings of increased AMPAR endocytosis with LTP are inconsistent with the assumption that AMPAR recycling occurs locally at synapses . Instead , AMPARs added to LTP-stimulated synapses may be trafficked into these synapses from outside the local synaptic pool .
We first examined whether the small GTPase , TC10 , had a role in AMPAR trafficking by knocking down TC10 expression ( Figure 1 ) . Knockdown was achieved using a short hairpin RNA construct ( shRNA ) that expressed a GFP reporter to identify neurons with the shRNA . Using real-time PCR , we observed that endogenous TC10 mRNA was reduced by 90% in cortical neurons ( Figure 1—figure supplement 1 ) . Neurons were co-transfected with AMPAR subunits , GluA1 , with a fluorescent mCherry tag at the extracellular N-terminus ( mCherry-GluA1 ) to simultaneously monitor GluA1 surface fluorescence and total GluA1 mCherry fluorescence . We found that the ratio of cell surface to total mCherry-GluA1 was similar from neuron to neuron ( SEM = 14% ) . With the TC10 knockdown , the surface/total mCherry-GluA1 ratio for neurons was significantly reduced by 79% ( Figure 1A , D ) compared to control neurons that expressed GFP without the shRNA . 10 . 7554/eLife . 06878 . 003Figure 1 . Disrupting TC10 level or function reduced AMPARs on the cell surface . ( A ) Representative somata of cultured rat hippocampal neurons ( E18 ) expressing mCherry-tagged GluA1 subunits , free Venus ( Control ) , shTC10RNA or Venus-tagged TC10WT , TC10DN or TC10CA . Intact , live neurons ( DIV18 ) were stained with anti-RFP antibody to visualize surface AMPA-type glutamate receptors ( AMPARs ) and then fixed and processed . Images of the whole neuron for each of the shown somata are displayed in Figure 1—figure supplement 2 . ( Scale bar = 10 μm ) . ( B ) Representative dendrites of neurons expressing mCherry-GluA1 without TC10 or with TC10WT , TC10DN , or TC10CA . Arrows indicate dendrites expressing TC10 mutants ( TC10DN or CA ) with weak surface expression of GluA1; arrowheads mark dendrites expressing mCherry-GluA1 without TC10 mutant expression and normal GluA1 surface expression level . ( Scale bar = 5 μm ) . ( C ) Representative dendrites of intact neurons expressing either free Venus ( Control ) , TC10WT , TC10DN , or TC10CA and labeled with an N-terminal , anti-GluA1 mAb to visualize endogenous , surface GluA1 receptors . ( Scale bar = 5 μm ) . ( D ) Quantification of the surface/total ratio of mCherry-GluA1 at the somata of transfected cells in ( A ) . Data are shown as ± SEM; control cells ( YFP ) 100% ± 14% , n = 32; TC10WT cells 118% ± 22% , n = 17; TC10DN cells 53% ± 9% , n = 23; TC10CA cells 47% ± 9% , n = 14; and shTC10 cells 21% ± 7% , n = 10 . ( **p < 0 . 03 relative to TC10WT ) . ( E ) Quantification of the surface/total ratio of mCherry-GluA1 in dendrites of transfected cells in ( B ) . Data are shown as ± SEM; control cells ( YFP ) 100% ± 12% , n = 23; TC10WT cells 133% ± 11%% , n = 23; TC10DN cells 53% ± 5% , n = 23; and TC10CA cells 54% ± 15% , n = 8 . ( *p < 0 . 03 relative to YFP;**p < 0 . 02 relative to TC10WT ) . ( F ) Quantification of endogenous surface GluA1 on dendrites in ( C ) . Data are shown as mean ± SEM; control cells ( YFP ) 100% ± 11%; TC10WT cells 130% ± 17%; TC10DN cells 59% ± 8%; and TC10CA cells 62% ± 10% ( n = 7–10 cells per group; **p < 0 . 003 relative to TC10WT ) . ( G ) Representative whole-cell patch-clamp paired recordings showing the evoked presynaptic action potential ( top ) and the post-synaptic AMPAR-mediated current response ( below ) . ( H ) Average AMPAR-mediated EPSC amplitudes of untransfected neurons ( control ) 365 . 2 ± 67 pA , n = 16 and neurons transfected with TC10WT 180 . 3 ± 36 pA , n = 8; TC10DN 144 . 5 ± 42 pA , n = 14; or TC10CA 139 . 2 ± 24 pA , n = 11 . ( *p < 0 . 04 relative to control;**p < 0 . 02 relative to control ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00310 . 7554/eLife . 06878 . 004Figure 1—figure supplement 1 . TC10 RNAi ( shRNA ) knocked down endogenous TC10 expression in neurons . Cortical neurons were infected with lentivirus encoding FUGW empty vector or TC10 short hairpin RNA construct ( shRNA ) on DIV7 and lysed for real-time PCR on DIV11 . shTC10 specifically knocked down TC10 mRNA level by 90% , but not the expression of Cdc42 and Rab11 . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00410 . 7554/eLife . 06878 . 005Figure 1—figure supplement 2 . Low-magnification images of cultured rat hippocampal neurons ( E18 ) corresponding to the somata shown in Figure 1A . Neurons express mCherry-tagged GluA1 subunits and either free Venus ( Control ) , TC10WT , TC10DN , or TC10CA . Intact , live neurons ( DIV18 ) were stained with anti-RFP antibody to visualize surface AMPARs and then fixed and imaged . ( Scale bar = 10 um ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00510 . 7554/eLife . 06878 . 006Figure 1—figure supplement 3 . TC10 mutants do not change synaptic density . The density of synapsin ( SYN ) puncta ( number of puncta/um ) remained unchanged when TC10WT or TC10DN/CA mutants were expressed ( n . s . = not significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00610 . 7554/eLife . 06878 . 007Figure 1—figure supplement 4 . TC10 mutants do not change the expression level of GluA1 . ( A ) Cortical neurons were infected with lentivirus encoding Venus-TC10WT or TC10DN/CA mutants on DIV1 and assayed for GluA1 expression with Western blot on DIV18 . The expression of GluA1 remained unchanged . ( B ) Quantification of GluA1 expression level in ( A ) , normalized to tubulin ( n=3 experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00710 . 7554/eLife . 06878 . 008Figure 1—figure supplement 5 . A comparison of somatic ER and the distribution of exogenous mCherry-GluA1 subunits and endogenous GluA1 subunits . Cultured neurons expressing mCherry-tagged GluA1 ( top panel ) were stained with an antibody against the endoplasmic reticulum ( ER ) marker protein , protein disulfide isomerase , and compared against endogenous GluA1 in untransfected cells ( bottom panel ) . Both exogenous and endogenous GluA1 exhibited similar distributions in somatic ER . ( Scale bar = 10 um ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 00810 . 7554/eLife . 06878 . 009Figure 1—figure supplement 6 . The effects of TC10 mutants on AMPARs in the somatic Golgi and in dendritic shafts . ( A ) Cultured hippocampal neurons expressing mCherry-GluA1 with Venus , TC10WT , TC10DN , or TC10CA were fixed and stained for the Golgi marker , GM130 on DIV18 . ( Scale bar = 10 um ) . ( B ) Quantification of the intensity of mCherry-GluA1 co-localizing with GM130 in the somata . Expression of TC10 WT or mutants did not change its distribution: Control cells ( normalized ) 100% ± 2% , n=11; TC10WT cells 103% ± 3% , n=6; TC10DN cells 109% ± 10% , n=19; TC10CA cells 104% ± 3% , n=15 . ( C ) Line profile of mCherry-GluA1 intensity along the dendrite when co-expressed with Venus or TC10DN . TC10DN expression results in mCherry-GluA1 accumulation in the dendritic shafts and not in the somata . ( Scale bar = 5 um , n=4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 009 The surface levels of GluA1-containing AMPARs at somata were also reduced when we expressed TC10 mutants in the neurons . The T31N mutation or the ‘dominant-negative’ mutant ( TC10DN ) keeps TC10 in its inactive , GDP-bound state . The Q75L mutation or the ‘constitutively active’ mutant ( TC10CA ) is kept in its GTP-bound state ( Neudauer et al . , 1998 ) . Expression of wild-type TC10 ( TC10WT ) did not significantly alter surface levels at somata assayed by mCherry-GluA1 transfection , whereas TC10DN and TC10CA mutants reduced GluA1-containing AMPAR surface levels by ∼50% ( Figure 1A , B; statistical details in the legend; images of the corresponding whole neurons are displayed in Figure 1—figure supplement 2 ) . While TC10 knockdown with shRNA did not alter somata morphology , dendritic morphology was altered and synapse numbers reduced ( not shown ) . In contrast , TC10DN , TC10CA mutants , and TC10WT did not alter dendrite morphology ( not shown ) or synapse number ( Figure 1—figure supplement 3 ) . In dendrites , GluA1 surface levels were again reduced ∼50% by TC10DN and TC10CA and a small but significant increase of GluA1 surface levels ( 33% ± 10% ) was observed with TC10WT expression ( Figure 1A ) . We obtained similar results in dendrites with TC10WT , TC10DN , and TC10CA , when a GluA1-specific antibody that recognizes an extracellular epitope was used to quantify the endogenous surface AMPAR levels ( Figure 1F ) . We also performed paired whole cell recordings from synaptically coupled cultured hippocampal neurons to assay levels of functional AMPARs at the synapses . TC10DN and TC10CA reduced the synaptic AMPAR EPSC amplitudes by ∼60% ( Figure 1G , H ) . Thus , the number of functional AMPARs at synapses was reduced to approximately the same extent as the total number of AMPARs on the cell surface as measured by immunostaining . Our findings with immunofluorescence and electrophysiology that expression of TC10DN and TC10CA in neurons caused the same effects on AMPAR levels in neurons are consistent with previous studies characterizing the role of TC10 in the secretory pathway . Assaying depolarization-induced secretion of neuropeptide Y ( NPY ) in PC12 cells , both TC10DN and TC10CA reduced secretion in the range of 40–60% ( Kawase et al . , 2006 ) similar to our findings with AMPARs ( Figure 1 ) . Using a TC10 FRET sensor construct to assay whether TC10 is in the GTP-TC10 or GDP-TC10 state , they concluded that the TC10 GTPase hydrolysis cycle is required for NPY secretion . TC10DN and TC10CA both reduced secretion by blocking the GTP hydrolysis cycle at different steps . Similar results were obtained assaying nerve growth factor-induced neurite outgrowth in PC12 cells ( Fujita et al . , 2013 ) . Both papers suggested that during exocytosis TC10CA allowed cargo to load into transport vesicles to be delivered to target membranes . TC10CA blocked exocytosis by preventing the GTP-TC10 to GDP-TC10 transition required for transport vesicle docking and/or fusion . In contrast , the results suggest that TC10DN blocked the GDP-TC10 to GTP-TC10 transition , which blocked a different step , cargo loading onto vesicles thus preventing vesicle delivery to target membranes . Unexpectedly , in the paired whole-cell recordings from synaptically coupled cultured hippocampal neurons , TC10WT reduced synaptic currents by 51% ( Figure 1E , F ) even though TC10WT increased cell-surface AMPARs by 33% ( Figure 1B , C ) . This differential effect of TC10WT suggests that TC10 function somehow distinguishes between AMPAR trafficking to and/or from synaptic sites compared to other sites on the cell surface of dendrites . To explore how perturbing TC10 function reduced AMPAR surface levels , we examined whether TC10 mutants altered endogenous GluA1 subunit levels . Lentiviral infection of ∼90% of the cultured neurons with TC10WT or the TC10 mutants did not alter endogenous GluA1 subunit levels , indicating that TC10 mutants do not alter AMPAR subunit synthesis or degradation ( Figure 1—figure supplement 4 ) . Nor did the TC10 mutant expression appear to alter AMPAR trafficking through the secretory pathway in the soma ( Figure 1—figure supplement 5 ) . Previously , we had found that AMPAR loss in dendrites correlated with AMPAR accumulation in the Golgi , which blocked AMPAR transport from somata to dendrites of cultured neurons ( Jeyifous et al . , 2009 ) . Consistent with this possibility , TC10WT , TC10DN , and TC10CA all strongly co-localized with the Golgi marker GM130 in somata ( Figure 1—figure supplement 6A ) . However , little mCherry-GluA1 was found in the Golgi , and most GluA1 subunits co-localized with endoplasmic reticulum ( ER ) markers ( Figure 1—figure supplement 5 ) , as previously observed for newly synthesized AMPAR subunits ( Greger et al . , 2002 ) . We observed a similar distribution for the native GluA1 subunits in the ER ( Figure 1—figure supplement 5 ) indicating that heterologous express of mCherry-GluA1 did not greatly increase levels of GluA1 in the ER . The small amount of mCherry-GluA1 that co-localized with the Golgi marker did not significantly change with TC10DN and TC10CA expression compared to that of TC10WT ( Figure 1—figure supplement 6A , B ) . In fact , when TC10 mutants were expressed , there was a significant increase in mCherry-GluA1 levels in the shafts of dendrites ( Figure 1—figure supplement 6C ) , suggesting that the TC10 mutants cause an accumulation of AMPAR intracellular levels in the dendritic shafts , but not in the somata , that results in the decreases in surface levels . We next examined whether the TC10 mutants caused AMPARs to accumulate during their recycling in dendritic shafts . It is well established that AMPARs enter REs in dendritic shafts after synaptic stimulation with AMPA or NMDA ( Malinow and Malenka , 2002; Newpher and Ehlers , 2008 ) . To test for accumulation of AMPARs in REs , we assayed whether TC10 mutants increased endogenous AMPAR co-localization with TfRs , which largely exist in REs in dendritic shafts where they co-localize with GluA1-containing AMPARs ( Carroll et al . , 1999; Ehlers , 2000 ) . Surprisingly , there was little co-localization , ( 4–6% ) , between endogenous GluA1 and TfR in dendrites under these conditions . The percentage of co-localization did not change for non-transfected neurons or for neurons expressing Venus-tagged TC10WT , TC10DN , or TC10CA ( Figure 2A , B ) . TC10WT , TC10DN , or TC10CA expression also did not alter the distribution of TfRs in the dendrites and did not significantly affect the recycling of TfRs ( data not shown ) . To confirm the validity of TfRs as a recycling endosome marker , we tested another recycling endosome protein , Rab11 , and found significant co-localization between mCherry-tagged Rab11 and TfRs confirming that TfRs are largely in REs in dendrites ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 06878 . 010Figure 2 . TC10 regulates AMPAR trafficking through an Arf6-containing endocytosis pathway in dendrites . ( A ) Effects of TC10 constructs on the co-localization between endogenous GluA1 subunits and transferrin receptors ( TfRs ) . Cultured neurons were transfected with Venus-tagged TC10WT or TC10DN/CA mutants and permeabilized cells stained for total GluA1 and TfR ( Tf-Alexa 647 ) . GluA1 showed little co-localization with Tf-labeled endosomes in dendrites . ( Scale bar = 5 μm ) . ( B ) Quantification of GluA1 and TfR co-localization . Images in A were analyzed to measure the percent of GluA1 puncta co-localizing with TfR puncta . Expression of TC10 constructs did not alter the degree of co-localization . Data are shown as mean ± SEM; control cells 4 . 6% ± 0 . 3%; TC10WT cells 4 . 9% ± 0 . 8%; TC10DN cells 4 . 2% ± 0 . 8%; TC10CA cells 5 . 6% ± 1% , n = 5 for all groups . ( C ) Effects of TC10 constructs on the co-localization between GluA1 subunits and Arf6 . Neurons were transfected with Arf6-HA and Venus-TC10WT or TC10DN/CA mutants . Cells were permeabilized and stained for Arf6-HA ( Rb anti-HA ) and total GluA1 ( anti-GluA1 mAb ) . ( Scale bar = 5 μm ) . ( D ) Quantification of the overlap between GluA1 and Arf6 . Images in C were analyzed to measure the Pearson's correlation coefficients ( Rr ) of GluA1 co-localization with Arf6 . Data are shown as mean ± SEM; TC10WT 0 . 46 ± 0 . 04; TC10DN 0 . 66 ± 0 . 04; TC10CA 0 . 31 ± 0 . 04 ( n = 7-10 cells per group; *p < 0 . 02 relative to TC10WT; **p < 0 . 05 relative to TC10WT ) . ( E ) Effects of TC10 constructs on GluA1 puncta density ( number of puncta per 10 μm ) in dendrites of neurons expressing Venus-TC10WT or TC10DN/CA mutants . Data are shown as mean ± SEM; TC10WT 6 . 9 ± 0 . 4; TC10DN 5 . 6 ± 0 . 7; TC10CA 41 . 7 ± 4 . 8 ( n = 5 fields per group; *p < 0 . 0002 relative to TC10WT ) . Inset , histogram showing the distribution of small ( diameter <200 nm ) and large ( diameter 200–1200 nm ) GluA1 puncta ( n = ∼100 puncta per group ) . ( F ) Quantification of the overlap ( Pearson's correlation coefficients , Rr ) between TC10 and Arf6 in ( C ) . Data are shown as mean ± SEM; TC10WT 0 . 77 ± 0 . 03; TC10DN 0 . 60 ± 0 . 02; TC10CA 0 . 84 ± 0 . 02 ( n = 7–10 cells per group; *p < 0 . 0004 relative to TC10WT ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 01010 . 7554/eLife . 06878 . 011Figure 2—figure supplement 1 . Co-distribution of TfR staining and mCherry-Rab11 . Hippocampal cultures were stained for TfR and Rab11 on DIV18 . The two markers showed extensive overlap . ( Scale bar = 5 um ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 011 Since few GluA1-containing AMPARs accumulated in TfR-labeled REs , it is possible that AMPAR recycling occurs via a different endocyctosis pathway not taken by TfRs . Several other endocytosis pathways exist ( Doherty and McMahon , 2009 ) . In particular , one endocyctosis pathway that utilizes the small GTPase , Arf6 , occurs in dendrites ( Gong et al . , 2007; Lavezzari and Roche , 2007 ) and may be involved in AMPAR endocytosis ( Scholz et al . , 2010 ) . In contrast to the ∼5% co-localization observed between GluA1 and TfR puncta , we observed a significantly higher degree of overlap ( Pearson's correlation coefficient , PCC = 0 . 46 ) between endogenous GluA1 subunits and transfected HA-Arf6 ( Figure 2C , D ) , an established marker of endosomes in the Arf6 , clathrin-independent pathway ( Donaldson et al . , 2009 ) . Importantly , there was a high degree of co-localization between TC10WT and HA-Arf6 ( PCC = 0 . 77 ) , indicating that TC10 is largely found in Arf6 endosomes in dendrites ( Figure 2E ) . Consistent with TC10 having a role in AMPAR recycling through Arf6 endosomes , the TC10 mutants caused significant changes in the distribution of AMPAR intracellular puncta that co-localized with Arf6 ( Figure 2C , D ) . TC10DN significantly increased GluA1 co-localization with Arf6 , changing from a PCC of 0 . 46 to 0 . 66 ( Figure 2C , D ) . Increases in AMPARs co-localizing with Arf6 appeared to be at subdomains within Arf6 endosomes ( Figure 2C ) , consistent with TC10CA blocking AMPAR exit , and their accumulation in Arf6-endosomes . The TC10CA mutant had the opposite effect . TC10CA caused GluA1 co-localization with Arf6 to significantly decrease from a PCC of 0 . 46 to 0 . 31 , consistent with increased AMPAR exit from Arf6 endosomes ( Figure 2C , D ) . Because TC10CA caused a ∼50% decrease in surface AMPAR , similar to the effects of TC10DN ( Figure 1B , C and F ) , increases in AMPARs exiting from Arf6 endosomes were not inserted into the cell surface . Expression of TC10CA changed the size and number of the GluA1-containing puncta in dendritic shafts . GluA1-containing puncta were much more numerous compared to TC10WT , with an increased number of smaller puncta and a loss of most of the larger puncta in dendrites ( Figure 2E ) . This result suggests that the AMPARs exiting from Arf6 endosomes remained in smaller transport vesicles because TC10CA blocked their exocytosis at the plasma membrane . These interpretations of the effects of the TC10DN and TC10CA mutants are supported by the conclusions of previous studies of the effects of the TC10 mutants on the distribution of different cargo during secretion ( Kawase et al . , 2006; Fujita et al . , 2013 ) ; specifically that TC10DN blocked the GDP-TC10 to GTP-TC10 transition and cargo loading onto the transport vesicle , while TC10CA blocked the GTP-TC10 to GDP-TC10 transition and the transport vesicle exocytosis . The clathrin-dependent endocytosis pathway taken by TfRs requires dynamin function , while Arf6-dependent clathrin-independent endocytosis is independent of dynamin function ( Doherty and McMahon , 2009 ) . As another test of the AMPAR endocytosis pathway , we used the reagent , dynasore , to block dynamin activity in neurons and assayed TfR and AMPAR endocytosis . In these experiments , we performed two different sets of assays to quantitatively measure how dynasore affected AMPAR and TfR internalization . In the first assay , mCherry-tagged GluA1 was expressed as in Figure 1 , and an ‘antibody feeding’ assay was used to label only the internalized mCherry-tagged , GluA1 AMPARs , and TfRs ( Figure 3A , B ) . After dynasore treatment , we compared the distribution of internalized mCherry-GluA1 subunits to that of TfRs surface labeled with Alexa Fluor 647-conjugated transferrin ( Tf-Alexa 647; Figure 3A ) . Consistent with a block of its endocytosis , dynasore treatment appeared to cause TfRs to buildup at the plasma membrane ( Figure 3A ) . Tf-Alexa 647 uptake was inhibited by 70% ± 18% in the presence of dynasore ( Figure 3B , C ) . However , dynasore treatment did not significantly alter the levels ( Figure 3C ) or distribution ( Figure 3A ) of internalized mCherry-GluA1 subunits . 10 . 7554/eLife . 06878 . 012Figure 3 . AMPARs undergo dynamin-independent endocytosis . ( A ) mCherry-GluA1 and TfR internalization and block of dynamin function . Hippocampal neurons were transfected with mCherry-GluA1 , and 1 day post-transfection , treated with 1% DMSO ( control group ) or 80 μM dynasore ( to block dynamin function ) for 30 min at 37°C . Neurons were then incubated with anti-RFP antibodies and Tf-Alexa 647 at 37°C for 40 min , acid washed to remove surface antibodies/dye , fixed , permeabilized , and imaged . With sham treatment ( control ) , surface-labeled GluA1 and TfR were both internalized in the dendritic shaft ( first row ) . Dynasore treatment had no effect on GluA1 internalization , but blocked TfR endocytosis ( Tf-Alexa 647 labeling confined to surface and not present intracellularly , second row ) . ( B ) Intensity plots of internalized TfR ( Tf-Alexa 647 ) with and without ( sham ) dynasore treatment as in ( A ) . Dynasore greatly inhibited TfR internalization . ( C ) Quantification of GluA1 and Tf-Alexa647 internalization and block of dynamin function . Data are shown as mean ± SEM; GluA1 endocytosis , sham 100% ± 16%; dynasore 112% ± 12% . For Tf-Alexa647 endocytosis , sham 100% ± 23%; dynasore 30% ± 18% , ( n = 5 for all groups; ***p < 0 . 001 ) . ( D ) Endogenous GluA1 and TfR internalization and block of dynamin function . As an alternative approach to measure GluA1 and TfR Internalization , cortical neurons were sham treated ( with DMSO ) or with dynasore as in ( A ) . After , surface proteins were labeled with Sulfo-NHS-SS-biotin and cultured for 40 min at 37°C to allow for endocytosis . Biotin on proteins remaining on the cell surface was removed by glutathione treatment and cells solubilized . Internalized proteins were pulled down with streptavidin beads and analyzed by Western blotting . Displayed are GluA1 , TrR , and actin bands ( loading control ) from whole cell lysates to estimate inputs for sham- or dynasore-treated neurons ( left ) and GluA1 and TfR bands from the streptavidin pull-downs to estimate internalized receptors for sham- or dynasore-treated neurons ( right ) . ( E ) Quantification of GluA1 and TfR internalization and block of dynamin function . The levels of GluA1 endocytosis were reduced to 90% ± 9% vs sham treated by dynasore treatment . The levels of TfR endocytosis were reduced to 61% ± 6% ( n = 3 experiments; ***p < 0 . 01 ) vs sham treated by dynasore treatment . Total protein levels ( inputs ) for GluA1 and TfR were not affected by dynasore treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 012 In the second assay , endogenous AMPAR internalization was assayed by biotinylating cell-surface proteins with a cleavable biotinylation cross-linking reagent and treating intact neurons with glutathione , a membrane impermeable reagent after internalization . Dynasore treatment reduced the levels of internalized TfR by 61% ± 6% ( Figure 3D , E ) , consistent with endocytosis predominantly through clathrin-dependent endocytosis pathway . In contrast , dynasore treatment reduced internalized GluA1 subunits by only 10% ± 9% ( Figure 3D , E ) , consistent with endocytosis predominantly through a different endocytosis pathway . Based on these results together with previous data ( Figure 2 ) , we conclude that TC10 has a role in AMPAR recycling through an Arf6-dependent recycling pathway that is different from the TfR recycling pathway . Evidence that AMPARs are in REs and clathrin-coated pits comes almost exclusively from studies examining AMPAR activity-dependent recycling . Our findings of AMPAR recycling via a TC10- and Arf6-dependent , dynamin-independent pathway occurred during constitutive AMPAR recycling . We , therefore , investigated whether the AMPAR endocytosis pathway changes under conditions where activity-dependent recycling occurs . First , we specifically tested how expressing TC10DN or TC10CA in neurons altered LTP and LTD . Chemically inducing LTP ( cLTP ) with glycine stimulation to activate synaptic NMDARs increased the AMPAR surface/total ratio by 39% ± 10% ( p < 0 . 1 , n = 10 , 14 ) , while chemically inducing LTD ( cLTD ) with acute NMDA treatment , decreased the AMPAR surface/total ratio by 62% ± 3% ( Figure 4A , B ) . 10 . 7554/eLife . 06878 . 013Figure 4 . Synaptic activity alters the endocytosis pathway taken by AMPARs . ( A ) Effects of TC10 constructs on chemically inducing LTP ( cLTP ) and chemically inducing LTD ( cLTD ) . Hippocampal neurons were transfected with mCherry-GluA1 plus Venus , TC10DN , or TC10CA . 1 day post-transfection , cells were stimulated with a mixture of the NMDA receptor agonist , glycine , and GABAA receptors antagonists to induce long-term potentiation ( LTP ) chemically in hippocampal cultures . Then , surface exposed mCherry-GluA1 was stained live with anti-RFP antibody . In control cells , cLTP treatment caused a 39% increase in surface GluA1 ( 100% ± 9% , n = 10–139% ± 10% , n = 14 ) . In cells expressing TC10DN and TC10CA , cLTP treatment also caused a 78% ( 37% ± 7% , n = 13–66% ± 9% , n = 12 ) and 54% ( 41% ± 4% , n = 21–63% ± 6% , n = 15 ) increase in surface GluA1 , respectively . In parallel , transfected cells were treated with NMDA to chemically induce long-term depression ( LTD ) as well . Control cells showed a 62% decrease of surface GluA1 ( 100% ± 10% , n = 10–38% ± 3% , n = 11 ) , while TC10DN and TC10CA expressing cells showed a 31% ( 67% ± 7% , n = 13–46% ± 3% , n = 10 ) and 55% ( 60% ± 5% , n = 5–27% ± 4 . 2% , n = 6 ) reduction , respectively . ( B ) Effects of TC10 constructs on cLTP and cLTD . Data in ( A ) replotted and normalized to respective sham values in order to compare the changes with cLTP and cLTD . ( C ) Effects of TC10 constructs on synaptically induced LTD . Left: AMPAR-EPSC amplitudes , expressed as a percentage of baseline AMPAR-EPSC amplitude ( 10 min ) , before and after the induction of LTD by electrical LFS for 5 min ( gap ) . Right: Bar graph of the average AMPAR-EPSC amplitude depression measured 20 min after LFS . Control cells showed a 52 . 4% ± 0 . 5% reduction in AMPAR-EPSC amplitude after LTD induction; TC10DN and TC10CA showed a 27 . 5% ± 1 . 0% and 48 . 3% ± 1 . 0% reduction in AMPAR-EPSC amplitude , respectively . ( D ) mCherry-GluA1 internalization after cLTP . Neurons were sham or cLTP treated and imaged for internalized mCherry-GluA1 and either TfR or HA-Arf6 . GluA1 internalization increased by 24% ± 5% when treated with cLTP ( n = 6; **p < 0 . 01 ) . ( E ) mCherry-GluA1 co-localization with TfR ( left ) and Arf6 ( middle ) after cLTP . Co-localization of internalized GluA1 with TfR increased with cLTP: sham-treated 21% ± 2% , n = 15; cLTP 56% ± 2% , n = 23 ( ***p < 0 . 0001 ) . Co-localization of internalized GluA1 with Arf6 did not significantly change: sham treated 52% ± 2% , n = 12; cLTP 47% ± 3% , n = 5 . Left panel displays the distribution of GluA1 co-localization with TfR , Arf6 , or neither marker for sham treated and cLTP conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 01310 . 7554/eLife . 06878 . 014Figure 4—figure supplement 1 . Synaptic activity alters the endocytosis pathway taken by AMPARs . Representative images for the quantification displayed in Figure 4D , E . Neurons were sham or cLTP treated and imaged for internalized mCherry-GluA1 and either TfR ( A ) or Arf6-HA ( B ) . cLTP treatment caused an increase in the co-localization between internalized GluA1 and TfR , while co-localization with Arf6 didn't change ( arrows indicate co-localizing puncta ) . ( Scale bar = 5 um ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 014 Neurons expressing TC10DN or TC10CA showed increased AMPAR surface/total ratios with cLTP ( Figure 4A ) . The percent increase in surface AMPARs was even larger than controls ( Figure 4B ) . The larger percent increase may be explained by the fact that the cLTP-induced increases in surface AMPARs were about the same size as controls , while the levels of surface AMPARs were reduced by TC10 variants ( Figure 1 ) . The finding that TC10DN or TC10CA expression had no negative effect on cLTP-induced increases but reduced surface AMPAR levels prior to cLTP is again consistent with AMPAR recycling occurring through multiple endocytosis pathways , one pathway providing AMPARs inserted during cLTP and the other pathway primarily involved in constitutive AMPAR recycling . In neurons expressing TC10 mutants during cLTD , we observed only half the cLTD-induced decrease in AMPARs with TC10DN , while TC10CA did not alter the cLTD-induced decrease ( Figure 4A , B ) . These results are different from what occurs during LTP and suggest that the TC10- and Arf6-dependent endocytosis pathway is involved to some degree in the AMPAR trafficking during LTD that removes AMPARs from synapses . To more directly test how TC10DN and TC10CA affect the trafficking of synaptic AMPARs during LTD , we induced LTD by low-frequency synaptic stimulation ( Montgomery and Madison , 2002; Montgomery et al . , 2005 ) ( Figure 4C ) . Similar to what we observed with cLTD , the AMPAR-mediated synaptic currents were reduced by 52 . 4% ± 0 . 5% by LTD . During LTD , AMPAR EPSCs were reduced by 27 . 5% ± 1 . 0% in the presence of TC10DN , and TC10CA had no effect on LTD ( a 48 . 3% ± 1 . 0% reduction ) ( Figure 4C ) . We could not perform similar experiments with LTP because LTP cannot be induced by electrical stimulation in dissociated cultures ( Shi et al . , 1999; Li et al . , 2011 ) . However in a previous study , we found that the same cLTP protocol applied to dissociated hippocampal neuronal cultures caused a LTP of EPSC amplitudes ( Li et al . , 2011 ) . Our results with TC10DN or TC10CA expression on AMPAR trafficking during cLTP suggest that constitutive and activity-dependent AMPAR recycling occurs through different pathways . To further test this possibility , we assayed how cLTP altered the levels of internalized GluA1 subunit co-localization with TfRs or with Arf6-HA . We observed that 21% ± 2% of the internalized AMPARs co-localized with TfR ( Figure 4D , E ) , consistent with a previous estimate of the amount of internalized AMPARs that co-localized with TfR during constitutive recycling ( Park et al . , 2004 ) . 52% of the internalized AMPAR co-localized with Arf6-HA and 27% of internalized AMPARs did not co-localize with TfRs or with Arf6-HA ( Figure 4E ) . This pool of internalized AMPARs is likely in a different endosomal pool such as early endosomes , late endosomes , or lysosomes ( Fernández-Monreal et al . , 2012 ) . After cLTP induction , the distribution of internalized AMPARs in the different pools changed: cLTP increased the amount of internalized GluA1 that co-localized with TfR from 21% to 56% ( Figure 4E; Figure 4—figure supplement 1A ) , while the internalized GluA1 that co-localized with Arf6 was unchanged ( 52% vs 47%; Figure 4E; Figure 4—figure supplement 1B ) . cLTP increased the total amount of internalized GluA1 by 24% ± 5% ( Figure 4D ) , while the pool of internalized AMPARs that did not co-localize with TfRs or with Arf6-HA was no longer measurable . Our results are consistent with significant changes in AMPAR endocytosis and recycling that occurs during cLTP . During constitutive recycling , AMPARs appear to be recycling primarily through a dynamin-independent , Arf6- , and TC10-dependent pathway though some recycle through a dynamin-dependent , TfR-containing pathway . After cLTP , a much larger fraction of the AMPARs recycle through the dynamin-dependent pathway taken by TfRs . Our results further suggest that after LTP , fewer AMPARs are trafficked for degradation via a late endosomal/lysosomal pathway as previously described ( Fernández-Monreal et al . , 2012 ) .
The findings of this study are consistent with constitutive AMPAR recycling occurring largely through a pathway different from the clathrin-dependent pathway used by TfRs . The data in support of an alternative recycling pathway are that internalized AMPARs co-localize much more with Arf6 than with TfRs ( Figure 2 ) , and that inhibition of dynamin function blocks TfR endocytosis without significantly altering AMPAR endocytosis ( Figure 3 ) . The features of the alternative AMPAR recycling pathway , for example , presence of Arf6 and independence from dynamin function , are consistent with a clathrin-independent endocytosis pathway ( Doherty and McMahon , 2009; Donaldson et al . , 2009 ) . The Arf6-mediated , clathrin-independent pathway is a separate endocytosis pathway that exists in neurons and mediates the endocytosis and recycling of a number of different receptors and transport proteins . These include the metabotropic glutamate receptors ( mGluR ) , mGluR5 ( Fourgeaud et al . , 2003 ) and mGluR7 ( Lavezzari and Roche , 2007 ) , and the potassium channel Kir3 . 4 ( Gong et al . , 2007 ) . AMPARs were shown to undergo a clathrin-independent recycling pathway , in addition to clathrin-mediated pathway in Caenorhabditis elegans ( Glodowski et al . , 2007 ) . Another study using electron microscopy ( EM ) ( Tao-Cheng et al . , 2011 ) also suggests that constitutive AMPAR recycling occurs through a different pathway . The study found that all intracellular structures with the features of REs were labeled for TfRs in dendritic shafts of cultured rat hippocampal neurons but only 28% of these REs were labeled for AMPARs . If AMPAR endocytosis occurs through a single , clathrin-dependent pathway , AMPARs would all enter clathrin-coated pits during constitutive AMPAR recycling . Entry into clathrin-coated pits can only be unambiguously resolved at the EM level . EM studies assaying AMPAR subunit localization in clathrin-coated pits observed few AMPARs in clathrin-coated pits ( Petralia et al . , 2003; Tao-Cheng et al . , 2011 ) . Tao-Cheng et al . found that 76% of the clathrin-coated pits contained TfRs but only 24% of the pits contained AMPAR subunits . Clathrin-coated pits near PSDs have been proposed to be specialized endocytic zones ( EZs ) ( Blanpied et al . , 2002; Racz et al . , 2004 ) that mediate endocytosis of AMPA receptors for local recycling in spines ( Lu et al . , 2007; Petrini et al . , 2009; Kennedy et al . , 2010 ) . Both EM studies failed to detect any AMPAR labeling in the EZs at synapses under conditions where constitutive AMPAR recycling was occurring ( Petralia et al . , 2003; Tao-Cheng et al . , 2011 ) . In this study , we have also characterized in detail the role of the small Rho GTPase , TC10 , in AMPAR recycling through the Arf6-mediated , clathrin-independent pathway . We found that altering TC10 expression and function in neurons reduced levels of cell-surface AMPARs . The TC10 mutants , TCDN and TC10CA , equally reduced cell-surface AMPARs by ∼50% but did not significantly affect AMPAR trafficking through the secretory pathway . Normal levels of AMPARs departed from the somatic Golgi and were transported to dendrites and synapses . However , the TC10 mutants had differential effects on where AMPARs accumulated in dendritic shafts . TC10DN reduced surface AMPARs by causing increased AMPAR accumulation in Arf6 endosomes apparently by blocking their exit from the endosomes . TC10CA reduced surface AMPARs by increasing their exit from Arf6 endosomes and blocking their exocytosis , thereby increasing what appear to be AMPAR transport vesicles in the dendritic shafts . Results from a previous study suggest that the associations between TC10 and AMPARs are indirect , requiring an adaptor protein , nPIST , which interacts with the AMPAR TARP subunit ( Cuadra et al . , 2004 ) . nPIST , like TC10 , primarily co-localizes with Golgi markers in the somata of cultured hippocampal neurons . But it is also found in puncta in dendritic shafts , not in spines , and the puncta do not co-localize with Golgi membranes ( Chen et al . , 2012 ) . nPIST interactions with TC10 in dendrites , thus , are likely at the Arf6 endosomes where we observed most of TC10 in dendrites . One possibility is that TC10 acts to regulate interactions between nPIST and AMPAR TARP subunits when present together in Arf6 endosomes , and thereby , regulate the trafficking of AMPARs from Arf6 endosomes to dendritic exocytosis sites . Our findings that AMPARs recycle through two different pathways provide new insights into how AMPAR recycling is altered in response to changes in synaptic activity . The increase in AMPARs in REs after cLTP reflects a redistribution of trafficking AMPARs in dendrites such that AMPAR receptor recycling via REs is increased , while recycling via the Arf6-TC10-containing endosomes was unchanged . We also observed a third pool of endocytosed AMPARs that did not co-localize with either TfR or Arf6 . This third pool , which should include AMPARs in early endosomes , late endosomes and lysosomes , decreased from 27% of the total to essentially 0% . A previous study suggested that during LTD an activity-dependent switch occurs such that more endocytosed AMPARs in early endosomes were routed to the Rab7-dependent pathway to lysosomes and less were routed to the recycling endosome pathway ( Fernández-Monreal et al . , 2012 ) . Our data suggest that the opposite is occurring during cLTP . That is , an activity-dependent switch occurs such that few to none of the endocytosed AMPARs in early endosomes are routed via the Rab7-dependent pathway to lysosomes . Instead , virtually all AMPARs in early endosomes are routed to the REs . However , this switch can only explain part of the increase in AMPARs in REs during cLTP because the total number of endocytosed AMPARs increased by an additional 24% . Thus , part of the increase in AMPARs in REs during cLTP appears to be caused by an increase in AMPAR endocytosis , presumably at clathrin-coated pits Expression of TC10DN increased AMPARs in Arf6-containing endosomes , while TC10CA decreased AMPARs in these endosomes ( Figure 2C , D ) . The differential effects of TC10DN and TC10CA on AMPAR localization in Arf6-containing endosomes provide a potential explanation for the differential effects of TC10DN and TC10CA during LTD . It is possible that AMPARs excluded from Arf6-containing endosomes by TC10CA have access to the endosomal pathway used during LTD . The AMPARs added to the Arf6-containing endosomes with TC10DN expression do not have access to the endosomal pathway used during LTD . Previous studies have found that synaptic stimulation increases dynamin- and clathrin-dependent AMPAR endocytosis ( Carroll et al . , 1999; Lüscher et al . , 1999; Ehlers , 2000; Lee et al . , 2002; Scholz et al . , 2010; Tao-Cheng et al . , 2011 ) . In some of these studies , blocking dynamin/clathrin-dependent endocytosis prevented LTD but did not alter constitutive AMPAR endocytosis . It was also found that blocking constitutive AMPAR endocytosis by interfering with NSF binding to GluA2 subunits does not alter LTD or activity-dependent AMPAR endocytosis ( Carroll et al . , 1999; Lüscher et al . , 1999; Ehlers , 2000; Lee et al . , 2002 ) . Altogether , these studies demonstrate that activity-dependent and constitutive AMPAR recycling can be uncoupled , and thus , appears to be independently regulated , consistent with separate processes underlying activity-dependent and constitutive AMPAR recycling . Our results with TC10 mutants are also consistent with separate processes underlying activity-dependent and constitutive AMPAR recycling . The results from this study suggest that the two different AMPAR recycling pathways serve different functions . The Arf6-dependent recycling pathway , also dependent on TC10 function , predominates during AMPAR constitutive recycling . Another recycling pathway , which is dynamin and clathrin dependent , increases during AMPAR activity-dependent recycling ( cLTP and cLTD ) . The simplest model of AMPAR recycling based on our data is displayed in Figure 5B . To explain the effects of dynamin inhibition , we propose that the two recycling pathways originate at separate endocytosis sites , either clathrin-coated pits or clathrin-independent sites . As proposed by others ( Donaldson et al . , 2009 ) but not shown in the model , the two recycling pathways merge at early endosomes , where AMPARs would be sorted for trafficking into the various pools , either the REs used by TfRs , the Arf6-depedent REs or late endosomes for lysosomal degradation . As proposed in the model , the two recycling pathways end when AMPARs are exocytosed at separate sites at the cell membrane . The presence of two different AMPAR recycling exocytosis sites may help to explain recent data describing different AMPAR recycling exocytosis sites ( Kennedy et al . , 2010; Ahmad et al . , 2012 ) . We envision that the Arf6/TC10-dependent recycling pathway has largely a caretaker role , delivering AMPARs to early endosomes where a decision is made to degrade internalized AMPARs or return them to the plasma membrane . The dynamin/clathrin-dependent recycling functions predominantly during synaptic activation and appear to have a different role than the Arf6/TC10-depedent recycling pathway ( Figure 5 ) . Other receptors are regulated in a similar way by the same two recycling pathways . β2-adrenergic and M3 muscarinic receptors undergo constitutive recycling when not activated by ligand via an Arf6-dependent , clathrin-independent pathway . After ligand activation , their recycling pathway switches and recycling occurs via the clathrin-dependent pathway ( Scarselli and Donaldson , 2009 ) . The recycling pathway of α1-integrin receptors also switches after their activation ( Arjonen et al . , 2012 ) . 10 . 7554/eLife . 06878 . 015Figure 5 . New model for AMPAR constitutive and activity-dependent recycling . ( A ) Conventional ‘single-synapse’ AMPAR recycling model . AMPARs exit post-synaptic densities ( PSDs ) by lateral diffusion and are endocytosed into clathrin-coated pits at sites near PSDs . After , endocytosed AMPARs are sorted into recycling endosomes ( REs ) , the same pathway used by TfR for recycling back to the cell surface , where AMPARs diffuse to and could be trapped in PSDs . An underlying assumption of the model is that AMPAR recycling is restricted to single spines and the endosomal membranes in the spine and dendrite neighboring the spine . Activity during LTP increases AMPAR exocytosis and transport from REs without altering its endocytosis thereby decreasing AMPAR levels in REs . ( B ) AMPAR recycling model with two AMPAR-recycling pathways under constitutive conditions . Based on evidence in this study , at least two different AMPAR-recycling pathways exist for AMPAR recycling . AMPARs largely recycle through the Arf6- and TC10-dependent recycling pathway , which originates at sites near the PSD and endocytose at clathrin- and dynamin-independent sites . This recycling pathway acts to move AMPARs from sites near the PSD to sites distant from the synapse such that AMPARs cannot return via membrane diffusion . Smaller numbers of AMPARs endocytose at sites distant from the spine via clathrin-coated pits using dynamin . Endocytosed AMPARs traffic into REs in the same pathway as TfRs . This recycling pathway acts to move AMPARs from distant sites to sites accessible to PSDs so that AMPARs can diffuse into PSDs to balance the loss via the Arf6-dependent recycling pathway . ( C ) AMPAR recycling after cLTP . Activity-dependent events during cLTP increase AMPAR recycling through the dynamin-dependent pathway , trafficking AMPARs from clathrin-coated pits distant from the stimulated synapse to sites accessible to the stimulated PSD . Endocytosed AMPARs are transported in the REs with the net effect of trafficking more AMPARs into the stimulated PSD . AMPAR recycling through the Arf6- and TC10-dependent recycling pathway continues unchanged after cLTP . ( D ) AMPAR recycling after cLTD . Similar to what is observed after cLTP , activity-dependent events during cLTP increase AMPAR recycling through the dynamin-dependent pathway except with the net effect of trafficking AMPARs out of PSDs and away from the stimulated spines . AMPAR endocytosis occurs at clathrin-coated pits near stimulated PSD and AMPARs transported in REs away from the cLTD spines and recycling to the cell surface at distant sites . The Arf6- and TC10-dependent recycling pathway , shown in ( B ) , is also unchanged during cLTD , and overall traffic AMPARs out of synaptic spines . DOI: http://dx . doi . org/10 . 7554/eLife . 06878 . 015 Our finding that AMPAR recycling through the dynamin/clathrin-dependent pathway increased during cLTP provides insights into the function of this pathway during AMPAR recycling . Previously , it was assumed that AMPAR recycling was largely confined to a single synaptic spine and the dendrite area nearby ( Figure 5A ) . Increases in synaptic AMPARs during NMDAR-dependent LTP were thought to increase AMPAR exocytosis without increasing AMPAR endocytosis causing a decrease in the level of AMPARs in REs ( Malinow and Malenka , 2002; Bredt and Nicoll , 2003; Collingridge et al . , 2004; Park et al . , 2004; Shepherd and Huganir , 2007 ) . However , we found that AMPAR levels in REs increased after LTP , and endocytosis increased in parallel with increased AMPAR levels at synapses during cLTP . Our results suggest that AMPAR recycling is not limited to trafficking AMPARs into and out of the same synaptic spine . Instead , we suggest that AMPAR recycling has the additional function of transporting AMPARs to sites distant from where they originate . Increased AMPAR endocytosis during cLTP occurs through the dynamin-dependent recycling pathway starting at clathrin-coated pits at sites distant from synapses undergoing cLTP ( Figure 5C ) . Consistent with this idea , AMPARs were not found in clathrin-coated pits at spines but in clathrin-coated pits well outside spines along the dendritic shafts ( Tao-Cheng et al . , 2011 ) . After endocytosis , we propose that AMPARs are trafficked in REs outside the synaptic spine region ( Figure 5C ) . Consistent with this trafficking role are studies demonstrating that different kinds of endosomes , including REs , travel long distances to new locations during recycling in dendrites and axons ( Yap and Winckler , 2012 ) . Furthermore , an EM study using three-dimensional reconstruction analysis found that in dendrites of rat , hippocampal neurons independent REs were not maintained at each spine . Instead , up to 20 spines shared a single recycling endosome ( Cooney et al . , 2002 ) . In our model ( Figure 5C ) , AMPAR exocytosis sites for the dynamin/clathrin-dependent pathway during cLTP are placed within the spine to deliver the AMPARs within the diffusible pool near the PSD of the synapse . In short , we propose that the role for the dynamin/clathrin-dependent recycling pathway during cLTP is to move AMPARs from sites distant from synaptic activation to sites near the activated synapse . During cLTD , we propose that AMPAR recycling via the dynamin/clathrin-dependent recycling pathway is also increased similar to what occurs during cLTP . Recycling AMPARs in TfR-containing REs are trafficked in and out of synapses and along dendrites except in the opposite direction ( Figure 5D ) . This is consistent with many studies that have found that AMPAR endocytosis increases via the dynamin/clathrin-dependent pathway . However , it has been assumed that during LTD , increased AMPAR endocytosis occurs without increased AMPAR exocytosis resulting in larger local endosomal stores of AMPARs . Instead , we suggest that during cLTD , AMPARs are transported away from synapses . In support of this idea , Tao-Cheng et al . , ( 2011 ) reported the appearance of AMPARs in clathrin-coated pits within synaptic spines , which was not observed during constitutive conditions or cLTP . This finding suggests that AMPARs near synapses are only removed locally via clathrin-coated pits during cLTD . In our model , AMPARs are trafficked during cLTD from clathrin-coated pits at activated spines via REs to distant sites . Overall , AMPARs flow out of synaptic spines during cLTD and into synaptic spines during cLTP . The function of the dynamin/clathrin-dependent recycling pathway is , thus , to traffic AMPARs from regions of low activity to regions of high activity and in this way the pathway underlies a Hebbian redistribution of AMPARs . Our finding that TC10WT expression had opposite effects on synaptic AMPAR currents , a measure of functional synaptic AMPARs , and the cell-surface levels of dendritic AMPARs suggests that the Arf6/TC10-dependent recycling pathway traffics AMPARs out of synaptic spines . In this study , we conclude that TC10 regulates AMPAR recycling through the Arf6/TC10-dependent pathway . The simplest explanation for why TC10WT expression acts to increase cell-surface AMPARs levels is that it increases exocytosis from Arf6-dependent pathway REs so that fewer AMPARs are in the Arf6 endosomes and more AMPARs are in the plasma membrane . The decrease in the levels of synaptic AMPARs with TC10WT expression can be explained if the Arf6-dependent recycling pathway acts to traffic AMPARs via endocytosis and exocytosis from synapses to distant plasma membrane sites . For this reason , in our model , we have placed the endocytosis site of the Arf6-dependent recycling pathway near the PSD and the exocytosis site outside of the spine ( Figure 5B ) . During constitutive AMPAR recycling , if AMPAR recycling via the Arf6-dependent recycling pathway acts to move synaptic AMPARs out of the spine and away from the synapse , then counteracting processes must be bringing other AMPARs back to the synapse in order to keep the levels of AMPAR constant . During cLTD ( Figure 5D ) , we suggest that the Arf6/TC10-dependent recycling pathway acts together with the dynamin/clathrin-dependent pathway to traffic AMPARs away from synaptic spines . If the Arf6/TC10-dependent recycling pathway traffics AMPARs out of the spines then blocking it would act to reduce LTD , as we observed with TC10DN expression ( Figure 4A–C ) . As explained above , the differential effects of TC10DN and TC10CA during LTD may arise from the differences in how they act during the Arf6/TC10-dependent recycling pathway . During constitutive conditions ( Figure 5B ) and during cLTP ( Figure 5C ) , the Arf6/TC10-dependent recycling pathway would act counter to the dynamin/clathrin-dependent pathway , which in our model would traffic AMPARs into synaptic spines under these conditions . This aspect of our model explains why blocking Arf6/TC10-dependent recycling pathway with the TC10DN expression may actually increase levels of LTP ( Figure 4A–C ) .
The following primary antibodies were used: anti-RFP ( MBL , rabbit , #PM005 ) , anti-GM130 ( BD Biosciences , mouse , 1:500 ) , anti-transferrin receptor ( Zymed , Thermo Scientific Pierce , Waltham , MA , #13-6800 , 1:300 ) , anti-synapsin ( Chemicon , EMD Millipore , Temecula , CA , mouse , 1:500 ) , anti-Bassoon ( Synaptic Systems , Goettingen , Germany , Guinea Pig , 1:300 ) , anti-GFP ( Sigma-Aldrich , St . Louis , MO , rabbit , 1:5000 ) , anti-GluR1 ( Millipore , EMD Millipore , Billerica , mouse; CalBioChem , Merck Millipore , Temecula , CA , rabbit ) . Dynasore was from Sigma ( D7693 ) ; Sulfo-NHS-SS-biotin was from Pierce . Human TC10 constructs were obtained from Dr . J . Marshall ( Brown University ) , and then subcloned into the pSP2 vector , containing a CMV promoter and Venus tag ( a brighter and more photostable YFP variant ) to generate fusion proteins . mCherry-GluA1 construct was obtained from Dr . C . Garner ( Stanford University ) . Rat TC10 , Cdc42 , and Rab11 genes were cloned from mRNA of 20 DIV cortical neuron cultures by RT-PCR with the following primers: TC10: FWD ( with EcoRI ) 5′-CCTTACATAGAATTCATGGCTCACGGGCCC-3′ , REV ( with BamHI ) 5′-GGCCCAGTGGATCCTCACGTAATCAAACAACAGTTTATAC-3′; Cdc42: FWD ( with EcoRI ) 5′-CGTTACTAAGAATTCATGGGCACCCGCGAC-3′ , REV ( with BamHI ) 5′-GGCCTCGACGGATCCTTAGATGTTCTGACAGCACTGC-3′; Rab11: FWD ( with EcoRI ) 5′-CGTTACTAAGAATTCATGGGCACCCGCGAC-3′ , REV ( with BamHI ) 5′-GGCCTCGACGGATCCTAAGATGTTCTGACAGCACTGC-3′ . The genes were then inserted into a customized vector ( originated from pEYFP from Clontech , Takara , Mountain View , CA ) with a CMV promoter and mCherry tag . The lentiviral vector , FUGW , and the helper plasmids , Δ8 . 9 and VSVg , were obtained from Dr . C . Garner ( Stanford University ) and were used to clone all genes listed above for production of lentiviruses . The 3 candidate RNAi constructs for TC10 were purchased from Sigma–Aldrich . Rat E18 hippocampal neurons were cultured on poly-L-lysine treated coverslips in Neurobasal medium supplemented with NS21 and GlutaMAX . At 14–17 DIV , neurons were transfected by Lipofectamine 2000 with serum-free Neurobasal medium . The amount of cDNA transfected ranges from 1 to 2 μg per coverslip ( d = 12 mm ) as needed . Hippocampal cultures were prepared using Neurobasal medium , 2% ( vol/vol ) B27 , and GlutaMAX . Briefly , hippocampi from embryonic ( E18–19 ) Sprague Dawley rats of either sex were dissected , dissociated in 0 . 05% trypsin ( vol/vol , Life Technologies , Thermo Fisher Scientific , Grand Island , NY ) , and cells were plated at a density of ∼4 × 105 cells/mL on poly-L-lysine-coated 12-mm coverslips . Coverslips were maintained in Neurobasal medium containing B27 and GlutaMAX ( all from Life Technologies ) . Neuronal cultures were transfected at 14–17 DIV with the Lipofectamine 2000 transfection reagent ( Life Technologies ) according to manufacturer's recommendations , with the exception that 1–2 . 5 μg of each cDNA in 62 . 5 μl of Neurobasal media and 2 . 0 μl of Lipofectamine 2000 in 62 . 5 μl of Neurobasal media were mixed and added to coverslips in 12-well plates . Freshly thawed HEK cells were cultured in DMEM + 10% fetal bovine serum ( FBS ) medium . Lentiviral plasmid and the helper plasmids were transfected using Ca3 ( PO4 ) 2 method . ∼50 hr after transfection , supernatant containing the virus particles was collected . After a brief centrifugation to remove cell debris , the virus solution was then mixed with PEG 8000 to incubate at 4°C overnight , followed by centrifugation at 4000 rpm for 30 min at 4°C . The virus pellect was then resuspended in cold phosphate buffered saline ( PBS ) . At 0–1 DIV , neurons were infected with high-titer lentiviruses , and 2 days post-infection , culture medium was changed with fresh Neurobasal , B27 , and GlutaMAX . For surface labeling , primary antibody was added into the culture medium and incubated at room temperature for 30 min , or for 20 min at 37°C . Cells were then washed with PBS and fixed with 4% PFA/4% sucrose/PBS for 10 min , then incubated with blocking solution ( 2% glycine , 1% BSA , 0 . 2% gelatin , 0 . 5M NH4Cl , PBS ) for 1 hr and secondary antibody in blocking solution at room temperature for 1 hr . For internalized staining ( antibody-feeding assay ) , primary antibody was added into the culture medium and incubated at 37°C for 40 min , then cells were washed with acid wash buffer ( 0 . 5M NaCl , 0 . 5% acetic acid , pH2 ) for 30 s , and fixed . Cells were then incubated with blocking solution and secondary antibody . For permeabilized staining , after fixation , the cells were permeablized in 0 . 1% Triton/PBS for 5–10 min , blocked and stained with primary and secondary antibodies . The stained coverslips were mounted to glass slides with ProLong Gold ( Life Technologies ) mounting media and left in dark to harden overnight . For internalized staining of Tf-Alexa 647 , Tf-Alexa 647 was added into the culture medium and incubated at 37°C for 1 hr . Then , cells were washed with acid wash buffer and fixed for imaging . All images were taken on either an Olympus DSU or Marianas Yokogawa type spinning disk confocal microscope with back-thinned EMCCD camera . Z-stack slices were taken with 0 . 2-μm step size , with 5–10 slices for each cell . For the surface staining experiments , z-plane limits for acquisition were determined by the surface staining on dendrites; for the assay to measure changes in Golgi/endosomal/synaptic localization of GluA1 , the z-plane limits were set according to organelle marker staining ( GM130 , transferrin receptor or Arf6 , and synapsin , respectively ) . To induce chemical LTP , DIV17 neurons were washed in Mg2+ free buffer ( in mM: NaCl 150 , CaCl2 2 , KCl 5 , HEPES 10 , glucose 30 , strychnine 0 . 001 , bicuculline 0 . 02 ) 3 times , and incubated in glycine buffer ( Mg2+ free buffer with 0 . 2 mM glycine ) at 37°C for 15 min . Then , Mg2+ buffer ( Mg2+ free buffer with 2 mM MgCl2 ) was added to block NMDARs and cells were incubated at 37°C for 30 min before live surface labeling with anti-RFP . To induce chemical LTD , 14 DIV neurons were washed in Mg2+ free buffer 3 times , and incubated in NMDA buffer ( Mg2+ free buffer with 0 . 02 mM NMDA ) at 37°C for 5 min . Then , Mg2+ buffer was added and cells were incubated at 37°C for 1 hr before live surface labeling . Image quantification was performed using NIH ImageJ software . To calculate the surface/total ratio for exogenously expressed GluA1 , all images of the same channel were first background subtracted using the same averaged value , which was measured manually across images ( with variation <0 . 5% ) . The sum of pixel intensity for the z-stack was calculated using the ‘sum of slices’ and ‘histogram’ functions , excluding zero-intensity pixels . The surface/total ratio was then calculated as the ratio of the intensity of surface channel to total channel . Surface expression of endogenous GluA1 was quantified using ‘sum of slice’ z-projections of images . Each field was background subtracted , and mean intensities were normalized to YFP control . To measure the Golgi localization of GluA1 , the images were background subtracted with the same method described above , and then , the GM130 channel image was thresholded and transformed into a binary mask used to measure the pixel intensity of GluA1 in each slice of the stack . The average intensity of the processed z-stack image was then measured by selecting the cell of interest manually ( if more than one cell was present in the image ) and applying the ‘measure’ function . To measure the degree of co-localization of GluA1 with TfR-positive endosomes , images were assigned a random number and analyzed blindly . Analysis of the co-localization of endogenous GluA1 with TfR was carried out by background subtracting and thresholding image fields so that only puncta that were twofold greater than background were selected . Co-localizing puncta were evaluated using the Analyze Particles function in ImageJ . To measure the degree of overlap between endogenous GluA1 and Arf6 sub-compartments , we compared fluorescence signals above background in both channels along manually outlined segments of dendrites . Pearson's correlation coefficients were generated for background-subtracted image pairs using the Intensity Correlation Analysis plugin in ImageJ . A similar approach was used to measure the degree of overlap between TC10 and Arf6 . Thresholded GluA1 punctal size and density in dendritic shafts was analyzed using the Analyze Particle function in ImageJ . Statistical comparisons were made using two-tailed Student's t tests or ANOVA/Tukey post hoc analysis as indicated . Statistical graphs were generated with Graphpad Prism or StatPlus software . Dual whole-cell recordings were performed at DIV12–15 on primary dissociated hippocampal cultures transfected with Venus-TC10DN or Venus-TC10CA . Neurons were bathed in carbogen ( 95% O2 , 5% CO2 ) bubbled ACSF ( in mM: 120 NaCl , 3 KCl , 2 CaCl2 , 1 . 25 NaH2PO4 , 2 MgSO4 , 20 D- ( + ) -glucose , 26 NaHCO3 ) . The internal solution consisted of ( in mM ) : 120 K gluconate or Cs gluconate , 40 HEPES , 5 MgCl2 , 2 NaATP , 0 . 3 NaGTP ) . Recordings were performed at room temperature ( 21°C ) . Hippocampal cultures were mounted on an Olympus microscope ( BX51WI ) and visualized using differential interference microscopy . Transfected neurons were visualized via excitation at 530–550 nm . Electrode resistance was between 5 and 10 MΩ . Patch-clamp recordings were obtained using a MultiClamp 700B Commander ( Molecular Devices , Sunnyvale , CA ) . Data acquisition and analysis were performed using AxoGraph X ( AxoGraph Scientific , Sydney , Australia ) and pCLAMP 9 ( Molecular Devices ) software . Events were sampled at 10 kHz and low-pass filtered at 2 kHz . Series resistance ( Rs ) was monitored throughout all experiments , and results were not included if significant variation ( >20% ) occurred during any experiment . Action potentials were induced in presynaptic neurons by a 20-ms current injection of 20–100 pA . AMPAR EPSCs in response to presynaptic action potentials were collected at 0 . 1 or 0 . 2 Hz . Statistical significance of changes in AMPAR EPSC amplitudes was tested using Student's t test with a level of significance set at p < 0 . 05 . Cortical neurons were pretreated with 80 μM dynasore or 1% DMSO for 40 min at 37°C . Cells were then washed , and incubated with 0 . 5 mg/mL Sulfo-NHS-SS-Biotin at 4°C for 30 min , and excessive Sulfo-NHS-SS-Biotin was washed off . Cells were then incubated at 37°C for 1 hr in the presence of 80 μM dynasore or 1% DMSO . Then Sulfo-NHS-SS-Biotin on the cell surface was cleaved with glutathione . Cells were then harvested and lysed; biotinylated proteins were pull down using streptavidin–sepharose beads and were analyzed on Western blot . | Cells called neurons transmit information around the brain in the form of electrical signals . At a junction between two neurons—called a synapse—an electrical signal triggers the release of small molecules called neurotransmitters . These molecules travel across the gap between the two neurons and trigger a new electrical signal in the second neuron . Memories can be stored in synapses: high levels of activity can ‘strengthen’ the synapse , which increases the transfer of information between the neurons . In many synapses , a molecule called glutamate is the neurotransmitter . Proteins called AMPARs , which are found on the surface of the neuron , can detect glutamate and transmit the signal along the second neuron . The strength of synapses is controlled by changes in AMPAR levels through ‘recycling’ , where AMPAR proteins are removed from synapses , internalized and later returned to synapses . It was thought that AMPARs are recycled via just one pathway at synapses . However , the amount of recycling is much higher when the synapses are active and it is not clear how this works . Now , Zheng et al . have used fluorescent tags to track the recycling of AMPARs in synapses from rats under a microscope . The experiments show that when the synapses are not active , most AMPARs are recycled via a pathway marked by a protein called Arf6 . However , when the synapses are active , most AMPAR is recycled via a different route marked by so-called ‘transferrin receptor’ proteins . The experiments also reveal that a protein called TC10 is involved in recycling AMPARs alongside Arf6 , but is not required for recycling when the synapses are active and being strengthened . Unexpectedly , AMPAR internalization—via the process involving transferrin receptors—increases during synapse strengthening . This suggests that some of the extra AMPAR proteins sent to the membrane have come from other parts of the neuron away from the synapse . Zheng et al . 's findings provide evidence that AMPARs are recycled through different routes depending on the activity of the synapse . The next challenge will be to directly test whether AMPARs are transported from other parts of the neuron to the strengthened synapse and to understand how this works . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2015 | Synaptic activity regulates AMPA receptor trafficking through different recycling pathways |
The mechanoreceptive sensory hair cells in the inner ear are selectively vulnerable to numerous genetic and environmental insults . In mammals , hair cells lack regenerative capacity , and their death leads to permanent hearing loss and vestibular dysfunction . Their paucity and inaccessibility has limited the search for otoprotective and regenerative strategies . Growing hair cells in vitro would provide a route to overcome this experimental bottleneck . We report a combination of four transcription factors ( Six1 , Atoh1 , Pou4f3 , and Gfi1 ) that can convert mouse embryonic fibroblasts , adult tail-tip fibroblasts and postnatal supporting cells into induced hair cell-like cells ( iHCs ) . iHCs exhibit hair cell-like morphology , transcriptomic and epigenetic profiles , electrophysiological properties , mechanosensory channel expression , and vulnerability to ototoxin in a high-content phenotypic screening system . Thus , direct reprogramming provides a platform to identify causes and treatments for hair cell loss , and may help identify future gene therapy approaches for restoring hearing .
Hearing loss is the most common sensory deficit with estimates of around 466 million people affected worldwide ( WHO , 2019 ) . Loss of sensory hair cells of the inner ear is the primary cause of sensorineural hearing loss ( Bohne and Harding , 2000; Hinojosa et al . , 2001; Géléoc and Holt , 2014; Wong and Ryan , 2015 ) . The highly structured sensory epithelium of the inner ear , known as the organ of Corti , develops from a post-mitotic , pro-sensory domain established in the developing cochlear duct between embryonic days E12 . 5 and E14 . 5 in mice ( Ruben and Sidman , 1967; Lowenheim et al . , 1999; Chen and Segil , 1999; Matei et al . , 2005; Lee et al . , 2006 ) . These post-mitotic cells are the progenitors for sensory hair cells and their adjacent supporting cells ( Fekete et al . , 1998; Kelley , 2006; Driver et al . , 2013 ) . Sensory hair cells function as the essential mechanoreceptors that convert sound vibrations into electrical signals , which are then transmitted to the brain via the spiral ganglion neurons that innervate the hair cells ( Géléoc and Holt , 2003 ) . Sensory hair cells are located in both the auditory and vestibular portions of the inner ear ( Figure 1A ) . The hair cells within the organ of Corti are precisely arranged into one row of inner hair cells and three rows of outer hair cells , interdigitating with a variety of supporting cells; inner border , inner phalangeal , pillar cells , Deiters’ cells and Hensen’s cells ( Figure 1A ) . Hair cells are susceptible to degeneration by a variety of genetic mutations and environmental stressors , such as exposure to loud noise , ototoxic drugs including cancer chemotherapy and aminoglycoside antibiotics , aging and over 200 known syndromic and non-syndromic genetic loci conferring predispositions to hearing loss ( Matsui et al . , 2004; Cheng et al . , 2005; Bodmer , 2008; Langer et al . , 2013; Atkinson et al . , 2015; Wong and Ryan , 2015; Vaisbuch and Santa Maria , 2018 ) . In mammals hearing and balance are dependent on the maintenance of hair cells present at birth ( Groves , 2010; Géléoc and Holt , 2014 ) , since hair cells do not spontaneously regenerate ( Roberson and Rubel , 1994; Chardin and Romand , 1995; Forge et al . , 1998 ) , and so their death leads to lifelong hearing loss and balance disorders . In contrast , non-mammalian species , such as birds and reptiles , are able to spontaneously regenerate lost hair cells from existing supporting cells , leading to full functional recovery ( Corwin and Cotanche , 1988; Ryals and Rubel , 1988; Stone and Cotanche , 2007; Brignull et al . , 2009 ) . Transcription factors regulate the temporal and spatial patterns of gene expression within the cells of complex tissues , establishing cell fate , and ultimately determining their morphological and functional properties ( Lemon and Tjian , 2000; Levine and Tjian , 2003; Zhang et al . , 2004 ) . Within the inner ear , expression of Atoh1 , a bHLH class transcription factor ( Lo et al . , 1991; Ross et al . , 2003 ) is both necessary and sufficient for the induction of sensory hair cells in the embryonic and neonatal cochlea , and ultimately plays an integral role in initiating the hair cell gene expression program ( Bermingham et al . , 1999; Zheng and Gao , 2000; Woods et al . , 2004; Kelly et al . , 2012; Chonko et al . , 2013; Cai et al . , 2013; Ryan et al . , 2015; Scheffer et al . , 2015; Stojanova et al . , 2016; Costa et al . , 2017 ) . However , previous studies have shown that Atoh1 expression alone is not sufficient to induce hair cell differentiation in somatic cells ( Izumikawa et al . , 2008; Costa et al . , 2015; Abdolazimi et al . , 2016 ) , or mature supporting cells of the organ of Corti ( Kelly et al . , 2012; Liu et al . , 2012b ) . The paucity and inaccessibility of primary inner ear hair cells have limited the identification of effective otoprotective and regenerative strategies . Recent studies have demonstrated the in vitro formation of hair cells from murine pluripotent stem cells and human embryonic stem cells by directed differentiation ( Oshima et al . , 2010; Koehler et al . , 2013; Li et al . , 2003; Ronaghi et al . , 2014 ) , or in a combination of directed differentiation to an ectodermal , non-neural , placodal cell type , followed by transcription factor induction to a hair cell-like state ( Costa et al . , 2015 ) . However , these elegant approaches require three-dimensional culture conditions that complicate high-throughput studies , for instance screening for otoprotectants . In contrast to morphogen-based directed differentiation of pluripotent stem cells , transcription factor ( TF ) -mediated lineage conversion of somatic cells enables the rapid production of neurons and other cell types in microtiter plates with ≥96 wells , allowing the reproducibility and homogeneity required for high-throughput phenotypic screening ( Xu et al . , 2015; Babos et al . , 2019 ) . Thus , the identification of a transcription factor cocktail that can convert somatic cells into sensory hair cells could enable screening for new otoprotective targets . Moreover , delivery of such a cocktail in vivo would enable regenerative medicine strategies for hair cell replacement in situ , which have thus far been ineffective ( Izumikawa et al . , 2005; Richardson and Atkinson , 2015; Roccio et al . , 2015 ) . To this end , we have identified a cocktail of four transcription factors , Six1 , Atoh1 , Pou4f3 , and Gfi1 ( SAPG ) , capable of converting mouse embryonic fibroblasts , adult tail tip fibroblasts , and postnatal mouse supporting cells into induced hair cells ( iHCs ) . iHCs are highly similar to primary hair cells in terms of global gene expression and chromatin accessibility profiles , morphological features , and electrophysiological properties . In addition , we established a robotic imaging platform with automated analysis to track iHC survival and show that like primary hair cells , iHCs are selectively sensitive to gentamicin toxicity . These findings show that iHCs make a valuable in vitro model to study hair cell regeneration , maturation , function and susceptibility to ototoxins .
To identify a group of TFs needed to convert somatic cells into induced hair cells , we analyzed the transcriptome of postnatal day 1 ( P1 ) cochlear hair cells that had been FACS-purified from a transgenic mouse expressing GFP in nascent hair cells under the control of an Atoh1 3’ enhancer ( Lumpkin et al . , 2003 ) . We compared the primary P1 cochlear hair cell transcriptome to a reference transcriptome of the FACS-purified GFP-negative cells from the same organ of Corti preparations ( Figure 1—figure supplement 1A ) . We identified 16 candidate TFs that were highly enriched in P1 hair cells ( Atoh1-nGFP+ ) , some of which are known to have essential roles in hair cell development ( Li et al . , 2003; Wallis et al . , 2003; Qian et al . , 2006; Hume et al . , 2007; Ahmed et al . , 2012; Chonko et al . , 2013; Liu et al . , 2014a; Cai et al . , 2015; Scheffer et al . , 2015 ) . Using retroviral delivery , we transduced the TFs into mouse embryonic fibroblasts ( MEFs ) from the Atoh1-nGFP reporter mouse ( Figure 1B ) . MEFs transduced with a control virus ( dsRed ) did not express the Atoh1-nGFP transgene after 14 days ( Figure 1—figure supplement 1B ) . In contrast , overexpression of all 16 TFs led to Atoh1-nGFP activation in 1 . 7% ( ± 0 . 3 ) of MEFs at 14 days post infection ( Figure 1—figure supplement 1C ) . Reprogramming efficiency was calculated as a percent of Atoh1-nGFP-positive MEFs out of the starting MEF number ( 5000 cells per well ) . This result indicated that within this initial group were individual transcription factors , or combinations thereof , able to reprogram MEFs to a hair cell-like state . The low level of reprogramming efficiency is expected when large numbers of factors are infected simultaneously , since only a subset of factors is expected to infect any given cell ( Phan and Wodarz , 2015; Mistry et al . , 2018 ) , and since using large numbers of factors , and/or virus , is likely to challenge cellular transcription/translational machinery , thus further reducing efficiency ( Babos et al . , 2019 ) . To identify the TFs critical for the Atoh1-nGFP reporter activation in MEFs , we tested the efficiency of Atoh1 and each of the other 15 TFs separately ( Figure 1—figure supplement 1C ) . We observed that Atoh1 alone activated the Atoh1-nGFP reporter in 5 . 8% ( ± 1 . 5 ) of starting MEFs , while Pou4f3-alone only did so in 0 . 15% ( ± 0 . 03 ) of the starting MEFs ( Figure 1—figure supplement 1C ) . None of the other 14 factors alone activated the Atoh1-nGFP reporter . We then tested the reprogramming efficiency of Atoh1 in combination with each of the other 15 TFs ( Figure 1—figure supplement 1C ) . The most significant reporter activation came from a combination of Atoh1 and Pou4f3 , which provided 17 . 5% ( ± 4 . 4 ) reprogramming efficiency ( Figure 1—figure supplement 1D ) . We then tested the addition of each remaining individual factor to the combination of Atoh1 and Pou4f3 ( AP ) ( Figure 1—figure supplement 1E ) . Gfi1 combined with AP ( APG ) increased the reporter activation to 26 . 9% ( ± 5 . 6 ) reprogramming efficiency ( Figure 1—figure supplement 1E ) . A subsequent round of addition of individual TFs to this three-factor combination showed that the addition of Six1 to Atoh1 , Pou4f3 , and Gfi1 ( SAPG ) further increased the reporter activation to reach 35 . 2% ( ± 1 . 8 ) reprogramming efficiency ( Figure 1—figure supplement 1F ) . Addition of the remaining individual factors to the cocktail of SAPG did not increase reprogramming efficiency ( Figure 1—figure supplement 1G ) . Since Atoh1 is expressed in other cell types and lineages ( Klisch et al . , 2011; Kim et al . , 2014; Ostrowski et al . , 2015 ) , we performed immunostaining for MyosinVIIa and Parvalbumin , two additional markers that are more specific to a hair cell fate ( Eybalin and Ripoll , 1990; Demêmes et al . , 1993; Pak and Slepecky , 1995; Hasson et al . , 1997; Sahly et al . , 1997; Richardson et al . , 1997; Richardson et al . , 1999; Boëda , 2002 ) . The majority of SAPG-transduced cells that activated Atoh1-nGFP also expressed MyosinVIIa and Parvalbumin ( 78 . 4% ± 1 . 9 ) ( Figure 1C , D ) . Overall , SAPG transduction activated Atoh1-nGFP with a 35% efficiency , and nearly 80% of all Atoh1-nGFP positive cells also expressed MyosinVIIa and Parvalbumin ( Figure 1D ) . In contrast , only 50% of the Atoh1-nGFP+ cells generated by AP or APG expressed MyosinVIIa and Parvalbumin , indicating that most Atoh1-nGFP+ cells generated from these alternative cocktails were not hair cell-like ( Figure 1D ) . Our results support the importance of Six1 , Atoh1 , Pou4f3 and Gfi1 in direct reprogramming of somatic cells to a hair cell-like state , with high efficiency , purity , and reproducibility . Direct lineage reprogramming relies on the forced expression of transcription factors to induce a molecular rewiring of the transcriptional programs that characterize specialized cells ( Takahashi and Yamanaka , 2006; Takahashi et al . , 2007 ) . This involves both upregulation of the target cell-specific genes , in this case of primary cochlear hair cells , and downregulation of the starting cell-specific genes , in this case of MEFs . To assay the extent to which induced hair cells replicate the mouse primary cochlear hair cell gene expression program , we performed RNA-sequencing on FACS-purified Atoh1-nGFP+ cells generated by overexpression of Six1 , Atoh1 , Pou4f3 , and Gfi1 ( SAPG ) at 14 days post infection ( dpi ) ( hereafter referred to as iHCs ) . We compared the gene expression of the iHCs to FACS-purified Atoh1-nGFP+ primary cochlear hair cells at postnatal day 1 ( P1; hereafter referred to as P1 HCs ) , and MEFs infected with a control retrovirus expressing a fluorescent protein ( dsRed; hereafter referred to as MEFs ) . We categorized the gene expression in iHCs as either ‘successfully reprogrammed’ , ‘not-reprogrammed’ or ‘inappropriately expressed’ and divided the categories into those genes that are normally expressed in P1 HCs , but not in MEFs ( P1 HC genes , black bar ) , and those that are normally expressed in MEFs , but not in primary hair cells ( MEF genes , black bar ) ( Figure 2A ) . From this analysis we determined that the iHCs had transcriptionally activated a hair cell-like signature by successfully upregulating 64% of P1 HC genes , while simultaneously becoming distinct from the starting MEF population by successfully downregulating 69% of MEF genes ( Figure 2A ) . These percentages are comparable to those achieved in the TF-induced direct conversion of MEFs into spinal motor neurons , as well as those attained in MEF-to-cardiomyocyte and hepatocyte-to-neuron direct conversion ( Ieda et al . , 2010; Marro et al . , 2011; Gopalakrishnan et al . , 2017; Ichida et al . , 2018 ) . These results suggest that iHCs largely resemble primary P1 cochlear hair cells at the transcriptional level . Nonetheless , a number of genes did not respond to the SAPG group of transcription factors used for reprogramming . Of the 1506 genes expressed in P1 HCs , but not MEFs , 36% were not successfully upregulated in the iHCs , and 118 genes were inappropriately upregulated . Of the 939 genes that are expressed in MEFs , but not in P1 HCs , and thus need to be downregulated during reprogramming , 31% failed to downregulate , and 142 genes were inappropriately downregulated . PCA analysis of bulk RNA-seq data from MEFs , compared to either primary P1 HCs or iHCs show the relative difference between these populations ( Figure 2—figure supplement 1B ) . Gene Ontology ( GO ) analysis showed that the genes successfully upregulated during reprogramming were significantly enriched for ‘sensory perception of sound’ and ‘detection of mechanical stimulus’ , which revealed as three clusters of genes ( Figure 2B ) . The first cluster was enriched for development-related GO terms such as ‘inner ear receptor cell development’ , ‘mechanoreceptor differentiation’ , and ‘hair cell differentiation’ ( Figure 2B ) . The second cluster was enriched for stereocilia-related GO terms such as ‘plasma membrane bound cell projection assembly’ , ‘cilium organization’ and ‘cilium movement’ ( Figure 2B ) . The third cluster was enriched for synaptic signaling GO terms such as ‘establishment of synaptic vesicle localization’ , ‘synaptic vesicle cycle’ and ‘neurotransmitter secretion’ ( Figure 2B ) . These three clusters of GO terms were used to generate cluster-specific gene sets driving the GO designation ( Figure 2C , Figure 2—source data 1 ) . Expression levels of each GO cluster-specific gene set , in each cell type , were plotted to visualize the statistically significant iHC divergence from MEFs , and iHC convergence towards a P1 HC expression profile ( Figure 2C ) . This analysis also revealed a significant difference in the level of gene expression ( p<0 . 05 ) between iHCs and P1 HCs ( Figure 2C ) . This difference may be explained by the maturity level of the iHCs as well as the presence of the residual MEF transcriptional profile . However , further investigation of the Rlog count values for several key genes in each GO cluster demonstrated that the iHCs efficiently upregulated important hair cell genes including Whirlin ( Whrn ) involved in cell polarity , Cadherin23 ( Cdh23 ) and Espin ( Espn ) important for stereocilia organization and functionality , as well as Bassoon ( Bsn ) and Otoferlin ( Otof ) required for synaptic scaffolding and synaptic vesicle signaling ( Figure 2D ) . After looking at the expression of key hair cell genes , we determined that the iHCs had also activated all of the initial transcription factors included in the set of 16 candidate factors , with the exception of Zfp503 ( Figure 2—figure supplement 1D ) . Together , the RNA sequencing results indicate that the iHCs generated by Six1 , Atoh1 , Pou4f3 , and Gfi1 ( SAPG ) overexpression are capable of repressing most of the initial MEF gene signature , while simultaneously adopting a gene expression signature similar to primary P1 HCs . Interestingly , the reprogramming transcription factors ( SAPG ) , are normally expressed in both cochlear and utricular differentiating hair cells . We compared postnatal day one cochlear hair cells ( P1 cHCs ) and utricular hair cells ( P1 uHCs ) to iHCs . PCA analysis shows that iHCs have not become more similar to one of the two primary hair cell populations ( Figure 2—figure supplement 1A ) . The PCA also demonstrates that at this early developmental time point the P1 cHCs and P1 uHCs are themselves immature , and have similar transcriptional profiles . Nonetheless , the expression profile of iHCs has drastically shifted away from the MEF signature and towards an immature hair cell-like signature . Expression of Atoh1 is necessary and sufficient for hair cell differentiation in the context of the inner ear primordium ( Bermingham et al . , 1999; Chen et al . , 2002; Woods et al . , 2004; Chonko et al . , 2013; Cai et al . , 2013 ) , however several other lineages including cerebellar granule cell progenitors ( Klisch et al . , 2011 ) , Merkel cells ( Ostrowski et al . , 2015 ) , and the secretory cell lineage of the gut ( Kim et al . , 2014 ) rely on Atoh1 expression for differentiation . To characterize the specificity of our reprogramming to the hair cell-like state , we analyzed RNA sequencing data from FACS-purified iHCs relative to other Atoh1-dependent lineages including cerebellar granule precursors ( CGP ) , secretory cells of the gut ( GUT ) , and Merkel cells ( MC ) . Principle component analysis ( PCA ) indicated that iHCs were more similar to primary P1 cochlear hair cells ( P1 HC ) than to either cerebellar granule cell precursors ( CGP ) ( Figure 3A ) , secretory cells of the gut ( GUT ) ( Figure 3B ) , or Merkel cells ( MC ) ( Figure 3C ) , showing that they established a hair cell-specific transcriptional program and have not adopted the transcriptional profile of other Atoh1-dependent lineages . As an additional test , we compared iHCs to the other Atoh1-dependent lineages using Gene Set Enrichment Analysis ( GSEA ) ( Subramanian et al . , 2005 ) . By comparing the transcriptomes in MEFs , P1 hair cells ( HC ) , P1 Cerebellar granule precursors ( CGP ) , adult gut secretory cells ( GUT ) , and P1 Merkel cells ( MC ) , we defined groups of genes as part of a specific signature for each cell type ( Figure 3—source data 1 ) . The GSEA program identified gene-lists exclusive to each cell type; these gene-lists included only genes which were not expressed in any two cell types . We calculated Normalized Enrichment Scores ( NES ) ( Subramanian et al . , 2005 ) for each cell type in comparison to iHCs . The largest NES was for the comparison of iHCs to P1 hair cells ( HC ) , indicating an enrichment for the HC gene signature , while showing lower enrichment scores , and even negative enrichment scores , for the other cell type comparisons ( Figure 3D ) . Thus , reprogramming with SAPG establishes a hair cell-like transcriptional program without adopting the transcriptional profiles of other Atoh1-dependent lineages . Chromatin structure controls the accessibility of genes for either activation or repression in response to developmental and environmental signaling ( Volpi et al . , 2000; Kozubek et al . , 2002; Goetze et al . , 2007; Buenrostro et al . , 2015b; Chen et al . , 2016; Sijacic et al . , 2018 ) , and as such , is an important regulator of cell type-specific gene expression . We used an Assay of Transposase Accessible Chromatin ( ATAC ) sequencing ( Buenrostro et al . , 2015a; Chen et al . , 2016 ) to analyze the regions of open/accessible chromatin in MEFs ( MEF peaks ) , primary P1 hair cells ( P1 HC peaks ) , and iHCs ( Figure 4A ) . As in our analysis of gene expression ( Figure 2 ) , we characterized the open chromatin regions into those that are present in primary P1 HCs , but not in MEFs ( P1 HC peaks , black bar ) , and those that are open in MEFs , but not primary HCs ( MEF peaks , black bar ) , as analyzed by ATAC-seq accessibility ( Figure 4A ) . We defined these groups , as in Figure 2 , as either ‘successfully reprogrammed’ , ‘not reprogrammed’ , and ‘inappropriately opened/closed’ ( i . e . not matching either P1 HC peaks or MEF peaks ) . The iHCs show robust opening of de novo distal element regions of the chromatin that are open in P1 HCs , as well as large-scale chromatin closing in regions of the genome that were accessible in the starting MEF population . The hair cell-appropriate changes in chromatin accessibility of iHCs are also accompanied by a proportion of inappropriate opening or closing of chromatin regions . Of the 13 , 390 peaks present uniquely in P1 HCs , 73% were successfully opened during reprogramming , while 27% were not opened during reprogramming , and an additional 18 , 084 peaks opened inappropriately in iHCs ( Figure 4A ) . Conversely , of the 26 , 847 peaks unique to MEFs , 84% of peaks were successfully closed during reprogramming , 16% were not closed during reprogramming , and an additional 833 peaks were inappropriately closed during reprogramming ( Figure 4A ) . Since most distal accessible elements are not active enhancers in a given cell type ( Heintzman et al . , 2007 ) , we analyzed the H3K27ac-state of the distal elements present in P1 HCs , a marker of active enhancers ( Creyghton et al . , 2010; Figure 4B ) . These results show that most of the enhancers identified in P1 HCs are opened in iHCs . Global enhancer targets have not been analyzed in these cell types due to small numbers , so we arbitrarily assigned putative gene targets to each P1 HC enhancer by identifying the closest transcriptional start site . This is expected to identify 27–47% of genuine targets , based on chromosome conformation capture experiments performed in other cell types ( Sanyal et al . , 2012 ) . Based on our RNA-seq data , we found that the genes defined as putative targets of P1 HC-specific enhancers had significantly higher expression in both P1 HCs and iHCs compared to MEFs ( Figure 4C ) . To visualize the ATAC-seq and ChIP-seq data at specific loci we used the Integrative Genomics Viewer ( IGV ) ( Robinson et al . , 2011 ) . We chose four known hair cell loci , Pou4f3 , Mreg , Rasd2 , and Barhl1 , that exemplify the changes in chromatin structure at H3K27ac-defined enhancers between P1 HCs , iHCs and MEFs ( Figure 4D ) . These results indicate that robust and hair cell-appropriate global changes in chromatin accessibility accompany the large shift in the transcriptional profile of iHCs . We have used mouse embryonic fibroblasts ( MEFs ) as a starting cell type for our reprogramming efforts in the experiments described thus far . However , MEFs are an embryonic and relatively heterogeneous cell population ( Singhal et al . , 2016 ) . To test the ability of the SAPG transcription factors to reprogram a mature somatic cell from a heterologous lineage , we chose as proof-of-principle , to reprogram Atoh1-nGFP transgenic adult tail tip fibroblasts ( TTFs ) . We virally transduced TTFs with Six1 , Atoh1 , Pou4f3 , and Gfi1 ( Figure 5A ) . While TTFs are known to be infected by retrovirus at a lower efficiency than MEFs ( Liu et al . , 2011; Lalit et al . , 2016 ) , SAPG activated the Atoh1-nGFP reporter in adult tail tip fibroblasts ( Figure 5A ) . In addition , SAPG catalyzed the expression of MyosinVIIa and Parvalbumin , indicating that as in MEFs , these transcription factors can convert adult tail tip fibroblasts into Atoh1-nGFP+/MyosinVIIa+/Parvalbumin+ iHCs ( Figure 5A , B ) . Supporting cells are an attractive target for gene therapy approaches to hair cell regeneration , due to their known role in regeneration in non-mammalian vertebrates ( Corwin and Cotanche , 1988; Ryals and Rubel , 1988; Stone and Cotanche , 2007; Brignull et al . , 2009 ) , their having a common progenitor with hair cells ( Fekete et al . , 1998; Kelley , 2006; Driver et al . , 2013 ) , and their survival in long-deafened mice ( Oesterle and Campbell , 2009 ) . Although hair cells do not regenerate in the mature mammalian cochlea , perinatal supporting cells have been shown to have a transient ability to directly transdifferentiate into hair cells in response to Atoh1 ( Kelly et al . , 2012; Liu et al . , 2012b ) , or loss of Notch-mediated lateral inhibition ( Mizutari et al . , 2013; Maass et al . , 2015 ) , however this potential is lost at very early postnatal stages ( White et al . , 2006; Takebayashi et al . , 2007; Doetzlhofer et al . , 2009; Liu et al . , 2012a; Cox et al . , 2014; Bramhall et al . , 2014 ) . One plausible route to the in vivo regeneration of hair cells in the organ of Corti would be the conversion of mature supporting cells into hair cells in long deafened individuals . Atoh1 alone can convert perinatal supporting cells into hair cells ( Kelly et al . , 2012; Liu et al . , 2012b; Yang et al . , 2013 ) , but transdifferentiation potential decreases rapidly thereafter , such that by two weeks of age , neither Atoh1 expression , nor induction of transdifferentiation by Notch-inhibition , can induce the transdifferentiation of supporting cells to a hair cell fate ( Maass et al . , 2015; Jiang et al . , 2016 ) . To determine if Six1 , Atoh1 , Pou4f3 , and Gfi1 ( SAPG ) are together able to convert supporting cells into hair cells from organs that had passed this transdifferentiation-permissive stage , we labeled P1 supporting cells using a transgenic cross , Lfng-CreERt2::Rosa26tdTomato , which allows for permanent labeling of supporting cells with a tdTomato fluorescent marker ( Figure 5—figure supplement 1B ) . We dissociated organs of Corti from Lfng-CreERt2::Rosa26tdTomato mice at P8 , a time when induced-transdifferentiation is no longer possible , and transduced them with virus encoding Atoh1 alone , or the combination of four factors , SAPG . Cells were infected and allowed to reprogram for two weeks before immunostaining for MyosinVIIa and Parvalbumin ( Figure 5C ) . The lineage traced cells , from now on referred to as Lfng-tdTomato-positive P8 supporting cells ( P8 SC ) transduced with the SAPG produced significantly more cells that activated MyosinVIIa and Parvalbumin than Atoh1-transduced supporting cells ( Figure 5D ) . Since the Lfng-tdTomato reporter is independent of the viral SAPG , the percent of triple positive ( Lfng-tdTomato+/MyosinVIIa+/Parvalbumin+ ) iHCs was calculated from the total number of Lfng-tdTomato-positive cells per well . These results indicate that the combination of Six1 , Atoh1 , Pou4f3 , and Gfi1 can convert adult tail tip fibroblasts and P8 supporting cells into induced hair cells at a significantly greater rate than Atoh1 alone . Sensory hair cells have a very distinct morphology . As their name suggests , these specialized cells possess hair-like actin-based apical membrane protrusions , called stereocilia , that contain at their tips the mechanically gated ion channels required for mechanotransduction ( Kawashima et al . , 2011; Pan et al . , 2013; Holt et al . , 2014 ) . Development of stereocilia involves the elaboration of a single primary , tubulin-based , cilium , known as the kinocilium , centered on a cuticular plate of F-actin filaments from which the stereocilia arise as elongated bundles of microvilli ( Cotanche and Corwin , 1991; Troutt et al . , 1994; Leibovici et al . , 2005; Wang et al . , 2005; Tarchini et al . , 2016; McGrath et al . , 2017 ) . To assess the morphological properties of iHCs we performed immunostaining at 14 days post-infection following SAPG reprogramming , which is approximately 10 days after initial Atoh1-nGFP detection . At this time , iHCs exhibited highly polarized F-actin staining as observed by well-defined labeling of Phalloidin-Rhodamine near the apical surface and a primary cilium that labels with antibody to acetylated tubulin , and is centered on the nascent cuticular plate ( Figure 6A ) . This highly polarized pattern is reminiscent of hair cells in both the developing cochlear and vestibular systems ( Cotanche and Corwin , 1991; Troutt et al . , 1994; Leibovici et al . , 2005; Wang et al . , 2005; Tarchini et al . , 2016; McGrath et al . , 2017 ) . Previous work has demonstrated that mixing dissociated cells from embryonic or perinatal organ of Corti with periotic mesenchyme , allows them to rapidly self-organize and differentiate in vitro into epithelial island-like structures ( Doetzlhofer et al . , 2004; White et al . , 2006 ) . To determine if iHCs are capable of integrating appropriately into these sensory epithelial-like structures which contain both primary hair cells and supporting cells , we FACS-purified Atoh1-nGFP+ iHCs and mixed them with dissociated primary embryonic ( E13 . 5 ) sensory epithelium , containing primary hair cells , primary supporting cells , and a portion of the surrounding periotic mesenchyme ( Figure 6—figure supplement 1A-C ) . After two weeks of co-culture , iHCs contained polarized cuticular plates ( F-actin ) and stereocilia ( espin-positive ) , and were found incorporated into the epithelial islands containing native hair cells and supporting cells ( Figure 6B ) . Greater than 80% of iHCs that engrafted in epithelial islands exhibited highly polarized F-actin staining ( data not shown ) . Thus , iHCs morphologically resemble primary hair cells and possess properties required for the proper structural integration with primary hair cells and supporting cells . To determine if iHCs possess electrophysiological properties similar to those of primary hair cells , we performed whole-cell patch-clamp recordings . We measured the biophysical properties of our cells in voltage-clamp and current-clamp to analyze the voltage-gated currents and passive membrane properties of these cells . We compared primary hair cells ( n = 5 ) with Atoh1-nGFP+ iHCs in two experimental conditions: monolayer-cultured iHCs ( n = 10 ) and iHCs co-cultured with dissociated organ of Corti ( n = 10 ) . Within co-cultures , the presence of the Atoh1-nGFP reporter enabled specific patch clamp analysis of iHCs . Current-clamp was used to measure the passive membrane properties of primary hair cells , co-cultured iHCs and monolayer-cultured iHCs ( Figure 6C ) . The properties measured included the resting potential , membrane capacitance , and input resistance ( Figure 6D ) . As a negative control we patch-clamped mouse embryonic fibroblasts . The MEFs showed gross electrophysiological properties that did not overlap with that of primary hair cells or induced hair cells ( data not shown ) . The mean resting potentials for primary hair cells , cocultured iHCs , and monolayer-cultured iHCs were −58 . 6 mV ( ± 6 . 9 ) , –54 . 8 mV ( ± 4 . 1 ) , and −50 . 8 mV ( ± 2 . 4 ) , respectively ( Figure 6D ) . These values are comparable to previously reported primary hair cell resting potentials ( Dallos , 1985; Oliver et al . , 2003 ) . The input resistances were measured to infer the total ion channel composition of the cell . Higher input resistance values indicate the cell may have fewer ion channels to allow current to flow in and out of the plasma membrane . The input resistance was highest in monolayer-cultured iHCs ( 3878 ± 557 MΩ ) . However , the input resistance of co-cultured iHCs ( 1432 ± 327 MΩ ) was comparable to that of primary hair cells ( 1950 ± 755 MΩ ) ( Figure 6D ) . Lastly , the capacitance , which can be used to infer the surface area of the cell , was highest in primary hair cells ( 8 . 4 ± 3 . 1 pF ) , followed by co-cultured iHCs ( 5 . 6 ± 2 . 2 pF ) and then monolayer-cultured iHCs ( 4 . 2 ± 1 . 2 pF ) ( Figure 6D ) . In addition , we performed voltage-clamp to measure the magnitude and time dependent activity of the whole-cell currents in primary hair cells , co-cultured iHCs and monolayer-cultured iHCs ( Figure 6E ) . In response to the applied voltage , both primary hair cells and iHCs produced positive-outward currents ( Figure 6E ) . However , the monolayer-cultured iHCs produced relatively small whole-cell currents that rapidly inactivated ( Figure 6E ) . In contrast , primary hair cells and co-cultured iHCs displayed robust outward currents that more slowly inactivated over the course of the protocol ( Figure 6E ) . We measured the steady-state outward current as a function of the voltage-clamp potential and normalized the current magnitude by the cell’s capacitance to analyze current densities . Monolayer-cultured iHCs showed small current densities while the co-cultured iHCs and primary hair cells displayed overlapping magnitudes of voltage-dependent current densities ( Figure 6F ) . A prominent voltage-clamp feature in primary hair cells is a delayed onset of a slow-activating outward current ( Housley and Ashmore , 1992; Marcotti and Kros , 1999 ) . In order to measure the kinetic properties of this slow-activating outward current , we fit a single exponential at the onset of the current ( Figure 6G ) to compare the mean time constants when the cells were clamped from −120 mV to 70 mV ( Figure 6H ) . The delayed onset current of monolayer-cultured iHCs displayed fast time constants ( Figure 6H ) . In contrast , the co-cultured iHCs and primary hair cells showed similarly longer time constants , indicating that their outward currents have similar activation kinetics ( Figure 6H ) . Together , these electrophysiological data suggest that both monolayer and co-cultured iHCs possess voltage dependent currents , however , when iHCs are co-cultured with dissociated organ of Corti , their size , passive membrane properties and ion channel function are more similar to those of primary hair cells . Primary sensory hair cells acquire distinct functional properties early in development in order to properly convert mechanical sound waves into neurotransmitter signaling ( Wu et al . , 2017 ) . Mechanotransduction relies on the organization of stereocilia , the assembly of tip links , and insertion of mechanically gated ion channels at the tip of each stereocilia ( Kawashima et al . , 2011; Pan et al . , 2013 ) . Mechanotransduction channels are highly permeable to styryl dyes , and their accumulation in hair cells occurs with much faster kinetics than most other cells ( Gale et al . , 2001 ) . Primary hair cells within the intact organ of Corti rapidly and selectively accumulate the styryl dye FM4-64 within seconds , a time frame consistent with entry of the dye through mechanotransduction channels rather than endocytosis ( Lelli et al . , 2009; Figure 6—figure supplement 1D ) . In contrast , MEFs failed to incorporate FM4-64 within the 30 s time frame ( Figure 6—figure supplement 1E ) . However , iHCs rapidly incorporated FM4-64 to high levels within a 30 s time course ( Figure 6I ) demonstrating that iHCs possess rudimentary mechanotransduction channels with similar styryl dye uptake as in primary hair cells within the intact organ of Corti ( Figure 6—figure supplement 1D ) and primary hair cells from dissociated organ of Corti preparations ( Figure 6J ) . Environmental and pharmacological ototoxins that cause selective degeneration of hair cells are major contributors to hearing loss worldwide ( Al-Malky et al . , 2015; Sagwa et al . , 2015; Knight et al . , 2017 ) . Gentamicin is representative of a large class of highly effective aminoglycoside antibiotics that result in significant hair cell degeneration ( Alharazneh et al . , 2011 ) . Unfortunately , a lack of mammalian models suitable for large scale screening of ototoxins and otoprotectants has restricted the development of small molecules to reduce ototoxicity and identification of compounds that can protect against known ototoxins . To determine if iHCs are sensitive to ototoxic compounds , we tested their ability to accumulate gentamicin in a similar manner to primary hair cells . Primary hair cells of the organ of Corti specifically accumulated Texas-Red conjugated gentamicin ( GTTR ) , but not Texas Red ( TR ) alone when treated with 0 . 5 mM of either compound for 3 hr ( Figure 7—figure supplement 1A , B ) . MEFs transduced with a GFP-expressing control virus did not accumulate GTTR ( Figure 7A ) . The iHCs , similarly to primary hair cells , robustly accumulated gentamicin-Texas Red ( GTTR ) ( Figure 7B ) , but not Texas Red ( TR ) alone ( Figure 7—figure supplement 1C ) , after a 3 hr treatment at 0 . 5 mM . To assess whether the gentamicin accumulation seen by GTTR treatment could cause iHCs to degenerate , we established a longitudinal survival assay using robotic imaging and automated tracking of iHC survival ( Figure 7C ) . To selectively identify and track survival of Atoh1-nGFP+ iHCs from daily whole-well images , we customized a time-lapse nuclei count recipe running on SVCell RS 4 . 0 ( a product of DRVision Technologies that has been rebranded to Aivia ) . The software can automatically detect and count iHCs based on nuclei morphology and the Atoh1-nGFP fluorescence , with comparable results to manual counting ( p=0 . 53 ) . We performed the survival assay with iHCs , dissociated primary P1 cochlear hair cells ( HC ) as a positive control , and induced motor neurons ( iMNs ) as a negative control . While gentamicin caused primary hair cell degeneration in a dose-dependent manner , it caused little-to-no toxicity to Hb9-RFP+ induced motor neurons generated from MEFs by transduction with Ngn2 , Isl1 , Lhx3 , Ascl1 , Brn2 , and Mty1l ( Son et al . , 2011; Figure 7D , E ) . Similar to primary hair cells , iHCs treated with gentamicin showed rapid , dose-dependent degeneration ( Figure 7D , E ) . These data indicate that iHCs possess functional properties of primary hair cells and display selective vulnerability to the known ototoxin , gentamicin . Moreover , these results suggest that iHCs provide a scalable platform for detecting agents that protect against gentamicin ototoxicity .
As confirmation of the importance of the four transcription factors identified in our unbiased reprogramming investigation , three of the four transcription factors , Atoh1 , Pou4f3 , and Gfi1 , were previously used to induce a hair cell like-fate from mouse ES cells that had been partially differentiated into ectodermal organoids ( Costa et al . , 2015 ) . In our hands , these factors are able to activate reporter expression in MEFs from Atoh1-nGFP mice , but yield a mixed population in which many GFP-positive cells failed to upregulate the hair cell markers MyosinVIIa and Parvalbumin . We hypothesize that the addition of Six1 increases reprogramming efficiency by pushing cells towards the sensory ectodermal lineage . Six1 has been reported previously to promote competency and progenitor-like state , as well as being an essential determinant of early sensory inner ear lineage ( Zheng et al . , 2003; Ozaki et al . , 2004; Zhang et al . , 2017 ) . The expression gradient of Six1 in the developing sensory epithelium precedes the activation of Atoh1 ( Ahmed et al . , 2012 ) . Six1 directly targets the Atoh1 autoregulatory enhancer , as well as other essential hair cell enhancers for Pou4f3 and Gfi1 , with an increase in Six1 occupancy as hair cells differentiate with in the sensory epithelium ( Li et al . , 2020 ) . Additionally , Six1 has been shown to play a role in the maturation of hair cells by regulating key genes involved in the establishment of planar cell polarity and hair-bundle orientation ( Li et al . , 2020 ) . These studies support our findings that , in addition to Atoh1 , Pou4f3 and Gfi1 , Six1 is an important upstream transcription factor in establishing a hair cell-specific gene expression program in our direct reprogramming . At perinatal times , only modest differences in gene expression are able to distinguish inner and outer hair cells of the cochlea ( Liu et al . , 2014b ) , as well as vestibular vs . cochlear hair cells ( Burns et al . , 2015 ) . The transcriptional profiles of postnatal day one cochlear hair cells and utricular hair cells are very similar ( Figure 2—figure supplement 1A ) . This time point represents an immature hair cell , and the transcriptional profiles are known to change as the hair cells mature and acquire subtype specificity and functionality ( Burns et al . , 2015; Zhu et al . , 2019; Hoa et al . , 2020 ) . As a result of comparing iHCs to postnatal day one primary hair cells , we are unable to statistically classify iHCs as being more similar to one or another hair cell type based on the bulk RNA sequencing . While many of the specialized gene characteristics of the cochlear hair cells are clearly upregulated during reprogramming ( Figure 2 ) , the iHCs fail to activate other important genes essential for the functional maturation of the sensory receptors in the cochlea , such as Prestin and Gata3 ( Liberman et al . , 2002; Bardhan et al . , 2019 ) . The imperfections of iHC reprogramming could have several causes . First , as noted , we could be lacking additional , hair cell-type transcription factors to drive cells to a more mature state . For example , the transcription factor Zfp503 identified in the initial set of 16 factors for reprogramming , does not get activated in response to the SAPG reprogramming cocktail . We saw that the addition of Zfp503 to the core group SAPG negatively impacted the efficiency of activation of the Atoh1-nGFP reporter ( Figure 1—figure supplement 1G ) , however it remains to be investigated whether the addition of Zfp503 can confer a more mature induced hair cell phenotype . Zfp503 is highly expressed in postnatal day one cochlear hair cells , but not in utricular hair cells , suggesting it could play an important role in conferring subtype specificity in the iHC reprogramming . Additional bioinformatic analysis has shown that some of the relatively small group of distal regulatory elements that are present in P1 hair cells , but fail to open in iHCs , associate with hair cell-specific genes that also fail to be robustly expressed during reprogramming ( data not shown ) . This further suggests additional factors , perhaps such as Gata3 , that may improve the quality and maturity of the iHCs . In addition , transcriptional characterization of older hair cells will allow for identification of additional TFs , which may improve the reprogramming strategy . Second , our current strategy relies on constitutive expression of the reprogramming factors , while continuous expression of Atoh1 is known to halt hair cell maturation ( Liu et al . , 2012b; Liu et al . , 2014a ) . We plan to overcome this limitation by using inducible gene expression constructs to drive reprogramming in the future . Finally , lack of organ-specific context in vitro may not provide additional signals for maturation . In fact , we demonstrate that iHCs co-cultured with dissociated organ of Corti cells , promoted morphological and electrophysiological maturation of iHCs ( Figure 6 ) . This functional maturation may be mediated , at least in part , by improved trafficking and assembly of ion channel subunits conferred by the co-cultures . We plan to examine the transcriptional profile changes that occur in iHCs after co-culture in order to understand what genes may be contributing to the morphological and functional maturation . As has been documented in other reprogramming experiments ( Wapinski et al . , 2013; Wapinski et al . , 2017; Rhee et al . , 2017 ) , the chromatin landscape was also drastically remodeled during reprogramming of MEFs to a hair cell-like state , readily opening de novo distal elements to change their chromatin to resemble P1 cochlear hair cells . Similar to the RNA sequencing results , there is a residual MEF signature of the chromatin landscape , and the failure to close down these chromatin regions may be acting as a barrier for more efficient and/or faithful reprogramming . In addition , the opening of a large number of peaks inappropriately ( peaks that occur in neither MEFs of primary P1 hair cells ) suggests that our cocktail may lack a transcriptional repressor , and that these inappropriately open distal elements may provide some explanation for the low level of inappropriate gene expression ( Figures 2 and 4 ) . Despite these pitfalls , common to most if not all reprogramming strategies published to date ( Ieda et al . , 2010; Son et al . , 2011; Treutlein et al . , 2016; Kaminski et al . , 2016; Van Pham et al . , 2017 ) , induced hair cells are highly similar to their primary counterparts on transcriptional and epigenetic levels , as well as functional levels . Recently the use of Anc80-based adeno associated vectors made inner ear gene delivery feasible ( Suzuki et al . , 2017; Tao et al . , 2018; Tan et al . , 2019 ) . The proof of concept studies have demonstrated functional recovery after administration of TMC1 gene therapy in animals carrying a mutation in the gene ( Yoshimura et al . , 2019 ) . Yet , the genetic causes of deafness are often unknown in patients , and loss of hair cells remains the leading contributor to hearing loss worldwide . We found that the SAPG four transcription factor combination is significantly more effective at activating the expression of hair cell genes MyosinVIIa and Parvalbumin in adult tail tip fibroblasts , and postnatal ( P8 ) supporting cells , when compared to Atoh1 alone ( Figure 5 ) . Supporting cells have been shown to be maintained in long deafened mice ( Oesterle and Campbell , 2009 ) and humans ( Johnsson et al . , 1981 ) . The demonstration that SAPG is able to convert P8 supporting cells to a hair cell-like fate , highlights the potential for future gene therapy approaches for hearing loss . The ability to reprogram cells to a hair cell fate provides new opportunities to target hearing loss through the development of disease-specific drug screens and personalized medicine . Primary human fibroblasts taken from patients , can be reprogrammed to induced pluripotent stem cells that can then be expanded and reprogrammed to a hair cell fate for patient- and disease-specific studies , opening the opportunity to study hearing loss mutations , ototoxicity and regenerative medicine approaches ( Koch et al . , 2011; Lim et al . , 2016a; Lim et al . , 2016b; Shi et al . , 2018; Huang et al . , 2019; Villanueva-Paz et al . , 2019; Lee et al . , 2019 ) . The evaluation of ototoxic compounds , and the identification of new otoprotectants has been severely limited by the lack of sufficient numbers of mammalian hair cells available for study . Although directed differentiation of ESCs towards the sensory hair cells have been described ( Li et al . , 2003; Oshima et al . , 2010; Koehler et al . , 2013; Ronaghi et al . , 2014 ) , these three-dimensional protocols are time consuming and the outcomes are ESC-line dependent with variable efficiency across protocols ( Hiler et al . , 2015; Mellough et al . , 2019; Yoon et al . , 2019 ) . Direct reprogramming allows for a much faster and more reliable approach to produce large quantities of induced hair cell-like cells to scale , and the monolayer culture used here overcomes several of the short-comings of directed differentiation systems presented so far , such as long culture periods ( Koehler et al . , 2013 ) , and 3D cultures that are more difficult to image robotically ( Breslin and O'Driscoll , 2013; van Vliet et al . , 2014 ) . The efficiency and reproducibility of direct reprogramming is essential for high throughput screening approaches . We demonstrated that iHCs are selectively vulnerable to gentamicin , and can be reprogrammed and cultured in microtiter plate format , providing a robotic-imaging platform for scalable monitoring of iHC survival and small-molecule screening . Importantly , reprogramming does not require an iPSC intermediate , thus generating iHCs from human patients with known or novel genetic mutations associated with hearing loss will enable screening for new therapeutic targets and agents for the treatment of genetic causes of deafness . The direct lineage conversion of somatic cells to a hair cell-like fate provides the means to study many outstanding questions in inner ear biology . For instance , the nature of gene regulatory logic underlying the differentiation of the variety of cochlear and vestibular hair cell types , as well as mechanisms underlying hair cell degeneration caused by ototoxins , and the numerous mutations responsible for the many types of syndromic and non-syndromic hearing loss .
Further information and requests for resources and reagents used in this study should be directed to the Lead Contacts , Dr . Justin Ichida ( ichida@usc . edu ) or Dr . Neil Segil ( nsegil@med . usc . edu ) . All experiments were performed at the University of Southern California . All animal experiments were conducted according to the National Institutes of Health Guide for Care and Use of Laboratory Animals . Protocols and experiments using animals were approved by the Institutional Animal Care and Use Committee at the University of Southern California . Mice were housed with free access to chow and water and a 12 hr day/night cycle . Breeding and genotyping of the mice was performed according to USC standard procedures . Atoh1-nGFP ( previously known as Math1-GFP ) transgenic line was obtained from Jane Johnson . Atoh1-nGFP transgenic mice were mated with wild type CD1 mice to obtain litters for mouse embryonic fibroblast isolations and tail tip fibroblast isolations . Wild type CD1 mice were used for harvesting wild type organs of Corti in co-culture experiments . Lfng-CreERt2::Rosa26tdTomato transgenic mice were used for harvesting organs of Corti with lineage traced supporting cells . Complimentary DNAs for the 16 candidate transcription factors were each cloned into viral expression vectors using the Gateway cloning ( Invitrogen ) . Retroviral and lentiviral plasmids were constructed into the entry vector pDONR221 . Entry clones were recombined into destination vectors via LR reaction into the pMXS-DEST ( retro ) or FUWO-tetO-DEST ( lenti ) . Plat-E cells and HEK293 cells were cultured in MEF medium ( DMEM containing 10% fetal bovine serum ) and used to produce retroviruses and lentiviruses respectively . Cells were transfected at 90% confluency with viral vectors containing genes of interest and viral packaging plasmids ( PIK-MLV-gp and pHDM for retrovirus; pPAX2 and VSVG for lentivirus ) using linear polyethylenimine ( PEI ) ( Sigma-Aldrich ) . After 24 hr of incubation with the plasmid DNA and PEI , the medium was replaced with fresh MEF medium and the culture was continued . Supernatants from the transfected cells were collected at 24 hr and 48 hr after medium replacement , filtered through 0 . 45 um filters and used immediately if generated from Plat-E or concentrated using Lenti-X concentrator ( Clontech ) and stored at −80°C if generated from HEK293T . MEFs were transduced by mixing virus with MEF media . The virus containing media was removed from the MEFs after 24 hr and replaced with MEF media . The next day , medium was switched to hair cell medium ( HCM: DMEM/F-12 ( supplier ) , N2 and B27 supplements ( supplier ) , EGF ( 2 . 5 ng/ml ) , and FGF ( 5 ng/ml ) ) . Mouse embryonic fibroblasts ( MEFs ) were obtained from E13-14 embryos taking care to exclude contamination with other Atoh1 expressing tissues ( kidney , brain , spinal cord and webbing between digits ) . Tissue was minced with a razor blade and enzymatically dissociated with 0 . 25% trypsin-EDTA for 30 min at 37°C . Trypsinization was quenched by addition of MEF media ( previously described ) . The isolated cells were centrifuged ( 800 g for 10 min ) and the pellet was resuspended in MEF media before plating onto gelatin coated T75 tissue culture flasks . We found that plating two embryos per T75 gave optimal survival post-dissection . The MEFs were cultured until confluency was reached and then cryopreserved in liquid nitrogen using freezing media ( 1:1 mixture of MEF media and Freezing media ( FM; 80% fetal bovine serum and 20% DMSO ) ) . MEFs were used without further passaging for reprogramming experiments . All cells were tested for mycoplasma contamination and came back negative . Tail tip fibroblasts ( TTFs ) were obtained from 6 month old adult Atoh1-nGFP transgenic mice . The tail was harvested from the sacrificed mouse by removing the skin and dissecting the remaining tail tissue into small segments . After plating on gelatin coated dishes , the tissue adhered to the dish and the expanding cells eventually covered the dish . The TTFs were harvested for reprogramming by simply moving the segments to a new dish and then collecting the remaining adherent cells . TTFs were cultured in DMEM with 40% fetal bovine serum . TTFs were frozen in liquid nitrogen and used without further passaging for reprogramming experiments . All cells were tested for mycoplasma contamination and came back negative . Skin was obtained from postnatal day 1 ( P1 ) mice . Skin was incubated overnight at 4°C in Accutase . The epithelium ( epidermis and hair follicles ) was separated from the underlying dermis with forceps and the epidermal cells were dissociated with trypsin for 10 min at 37°C then dissociated to single cell suspension . The freshly isolated epidermal cell suspension was then FACS purified to sort for Atoh1-nGFP positive Merkel cells . Small intestines were obtained from adult mice . Intestinal villi were scraped away , crypt epithelium was collected by shaking in 5 mM EDTA for 50 min at four degrees Celsius , and single cell suspensions were prepared by digestion in 4x TrypLE ( Invitrogen ) for 50 min at 37°C . The freshly isolated cell suspension was then FACS purified to sort for Atoh1-nGFP positive cells gut secretory cells . Cerebellums were obtained from postnatal day 1 ( P1 ) mice . Tissue was minced and enzymatically digested using 0 . 25% trypsin for 10 min at 37°C then dissociated to single cell suspension . The freshly isolated cerebellar cell suspension was then FACS purified to sort for Atoh1-nGFP positive cells cerebellar granule precursors . The primary hair cell culture was established by dissecting the organs of Corti from P1 transgenic Atoh1-nGFP mice . The cells were dissociated to a single cell suspension and plated onto laminin coated tissue culture plates or cover slips . The primary cultures from Lfng-CreERt2::Rosa26tdTomato transgenic mice were done at postnatal day 8 ( P8 ) . Lfng-CreERt2::Rosa26tdTomato transgenic mice were injected with tamoxifen at postnatal day three for lineage tracing of the Lfng+ supporting cell population . The organs of Corti were harvested at P8 , dissociated to single cell suspension in HCM and the reprogramming factors were added to the cell suspension . The cells were then plated onto laminin coated tissue culture treated cover slips . The virus containing media was replaced after 24 hr with fresh HCM . All primary cultures were plated using ROCK Inhibitor ( Y-27632 ) ( Sigma-Aldrich ) for the first 24 hr to help promote survival . Induced hair cells were FACS purified to obtain the Atoh1-nGFP positive cells and collected in HCM . The primary organ of Corti was dissected from wild type mice at E13 . 5 and enzymatically dissociated to a single cell suspension containing primary hair cells , primary supporting cells and a portion of the surrounding periotic mesenchmye . The iHCs were then mixed with the dissociated organs of Corti . The ratio of iHC to cells of the organ of Corti was kept at about 1:33 . This ratio was determined from the fact that the primary organ of Corti contains approximately 3000 hair cells and upon dissociation gives approximately 100 , 000 total cells . Co-cultures were grown on tissue culture treated coverslips in wells of 24 well plates that had been coated with a 20 ul drop of matrigel ( 10% in HCM ) at the center of the coverslip . The co-culture cell suspension was plated as 30 ul drops ( 2 , 500–3 , 000 cells per ul ) in the center of the matrigel coated drop on the cover slip . 12–24 hr after plating the drops the well was flooded with 500 ul of HCM . All cocultures were maintained in HCM . Cells for staining were washed with PBS and fixed using 4% paraformaldehyde ( PFA ) in phosphate-buffered saline ( PBS ) for 15 min at room temperature . For permeabilization and blocking the cells were incubated in PBST ( 0 . 1% Triton-X 100 in PBS ) with 10% fetal bovine serum for 2 hr at room temperature or overnight at 4°C . After blocking , the cells were washed 3 times for 5 min with PBS . Cells were then incubated with the primary antibody overnight at 4°C . Then the cells were washed with PBS again before incubation with the secondary antibody for one hour at room temperature or overnight at 4°C . Primary and secondary antibodies were diluted in PBST with 10% serum . The DNA was stained with Hoechst diluted 1:1000 in PBS for 10 min at room temperature . Immunostaining images of adherent cell cultures were acquired on an LSM780 confocal microscope using Carl Zeiss Zen blue/black software and processed using Adobe Illustrator CS6 software . For quantification of reprogramming efficiency in adherent cultures , images were acquired at 10x using the Molecular Devices ImageExpress . The images were either processed manually using ImageJ software and the Cell Counter plug in or automatically using SVCell RS ( described below ) . Counts are represented as reprogramming efficiency ( percent of Atoh1-nGFP+ cells per well of 5000 MEFs infected ) . Automated cell counting used thresholds for size , intensity and roundness of the Atoh1-nGFP signal . The imaging was done at 10x . For each time frame , the customized time-lapse nuclei count recipe of SVCell RS is applied to first reduce noise with image smoothing . Objects are detected by performing background removal followed by adaptive thresholding . A size filtering is then applied to remove objects that are either too large or too small . The count of remaining objects is measured for the time point . Batch processing is available for applying the recipe to multiple time-lapse images and saving results . To ensure the reliability of the automated counting a comparison was done of 20 wells counted both manually and automatically ( p=0 . 53 ) . Primary hair cells were harvested from Atoh1-nGFP transgenic mice . The cochleas were incubated in 0 . 25% trypsin for 8 min and gently triturated to single cell suspension . Media ( DMEM with 10% FBS ) was added to the dissociated cells and then spun down at 1000 rpm for 5 min , resuspended in Hair Cell Media , passed through a 70 um cell strainer and then FACS-purified ) . The same procedure was used to FACS-purify dsRED MEFs and Atoh1-nGFP+ cells from the reprogramming cultures . Total RNA was extracted from primary mouse hair cells ( at postnatal day 1 ) , mouse iHCs ( at day 14 post infection with reprogramming factors ) and MEFs ( at 14 days post transduction with dsRed ) . For each replicate 20 , 000 FACS-sorted cells were used as input for RNA-seq . Total RNA was extracted with either Quick-RNA Microprep kit ( Zymo Research ) , quantified by bioanalyzer and then processed for libraries with either QIAseq FX Single Cell RNA Library Kit ( Qiagen ) or TruSeq RNA Library Prep Kit v2 ( Illumina ) . Specific sequencing parameters and instrument models were submitted with GEO datasets . At least three replicates were collected for each condition and sequenced to a depth of at least 20 million reads . Reads were mapped to the mouse reference genome ( Gencode Mm10v11 ) using STAR . Read counts were quantified by RSEM . Only protein coding polyA tail transcripts and autosomal genes were kept . Transcript counts were collapsed to gene counts . Differentially expressed genes were identified using the DESeq2 package . Genes with a log fold change threshold greater than one and adjusted P-value of less than 0 . 1 were considered significant . Principle component analysis and unsupervised hierarchical clustering of RNA-seq was performed using counts transformed by DESeq2’s regularized logarithm ( Rlog ) . GEO accession number: GSE149260 . Gene ontology analysis was performed on categorized gene sets using R clusterProfiler package . GO results were visualized using the R enrichplot package . Gene Set Enrichment Analysis was performed using the R package fgsea . The Wald statistic from the differential comparison of reprogrammed cells versus MEFs was used to pre-rank genes for subsequent GSEA analysis . Gene sets representing unique signatures for each Atoh1 positive cell-type were tested for enrichment in SAPG . To determine signature gene sets for each Atoh1 cell type , only genes with a log2 foldchange greater than or equal to two with adjusted P-value less than 0 . 01 compared between each profiled Atoh1 positive cell-type were used . Utricle and cochlear hair cells were treated as a single cell type due to small number of unique genes at the postnatal day one developmental stage used . Cells were collected by FACS purification into cold PBS , and centrifuged 500 xg for 15 min . Cell pellet were resuspended with 50 ul transposition buffer consisting of 10 mM Tris-HCL pH8 . 0 , 5 mM MgCl2 , 10% DMF , 0 . 2% NP40 , and home-made transposase Tn5 . Transposition was performed at 37°C for 20 min . DNA was collected immediately after transposition using Qiagen Mini-elute kit . Encode pipeline was adapted for alignment and QC for ATAC-seq and ChIP-seq data . Paired-end reads were quality trimmed with cutadapt ( v1 . 18 ) and aligned to mouse reference genome ( Gencode Mm10v11 ) with bowtie2 ( v2 . 2 . 6 ) using parameters -X2000 -mm –local . PCR duplicates were removed based on genomic coordinates . Only autosomal chromosomes were selected and used for downstream analysis . Specific sequencing parameters and instrument models were submitted with GEO datasets . Histone ChIP-seq protocol was developed by us based on μChIP-PCR protocol published previously ( Stojanova et al . , 2016 ) with additional Tn5 tagmentation step . Briefly , chromatin was cross-linked with 1% formaldehyde ( Thermo Fisher ) for 8 min , quenched with 125 mM Glycine ( Sigma ) for 5 min at room temperature , sonicated using the microtip of a High Intensity Ultrasonic Processor ( Sonics and Materials , Newtown , CT; amplitude 50 , power 50 ) for 8 × 30 s with 30 s pause , tagmentated with Tn5 transposase for 30 min at 37°C , incubated with antibody complexed with Dynabeads Protein A ( Thermo Fisher ) overnight at 4°C , precipitated and washed three times on magnetic rack , and finally PCR amplified with primers matching Tn5 adapters . Encode pipeline was adapted for alignment and QC for ChIP-seq data . Paired-end reads were quality trimmed to 36 bp with cutadapt ( v1 . 18 ) and aligned to mm10 reference genome ( Gencode Mm10v11 ) with STAR aligner using parameters end-to-end and alignIntronMax = 1 for DNA alignment . PCR duplicates were removed with STAR . Only autosomal chromosomes were selected and used for downstream analysis . Specific sequencing parameters and instrument models were submitted with GEO datasets . ATAC peaks and H3K27ac peaks were identified using the R package chromstaR ( parameters: binsize = 500 bp , stepsize = 250 bp , mode = full ) . An equal number of reads were randomly sampled for H3K27ac replicates ( 17 . 5 million ) and ATAC replicates ( 15 million reads ) as input for subsequent chromatin analysis . For peak calling , a false discovery rate ( FDR ) cutoff of 0 . 01 and 0 . 001 was used for ATAC and H3K27ac respectively and an RPKM cut off >2 . Promoter regions were defined by 2 kb upstream of 500 bp downstream of protein coding transcription start sites; all remaining regions were considered distal . Enhancers were defined by cooccurrence of an ATAC peak and H3K27ac peak at distal regions . Differential ATAC peak analysis was performed between P1HC , SAPG iHCs , and MEFs using chromstaR . Regions with a differential score of at least 0 . 999999 were considered differentially accessible . Regions with differential score less than 1E-06 were considered non-differentially accessible . Deeptools was used to average replicates and calculate coverage tracks and for ATAC-seq and ChIP-seq data for visualization on IGV and heatmaps . Whole cell patch clamping was performed on three different preparations of cells . The first preparation was iHCs in the monolayer culture of MEFs at D14-15 post infection with SAPG . The second preparation was iHCs FACS purified and replated with dissociated wild type organ of Corti . The third preparation was postnatal day one primary hair cells from the dissociated transgenic Atoh1-nGFP organ of Corti . Preparations were viewed at X630 using a Zeiss Axios Examiner D1 microscope fitted with Zeiss W Plan-Aprochromat optics . Signals were driven , recorded , and amplified with an Multiclamp 700B amplifier , Digidata 1440 board and pClamp 10 . 7 software ( pClamp , RRID:SCR_011323 ) . Recording and cleaning pipettes were fabricated using filamented borosilicate glass . Pipettes were fired polished to yield an access resistance between 4–8 MΩ . Each recording pipette was covered in a layer of parafilm to reduce pipette capacitance . Recording pipettes were filled with standard internal solution . The contents of the standard internal solution are ( in mM ) : 135 KCl , 3 . 5 MgCl2 , 3 Na2ATP , 5 HEPES , 5 EGTA , 0 . 1 CaCl2 , 0 . 1 Li-GTP , and titrated with 1M KOH to a pH of 7 . 35 and an osmolarity of about 300 mmol/kg . The voltage clamp protocol was performed by holding the cell at −60 mV followed by a stimulus of voltage steps ( −120 to +70 mV , by intervals of 10 mV ) . The current response of the cell was recorded along with measures of ionic current peak amplitudes , channel conductance values , and current activation kinetics . Analysis of the data was performed using a combination of pClamp ( pClamp , RRID:SCR_011323 ) , Matlab ( MATLAB , RRID:SCR_001622 ) , JMP ( JMP , RRID:SCR_014242 ) , Origin Pro ( OriginPro , RRID:SCR_015636 ) , and Imaris ( Imaris , RRID:SCR_007370 ) . pClamp software was be used to gather and quantify raw data from electrophysiological recordings . Cells were incubated with 1 uM FM 4-64FX , the fixable analog of FM4-64 ( Life Technologies , catalog# F34653 ) . Prior to incubation the FM 4-64FX was resuspended in ice cold HBSS at a 1 mM concentration . The cells were incubated with a final concentration of 1 uM FM 4-64FX in ice cold HBSS for 30 s . After incubation , the cells were rinsed in HBSS three times and then immediately imaged . Using the software ImageJ , the images were filtered on minimum background intensity in order to reduce the amount of background signal . The filter measures the minimum signal intensity found in the image and applies the filter to remove the minimum signal across the entire image . This image enhancement was used uniformly on all images and all channels for each cell type . Gentamicin sulfate salt ( Sigma Aldrich catalog# G3632-5G , 50 mg/ml in K2CO3 , pH0 ) and Texas-Red ( Thermo Fisher Scientific catalog# T20175 , 2 mg/ml in dimethyl formamide ) were agitated together overnight to produce gentamicin-Texas Red conjugate ( GTTR ) . The mixture contained 4 . 4mls of 50 mg/ml gentamicin ( GT ) with 0 . 6mls of 2 mg/ml Texas Red ( TR ) to produce approximately 300:1 molar ratio of GT:GTTR . A high ratio of gentamicin ensures a minimum of unbound Texas Red molecules . The molecular weight of GT is 477 . 6 g/mol and the molecular weight of TR is 816 . 94 g/mol . The GTTR was made at a stock concentration of 100 mM . The cells were incubated with HCM containing 0 . 5 mM or 1 mM GTTR for three hours . After incubation the cells were washed three times with PBS and then immediately fixed using 4% PFA in PBS for 15 min at room temperature . The cells were cultured ( for primary hair cells ) or reprogrammed ( for iHCs and iMNs ) in a 96 well tissue culture plate . The primary hair cells were used 24 hr post dissociation of the organ of Corti and plating . The iHCs and iMNs were reprogrammed for 14 days prior to starting the survival assay . The gentamicin was dissolved in HCM at a concentration of 100 mM and subsequently diluted to 8 mM in HCM . The stock at 8 mM in HCM was used for serial dilution to get the desired range of concentrations ( 8 mM , 4 mM , 2 mM , 1 mM , 0 . 5 mM , 0 . 25 mM , and 0 . 125 mM ) . The control wells received only HCM . The assay was performed over a period of 5 days . The HCM containing with or without gentamicin was added to the cells on Day one and the cells were imaged every 24 hr after the initial treatment . Subsequent media changes were performed every other day ( Day 3 ) . The HCM for gentamicin treated wells and control wells was made fresh for each media change . The assay ended on Day 5 . The Molecular Devices ImageExpress was used for imaging the plate robotically every 24 hr . The images were taken at 10x . Sample numbers , experimental repeats and statistical test used are indicated in figure legends . Unless otherwise stated , data presented as mean + SEM of at least three biological replicates . Significance summary: p>0 . 05 ( ns ) , ∗p≤0 . 05 , ∗∗p≤0 . 01 , ∗∗∗p≤0 . 001 , and ∗∗∗∗p≤0 . 0001 . | Worldwide , hearing loss is the most common loss of sensation . Most cases of hearing loss are due to the death of specialized hair cells found deep inside the ear . These hair cells convert sounds into nerve impulses which can be understood by the brain . Hair cells naturally degrade as part of aging and can be damaged by other factors including loud noises , and otherwise therapeutic drugs , such as those used in chemotherapy for cancer . In humans and other mammals , once hair cells are lost they cannot be replaced . Hair cells have often been studied using mice , but the small number of hair cells in their ears , and their location deep inside the skull , makes it particularly difficult to study them in this way . Scientists are seeking ways to grow hair cells in the laboratory to make it easier to understand how they work and the factors that contribute to their damage and loss . Different cell types in the body are formed in response to specific combinations of biological signals . Currently , scientists do not have an efficient way to grow hair cells in the laboratory , because the correct signals needed to create them are not known . Menendez et al . have now identified four proteins which , when activated , convert fibroblasts , a common type of cell , into hair cells similar to those in the ear . These proteins are called Six1 , Atoh1 , Pou4f3 and Gfi1 . Menendez et al . termed the resulting cells induced hair cells , or iHCs for short , and analyzed these cells to identify those characteristics that are similar to normal hair cells , as well as their differences . Importantly , the iHCs were found to be damaged by the same chemicals that specifically harm normal hair cells , suggesting they are useful test subjects . The ability to create hair cells in the laboratory using more easily available cells has many uses . These cells can help to understand the normal function of hair cells and how they become damaged . They can also be used to test new drugs to assess their success in preventing or reversing hearing loss . These findings may also lead to genetic solutions to curing hearing loss . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"developmental",
"biology"
] | 2020 | Generation of inner ear hair cells by direct lineage conversion of primary somatic cells |
Spatial attention changes the sampling of visual space . Behavioral studies suggest that feature-based attention modulates this resampling to optimize the attended feature's sampling . We investigate this hypothesis by estimating spatial sampling in visual cortex while independently varying both feature-based and spatial attention . Our results show that spatial and feature-based attention interacted: resampling of visual space depended on both the attended location and feature ( color vs . temporal frequency ) . This interaction occurred similarly throughout visual cortex , regardless of an area's overall feature preference . However , the interaction did depend on spatial sampling properties of voxels that prefer the attended feature . These findings are parsimoniously explained by variations in the precision of an attentional gain field . Our results demonstrate that the deployment of spatial attention is tailored to the spatial sampling properties of units that are sensitive to the attended feature .
The resolution of the visual system is highest at the fovea and decreases gradually with increasing eccentricity . But the visual system’s resolution is not fixed . Attention can be directed to a location in space and/or a visual feature , which temporarily improves behavioral performance ( Posner et al . , 1980; Rossi and Paradiso , 1995; Found and Müller , 1996; Carrasco and Yeshurun , 1998; Yeshurun and Carrasco , 1999; Kumada , 2001; Sàenz et al . , 2003; Wolfe et al . , 2003; Theeuwes and Van der Burg , 2007 ) at the cost of reduced sensitivity for non-attended locations and features ( Kastner and Pinsk , 2004; Pestilli and Carrasco , 2005; Wegener et al . , 2008 ) . Attending a location in space increases activity in units representing the attended location , as shown by both electrophysiological ( Luck et al . , 1997; Reynolds et al . , 2000 ) and fMRI studies ( Tootell et al . , 1998; Silver et al . , 2005; Datta and DeYoe , 2009 ) . In addition , spatial receptive fields were shown to shift toward an attended location in macaque MT+ ( Womelsdorf et al . , 2006 ) and V4 ( Connor et al . , 1997 ) . Using fMRI to measure population receptive fields ( pRFs; Dumoulin and Wandell , 2008; Dumoulin and Knapen , 2018 ) , it was found that pRF shifts induced by spatial attention occur throughout human visual cortex ( Klein et al . , 2014; Kay et al . , 2015; Sheremata and Silver , 2015; Vo et al . , 2017 ) , a process thought to improve visual resolution at the attended location ( Anton-Erxleben and Carrasco , 2013; Kay et al . , 2015; Vo et al . , 2017 ) . Such spatial resampling is understood to be the result of an interaction between bottom-up sensory signals and a top-down attentional gain field ( Womelsdorf et al . , 2008; Klein et al . , 2014; Miconi and VanRullen , 2016 ) . Feature-based attention , for example directed toward color or motion , selectively increases activity in those units that represent the attended feature , as evidenced by electrophysiological ( Treue and Maunsell , 1996; Treue et al . , 1999; McAdams and Maunsell , 2000; Maunsell and Treue , 2006; Müller et al . , 2006; Zhang and Luck , 2009; Zhou and Desimone , 2011 ) , fMRI ( Saenz et al . , 2002; Serences and Boynton , 2007; Jehee et al . , 2011 ) , and behavioral reports ( Sàenz et al . , 2003; White and Carrasco , 2011 ) . These studies consistently show that feature-based attention modulates processing irrespective of the attended stimulus's spatial location . In addition , feature-based attention also appears to shift featural tuning curves toward the attended value , as reported by both electrophysiological ( Motter , 1994; David et al . , 2008 ) and fMRI studies ( Çukur et al . , 2013 ) . The similarity in the effects of feature-based and spatial attention on affected neural units suggests a common neural mechanism for both sources of attention ( Hayden and Gallant , 2005; Cohen and Maunsell , 2011 ) . Yet spatial attention necessitates retinotopically precise feedback ( Miconi and VanRullen , 2016 ) , whereas feature-based attention operates throughout the visual field ( Maunsell and Treue , 2006 ) . Studies investigating whether one source of attention potentiates the other generally find that interactions are either nonexistent or very weak at the earliest stages of processing ( David et al . , 2008; Hayden and Gallant , 2009; Patzwahl and Treue , 2009; Zhang and Luck , 2009 ) , but emerge at later stages of visual processing ( Hillyard and Münte , 1984; Handy et al . , 2001; Bengson et al . , 2012; Ibos and Freedman , 2016 ) , and ultimately influence behavior ( Kingstone , 1992; Kravitz and Behrmann , 2011; Leonard et al . , 2015; White et al . , 2015; Nordfang et al . , 2018 ) . In addition , the effects of feature-based compared to spatial attention arise earlier in time . This occurs both when only feature-based attention is endogenously cued and subsequently guides spatial attention towards the attended feature's location ( Hopf et al . , 2004 ) , and when both types of attention are endogenously cued ( Hayden and Gallant , 2005; Andersen et al . , 2011 ) . This supports the idea that feature-based attention can direct spatial attention toward or away from specific locations containing attended or unattended features ( Cohen and Shoup , 1997; Cepeda et al . , 1998; Burnett et al . , 2016 ) . The studies mentioned above investigated modulatory effects of feature-based attention on spatial attention by measuring changes in response amplitude ( e . g . ERP/firing rate ) . However , no study to date has investigated the effect of feature-based attention on spatial sampling . Yet , exactly this relationship is predicted by behavioral studies . Especially when attention is endogenously cued , feature-based attention has been argued to influence the spatial resampling induced by spatial attention to optimize sampling of visual features for behavior ( Yeshurun and Carrasco , 1998; Yeshurun and Carrasco , 2000; Yeshurun et al . , 2008; Barbot and Carrasco , 2017 ) . Specifically , these authors suggested that when attending features that are processed by neurons with smaller receptive fields , a greater degree of spatial resampling is required to resolve the required featural resolution at the attended location . In the current study , we put this hypothesis to the test by measuring the brain's representation of space under conditions of differential attention . Specifically , we measured pRFs under conditions of differential spatial attention ( i . e . toward fixation or the mapping stimulus ) and feature-based attention ( i . e . toward the mapping stimulus's temporal frequency or color content ) . One important reason for studying the effects of attention to color and temporal frequency is that they are known to be processed at different spatial scales . Specifically , color is generally processed at a finer spatial scale compared to temporal frequency . First , this is a result of color compared to temporal frequency information being processed predominantly by the parvocellular rather than the magnocellular pathways , which in turn pool differentially across visual space ( Schiller and Malpeli , 1978; Hicks et al . , 1983; Denison et al . , 2014 ) . Second , temporal frequency and color are processed across different cortical areas ( i . e . MT +compared to hV4; Liu and Wandell ( 2005 ) ; Brouwer and Heeger , 2009 , Brouwer and Heeger , 2013; Winawer et al . , 2010 ) that have differential spatial precision ( Amano et al . , 2009; Winawer et al . , 2010 ) . Third , preference for color compared to temporal frequency is generally greater in the fovea compared to the periphery ( Curcio et al . , 1990; Azzopardi et al . , 1999; Brewer et al . , 2005 ) , where receptive fields are generally smaller ( Dumoulin and Wandell , 2008 ) . The hypothesized relation between spatial scale of attended features and spatial resampling as proposed by the behavioral studies mentioned above thus implies that attending color compared to temporal frequency at a particular location should lead to stronger spatial resampling . In addition , as color and temporal frequency are differentially processed across cortical areas , studying these features allows us to investigate whether modulations of spatial resampling by feature-based attention are specific for areas that prefer the attended feature . We specifically chose temporal frequency and not coherent motion , as coherent motion signals have been shown to influence pRF measurements ( Harvey and Dumoulin , 2016 ) . It was previously shown that attention can be directed to both feature domains ( Wolfe and Horowitz , 2004; Cass et al . , 2011 ) . We characterized how spatial attention influences the sampling of visual space , and subsequently investigated how feature-based attention modulates this spatial resampling . In addition , an explicit gain-field interaction model allowed us to formally capture the pRF position changes resulting from our attentional manipulations ( Klein et al . , 2014 ) . Finally , we also performed a full-field stimulation experiment which allowed us to relate these attentional modulations to each voxel's bottom-up preference for color and temporal frequency . In brief , our results show that pRF changes are indeed stronger when attending the stimulus's color compared to temporal frequency content . These modulations occurred similarly throughout the visual system , regardless of an area's bottom-up feature preference . We show that these feature-based attentional modulations can be explained by changes in the precision of the attentional gain field . Together , this confirms the idea that the degree of spatial resampling is dependent on the spatial scale at which attended features are processed .
Figure 2A shows voxels' Attend Fixation location preferences , by depicting color-coded polar angle coordinates on an inflated cortical surface for one example participant's right hemisphere . We examined the relation between pRF eccentricity and size within each of the retinotopic regions , and performed further analyses on those regions that showed clear progressions of polar angle on the surface as well as positive size-eccentricity relations , as shown in Figure 2B . In addition , we created a combined ROI that pooled voxels across selected ROIs to evaluate pRF changes across the visual system . To quantify pRF changes resulting from differential allocation of spatial attention , we created an Attend Stimulus condition by averaging pRF parameters between the Attend Color and Attend TF conditions . To inspect how spatial attention affected pRF positions , we plotted a vector from the Attend Fixation to the Attend Stimulus pRF position . For visualization purposes , we created visual field quadrant representations by multiplying both the Attend Fixation and Attend Stimulus pRF x- and y-coordinates with the sign of pRF x- and y-coordinates in the Attend Fixation condition ( see Figure 3A ) . This means , for example , that pRFs in the upper-right quadrant were unaffected ( i . e . x- and y-coordinate multiplied by 1 ) , while pRFs in the lower-right quadrant were mirrored along the y-axis ( i . e . x-coordinate multiplied by 1 , y-coordinate multiplied by −1 ) . Note that mirroring based on the Attend Fixation condition preserves any pRF shifts across a meridian , allowing pRFs to shift outside the target visual field quadrant . Visual inspection of these pRF position shifts shows both increasing shift magnitude up the visual hierarchy and shifts occurring mainly along the radial dimension ( i . e . toward or away from the fovea; Figure 3B ) . This latter observation seems at apparent odds with a recent study reporting that pRFs shift mainly in the horizontal direction ( Sheremata and Silver , 2015 ) . To quantify the observed direction of pRF shifts we computed the ratio of shifts in the radial , horizontal and vertical directions ( see Figure 3C ) . In line with the data of Sheremata and Silver ( 2015 ) , we find that changes of pRF horizontal location consistently better describe the overall shifts than do changes of pRF vertical location in all ROIs except V1/2/3 ( p's < 0 . 05 , see Supplementary file 1 -Table 2 ) . We also find that pRF shifts are described even better by shifts in the radial dimension ( i . e . changes in eccentricity ) compared to shifts in the horizontal direction in all ROIs ( p's < 0 . 01 , see Supplementary file 1 -Table 2 ) . Figure 3D is intended to ease interpretation of these results . It depicts how different hypotheses regarding the underlying directionality of pRF shifts ( i . e . horizontal , vertical or radial - i . e . foveopetal/foveofugal ) , translate into changes in measured pRF x , y and eccentricity as a function of quarter visual field polar angle ( i . e . from vertical to horizontal meridian ) . For example , if pRFs shift primarily in the radial direction ( right hypothesis column , Figure 3D ) , this would result in the strongest pRF x-direction changes close to the horizontal meridian and the strongest pRF y-direction changes close to the vertical meridian . pRF eccentricity changes however , would show no dependence on polar angle . Figure 3D , right column , shows that the data ( combined ROI ) correspond most to the radial shift hypothesis . To quantify this visual intuition , we compared the slopes of the change in pRF x and y over polar angle by binning polar angle into three bins and comparing the first and last bins ( i . e . horizontal and vertical meridian , respectively ) . This showed that , compared to the slope of pRF y change over polar angle , the slope of pRF x change was more negative ( p < 0 . 001 , Cohen's d = 0 . 677 , N = 11946 ) . This pattern of results can only be explained by pRFs shifting in the radial direction . Visual field coverage is known to be non-uniform such that the horizontal meridian is overrepresented at both subcortical ( Schneider et al . , 2004 ) and cortical ( Swisher et al . , 2007; Silva et al . , 2018 ) levels and was also clearly present in our data ( Rayleigh tests for non-uniformity in ROIs separately , p's < 0 . 001 , see Supplementary file 1 -Table 3 ) . This means that shifts that occur exclusively in the radial dimension appear as a dominance of horizontal compared to vertical shifts when averaging over the visual field . To further inspect the attention-induced radial shifts described above , we plotted the difference between Attend Stimulus and Attend Fixation pRF eccentricity for each of four Attend Fixation pRF eccentricity bins ( Figure 4A ) . This returned varying patterns of pRF eccentricity changes across the different ROIs . The combined ROI shows that overall , parafoveal pRFs shifted away from the fovea , while peripheral pRFs shifted toward the fovea . These outward shifting parafoveal pRFs are found in all other ROIs except V1 and V2 , whereas the inward shifting peripheral pRFs are also present in V1 , V2 and V3 ( see Supplementary file 1 -Tables 4 and 5 ) . In addition to pRF position changes , we also inspected changes in pRF size induced by differences in spatial attention as a function of Attend Fixation pRF eccentricity ( Figure 4B ) . Overall , parafoveal pRFs increased in size , while peripheral pRFs decreased in size . These expanding parafoveal pRFs were present in all ROIs except V2/3 , whereas shrinking peripheral pRFs were found in all ROIs except V1 , MT+ and IPS0 ( see Supplementary file 1 -Tables 6 and 7 ) . Overall , this pattern of results is strikingly similar to the changes in pRF eccentricity described above . In fact , the changes in pRF size and eccentricity were strongly correlated in all ROIs ( Figure 4C , Pearson R over 20 5-percentile bins between 0 . 74 and 0 . 99 , p's < 0 . 001 , see Supplementary file 1 -Table 8 ) . Together , these results show that attention to the stimulus caused parafoveal pRFs to shift away from the fovea and increase in size , whereas peripheral pRFs shifted toward the fovea and decreased in size . To provide a mechanistic explanation for the complex pattern of pRF shifts described above , we modeled our results using a multiplicative Gaussian gain field model ( Womelsdorf et al . , 2008; Klein et al . , 2014 ) . We adapted this framework to work in conditions where attention shifted over space as a function of time ( see Materials and methods ) . In brief , this modeling procedure used the Attend Fixation pRF , one attentional gain field at fixation and another convolved with the stimulus to predict the Attend Stimulus pRF position . We determined optimal attentional gain field sizes by minimizing the difference between observed and predicted Attend Stimulus pRF positions in the quadrant visual field format of Figure 3B . Figure 5A illustrates that model predictions closely followed the data , thereby accurately reproducing radially shifting pRFs . Examining the predicted change in pRF eccentricity as a function of eccentricity ( i . e . the dominant pRF shift direction; Figure 5B ) showed that the model was able to capture widely varying eccentricity change profiles across ROIs using very similar attentional gain field sizes ( Figure 5C ) . This shows that a common attentional influence can result in very different pRF shift patterns , which then depend on differential spatial sampling properties across ROIs ( i . e . distribution of pRF sizes and positions ) . In sum , these results show that the attentional gain field model provides a parsimonious and powerful account for the variety of pRF shift patterns across ROIs . We further investigated how the model was able to reproduce the eccentricity-dependent eccentricity changes we reported above . For this , we inspected pRF shifts induced by attending either fixation or the stimulus relative to the stimulus drive ( i . e . the pRF outside the influence of attention derived from the model ) . For illustrative purposes , we display results for V2 , V3 and IPS0 as these areas showed marked differences in their eccentricity change profile ( Figure 5D–F ) . The left panels of each figure reveal the effects of attending fixation and the stimulus separately . This shows that both sources of spatial attention pull the measured pRFs toward the fovea , albeit with differing relative magnitudes across eccentricity . The right panel of each figure shows that the resulting difference between attending fixation and the stimulus constitutes the eccentricity-dependent patterns observed in the data ( Figure 5B ) . Together , these analyses show that existing multiplicative gain field models of attention can be extended to predict pRF shifts in situations where spatial attention shifts over time . Additionally , it confirms , extends and quantifies earlier reports showing that the precision of the attentional gain field is similar across the visual hierarchy ( Klein et al . , 2014 ) . Having established ( 1 ) the pattern of changes in spatial sampling ( i . e . changes in pRF size and eccentricity ) resulting from differential allocation of spatial attention , and ( 2 ) a mechanistic explanation of these changes , we next examined how this pattern was modulated by differences in feature-based attention . Figure 6A shows how pRF eccentricity and size are differentially affected by attending color or temporal frequency within the stimulus for the combined ROI . This illustrates that while both tasks caused similar pRF changes , these effects were generally more pronounced when attending color . To quantify the modulation of feature-based attention per voxel , we first set up a single robust index of the degree to which spatial attention resampled visual space , combining changes in pRF eccentricity and size ( as these were highly correlated , see Figure 4C ) . This Attentional Modulation Index ( AMI , see Materials and methods ) is depicted in Figure 6B for the combined ROI when attending color and TF . We then quantified the difference in this AMI between attending color and temporal frequency as a feature-based Attentional Modulation Index ( feature AMI , see Materials and methods ) . Positive values of feature AMI indicate that attending color induced greater pRF changes , while negative values indicate that attending TF led to stronger pRF changes . Figure 6C shows that this feature AMI was positive across eccentricity in the combined ROI . Inspecting the average feature AMI across voxels within each ROI ( Figure 6D ) reveals that attending changes in color compared to TF in the bar stimulus produced stronger spatial resampling in all ROIs ( p's < 0 . 01 , see Supplementary file 1 -Table 9 ) . Specifically , the feature AMI was around 0 . 05 on average across ROIs . As the feature AMI is a contrast measure where difference is divided by the sum , this corresponds to roughly 10% stronger pRF changes when attending color compared to temporal frequency . In some ROIs ( V3AB/IPS0 , see Figure 6—figure supplement 1 ) , the FAMI reached values of 0 . 20 , which corresponds roughly to 50% stronger pRF changes when attending color compared to temporal frequency . Computing the AMI with either pRF eccentricity or size changes separately ( i . e . not as a combined measure ) yields similar results ( see Figure 6—figure supplement 2 ) . The feature-based modulations we describe above are possibly related to differences in bottom-up preference for the attended features . Feature-based attention is known to increase activity of neurons selective for the attended feature , regardless of spatial location ( Treue and Maunsell , 1996; Treue et al . , 1999; McAdams and Maunsell , 2000; Maunsell and Treue , 2006; Müller et al . , 2006; Zhang and Luck , 2009; Zhou and Desimone , 2011 ) . This suggests that if activity in a given voxel is modulated more strongly by changes in a certain feature , this could lead to a greater apparent pRF shift when attending that feature . To test this hypothesis , we estimated the difference in response amplitude to the presence of color and temporal frequency within a full-field stimulus ( in a separate experiment , see Materials and methods ) . We then summarized each voxel's relative preference for color and temporal frequency by means of a feature preference index . Higher values of this feature preference index indicate greater preference for color compared to TF . Figure 6E displays the feature AMI as a function of feature preference , for each ROI . Note that feature preference was negative on average in most ROIs , suggesting that our TF manipulation ( 7 vs 0 Hz grayscale Gabors ) caused stronger response modulations compared to our color manipulation ( colored vs grayscale Gabors ) . Although this induced an offset across the brain , variations in this measure across ROIs replicate known specializations of the visual system with high precision ( Liu and Wandell , 2005; Brouwer and Heeger , 2009; Brouwer and Heeger , 2013 ) : while areas MT+ and V1 show the strongest preference for TF compared to color , areas V4 and VO show the strongest preference for color compared to TF . Importantly , regardless of these large variations in feature preferences between MT+/V1 and VO/hV4 , average feature AMI was nearly equal in these ROIs . In addition , there was no correlation between feature preference and feature AMI across all ROIs ( R = 0 . 20 , p = 0 . 608 , N = 9 , rho = 0 . 37 , p = 0 . 332 , N = 9 ) . These results show that the observed feature-based attentional modulations occur globally across the brain , and do not depend on bottom-up feature preference . How do we explain that attending color in the stimulus induced greater changes in spatial sampling ? Behavioral studies have suggested that the influence of spatial attention should be adjusted by feature-based attention to improve sampling of attended visual features ( Yeshurun et al . , 2008; Barbot and Carrasco , 2017 ) . Specifically , these authors suggested that attending features processed by relatively smaller receptive fields requires a greater degree of spatial resampling . This implies that if color-preferring voxels are relatively small , this could explain the greater degree of resampling observed when attending color . This is indeed predicted by the fact that color compared to temporal frequency information is predominantly processed by the parvocellular compared to the magnocellular pathway , where spatial sampling is generally more fine-grained ( Schiller and Malpeli , 1978; Hicks et al . , 1983; Denison et al . , 2014 ) . In addition , pRF size varies on average between visual regions , such that pRF size is generally larger in area MT+ ( preferring temporal frequency ) compared to hV4 ( preferring color ) . Finally , both pRF size ( Dumoulin and Wandell , 2008 ) and color compared to TF preference ( Curcio et al . , 1990; Azzopardi et al . , 1999; Brewer et al . , 2005 ) are known to be strongly eccentricity-dependent such that foveal voxels have relatively small pRFs and are relatively color-sensitive . We also clearly observe both effects in our data ( see Figure 2B and Figure 7 , correlation between feature-preference and eccentricity is negative except in LO and VO , see Supplementary file 1 - Table 10 ) . In sum , the greater amount of spatial resampling when attending color can be parsimoniously explained by color being sampled by relatively smaller pRFs . Smaller pRFs also require a more precise attentional gain field to shift a given distance ( a property of the multiplication of Gaussians ) . Combining this with our observation that pRFs experience greater shifts when attending color , we predict that attentional gain fields should be more precise in this condition . To test this , we repeated the attentional gain field modeling procedure described above , replacing the Attend Stimulus data with the Attend Color and Attend TF data in two separate fit procedures . Both our data ( Figure 5 ) and previous findings ( Klein et al . , 2014 ) showed that a single attentional gain field affects the different visual regions similarly . In addition , our results presented above showed that pRF modulation by feature-based attention was not related to feature preference across ROIs . We therefore analyzed feature-based attentional modulations of the attentional gain field both on data from all ROIs fitted together ( the 'combined ROI' ) , and as the median across individually fitted ROIs . This analysis returned smaller fitted stimulus attentional gain field sizes in the Attend Color compared to the Attend TF fit procedure ( Figure 8 ) both in the combined ROI ( 0 . 094 dva smaller over subjects when attending color , t ( 4 ) = 9 . 021 , p = 0 . 001 , Cohen's d = 4 . 511 ) and across ROIs ( median over ROIs on average 0 . 061 dva smaller over subjects when attending color , t ( 4 ) = 4 . 243 , p = 0 . 013 , Cohen's d = 2 . 121 ) . As the Attend Fixation data were used as input in both modeling procedures , we verified that the estimated fixation attentional gain field was not different between procedures ( p's of . 693 and . 224 and Cohen's d of −0 . 213 and −0 . 719 for across ROIs and combined ROI , respectively ) . These analyses show that the stronger influence of spatial attention when attending color is realized by a more precise attentional gain field located at the stimulus . In sum , our results suggest that ( 1 ) the attentional system adjusts its influence in accordance with the spatial sampling characteristics of units that prefer the attended feature and ( 2 ) that it does this equally across visual regions regardless of their bottom-up feature preference . To evaluate the stability of our results , we repeated all analyses for individual subjects . The figures and details for these results can be found in the supplements of the specific figures , and in additional statistical tables . This showed that although spatial attention resulted in somewhat different patterns of pRF shifts between subjects , these individual differences were well captured by the attentional gain field model . This suggests that individual differences in pRF changes are likely the result of the known individual differences in pRF parameter distributions ( e . g . eccentricity-size relations [Figure 2—figure supplement 1 , and see Dumoulin and Wandell , 2008] ) . In addition , these analyses showed that spatial resampling was consistently modulated by feature-based attention across subjects ( i . e . feature AMI was on average 0 . 059 greater than 0 ( F ( 1 , 4 ) = 18 . 868 , p = 0 . 012 , η2p = 0 . 394 ) , and was not different between ROIs ( F ( 8 , 32 ) = 0 . 631 , p = 0 . 746 , η2p = 0 . 066 ) ) . Finally , we verified that the pRF results were not affected by differences in fixation accuracy or behavioral performance ( see Figure 9 ) . To provide evidence in favor of these null hypotheses , we performed JZL Bayes factor analyses ( using JASP; Love et al . , 2015 ) as frequentist statistics are not capable of providing evidence for the null ( Altman and Bland , 1995 ) . We rotated recorded eye position to the direction of bar movement and computed the median and standard deviation of position along this dimension across bar passes per bar position ( Figure 9B ) . We next set up a model including the factor of attention condition ( 3 levels ) , bar position ( 24 levels ) and their interaction . We found that when predicting gaze position , the evidence was in favor of the null hypothesis with a Bayes Factor ( BF ) of 18620 . When predicting gaze variability , however , we found evidence against the null hypothesis with a BF of 5 . 980 . Evidence for including each of the factors ( condition , bar position and their interaction ) into the model returned BFs of 0 . 713 , 547 . 193 and 0 . 017 , respectively . The BF of 0 . 713 for the factor of condition means that we cannot determine whether gaze variability was different between conditions . However , even if this were the case , this could only lead to an offset in pRF size and not to changes in pRF position ( Levin et al . , 2010; Klein et al . , 2014; Hummer et al . , 2016 ) . In addition , any anisotropy in gaze position variability could potentially lead to offsets in pRF center positions . Nevertheless , these biases would be identical for all pRFs throughout the visual field . As we find that peripheral pRFs shift inwards and decrease in size and central pRFs shift outwards and increase in size , global offsets in pRF size and position cannot parsimoniously explain our results . More importantly , the analyses also showed that although bar position influenced gaze variability ( BF of 547 . 193 ) , it did not do so differently between attention conditions ( BF of 0 . 017 ) . Although we used a Quest procedure to equate difficulty across attention conditions and across different levels of eccentricity , it is possible that this procedure stabilized at a faulty difficulty level . To verify whether the Quest procedure successfully equated performance , we used a similar Bayesian approach , testing whether a model including attention condition ( three levels ) and stimulus eccentricity ( three levels ) influenced behavioral performance ( Figure 9A ) . This returned evidence for the null hypothesis with a BF of 6 . 25 . Together , these results show that differences in pRF parameters between conditions cannot be explained by either fixation accuracy or behavioral difficulty .
We investigated how spatial and feature-based attention jointly modulate the sampling of visual space . We found that directing covert spatial attention toward a moving bar stimulus altered the eccentricity and size of pRFs in concert . These changes in spatial sampling were parsimoniously explained by an attentional gain field model . Attending color changes within this stimulus induced stronger pRF changes compared to attending temporal frequency changes . These feature-based attentional modulations occurred globally throughout the brain , irrespective of a visual region's average feature preference . We suggest that the greater degree of spatial resampling when attending color is related to smaller pRF sizes in relatively color preferring voxels . In addition , we showed that the greater degree of spatial resampling when attending color is caused by a more precise attentional gain field on the stimulus . Previous behavioral reports suggested that the spatial scale at which an attended feature is processed influenced the degree of spatial resampling ( Yeshurun and Carrasco , 1998; Yeshurun and Carrasco , 2000; Yeshurun et al . , 2008; Barbot and Carrasco , 2017 ) . Specifically , features processed at a finer spatial scale require a greater degree of spatial resampling . This means that the greater degree of spatial resampling when attending color compared to temporal frequency could be explained by color being sampled at a finer spatial scale . This is a canonical difference between parvocellular small , color-sensitive receptive fields and magnocellular large , temporal frequency-sensitive receptive fields ( Schiller and Malpeli , 1978; Hicks et al . , 1983; Denison et al . , 2014 ) . In addition , both our data and previous findings show that color is preferentially processed by visual areas that on average have smaller receptive field sizes ( i . e . hV4 compared to MT+ in Liu and Wandell ( 2005 ) ; Amano et al . , 2009; Brouwer and Heeger , 2009 , Brouwer and Heeger , 2013; Winawer et al . , 2010 ) . Finally , both the current and previous studies show that pRF size ( Dumoulin and Wandell , 2008 ) and color compared to temporal frequency preference ( Curcio et al . , 1990; Azzopardi et al . , 1999; Brewer et al . , 2005 ) vary across eccentricity such that foveal voxels have smaller pRFs and are more color-sensitive . In sum , we suggest that the greater degree of spatial resampling when attending color compared to temporal frequency can be explained by the difference in spatial scale at which these features are processed . We therefore predict that our results should generalize to any other comparison of attended visual features as long as these features differ in their spatial scale . This includes attending different feature values such as high compared to low spatial frequency , or attending different feature dimensions such as faces ( broader spatial scale ) versus letters ( finer spatial scale ) . Electrophysiological studies on the interaction between feature-based and spatial attention generally measure overall response amplitudes rather than changes in spatial sampling . This implies that interactions between feature-based and spatial attention are weak to non-existent in relatively early stages of processing ( David et al . , 2008; Hayden and Gallant , 2009; Patzwahl and Treue , 2009; Zhang and Luck , 2009 ) , but develop at later stages of visual processing ( Hillyard and Münte , 1984; Handy et al . , 2001; Andersen et al . , 2011; Bengson et al . , 2012; Ibos and Freedman , 2016 ) , but see Egner et al . , 2008 ) . We add to this ( 1 ) that feature-based attention modulates the effects of spatial attention on spatial resampling , and ( 2 ) that these interactions occur globally throughout the brain , manifesting themselves in even the earliest cortical visual regions . Interactions between spatial and feature-based attention in the early stages of processing could be concealed when focusing on changes in response amplitude rather than changes in spatial sampling . However , it is important to note that measuring spatial sampling at the level of voxels does not allow us to determine whether observed changes in spatial sampling are the result of changes in spatial sampling of individual neurons , or rather the result of differential weighting of subpopulations of neurons within a voxel . Nevertheless , it has been shown that spatial attention does influence spatial sampling of individual neurons ( Connor et al . , 1997; Womelsdorf et al . , 2006 ) . Yet , future studies are required to extend our conclusions regarding the interactions between feature-based and spatial attention to the single neuron level . The greater degree of spatial resampling when attending color compared to temporal frequency occurred throughout the brain , irrespective of visual regions' preference for the attended features . In other words , while MT+ and hV4 differed greatly in their relative feature preference , both areas showed comparable pRF changes resulting from differences in feature-based attention . This stands in apparent contrast to previous studies reporting that feature-based attention selectively enhances responses in cortical areas specialized in processing the attended feature ( Corbetta et al . , 1990; Chawla et al . , 1999; O'Craven et al . , 1999; Schoenfeld et al . , 2007; Baldauf and Desimone , 2014 ) . However , attending a stimulus consisting of multiple features was shown to spread attentional response modulations of one of the object's feature dimensions to other constituent feature dimensions ( Katzner et al . , 2009; Çukur et al . , 2013; Kay and Yeatman , 2017 ) , albeit somewhat later in time ( ±60 ms; Schoenfeld et al . , 2014 ) . This could mean that the global pattern of pRF shifts we observed is caused by such an object-based attentional transfer mechanism . In addition , changing the sampling of visual space globally throughout the brain enhances stability in the representation of space . Different modifications of visual space per visual region would require an additional mechanism linking different spatial representations . Instead , the global nature of spatial resampling we observe supports a temporally dynamic but spatially consistent adaptation of visual space . An important remaining question pertains to the source of the interactions between feature-based and spatial attention . Signals of spatial selection are thought to originate from a network of frontal and parietal areas , identified using fMRI ( Shulman et al . , 2002; Silver et al . , 2005; Jerde et al . , 2012; Sprague and Serences , 2013; Szczepanski et al . , 2013; Kay and Yeatman , 2017; Mackey et al . , 2017 ) and electrophysiology ( Moore and Armstrong , 2003; Gregoriou et al . , 2009 ) . As we focused on careful measurement of spatial sampling in feature-sensitive visual cortex with a relatively small stimulus region , we did not include the frontoparietal regions containing large receptive fields in our analyses . A recent study suggested a central role for the ventral prearcuate gyrus for conjoined spatial and feature-based attentional modulations ( Bichot et al . , 2015 ) . Correspondingly , signals of feature selection in humans have been localized to a likely human homologue of this area , the inferior frontal junction ( IFJ; Zanto et al . , 2010; Baldauf and Desimone , 2014 ) . This region is therefore a possible candidate for controlling the interactions between feature-based and spatial attention . The average changes in pRF size and eccentricity for each visual region in our data are largely consistent with previous studies in which attention was devoted to a peripheral stimulus versus fixation ( Kay et al . , 2015; Sheremata and Silver , 2015 ) . Moreover , our analyses go beyond these average pRF changes by investigating the spatial structure of the complex pattern of pRF changes that resulted from such differential spatial attention . The resulting characterization details how the sampling of visual space by single voxel pRFs is affected by spatial attention , which is of specific relevance for future studies that determine spatial selectivity for voxel selections . First , we show that attending the stimulus compared to fixation caused pRFs to shift radially . Although a previous study reported a dominant horizontal shift direction ( Sheremata and Silver , 2015 ) , we suggest that the overrepresentation of the horizontal meridian ( Schneider et al . , 2004; Swisher et al . , 2007 ) made radially shifting pRFs appear as predominantly horizontal changes . Second , we report closely coupled pRF eccentricity and size changes that were dependent on pRF eccentricity . Specifically , we found that parafoveal pRFs shifted toward the periphery and increased in size , whereas peripheral pRFs shifted toward the fovea and decreased in size . This finding supports the resolution hypothesis of attention ( Anton-Erxleben and Carrasco , 2013 ) , which posits that spatial attention acts to reduce resolution differences between the fovea and periphery . We note that the functional implication of pRF size changes was recently questioned , as stimulus encoding fidelity was shown to be unaffected by pRF size changes ( Vo et al . , 2017 ) . However , the exact functional significance of changes in pRF size bears no consequence for the conclusions currently presented . As we observed a strong correlation between pRF eccentricity and size changes , we combined both measures into a single robust index . Our relevant quantifications are therefore agnostic to the potentially separable functional implications of changes in pRF size and eccentricity . The pattern of pRF shifts we observe is described well by an attentional gain field model ( Reynolds and Heeger , 2009; Klein et al . , 2014 ) . First , this highlights that a simple and well-understood mechanism underpins the apparent complexity of the observed pattern of pRF changes . Second , it extends the utility of such attentional gain field models to situations in which attention is dynamically deployed over space and time during the mapping of the pRF ( Kay et al . , 2015 ) . In agreement with earlier reports ( Klein et al . , 2014; Puckett and DeYoe , 2015 ) , we found that the best-fitting model implemented comparable attentional gain field sizes across visual regions . This strongly points to spatial attention being implemented as a global influence across visual cortex . We conclude that differences in pRF shift patterns between different visual regions depended primarily on differences in visual sampling ( i . e . differences in pRF center and size distributions ) rather than on differing attentional influences . Despite the broad correspondence between model fits and data , the model did not capture the observed decreases in pRF eccentricity of the most foveal pRFs in V1 . Two recent studies showed that in early visual areas , spatial attention shifted pRFs away from the attended location , but toward the attended location in higher visual areas ( de Haas et al . , 2014; Vo et al . , 2017 ) . Other studies showed that in precisely these visual regions , both the pRF and the attentional gain field are composed of a suppressive surround in addition to their positive peak ( Zuiderbaan et al . , 2012; Puckett and DeYoe , 2015 ) . We leave the question of whether these suppressive surrounds could explain such repulsive shifts in lower visual cortex for future research . As a more general aside , gain fields have been shown to influence visual processing at the motor stage ( Van Opstal et al . , 1995; Snyder et al . , 1998; Trotter and Celebrini , 1999 ) . Thus , our results further establish the close link between attentional and motor processes ( Rizzolatti et al . , 1987; Corbetta et al . , 1998 ) . In sum , we show that visuospatial sampling is not only affected by attended locations but also depends on the spatial sampling properties of units that prefer attended visual features . The global nature of these modulations highlights the flexibility of the brain’s encoding of sensory information to meet task demands ( Rosenholtz , 2016 ) .
Five participants ( two female , two authors , aged between 25 and 37 ) participated in the study . All gave informed consent , and procedures were approved by the ethical review board of the University of Amsterdam ( 2016-BC-7145 ) , where scanning took place . All statistical Tables referenced in this manuscript can be found in a separate file attached to this submission , termed ‘Supplementary file 1’ . These Tables specify the N , effect sizes and p-values for all ROIs and for the different statistical methods ( ‘super subject’ , ‘over subjects’ and ‘per subject’; see Materials and methods for definitions ) . | Much like digital cameras record images using a grid of tiny pixels , our own visual experience results from the activity of many neurons , each with its own receptive field . A neuron’s receptive field is the area of visual space – the scene in front of our eyes – to which that neuron responds . But whereas digital pixels have fixed locations , the receptive fields of neurons do not . If we switch our attention to a different area of the scene in front of us , visual neurons move their receptive fields to cover that area instead . We do not need to move our eyes for this to happen , just the focus of our attention . Moving receptive fields in this way enables the visual system to generate more detailed vision at the new attended location . Unlike a digital camera , the brain is thus much more than a passive recording device . But does the movement of receptive fields also depend on what we are attending to at a given location ? Paying attention to tiny details , for example , might require many receptive fields to move by large amounts to produce vision with high enough resolution . Van Es et al . have now answered this question by using a brain scanner to measure receptive fields in healthy volunteers . The volunteers focused on different visual features , such as color or motion , and to various visual locations . When the volunteers attended to color , their attention was more tightly focused than when they attended to motion . This might be because processing color requires fine-detail vision , whereas we can detect movement with our attention spread over a larger area . As a result , receptive fields moved more when the volunteers attended to color than when they attended to motion . Movement of visual receptive fields thus depends on what we attend to , as well as where we focus our attention . This adds to our understanding of how the brain filters the information bombarding our senses . This might lead to better diagnosis and treatment of disorders that include attentional problems , such as autism and ADHD . The results could also help develop artificial intelligence systems that , like the visual system , can process information flexibly to achieve different goals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Spatial sampling in human visual cortex is modulated by both spatial and feature-based attention |
The success of antimicrobial treatment is threatened by the evolution of drug resistance . Population genetic models are an important tool in mitigating that threat . However , most such models consider resistance emergence via a single mutational step . Here , we assembled experimental evidence that drug resistance evolution follows two patterns: ( i ) a single mutation , which provides a large resistance benefit , or ( ii ) multiple mutations , each conferring a small benefit , which combine to yield high-level resistance . Using stochastic modeling , we then investigated the consequences of these two patterns for treatment failure and population diversity under various treatments . We find that resistance evolution is substantially limited if more than two mutations are required and that the extent of this limitation depends on the combination of drug type and pharmacokinetic profile . Further , if multiple mutations are necessary , adaptive treatment , which only suppresses the bacterial population , delays treatment failure due to resistance for a longer time than aggressive treatment , which aims at eradication .
The rapid rise and spread of antimicrobial resistance severely curb the efficacy of drug treatments against pathogen infections . Treatment strategies are designed to maximize efficacy and limit toxicity , but their long-term applicability depends on the risk of resistance evolution ( Nielsen and Friberg , 2013; Foo and Michor , 2009; Bonhoeffer et al . , 1997 ) . This highlights the importance of careful consideration of drug type , dose , and duration to guarantee the desired patient outcome whilst also reducing the risk of resistance evolution ( Nielsen and Friberg , 2013; Drusano , 2004 ) . In order to prevent drug resistance and preserve drug efficacy , treatment strategies should also be guided by an understanding of resistance evolution and the ability to assess its risk ( Andersson et al . , 2020; Read et al . , 2011 ) – a task that is substantially facilitated through mathematical modeling ( Nielsen and Friberg , 2013; Andersson et al . , 2020; Read et al . , 2011; Clarelli et al . , 2020 ) . The main class of models used to predict drug action and treatment outcome are pharmacokinetic and pharmacodynamic ( PKPD ) models ( Clarelli et al . , 2020; Drlica , 2003; Abel Zur Wiesch et al . , 2017; Chakrabarti and Michor , 2017 ) , which describe the change in drug concentration over time ( PKs ) and the corresponding effect on a pathogen population ( PDs ) . PKPD approaches are most commonly employed to study the efficacy of treatment without considering the possibility of resistance evolution , but coupled with bacterial population models , they can be used to investigate drug resistance evolution over time ( Yu et al . , 2018 ) . One severely understudied aspect in such approaches is that there are two fundamentally different patterns of de novo antibiotic ( AB ) resistance evolution: ( i ) ‘single-step’ resistance: a single mutation provides high drug resistance ( Nielsen and Friberg , 2013; Drlica , 2003; Yu et al . , 2018 ) ; or ( ii ) ‘multi-step’ resistance: the accumulation of several mutations of low individual benefit is necessary for high-level resistance ( where high resistance here means higher than a given treatment dose ) . The availability of either pattern to a pathogen population under drug selection will affect the potential for resistance evolution and therefore the evolutionary dynamics in response to various treatment strategies . We focus on resistance by de novo mutations as long-lasting infections such as those caused by Pseudomonas aeruginosa become hard to treat due to resistance evolving via mutations within the host during the course of the treatment ( Oliver et al . , 2000 ) . Another example is tuberculosis ( TB ) , arguably the infectious disease that has caused the highest number of deaths globally ( Castro et al . , 2020 ) . During persistent TB infections , drug resistance evolves by chromosomal mutations while resistance by horizontal gene transfer ( HGT ) has not been observed ( Castro et al . , 2020 ) . HGT is a common path to resistance in hospital-acquired infections and in cases of shorter treatment durations , as exemplified by Staphylococcus epidermidis infections that became resistant by acquiring plasmids carrying genes for linezolid resistance ( Dortet et al . , 2018 ) . In this study , we will comprehensively study the influence of the mechanistic pattern of resistance evolution itself ( namely the benefits and costs of mutations ) by considering ‘single-step’ resistance vs . ‘multi-step’ resistance . The emergence of mutations and their selection depend on an interplay between various treatment factors like drug type , dose , and treatment duration . These factors have been studied before to various extent in isolation ( Nielsen and Friberg , 2013; Drusano , 2004 ) , although rarely how their interactions shape resistance evolution ( Martinez et al . , 2012; Olofsson and Cars , 2007 ) . We will first establish the existence of single-step and multi-step resistance patterns by reviewing evidence in the experimental literature , and then use the obtained parameter values to inform a stochastic PKPD model of multi-step resistance evolution , which we will explore under various treatment regimens . We will establish the fundamental differences between evolutionary dynamics emerging from these two patterns in one specific treatment setting , but also explore the impact of various clinically relevant treatment strategies . First , we will compare two types of drugs , ABs and antimicrobial peptides ( AMPs ) . AMPs are key components of innate defenses but also important new antimicrobial drugs , which work by disrupting the bacterial membrane ( Zasloff , 2002; Mookherjee et al . , 2020 ) – as opposed to ABs , which usually target intracellular structures . AMPs have been found previously to significantly reduce the risk of resistance evolution compared to conventional ABs ( Yu et al . , 2018; Lazzaro et al . , 2020 ) , partly explained by their distinct PDs like higher killing rates ( Yu et al . , 2018 ) . Second , we will consider three different shapes of drug PKs , which are all clinically relevant ( Nielsen and Friberg , 2013 ) , but have rarely been compared in a systematic manner ( Chakrabarti and Michor , 2017; Foo et al . , 2012 ) . These comprise fluctuating drug concentrations , increasing concentrations ( which are then maintained at the highest level ) , and finally constant ( which can be achieved in high-dose IV ( intravenous ) interventions ) . Third , as a number of recent studies have questioned the practice of ‘radical pathogen elimination’ ( Read et al . , 2011; Hansen et al . , 2017; Hansen et al . , 2020 ) , we will compare aggressive elimination treatment with adaptive suppression – a strategy where the drug concentration is regularly adapted to the pathogen load – in a multi-step mutational framework ( Hansen et al . , 2017; Gatenby et al . , 2009 ) . Lastly , depending on the drug type , resistance evolution can be shaped either by chromosomal mutations or HGT , or both ( van Hoek et al . , 2011; Woodford and Ellington , 2007 ) . Assuming a scenario where both options are available , we will study the relative importance of resistance resulting from de novo mutations as compared to HGT , which plays an important role in AB resistance evolution ( van Hoek et al . , 2011 ) , although likely not as much in AMP resistance ( Kintses et al . , 2019 ) . Taken together , this allows us to obtain an empirically informed modeling framework , which predicts evolutionary dynamics of ‘single-step’ resistance vs . ‘multi-step’ resistance in the context of drug type , PKs , and treatment strategies . We show how this framework provides critical insights into drug resistance emergence in clinically relevant treatment settings .
Experimental studies document single target mutations as well as a sequence of mutational steps to drug resistance evolution in bacterial populations ( Spohn et al . , 2019; Chevereau et al . , 2015; Melnyk et al . , 2015; Lofton et al . , 2013; Makarova et al . , 2018; Kubicek-Sutherland et al . , 2017 ) , but no systematic review of these patterns has been conducted so far . Here , we only selected studies that report on both parameters , benefit and costs of resistance ( see Materials and methods ) ( Spohn et al . , 2019; Chevereau et al . , 2015; Melnyk et al . , 2015; Lofton et al . , 2013; Makarova et al . , 2018; Kubicek-Sutherland et al . , 2017 ) , in order to obtain a complete picture of the mutational effects . We define the benefit and cost of a mutation as an increase in the minimum inhibitory concentration ( MIC ) and as a reduction in growth ( in the absence of drug ) , respectively . Despite differences in study setup and type of resistance mutations , we clearly found a wide range of effects , with a large number of benefits below typical clinical MIC breakpoint values ( defining whether a bacterial strain is resistant ) , which are often 10xMIC or higher ( EUCAST , 2020; Table 1 , Figure 1 ) – hence likely necessitating multiple mutations for high resistance . The corresponding fitness costs range from almost none to a 25% reduction of the population growth rate and show a very weak positive correlation ( R2 = 0 . 07 , p=0 . 09 ) with ( log ) benefit over all studies taken together ( Figure 1B , Figure 1—figure supplement 1 ) . In general , mutations seem likely to incur more costs than benefits . Notably , our literature search suggests a difference in mutational benefit available for two different antimicrobials: the average benefit of resistance mutations to AMPs is substantially lower than for commonly used ABs ( Table 1 , Figure 1—figure supplement 1 ) . In the following , we use the correlation observed with these assembled benefit and cost values to inform a PKPD model that reflects the two patterns of resistance evolution . We mainly investigated the rise of de novo resistance in a clonal , susceptible pathogen population , which is a common starting point for many clinically relevant infections ( Balmer and Tanner , 2011 ) , by extending a previously described stochastic PKPD model ( Materials and methods ) ( Yu et al . , 2018 ) . Specifically , we considered not only a single resistance mutation , but the potential emergence of a sequence of mutations , with each mutation conferring a certain ( additional ) benefit and cost ( Figure 1 ) . The number of mutations needed for ‘full’ resistance depends on the applied drug dose , but generally low mutational benefits are more likely to necessitate multi-step resistance evolution . To compare scenarios where a single mutation is sufficient to scenarios where several mutations have to arise in one cell , we ran the simulations over a range of mutational benefits ( 2–100 xMIC ) – and their correlated fitness costs ( Table 1 , Figure 1 ) – in combination with various drug doses ( 0 . 5–100 xMIC ) . Hence , the minimum number of mutations necessary for resistance was predetermined ( Figure 1—figure supplement 2 ) , and we investigated how this affects the potential for pathogen survival and mutational diversity under various treatment strategies ( PKs ) and for two different antimicrobials ( PDs ) as described below . Competition between the various mutant subpopulations was modeled by imposing a carrying capacity for bacterial growth and very low turnover as soon as this capacity is reached . First , we determined the probability of treatment failure by simulating change of the pathogen population over 200 hr under treatment with drugs ( PD ) parameters typical for bactericidal ABs ( Yu et al . , 2018; Supplementary file 1 ) being applied once every 24 hr ( PK ) . We assumed that the pathogen population initially consists of completely susceptible bacteria and defined a treatment as failed if the pathogen population was not eradicated after 200 hr . We found that the probability of treatment failure was always close to 1 for single-step resistance evolution , but decreased rapidly if multiple mutations were required . Notably , already if three mutations were necessary to overcome the applied dose , the probability of pathogen survival approached 0 ( Figure 2A , Figure 1—figure supplement 2 ) . The qualitative picture of these results was not dependent on the specific cost-benefit correlation that we are assuming for most of our simulations ( Figure 2—figure supplement 1 ) . One aspect of resistance evolution that is especially important when considering multiple mutations is the mutational diversity that arises in the pathogen population: high genetic diversity ( here meaning diversity in the resistance phenotype ) increases the probability that some individuals will be able to survive a given environment – such as treatment with other drugs ( Castro et al . , 2020 ) – and increases the adaptive potential overall ( Van Egeren et al . , 2018 ) . Using the Shannon index to determine the highest mutational diversity obtained in the population over the treatment period , we clearly observed higher diversity with single-step than multi-step resistance evolution ( Figure 2B ) , even if we increased the mutation rate proportionally to the number of mutations required ( Figure 2—figure supplement 2 ) . It can be shown analytically that a mutant strain can invade at the mutant-free equilibrium if the death rate of the sensitive strain is higher than the death rate of the mutant , where the mutant death rate is a combination of intrinsic and drug-induced death as well as the mutational cost ( Materials and methods ) . The observed higher diversity with single-step patterns seems counterintuitive as the need for multiple mutations should increase diversity ( Figure 2B ) , but can be explained as follows: at high drug doses and low benefits , this effect is due to extinction that effectively reduces genetic diversity , while at low doses and high benefits , high mutational costs inhibit the build-up of diversity . These findings agree with an experimental study showing that resistance alleles with low costs are favored ( Wichelhaus et al . , 2002 ) . Our results clearly show less resistance if multiple mutations are necessary , but the relative importance of the number of resistance mutations compared to other treatment considerations like the dose-response profile of a drug ( PD ) ( Yu et al . , 2018; Spohn et al . , 2019 ) or the administration mode ( PK ) required further investigation . Hence , we compared three different PKs: ‘peak’ ( fast absorption and exponential decay ) , ‘ramp’ ( slow , linear absorption and no decay ) , and ‘constant’ ( immediate absorption and no decay ) ( Figure 3A ) . Whereas constant PKs distinctly lowered the probability of treatment failure and the emergence of mutational diversity , peak and ramp PKs showed similar magnitudes of resistance evolution ( Figure 3B , C , Figure 3—figure supplements 1 and 2 ) . However , ramp PKs lead to more than twice the mutational diversity with multi-step resistance patterns ( Figure 3—figure supplement 1 ) , which suggests that treatment failure and pathogen diversity are connected in a non-trivial manner: while higher mutational diversity increases the risk of resistance evolution , neither its presence nor absence is obviously predictive of the treatment outcome ( Figure 2 , Figure 3—figure supplements 1 and 2 ) . The evolutionary dynamics can also be contrasted for different antimicrobial drugs , AMPs and ABs , by using two different PD parameter sets ( Materials and methods , Figure 1—source data 1 ) . Briefly , AMP treatments are characterized by higher killing rates , steeper dose-response curves ( Figure 1A ) , and lower mutation rates than AB ones ( Yu et al . , 2018 ) . Consistent with previous findings that AMPs lead to a lower risk of resistance evolution and a narrower mutant selection window ( MSW ) than ABs ( Yu et al . , 2018 ) , treatment failure and mutational diversity was lower for AMPs with peak and constant PK treatments ( Figures 2 and 3 , Figure 3—figure supplement 2 ) . Notably , in accordance with empirical studies ( Andersson and Hughes , 2014 ) , we generally see mutations accumulating at sublethal drug doses , but the maximal diversity is substantially lower in AMP treatments ( Figure 2B , Figure 2—figure supplement 1 ) . Interestingly , the steeper dose-response curve of AMPs seems to make their resistance dynamics more sensitive to the shape of the PK than those of ABs ( Figures 2 and 3 , Figure 3—figure supplements 1 , 2 , and 4 ) : in contrast to the other two PK profiles , ramp PKs lead to a drastic increase in treatment failure with AMPs , especially in multi-step scenarios ( Figure 3 , Figure 3—figure supplement 1 ) . Accordingly , for ramp PKs , AMPs did not perform better and under some conditions even worse than ABs ( Figure 3—figure supplement 4 ) . By varying the ramp duration ( or equivalently the rate of drug uptake ) , we found that there is an intermediate range ( 48–84 hr ) , which showed increased treatment failure with AMPs over ABs ( Figure 3—figure supplement 5A ) . Paradoxically , while a narrow MSW generally hinders the emergence of numerous mutations in the population , for ramp PKs it can lead to optimal selection conditions for the sequential emergence of increasingly higher resistance mutations due to the strong selection for the next mutation combined with sufficient time for its emergence . Hence , especially the risk of multi-step resistance is increased if AMPs are used with ramp treatments as compared to the other PKs ( Figure 3B , C ) . The broader selection windows in the presence of ABs , on the other hand , overlap and resistance mutations are less strongly favored ( Figure 3—figure supplement 5B ) . Overall , the number of resistance mutations was the main determinant of treatment outcome , but we also found a complex dependence on PK and PD characteristics . This complexity in resistance determinants raises the question in how far the type of drug action influences treatment outcome . Specifically , antimicrobials can have bactericidal action ( which we were modeling so far , through a drug-dependent death rate ) , but they can also act bacteriostatically , that is , decreasing bacterial growth . We would expect bacteriostatic antimicrobials to slow down the rise of mutations in comparison to bactericidal ones as the acquisition of mutations is also coupled to bacterial growth . However , we find that this is only true for ramp and constant PK treatments ( Figure 3—figure supplement 6A ) . Peak PKs allow for regrowth of bacterial cells due to drug decay , which increases bacterial survival and treatment failure , especially with multi-step resistance . Interestingly , mutational diversity only increased for AMP treatments ( Figure 3—figure supplement 6B ) . The conventional treatment goal is to ‘eradicate’ the pathogen population , but it has been suggested that under certain conditions ‘mitigation’ could be a superior strategy ( Hansen et al . , 2017; Hansen et al . , 2020; Gatenby et al . , 2009 ) , for example , if it is likely that a resistant subpopulation already exists at the beginning of the treatment . This strategy is called adaptive treatment as drug doses are adapted to keep the sensitive population as big as possible and the total pathogen burden below a given limit . ( In practice , this is challenging as it requires monitoring of the pathogen burden and adjusting drug doses accordingly , which is difficult to implement even for measurements of total within-patient loads . ) In adaptive treatment , the sensitive population provides a benefit by competitively inhibiting the resistant subpopulation , but also a risk by supplying mutational input ( Figure 4 ) . This trade-off creates a threshold for the size of the pre-existent resistant subpopulation above which adaptive treatment is more effective than aggressive ‘eradication’ in containing the infection ( Hansen et al . , 2017 ) . Previously , the threshold for adaptive treatment was derived in a single-step resistance scenario ( Hansen et al . , 2017 ) . When we incorporated adaptive treatment in our multi-step resistance framework ( Figure 4—figure supplement 1 ) , we found that the resistant subpopulation threshold above which adaptive treatment is more beneficial can be much lower in the multi-step scenario than in the single-step one ( Figure 4A ) . This can be intuitively explained by the fact that all ( partially ) sensitive bacteria serve as competitors for fully resistant cells , but only the subpopulation one mutation away from being fully resistant constitutes the risk population ( Figure 4B ) . Thus , with multi-step resistance there is a smaller population to supply resistant bacteria than with single-step resistance , changing the trade-off towards adaptive treatment . Additionally ( in scenarios where adaptive treatment is favorable ) , the difference between adaptive and aggressive treatment in the duration until treatment failure can be several-fold larger for multi-step than single-step resistance patterns ( Figure 4—figure supplement 2 ) . Hence , assuming either single- or multi-step evolution could lead to considerably different treatment strategy assessments with regard to treatment failure through resistant pathogens . In addition to chromosomal mutations ( Woodford and Ellington , 2007 ) , antimicrobial resistance can be conferred through HGT ( van Hoek et al . , 2011 ) , which could facilitate resistance in multi-step scenarios . To account for this possibility , we extended the model to allow for acquisition of a gene conferring full resistance , initially only at a low rate from the environment , and then at a density-dependent rate from other cells carrying the HGT gene ( for assumption and implementation details , see Materials and methods ) . The HGT gene always provided immediate resistance to the applied maximal dose , regardless of the benefit or costs conferred by mutations . In order to compare the population dynamics of these two main antimicrobial resistance acquisition mechanisms , we assumed that resistance through mutations or HGT can be acquired independently of each other and that their effects are multiplicative . Even though HGT carriers dominated the remaining pathogen population at the end of the treatment ( Figure 2—figure supplement 3 ) , the addition of HGT did not change the probability of treatment failure ( Figure 2—figure supplement 4 ) . This result holds true as long as the acquisition rate from the environment is lower than the mutation rate ( this constraint is examined further in the Discussion ) . Consequently , initial rescue of the population is due to mutations – and therefore dependent on the magnitude of the mutational benefit – whereas HGT resistance is acquired later during the infection , after which it spreads rapidly .
In this study , we compared the risk of drug resistance evolution patterns that either feature single resistance mutations with large costs and benefits or multiple steps involving mutations with smaller costs and benefits . We extended this comparison across a wide range of PD and PK profiles , which cover a multitude of antimicrobials and treatment strategies . We first showed that the single- and multi-step resistance patterns are relevant by gathering evidence of multi-step resistance patterns in the experimental literature ( Table 1 , Figure 1 ) . While it is intuitive that drug resistance requiring more than one mutation will arise more slowly , we find that it can be a surprisingly strong inhibitor of resistance evolution and mutational diversity , depending on the drug class and administration route ( Figure 3 ) . We demonstrated that the number of mutations necessary for resistance strongly affects predictions of treatment outcome and optimality with regard to antimicrobial resistance – in a manner that is robust to variations in mutation rates and in the cost per mutation ( Figure 2—figure supplements 2 and 5 ) . Experimental support for our simulation results comes from studies reporting that mutational input limited to low benefits ( Drlica , 2003 ) leads to decreased drug resistance evolution as compared to systems , in which high-benefit mutations are available ( Allen et al . , 2004 ) . Moreover , limited access to high-benefit mutations seems to curtail MIC increase beyond a certain threshold ( Chevereau et al . , 2015 ) . The pattern of resistance evolution ( single- and multi-step ) is likely to be associated with the molecular mechanisms of resistance for a given antimicrobial: as an overall rule , the magnitude of the resistance benefit correlates with the mechanism of resistance , for example , efflux pumps yield low benefits , whereas specific drug target mutations yield high benefits ( Hughes and Andersson , 2017 ) . Unfortunately , the specific mutations linked to the benefit and cost of mutations in our literature analysis ( Table 1 ) are generally not known . Overall , however , MIC increase was low for drugs , which typically show unspecific resistance mechanisms via two-component systems or lipopolysaccharide modifications – as generally seen for AMP resistance ( Lofton et al . , 2013; Makarova et al . , 2018; Kubicek-Sutherland et al . , 2017 ) – and high for drugs with typical resistance via specific target modifications , as seen for some AB classes ( e . g . , rifampicin resistance via RNAP subunit mutations ) ( Goldstein , 2014 ) . For most drugs , the prevalent resistance mechanisms are known ( van Hoek et al . , 2011 ) ; hence , this information can be used to determine drug and dosing regimens that minimize resistance evolution based on the inferred pattern of resistance evolution ( i . e . , using the probability that a single- or multi-step pattern is underlying resistance evolution ) . A recent study also suggests that resistance evolution in biofilms , which are often associated with clinical infections , is prone to occur through unspecific mechanisms , even if specific mechanisms are favored in planktonic cultures ( Santos-Lopez et al . , 2019 ) . Interestingly , the risk of resistance evolution does not seem to be related to the emerging mutational diversity in the population in a trivial manner as it is either limited by fast extinction or high mutational cost ( Figure 2 , Figure 3—figure supplements 1 and 2 ) . Reducing mutational diversity is however a worthwhile goal in its own right as mutational diversity can increase adaptation by fixing more mildly deleterious mutations , which can then act as stepping stones for multi-drug resistance evolution ( Van Egeren et al . , 2018 ) . Further , the diversity arising during the treatment period will help to determine if escalating the drug dose is expected to be beneficial or if , conversely , it would be detrimental because higher-resistance mutations are already present in the population and would be selected . The strength of this selection is determined by the MSW of the antimicrobial . Hence , the number of resistance mutations emerging over the treatment can be useful in estimating the width of the MSW – even though the diversity remaining at the end of the treatment will likely be lower ( Figure 3—figure supplement 3 ) . We find that mutational diversity arises from a combination of selection pressure , bacterial growth , and fitness costs and cannot be predicted from the mutational benefit or the probability of treatment failure alone . Diversity is also shaped in unexpected ways by interactions between the drug type and drug concentration changes , making drug choice not only dependent on the PD characteristics , but also the specific drug PK in the target body compartment . Notably , this can lead to more favorable assessment of a specific drug application mode for one type of drug ( e . g . , AMPs for bolus drug application ) , but a different mode for another drug ( e . g . , ABs for drug infusions ) . While we mostly focused on the action of bactericidal drugs in this study , we note that purely bacteriostatic effects can lead to different trends for PK and PD influence on treatment failure and mutational diversity , for example , making peak PKs the least favorable drug administration route ( Figure 3—figure supplement 6 ) . The unexpected complexity in predicting which treatment strategies will minimize resistance evolution highlights the need of critically evaluating assumptions such as single-step resistance made in current PKPD models . The role of specific drug characteristics in resistance evolution is exemplified by the steepness of the PD curve , κ . By analyzing the selection coefficients for various treatments , we find that κ governs not only the size of the MSW ( Yu et al . , 2018; Chevereau et al . , 2015 ) , but generally shapes the selection pressure for resistance evolution in a qualitative manner . ψmin , the minimal bacterial growth rate , on the other hand , leads to substantial quantitative changes in selection pressure , meaning that κ and ψmin shape the form and strength of drug selection independently ( Figure 3—figure supplement 5C ) . Ultimately , the interactions between PD and PK characteristics give rise to complex , and dynamic , fitness landscapes that are navigated by mutations of various benefit and cost sizes . Interestingly , AMP-like drugs show considerably more resistance evolution with ramp PKs than in the other PK scenarios . This is noteworthy as natural AMP expression patterns in the producing organisms resemble ramp PKs ( Johnston et al . , 2014; Haine et al . , 2008 ) . This finding could suggest another reason why natural AMP production in cocktails is favorable ( Zanchi et al . , 2017 ) as AMP cocktails will limit the selection pressure and potential for resistance evolution to individual components . Intuitively , we would expect that a gradual increase in drug concentration would facilitate the rise of multiple mutations and indeed we find that ramp PKs lead to the highest probability of treatment failure and mutational diversity ( Figure 3 ) . However , a high probability of treatment failure is still mostly observed with high mutational benefits ( Figure 3—figure supplement 1 ) , that is , limited with the small-benefit mutations likely associated with multi-step resistance ( Jochumsen et al . , 2016 ) . For clinical settings , our simulations caution that attention should be paid to the drug application mode when using AMPs . AMP-like colistin and daptomycin , for example , are typically applied as ( short ) IV treatments ( Liu et al . , 2011; Tsuji et al . , 2019 ) , which resemble peak PKs , and they are still active as last-resort drugs for multi-drug-resistant bacterial pathogens ( Liu et al . , 2011; Tsuji et al . , 2019 ) . Overall , our results agree with Yu et al . , 2018 in that AMP treatment lowers resistance evolution and mutational diversity . This is particularly notable as multi-step patterns seem to be the common mechanism by which AMP resistance evolves ( Table 1; Spohn et al . , 2019; Jochumsen et al . , 2016; Joo et al . , 2016 ) – thereby suggesting another advantage over ABs , for which single- and multi-step evolution is common ( Drlica , 2003; Weinreich et al . , 2006; Wistrand-Yuen et al . , 2018; Marcusson et al . , 2009; Jin and Gross , 1988 ) . Unfortunately , distributions of mutational effects have rarely been characterized experimentally for drug resistance , and even then only for a single mutational step ( Chevereau et al . , 2015 ) . We show , however , that this information is crucial as input for PKPD models to accurately predict resistance evolution and population diversity in response to drug treatment . Even between mutations involved in multi-step resistance to a single drug , benefit and costs of individual mutations are likely to vary ( Figure 1B ) . In addition , epistasis in benefit and or cost magnitude can facilitate or preclude certain evolutionary pathways ( Jochumsen et al . , 2016 ) . Both options can be easily included in our model , but empirical data in this regard is sparse , and we expect our main results with regard to PD and PK influence on single- and multi-step resistance to be robust to such changes . The empirical data that we used to inform our simulations did also not provide explicit information about potential compensatory mutations , which arguably can influence the dynamics of resistance evolution ( Andersson and Hughes , 2010 ) – although likely in a very complex manner , as recent studies suggest ( Dunai et al . , 2019 ) . According to our results , these mutations might even be a necessary means to allow multi-step resistance patterns to arise . If they emerge fast enough to compensate for the cost of the first mutation , they would increase the selection coefficient of this mutational subpopulation and thereby provide a stepping stone to high-level resistance . This might either be akin to crossing a fitness valley , if the first mutation does not provide a benefit , or it might facilitate climbing a fitness peak by making low-benefit mutations more favorable . For many antimicrobial drugs , resistance evolution can not only arise through chromosomal mutations , but also by acquisition of resistance genes through HGT ( van Hoek et al . , 2011 ) . Notably , our results highlight the importance of transfer rates as we find rescue of the pathogen population through HGT resistance only if the initial acquisition rate is higher than the mutation rate . HGT is not only dependent on the recipient population size but also on the donor population size , hence using typical experimentally measured conjugation rates of 10−11–10−13 ml cell−1 h-1 ( Licht et al . , 1999 ) , environmental donors have to be more abundant than 105 cells ml−1 to be faster than chromosomal mutation rates ( Rodríguez-Rojas et al . , 2014 ) of 10−6 , which might not always be the case at bacterial infection sites ( Stecher et al . , 2012 ) . This implies either ( i ) that HGT resistance is acquired after chromosomal mutations , ( ii ) that HGT spreads mostly at sublethal drug doses , or ( iii ) that acquisition rates from a pre-existent pool of HGT carriers are high . Plasmid transfer rates are likely increased at low AB doses ( Cairns et al . , 2018 ) , but generally they are highly variable , and even though they are biased towards spread between clone-mates , there seems to be no obvious correlation between transfer rates and genetic distance of donors and recipients ( Dimitriu et al . , 2019 ) . Hence , determining the relative importance of resistance evolution through HGT or chromosomal mutations is difficult , but for specific drugs like AMPs , for which spread of HGT resistance from the gut microbiota seems to be low ( Kintses et al . , 2019 ) , the risk of treatment failure is mainly shaped by the beneficial mutations available to the population . Most of our results assumed a completely susceptible pathogen population at onset of treatment , as seen in many bacterial infections ( Balmer and Tanner , 2011 ) . However , the fast growth and high mutation rates can lead to significant heterogeneity in bacterial populations and we would expect this ( neutral ) heterogeneity to increase treatment failure , even with multi-step resistance patterns , by giving the population a ‘head-start’ in the accumulation of mutations . This is indeed what we see with our model when we start from a heterogeneous population , but we still find on average less than 50% treatment failure in each multi-step resistance scenario ( considering various PKs and PDs ) ( Materials and methods , Figure 2—figure supplement 6 , Figure 3—figure supplement 7 ) . When starting from populations that likely already contain resistance mutations , aiming for ‘mitigation’ ( adaptive treatment ) can be more effective in reducing resistance spread than trying to completely ‘eradicate’ the pathogen population ( aggressive treatment ) . If multiple steps are necessary to obtain full resistance to the highest possible drug dose , we find that the threshold for choosing adaptive over aggressive treatment can be much lower than if only a single mutation were necessary ( Figure 4 ) . Additionally , in drug-free environments , we expect a lower frequency of resistant cells for multi-step patterns as it is less likely that neutral heterogeneity produces cells carrying all resistance mutations . Hence , the high competitive benefit is paired with a low risk for resistance evolution . Even though determination of the number and size of resistant subpopulations is very difficult in practice , this suggests that adaptive treatment is likely to be superior in containment of resistant infections for many drugs , for which multi-step patterns are the most common resistance mechanism . Further , the assumptions in our model are not specific to bacterial populations or antimicrobials , which makes them more broadly applicable to other drug treatments , like cancer therapy ( Chakrabarti and Michor , 2017 ) . Our results suggest a way forward to develop treatment strategies that – in addition to all other important considerations – explicitly account for the risk of drug resistance evolution .
We compounded a comprehensive set of experimental evolution studies ( or reviews thereof ) that measured both fitness costs ( usually growth rate reductions in the absence of drugs ) and benefits ( usually increases in MIC ) of AB or AMP resistance mutations within the same set of experiments . The studies used various bacterial species , including pathogenic isolates ( see Table 1 ) . From empirically measured data of sample replicates , we calculated costs as the arithmetic mean of 1- ( relative fitness to wildtype ) and the benefit as the geometric mean ( due to the logarithmic scale of MIC/IC50 measures ) of MIC or IC50 increase relative to the wildtype . ( Note: Chevereau et al . , 2015 used IC50 instead of MIC but our calculation of IC50 and IC90 – which is likely very close to MIC – in their data gave a good correlation [R2 = 0 . 45 , p<0 . 001] , which indicates that the benefits obtained from IC50 measurements are comparable to ones obtained from MIC measurements . ) As fitness measure , we considered only the measurements done in the same conditions ( media and temperature ) that was also used for experimental evolution , even if growth was also measured in different environments . We list the conditions of the various evolution experiments , MIC and fitness measurements in Table 1 , with the exception of Melnyk et al . , 2015 , where we only list the eight different pathogenic strains used , as this paper synthesizes 24 different studies , grown under various conditions . We obtained the type and average number of mutational events observed from supplemental data in most studies , but there was generally no possibility to link any individual resistance mutation with a specific cost and benefit . Hence , we divided the overall costs and benefits by the average number of observed ( adaptive ) mutations ( i . e . , mutations that were not observed in control lines ) , assuming that each mutation provides a similar share to the overall magnitude . As most studies have a very low number of mutational events linked to resistance , this assumption is not expected to lead to strong biases . Overall , the results from all of the studies gave only a very weak positive linear correlation between the log ( benefit ) and cost of a mutational event ( Figure 1B ) . Mutations seem to be more likely to incur costs than benefits . This result is largely determined by the large data set from Spohn et al . , 2019 , which gives a very weak correlation between cost and benefit ( Figure 1—figure supplement 1 ) , similar to the data points from Melnyk et al . , 2015 . The dataset from Spohn et al . , 2019 is the only one that fulfilled our criteria and directly compared AB and AMP mutational effects , which we summarize in Figure 1—figure supplement 1 . The calculated benefit and cost per mutation for each individual AB and AMP in the Spohn et al . , 2019 and Melnyk et al . , 2015 data is given in Figure 1—source data 1 . We combined a PD model , which connects the growth of bacterial ( mutant ) subpopulations to antimicrobial drug concentration ( Figure 1A; Nielsen and Friberg , 2013; Andersson et al . , 2020; Read et al . , 2011; Clarelli et al . , 2020; Yu et al . , 2018; Regoes et al . , 2004 ) , with a population model to predict the emergence of resistance mutations in individual bacterial cells . In the population model , bacteria can grow up to a certain carrying capacity and can accumulate mutations during replication at a certain rate . In order to simulate de novo mutation emergence , we started most of our simulations from a completely susceptible population M0 , but we also ran simulations starting from neutral diversity ( meaning that we ran the simulation for 50 hr without AB treatment and then started from the observed neutral heterogeneity ) ( Figure 2—figure supplement 6 , Figure 3—figure supplement 7 ) . We do not allow for reversion of resistance mutations , which has been found to be rare ( Dunai et al . , 2019 ) and likely does not play a role in multi-step resistance networks ( over the time frame of a single treatment period ) . The population dynamics is captured by the following deterministic equations ( which were implemented in a stochastic manner ) :dMidt=r∙1-ci-1∙u∙Mi-1∙1-MK+r∙1-ci∙1-u∙Mi∙1-MK-γ+γi∙Miwithi=0 , 1 , 2 , …M=∑iMi Here , Mi is the bacterial subpopulation carrying i mutations , r the wildtype growth rate ( set to 1 in our simulations ) , c the cost of each mutation , u the mutation rate , K the carrying capacity of the system , γ the natural death rate , and γi the death rate caused by drugs ( which captures the PD properties of a drug and the resistance level of the bacterial population via the mutational benefit ) . In our population model , cells die at a low intrinsic rate γ , whereas death due to antimicrobials ( γi ) is dependent on the properties of the antimicrobial applied , the benefit conferred by each mutation , and the PK profile . Specifically , γi is calculated from the maximal and minimal growth rates ψmax and ψmin ( note that ψmin can be negative in the presence of drugs , meaning that we generally consider bactericidal AB action ) , the ( time-dependent ) concentration of the drug a , the MIC of the mutation-free population ( set to 1 in our simulations ) , the benefit bi conferred by each mutation , and the sensitivity of the dose–growth relationship κ ( the Hill coefficient or steepness of the curve ) :γi= ( ψmax-ψmin ) ∙ ( a/ ( MIC∙bi ) ) κ ( a/ ( MIC∙bi ) ) κ-ψmin/ψmaxψmax=r∙1-ci-γ Considering bacteriostatic antimicrobial action can be achieved in our model by using a small ψmin value and incorporating antimicrobial effect into the growth , not the death term . Note that introducing antimicrobial action into the birth term here leads to density-dependent antimicrobial effects . This is not entirely unrealistic , considering persister bacteria , whose dormant state protects them from killing by ABs ( Kussell et al . , 2005 ) . However , bacteriostatic action in itself would result in a high bacterial presence at the end of the treatment – even if bacteria are fully susceptible to the antimicrobial – as intrinsic bacterial death is very low . Hence , we incorporated an extrinsic removal rate -γcl∙Mi , akin to immune system clearance of inert bacterial cells , with γcl=0 . 1h-1 being in a realistic range ( Roach et al . , 2017 ) . The model for bacteriostatic drug action is then given bydMidt= ( r∙1-ci-1-γi-1 ) ∙u∙Mi-1∙1-MK+ ( r∙1-ci-γi ) ∙1-u∙Mi∙1-MK-γ+γcl∙Miwithi=0 , 1 , 2 , …M=∑iMi The main interest of our study is the comparison of AB resistance evolution via ‘typical’ single mutations with complex , multi-step processes as shown in Jochumsen et al . , 2016 . The latter are characterized by a network of mutations of small benefits in multiple genomic resistance loci that create evolutionary pathways to high-level AB resistance ( Jochumsen et al . , 2016 ) . We model this mutation accumulation via sequential acquisition of mutations with a certain benefit and cost , that is , decreases in drug-induced death and decreases in the maximum growth rate . Benefits and costs of each mutation were taken from the positive correlation that was observed with literature values ( slope = 0 . 0087 ) – except for simulations testing the dependence of our results on this relationship , where we took a steeper correlation ( slope = 0 . 0467 ) ( Figure 2—figure supplement 1 ) . As benefits and costs are likely to vary , we also confirmed that our results are robust with regard to drawing benefits and costs of each mutation from a normal distribution . Similarly , we ran simulations with ‘peak PK’ , where only the first mutational benefit/cost was fixed ( i . e . , deciding if a single- or multi-step pattern was necessary ) and the other mutations were sampled from the whole range of benefits and costs obtained from the literature , independently of each other ( Figure 2—figure supplement 5 ) . We ignore the possibility of positive epistasis between these mutations ( which would speed up resistance evolution as fewer mutations would be required for higher levels of resistance ) , as well as the possibility of negative epistasis , which would limit access to some mutations and the available pathways to resistance ( thereby slowing down resistance evolution as ‘effective’ mutation rates might be lower than we assume in our model ) . Both of these processes are complex and not well understood , hence by ignoring these possibilities we aim to provide a more fundamental and intuitive comparison between single- and multi-step resistance evolution . Resistance mutation rates were generally kept the same for each simulation run ( i . e . , regardless of the benefit magnitude ) . In reality , there might be more mutations available that provide a low benefit – which are likely to be less specific and therefore have a larger genomic target size , but using higher mutation rates for mutations with lower benefits and costs – which was done proportional to the number of steps needed to obtain resistance – did not change our results noticeably ( Figure 2—figure supplement 2 ) . In our simulations , we used three different PK functions to evaluate resistance evolution dynamics . ‘Peak PK’ describes the intake of a drug with a certain period τ , which is absorbed instantaneously and then decays exponentially at rate k ( Yu et al . , 2018 ) :a ( t ) =∑nd∙ ( e-kt-n-1τ ) with n = 1 , 2 , … the number of times the treatment dose d is applied . For ‘constant PK’ , the drug concentration is independent of time and simplifies to a = d , whereas for ‘ramp PK’ , the drug concentration increases linearly over a time k2cmax ( hence the rate of drug concentration increase is given as d/k2cmax ) and then stays constant for the rest of the treatment period . The value for k2cmax used for most simulations ( 48 hr ) was taken from literature and describes an example of AMP production timing during a natural immune response ( Haine et al . , 2008 ) . The model was implemented in R using the package adaptivetau ( Johnson , 2019 ) for stochastic implementations via the Gillespie algorithm . We focused on stochastic simulations as we were particularly interested in the timing and probability of the de novo rise and fixation of multiple mutations . We used the package adaptivetau because it allows for time-varying reaction rates , which was necessary in order to incorporate drug-concentration-dependent death rates . It also allows for deterministic simulation of a subset of rates as we did not want the AB concentration to be stochastic . For increased accuracy , we changed the epsilon parameter ( which describes the tolerance of relative rate changes in step size selection ) to 0 . 01 , which we found gave the same results as exact simulations . We calculated treatment failure probability as the frequency of runs , in which bacteria were not eradicated at the end of the treatment period ( 200 hr ) . Mutational diversity was calculated using the Shannon index , which takes into account the richness and evenness of the distribution of mutant subpopulations , and either averaging the maximum per treatment period ( for most results ) or the end diversity ( Figure 3—figure supplement 3 ) over all simulation runs . Treatment failure probability and mutational diversity were plotted using the R function filled . contour , which , as far as we could ascertain , interpolates linearly between ( potentially irregularly ) spaced grid points . To increase the appeal of our figures , we increased the option nlevels from the default value of 20 to 50 . The values for treatment failure and diversity shown in the contour plots were then averaged over the whole multi- or single-step area ( colored triangles shown in Figure 2 ) in order to compare different treatment strategies . The difference in treatment failure and mutational diversity between the two antimicrobial classes ( PK profiles ) was obtained by subtracting the corresponding values after every simulation of an AMP treatment from the one obtained in a simulation for an AB treatment and plotting the individual resulting differences ( for 500 simulations ) as well as the density via violin plots . Model parameters other than benefit , cost , and drug dose are taken from Yu et al . , 2018 ( Supplementary file 1 ) . The two different antimicrobial drug classes were defined based on previous experimental and theoretical work ( Yu et al . , 2018; Rodríguez-Rojas et al . , 2014; Yu et al . , 2016 ) by using two parameter sets: for the AB class , the mutation rate was 3 * 10−6 per division , κ was 1 . 5 and ψmin was –5 h−1; whereas for the AMP class , the mutation rate was 10−6 , κ was 5 and ψmin was –50 h−1 . The described code has been made available as an R package ( Source code 1 ) . Selection coefficients for our PD model were calculated under the assumption that the sensitive population is very small compared to the carrying capacity , which is a good approximation to the selection pressure at the start of an infection . This means that we can neglect the logistic growth term in our calculations . As the results were very similar to assuming a population at the carrying capacity ( which is an approximation for an infection that has had time to establish itself ) , we will focus on the selection coefficients with a small starting population . Selection coefficients were determined through eigenvalues obtained from the Jacobi matrix given by dMdt=dM0dt⋮dMndt=N ( i ) *M ( t ) :N ( i ) =r∙1-u-γ+γ00r∙ur∙1-c1∙1-u-γ+γ1⋯⋯⋯0⋯⋮0r∙1-c1∙u⋱⋮⋮⋱0⋯⋱r∙1-cn-1∙1-u-γ+γn-10r∙1-cn-1∙ur∙1-cn∙1-u-γ+γn The eigenvalues of N ( i ) are its diagonal entries , which correspond to the net growth of each population . We calculated selection coefficients for each of the mutational subpopulations in our model as the difference in growth rates between bacteria with i mutations and bacteria with i–1 mutations ( i . e . , the difference between their eigenvalues ) :si=growthMi-growthMi-1= ( r∙1-ci∙1-u-γ+γi ) - ( r∙1-ci-1∙1-u-γ+γi-1 ) The AB concentration over time was calculated deterministically using the R package deSolve ( Soetaert et al . , 2010 ) in order to calculate the death rates due to antimicrobial treatment . The difference between the parameter sets for the two antimicrobial classes used here lies in the higher mutation rate u , lower κ , and higher ψmin for AB treatments ( Yu et al . , 2018 ) . Hence , we investigated the importance of the two PD parameters κ and ψmin by calculating the selection coefficients using the AMP parameter set and swapping either κ or ψmin with that of the AB parameter set . More generally , we can consider the Jacobi matrix for the resistant populations invading at the mutant-free equilibrium:N ( i ) = ( r⋅ ( 1−c ) 1⋅ ( 1−u ) ⋅ ( 1−M0∗K ) − ( γ+γ1 ) 0r⋅ ( 1−c ) 1⋅u⋅ ( 1−M0∗K ) r⋅ ( 1−c ) 2⋅ ( 1−u ) ⋅ ( 1−M0∗K ) − ( γ+γ2 ) ⋯0⋯0⋯⋮0⋯⋱⋱⋮⋮⋱0⋯⋯r⋅ ( 1−c ) n−1⋅ ( 1−u ) ⋅ ( 1−M0∗K ) − ( γ+γn−1 ) 0r⋅ ( 1−c ) n⋅u⋅ ( 1−M0∗K ) r⋅ ( 1−c ) n⋅ ( 1−u ) ⋅ ( 1−M∗K ) − ( γ+γn ) ) The criteria for invasion of a mutant into the susceptible population is then that the eigenvalue of the mutant has to be bigger than zero , that is , λn=r⋅ ( 1−c ) n⋅ ( 1−u ) ⋅ ( 1−M0∗K ) − ( γ+γn ) >0 Inserting the mutant-free equilibrium M0*=K∙1-γ+γ0r∙1-u yieldsγ+γ0>γ+γn ( 1−c ) nwhich means that bacterial cells with n mutations can invade if the death rate of the sensitive strain is higher than the death rate of the mutant normalized by the cost of the mutation ( s ) . We added HGT to the model by allowing for an additional resistance gene ( with benefit bp and cost cp ) to be acquired , which gives resistance in a single step . Hence , the benefit bp and the corresponding cost cp were adjusted with respect to the drug dose applied by using a benefit that would increase the MIC 20% above the applied drug dose and calculating the cost accordingly through the linear correlation obtained from Table 1 . We assume that the bacterial population under investigation has not yet acquired the HGT element , and initial transfer has to come from the environment , that is , initial conditions were the same as for simulations without HGT and Mp ( 0 ) = 0 . This gene can be acquired at a low rate α from the environment or at a density-dependent rate β , which we assumed to be on the same order of magnitude as the mutation rate ( Bakkeren et al . , 2019 ) . The horizontally transferred gene can be acquired by sensitive or mutant bacterial populations , and cells containing HGT resistance can still acquire further mutations ( but not further HGT resistance ) . Hence , we assume that HGT resistance is , for example , acquired via a specific resistance gene on a plasmid ( typically a plasmid can only be acquired once per cell ) and that the resistance gene from this plasmid ( e . g . , using enzymatic drug inactivation ) acts through a different mechanism than resistance by chromosomal mutation ( e . g . , modification of the drug target ) ( van Hoek et al . , 2011 ) . The equations were modified as follows:dMidt=r∙1-ci-1∙u∙Mi-1∙1-MK+r∙1-ci∙1-u∙Mi∙1-MK-γ+γi∙Mi- ( α+β∙Mp ) ∙MidMpidt=r⋅ ( ( 1−c ) i−1∗cp ) ⋅u⋅Mpi−1⋅ ( 1−MK ) +r⋅ ( ( 1−c ) i∗cp ) ⋅ ( 1−u ) ⋅Mpi⋅ ( 1−MK ) − ( γ+γpi ) ⋅Mpi+ ( α+β⋅Mp ) ⋅MiM=∑iMi+∑iMpiMp=∑iMpi Here , Mpi is the bacterial subpopulation carrying the HGT gene and i mutations , Mp the total number of HGT subpopulations , and M the total number of all bacterial populations . Relative population frequencies were calculated at the end of the treatment period by dividing the cell number of each subpopulation through the whole population size . In adaptive treatment , the goal is not to eradicate the bacterial population entirely but to adjust the treatment dose continuously in order to keep the pathogen level below a certain upper limit . Hansen et al . , 2017 calculated the threshold of resistant cells that are necessary at the beginning of the treatment for adaptive treatment to outperform aggressive treatment ( i . e . , giving the full dose right away ) , which is based on the idea that sensitive cells provide a risk for becoming resistant through mutation and a benefit through growth competition with the resistant cells at the same time . Hansen et al . , 2017 only considered one mutation to resistance , which means that their risk subpopulation and competitor subpopulation was the same . If we consider however sequential mutational steps , then the risk population only consists of the subpopulation one mutation away from full resistance ( which will be the mth mutation ) , whereas the competitor population for the fully resistant strain contains all ( partially ) sensitive bacteria ( i . e . , including mutant strains , which are not fully resistant to the highest possible treatment dose ) . Therefore , the threshold of resistant bacteria is given by ( compare to [4] in Hansen et al . , 2017 ) :Competitivebenefitof ( partially ) sensitives=MutationalriskfromMm−1r∙1-cm∙Mm∙δ∙Pmax-Mm=u∙r∙1-cm-1∙1-δPmax∙Mm-1Mm=u⋅ ( 1−δPmax ) ⋅Mm−1 ( 1−c ) ⋅δ⋅∑j=0m−1Mjwith ( Pmax−Mm ) =∑j=0m−1Mj Here , δ describes the strength of competition and Pmax the upper limit of acceptable pathogen burden . Note that here it is assumed that all pathogens ( regardless of drug sensitivity ) contribute equally to competition ( Hansen et al . , 2017 ) . This leads to a quadratic equation for the subpopulation with m mutations , Mm , which we used to calculate how the resistant population threshold for adaptive treatment ( i . e . , the initial density Mm0 above which adaptive treatment is more favorable ) differs between multi- and single-step resistance patterns ( Figure 4 ) . We implemented adaptive treatment in our model by setting a defined upper bound of acceptable pathogen cells , which was equal to the starting density in these simulations ( Hansen et al . , 2017 ) ( i . e . , assuming that the bacterial infection already progressed to a level at which treatment becomes necessary ) . We used a relatively low acceptable burden of 105 CFU , which is supported by bacterial loads in , for example , urinary tract infections ( Schmiemann et al . , 2010 ) . Note that defining an acceptable limit of pathogen burden in clinical settings is far more intricate as a patient’s individual biology will play a significant role and is beyond the scope of this paper . We adjusted the treatment dose in order to keep the pathogen load at or below this threshold value but the subpopulations of at least partially sensitive cells as big as possible ( Figure 4—figure supplement 1 ) : specifically , we increased the treatment dose to the MIC of the highest resistant subpopulation when its frequency exceeded 1% of the total population and the total pathogen load was higher than our set acceptable burden – until the maximum dose set for a specific treatment simulation was reached; after which the maximum dose was applied continuously . The ( partially ) sensitive cells serve as competitors for the resistant strain that carries a mutational growth cost and can be outcompeted at low drug doses ( Hansen et al . , 2017 ) . At the same time , subpopulations that are one step away from the resistant population provide a risk population as they are likely to gain resistance . For simulations of adaptive and aggressive treatment , we started from a population with neutral heterogeneity , meaning that we calculated the steady-state number of cells with a specific number of mutations given a certain cost ( and benefit ) in the absence of drug selection . As we want to compare the time difference to treatment failure between adaptive and aggressive treatment for single- and multi-step patterns , we initially add to this ‘neutral population’ the predicted number of resistant cells necessary to make adaptive treatment superior to aggressive treatment . The drug dose in adaptive treatments was then adjusted to keep the number of pathogens below the acceptable burden as described above . The time of treatment failure was determined as the time where the total pathogen population crossed 108 CFUs . We compared adaptive and aggressive treatment by dividing the time to treatment failure obtained from the adaptive strategy by the one obtained with the aggressive strategy , yielding the fold difference in treatment success duration . | The rise in antibiotic resistance is threatening our ability to treat bacterial infections . Bacteria often evolve resistance by acquiring new genetic mutations during the treatment period . Understanding how resistance emerges and spreads through a bacterial population is crucial to prevent antibiotic drugs from failing . Mathematical models are a useful tool for exploring how bacteria will respond to antibiotics and assessing the risk of resistance . Usually , these models only consider instances where bacteria acquire one genetic mutation that makes them virtually impervious to treatment . But , in nature , this is not the only possibility . Although some mutations do give bacteria a high level of resistance , numerous others only provide small amounts of protection against the drug . If these mutations accumulate in the same bacterial cell , their effects can combine to make the strain highly resistant to treatment . But it was unclear how the emergence of multiple mutations affects the risk of treatment failure and the diversity of the bacterial population . To answer this question , Igler et al . devised a mathematical model in which each bacterium is able to mutate multiple times during the treatment period . The model revealed that if one mutation provides a high level of resistance on its own , the risk of bacteria surviving treatment is very high . But , if it takes more than two mutations to achieve a high level of resistance , the risk drops to almost nothing . Igler et al . also found that the chance of bacteria evolving high enough resistance is affected by the type of antibiotics used and how fast the drug decays . With low-level resistance mutations , adapting treatment to maintain an acceptable number of sensitive bacteria as competitors for ( a small number of ) resistant bacteria was more effective at delaying treatment failure than trying to kill all the bacteria at once . These findings suggest that adjusting the treatment strategy used for bacterial infections according to the proportion of low- and high-level resistance mutations could slow down the evolution of resistance . To apply these models in the real world , it will be important to measure the level of resistance conferred by single mutations . The type of models used here could also predict the response of other diseases that resist treatment , such as cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2021 | Multi-step vs. single-step resistance evolution under different drugs, pharmacokinetics, and treatment regimens |
Genomic imprinting is an epigenetic phenomenon causing parent-of-origin specific differential expression of maternally and paternally inherited alleles . While many imprinted genes have been identified in plants , the functional roles of most of them are unknown . In this study , we systematically examine the functional requirement of paternally expressed imprinted genes ( PEGs ) during seed development in Arabidopsis thaliana . While none of the 15 analyzed peg mutants has qualitative or quantitative abnormalities of seed development , we identify three PEGs that establish postzygotic hybridization barriers in the endosperm , revealing that PEGs have a major role as speciation genes in plants . Our work reveals that a subset of PEGs maintains functional roles in the inbreeding plant Arabidopsis that become evident upon deregulated expression .
Genomic imprinting is an epigenetic phenomenon occurring in mammals and flowering plants that leads to parent-of-origin specific differential expression of maternally and paternally inherited alleles ( Gehring , 2013 ) . Recent screening of the seed transcriptome in various plant species revealed dozens to several hundreds of novel candidate imprinted genes in maize , rice , castor bean , and Arabidopsis thaliana ( Gehring et al . , 2011; Hsieh et al . , 2011; Luo et al . , 2011; Waters et al . , 2011; Wolff et al . , 2011; Pignatta et al . , 2014; Xu et al . , 2014 ) . While few reports demonstrate genes to be temporally imprinted in the plant embryo ( Jahnke and Scholten , 2009; Raissig et al . , 2013 ) , the vast majority of imprinted genes has been observed in the endosperm , the ephemeral triploid tissue derived after fertilization of the diploid central cell with a haploid sperm cell . In most angiosperms the endosperm initially develops as a syncytium and cellularizes after a defined number of mitotic divisions ( Li and Berger , 2012 ) . The right timing of endosperm cellularization is crucial for proper seed development , its failure results in deficient nutrient supply , which causes embryo arrest and eventually seed abortion ( Hehenberger et al . , 2012 ) . In Arabidopsis , endosperm cellularization is regulated by , among others , the type I MADS-box transcription factor AGL62 . Loss of AGL62 leads to precocious endosperm cellularization ( Kang et al . , 2008 ) , whereas increased AGL62 expression correlates with delayed or failed cellularization ( Erilova et al . , 2009; Tiwari et al . , 2010 ) . Similar effects on endosperm development have been observed in response to interploidy hybridizations . While maternal excess hybridizations cause precocious endosperm cellularization and reduced seed size , the reciprocal cross leads to endosperm cellularization failure and seed abortion in an accession-dependent frequency ( Scott et al . , 1998; Dilkes et al . , 2008 ) . This phenomenon establishes a postzygotic reproductive barrier by preventing the formation of viable triploid seeds and has been termed ‘triploid block’ ( Marks , 1966 ) . Dosage-sensitivity of the endosperm has been proposed to be a consequence of deregulated imprinted genes that are responsible for interploidy hybridization failure ( Haig and Westoby , 1989; Gutierrez-Marcos et al . , 2003; Kinoshita , 2007 ) . Indeed , in response to interploidy hybridizations many imprinted genes are deregulated ( Jullien and Berger , 2010; Tiwari et al . , 2010; Wolff et al . , 2011 ) and the paternally expressed imprinted gene ADMETOS ( ADM ) has been identified to be a causative gene responsible for abortion of triploid seeds upon paternal excess hybridizations in Arabidopsis ( Kradolfer et al . , 2013 ) . While the identification of ADM provided first evidence that imprinted genes can establish reproductive barriers , the question whether this is a more general phenomenon applying to other imprinted genes as well , remained unresolved . In this study we investigated the functional role of 15 PEGs during seed development in Arabidopsis . None of the analyzed peg mutants caused qualitative or quantitative abnormalities of diploid seed development , revealing that many PEGs do either not have an important functional role in Arabidopsis seeds or act redundantly with non-imprinted genes . However , 3 out of ten tested peg mutants rescued triploid seed abortion , uncovering a major role of PEGs in establishing postzygotic interploidy hybridization barriers .
Genomic imprinting has been proposed to have a major impact on seed development ( Haig and Westoby , 1989 ) . We tested this hypothesis by investigating whether loss of PEG function would negatively impact on seed development and viability . We examined 15 PEGs that were shown to be imprinted at 4 days after pollination ( DAP ) in reciprocal crosses between Col and Bur-0 accessions ( Wolff et al . , 2011 ) . While At2g36560 ( PEG5 ) , At4g05470 ( PEG8 ) , At1g67830 ( FXG1 ) , At1g17770 ( SUVH7 ) , At1g57800 ( VIM5 ) and AT1g48910 ( YUC10 ) were also identified to be imprinted in Col and Ler accessions , At1g11810 ( PEG1 ) , At1g49290 ( PEG2 ) , At1g60400 ( PEG3 ) , At1g66630 ( PEG4 ) , At3g49770 ( PEG6 ) , At3g50720 ( PEG7 ) , At5g15140 ( PEG9 ) , At1g34650 ( HDG10 ) and At4g31900 ( PKR2 ) were not identified as being imprinted in Col and Ler accessions by other studies ( Gehring et al . , 2011; Hsieh et al . , 2011; Pignatta et al . , 2014 ) . Due to the lack of small nucleotide polymorphisms ( SNPs ) for PEG1 , PEG9 , and PKR2 we only tested the imprinting status of the 6 remaining PEGs in reciprocal crosses of Col and Ler accessions at 4 DAP . Parent-of-origin specific expression could be detected for all the genes tested ( Figure 1—figure supplement 1 ) , revealing that all tested PEGs are consistently imprinted in different accessions . To analyze the functional role of PEGs during seed development , we obtained T-DNA insertion mutants for all genes ( Figure 1—figure supplement 2 ) . We tested the mRNA levels in all mutants that were not yet previously investigated and found them strongly reduced compared to wild type ( Figure 1—figure supplement 3 ) . Nevertheless , none of the analyzed mutants had increased levels of unfertilized ovules or seed abortion compared to wild-type plants ( Figure 1A ) and neither transmission through the male gametophyte was significantly impaired ( Chi–Square test , 1:1 segregation hypothesis; p > 0 . 3 ) ( Figure 1—figure supplement 4 ) . We investigated the possibility that PEGs have a positive impact on seed size by pollinating wild-type plants with pollen from heterozygous peg mutants and analyzed the size of mature seeds . This analysis revealed that there was no significant ( F-test; α = 0 . 01 ) difference between wild-type and peg/+ mutant seeds ( Figure 1B ) . We furthermore tested the possibility whether PEG function could have a more prominent effect when fitness of the maternal parent was compromised . We completely removed rosette leaves of Col mother plants 2 days prior to emasculation and repeated the crosses with peg/+ mutants . In agreement with a previous report ( Akiyama and Agren , 2012 ) , substantial loss of source tissue caused a reduction of seed number in the majority of crosses ( Figure 1—figure supplement 5 ) . Nevertheless , there was no significant ( F-test; α = 0 . 01 ) difference in seed size between wild-type and peg/+ mutant seeds ( Figure 1—figure supplement 6 ) , revealing that there is no general , non-redundant role of PEGs in controlling seed development and seed size in Arabidopsis . 10 . 7554/eLife . 10074 . 003Figure 1 . Impact of PEG function on diploid seed development . ( A ) Percentage of normal , unfertilized , and aborted seeds from self-fertilized wild-type and homozygous peg mutant plants . A minimum of 300 seeds was analyzed for each genotype . ( B ) Size measurements of mature seeds derived from crosses of maternal Col plants pollinated with heterozygous peg pollen ( black line ) were normalized , plotted on a histogram and distribution was compared with wild type control crosses ( grey line ) . A minimum of 400 seeds was analyzed for each cross . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00310 . 7554/eLife . 10074 . 004Figure 1—figure supplement 1 . Imprinting of PEGs in Col/Ler . Allele-specific expression analysis of indicated PEGs in siliques derived from crosses of Col x Ler and Ler x Col . Siliques were harvested at 4 DAP and allele-specific expression was tested by PCR and subsequent DNA sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00410 . 7554/eLife . 10074 . 005Figure 1—figure supplement 2 . T-DNA insertions in peg mutants . Schematic presentation of T-DNA insertions ( red triangle ) in peg mutants that were analyzed in this study and not previously published . Dark grey bars indicate exons , light grey bars indicate UTRs , black lines indicate introns and promoter regions are indicated by a dotted line . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00510 . 7554/eLife . 10074 . 006Figure 1—figure supplement 3 . PEG expression in peg mutants . Expression of PEG genes in peg mutants and wild type in whole siliques at 4 DAP , measured by qRT-PCR . Expression of PEG9 and SUVH7 was analyzed in peg9-1 and suvh7-1 . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00610 . 7554/eLife . 10074 . 007Figure 1—figure supplement 4 . Transmission analysis of peg mutant alleles through the male germ line . Frequency of peg mutants among the progeny of crosses derived from wild-type plants pollinated with heterozygous peg pollen . 24 seedlings were genotyped for each cross . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00710 . 7554/eLife . 10074 . 008Figure 1—figure supplement 5 . Effect of stress on seed set in Col x peg/+ crosses . Number of seeds per silique in crosses of unstressed and stressed Col plants pollinated with Col pollen or heterozygous peg pollen . A minimum of 8 siliques was analyzed for each cross . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 00810 . 7554/eLife . 10074 . 009Figure 1—figure supplement 6 . Seed size analysis under stress conditions . Size measurements of mature seeds derived from crosses of stressed Col plants pollinated with heterozygous peg pollen ( black line ) were normalized , plotted on a histogram and distribution was compared with wild-type control crosses ( grey line ) . A minimum of 400 seeds was analyzed for each cross . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 009 We tested the hypothesis that in addition to ADM other PEGs are involved in building the triploid block . We randomly selected 10 PEGs that were , with the exception of VIM5 , strongly up-regulated in triploid seeds ( Figure 2—figure supplement 1 ) , and generated double mutants with the omission of second division1 ( osd1 ) mutant . The osd1 mutation causes the formation of unreduced ( 2n ) male gametes at a frequency of almost 100% ( d'Erfurth et al . , 2009 ) . Therefore , using osd1 as pollen donor leads to the formation of almost 100% triploid seeds . We used pollen from plants homozygous for both mutations for crosses with wild-type Col mothers . Strikingly , 3 out of ten peg mutants were able to rescue triploid seed abortion; while crosses of Col x osd1 gave rise to 8% non-collapsed seeds , crosses of Col with suvh7 osd1 , peg2 osd1 and peg9 osd1 increased levels of non-collapsed seeds to 53% , 86% and 54% , respectively ( Figure 2A ) . The majority of non-collapsed triploid seeds germinated ( Figure 2A ) and developed into viable seedlings ( Figure 2B ) . Independent mutant alleles for suvh7 and peg9 introduced into the osd1 background caused a similar rescue effect on triploid seeds ( Figure 2—figure supplement 2 ) . As for peg2 no second mutant allele could be identified , we complemented the mutant phenotype with a genomic PEG2::PEG2 construct that restored triploid seed inviability , revealing that the peg2 phenotype is indeed caused by failure of PEG2 function ( Figure 2—figure supplement 2 ) . To test whether maternal loss of PEG function would impact on triploid seed rescue , we pollinated either osd1 or peg osd1 mutant pollen onto maternal plants mutant for the corresponding peg . However , no maternal effect on triploid seed rescue could be detected ( Figure 2C ) , revealing no impact of the maternal alleles of SUVH7 , PEG2 and PEG9 on the triploid block . Consistently , SUVH7 and PEG2 remained imprinted in triploid seeds ( Figure 2D ) . PEG9 has no polymorphism between Col and Ler accessions and could not be tested . 10 . 7554/eLife . 10074 . 010Figure 2 . PEGs establish interploidy hybridization barriers . ( A ) Percentages of non-collapsed and germinated seeds of wild-type plants pollinated with osd1 and peg osd1 pollen . Numbers on top of bars correspond to number of analyzed seeds . ( B ) Triploid seedlings 14 days after germination . Scale bar , 1 cm . ( C ) Percentages of non-collapsed and germinated seeds of peg mutant plants pollinated with osd1 and peg osd1 pollen . Numbers on top of bars correspond to number of analyzed seeds . ( D ) PEG2 and SUVH7 remain imprinted in diploid and triploid seeds . Siliques of crosses of Ler plants pollinated with Col or osd1 pollen were harvested at 4 DAP and imprinted expression was tested by PCR and subsequent DNA sequencing . Siliques of Col and Ler plants pollinated with Col and Ler pollen were used as controls . ( E ) Sections of seeds derived from crosses of Col plants pollinated with Col , osd1 , peg2 osd1 , peg9-1 osd1 and suvh7-1 osd1 pollen at 8 DAP . Scale bar , 0 . 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 01010 . 7554/eLife . 10074 . 011Figure 2—figure supplement 1 . PEG expression in triploid seeds . Log2 fold changes of PEG expression in triploid seeds compared to PEG expression in diploid seeds at 6 DAP , measured by whole genome deep sequencing transcriptome analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 01110 . 7554/eLife . 10074 . 012Figure 2—figure supplement 2 . Analysis of independent mutant alleles for suvh7 and peg9 and genomic complementation of peg2 . Percentages of non-collapsed seeds of wild-type plants pollinated with osd1 , suvh7-2 osd1 , peg9-2 osd1 or peg2 osd1;PEG2::PEG2 pollen . Numbers on top of bars correspond to number of analyzed seeds . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 012 SUVH7 encodes for a putative histone-lysine N-methyltransferase that however did not have histone methyltransferase activity against H3K9 in in vitro assays ( Ebbs and Bender , 2006 ) . PEG2 encodes for an unknown protein with no predicted structural domains , while PEG9 encodes for a galactose mutarotase , which catalyzes the first step of the Leloir pathway , the conversion of beta-D-galactose to alpha-D-galactose ( Holden et al . , 2003 ) . Galactose is a major constituent of the Arabidopsis endosperm cell wall in the form of pectic homogalacturan ( Lee et al . , 2012 ) and altered PEG9 activity could impact on cell wall composition . We tested whether rescue of triploid seed viability by suvh7 , peg2 and peg9 was associated with restored endosperm cellularization . While the endosperm of triploid seeds was completely uncellularized at 8 DAP , endosperm cellularization in triploid suvh7 , peg2 and peg9 seeds was almost complete at 8 DAP; only the over-proliferated chalazal cyst remained uncellularized ( Figure 2E ) . Failure of endosperm cellularization is responsible for embryo arrest ( Hehenberger et al . , 2012 ) . Consistently , embryos of peg2 , peg9 , and suvh7 triploid seeds developed , albeit delayed compared to wild-type diploid embryos ( Figure 2E ) . Triploid seed rescue by adm is associated with strongly decreased mRNA levels of type I AGAMOUS-LIKE MADS-box genes ( AGLs ) as well as PEGs ( Kradolfer et al . , 2013 ) . To test whether triploid rescue is generally connected with decreased expression of AGLs and PEGs , we generated whole-genome transcriptome data of seeds from triploid adm , suvh7 and peg2 mutants and analyzed expression of AGLs and PEGs that had increased mRNA levels in triploid seeds . Triploid seed rescue by adm and suvh7 was associated with a similar decrease of mRNA levels of many AGLs and PEGs ( Figure 3A , B ) , suggesting that ADM and SUVH7 share a common mode of action . Strikingly , while peg2 had the strongest effect on triploid seed rescue ( Figure 2A ) , the effect on gene expression was weakest among the three mutants . In particular , AGLs and PEGs that were strongly affected in adm and suvh7 triploid seeds were only weakly affected in peg2 triploid seeds ( Figure 3A , B ) . This suggests that peg2-mediated triploid seed rescue occurs independently of normalized AGL and PEG expression and may affect a pathway downstream of either AGLs or PEGs . In support of this view , most genes down-regulated in triploid peg2 seeds were similarly down-regulated in triploid adm and suvh7 seeds ( Figure 3C , Figure 3—source data 1 ) , suggesting that all three mutants affect a common downstream pathway . Genes commonly down-regulated in all three mutants were significantly enriched for genes involved in carbohydrate metabolism and in particular genes encoding for polygalacturonases ( Figure 3—figure supplement 1 ) . Polygalacturonan is the backbone of the major primary cell wall component pectin , which is degraded by polygalacturonases ( Atmodjo et al . , 2013 ) . Pectin degradation is assumed to be a key step in the deconstruction of plant cell walls ( Xiao et al . , 2014 ) , therefore suppression of pectin hydrolysis in adm , suvh7 , peg2 , and peg9 may be the key mechanism to induce endosperm cellularization and restore triploid seed viability . In agreement with this notion , PEG9 encodes an enzyme that acts on galactose , the carbohydrate building pectin . 10 . 7554/eLife . 10074 . 013Figure 3 . Transcriptome analysis of AGL genes and PEGs in triploid adm-2 , suvh7-1 and peg2 seeds . ( A ) Log2 fold change expression of AGLs in triploid adm-2 , suvh7-1 and peg2 mutants compared to triploid wild-type seeds . Only AGLs were tested that were up-regulated in triploid wild-type seeds . ( B ) Log2 fold change expression of PEGs in triploid adm-2 , suvh7-1 and peg2 mutants compared to triploid wild-type seeds . Only PEGs were tested that were up-regulated in triploid wild-type seeds . ( C ) Left panel: Venn diagram showing overlap of genes being up-regulated in seeds derived from wild type x osd1 crosses ( signal log ratio [SLR] > 1 , p < 0 . 05 ) and down-regulated in wild type x adm-2 osd1 ( SLR < −1 , p < 0 . 05 ) , wild type x suvh7-1 osd1 ( SLR < −1 , p < 0 . 05 ) , and wild type x peg2 osd1 ( SLR < −1 , p < 0 . 05 ) Hypergeometric testing was used to test for significance of overlap; p = 5 . 853e-09 . Right panel: Venn diagram showing overlap of genes being down-regulated in seeds derived from wild type x osd1 crosses ( SLR < −1 , p < 0 . 05 ) and up-regulated in wild type x adm-2 osd1 ( SLR >1 , p < 0 . 05 ) , wild type x suvh7-1 osd1 ( SLR >1 , p < 0 . 05 ) , and wild type x peg2 osd1 ( SLR >1 , p < 0 . 05 ) . Hypergeometric testing was used to test for significance of overlap; p = 4 . 622 e−16 . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 01310 . 7554/eLife . 10074 . 014Figure 3—source data 1 . List of genes deregulated in seeds derived from interploidy crosses . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 01410 . 7554/eLife . 10074 . 015Figure 3—figure supplement 1 . Gene ontology analysis . GO analysis of genes commonly up- and down-regulated in triploid adm-2 , suvh7-1 and peg2 mutants compared to triploid wild-type seeds . Only GO categories with a p-value <0 . 05 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 015 To test the hypothesis that endosperm cellularization failure in triploid seeds is a consequence of disturbed pectin degradation pathways , we analyzed expression of 173 genes acting in pectin degrading pathways ( pectate lyases ( GO:0030570 ) , pectin methylesterases ( GO:0030599 ) and polygalacturonases ( GO:0004650 ) ) . Of those , 33 genes were upregulated in triploid seeds ( Figure 4A ) . In contrast , expression of pectin biosynthesis genes ( Atmodjo et al . , 2013 ) was not negatively affected in triploid seeds ( Figure 4B ) , suggesting that pectin degradation rather than pectin biosynthesis is disturbed in triploid seeds . Most of the 33 pectin degradation genes remained repressed until the heart stage of embryo development in wild-type seeds ( Figure 4C ) and became expressed at the cotyledon stage of embryo development , concomitantly with endosperm degradation ( Figure 4C ) . In concordance with restored endosperm cellularization in triploid peg mutants , expression of 27 out of 33 pectin degradation genes that were upregulated in triploid seeds was normalized in at least one of the peg mutants ( Figure 4D ) . Pectins are polymerized and methylesterified in the golgi and secreted into the cell wall as highly methylesterified forms . Subsequently , they can be modified by pectinases such as pectin methylesterases that catalyse the demethylesterification of homogalacturonans releasing acidic pectins , which can be visualized by ruthenium red binding ( Downie et al . , 1998; Micheli , 2001 ) . Before endosperm cellularization , the ruthenium red signal was similar between diploid and triploid seeds ( Figure 4E; Figure 4—figure supplement 1 ) , suggesting that there were no major differences in pectin synthesis and degradation between diploid and triploid seeds . At 6 DAP the wild-type endosperm was largely cellularized and only a weak ruthenium red signal could be detected ( Figure 4E; Figure 4—figure supplement 1 ) , in agreement with pectin being deposited in the cell wall in a highly methylesterified form that is less intensively stained by ruthenium red ( Micheli , 2001 ) . Consistent with the expression of genes encoding for pectin degrading enzymes in 6 DAP triploid seeds ( Figure 4D ) , the ruthenium red signal in the uncellularized endosperm at 6 DAP seeds was substantially weaker compared to the signal in 4DAP seeds ( Figure 4E; Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 10074 . 016Figure 4 . Analysis of pectin biosynthesis and degradation genes in diploid and triploid seeds . ( A ) Venn diagram showing overlap of pectin degradation genes and genes being upregulated in seeds derived from wild type x osd1 crosses ( signal log ratio [SLR] > 1 , p < 0 . 05 ) and genes being downregulated in either wild type x adm-2 osd1 ( SLR < −1 , p < 0 . 05 ) , wild type x suvh7-1 osd1 ( SLR < −1 , p < 0 . 05 ) , or wild type x peg2 osd1 ( SLR < −1 , p < 0 . 05 ) . Hypergeometric testing was used to test for significance of overlap , p = 4 . 048 e−10 . ( B ) Venn diagram showing overlap of pectin biosynthesis genes and genes being down-regulated in seeds derived from wild type x osd1 crosses ( SLR <1 , p < 0 . 05 ) . ( C ) Cluster analysis of pectin degradation genes that are upregulated in triploid wild-type seeds , based on their expression in endosperm during different stages of diploid seed development ( Belmonte et al . , 2013 ) . Each row represents a gene , and each column represents a tissue type . Tissue types are: micropylar ( MPE ) , peripheral ( PE ) , chalazal ( CZE ) and cellularized endosperm ( CES ) derived from seeds containing embryos of the preglobular stage to the mature stage . Red or green indicate tissues in which a particular gene is highly expressed or repressed , respectively . ( D ) Log2 fold change expression of pectin degradation genes in triploid wild-type seeds ( compared to diploid wild-type seeds ) and triploid adm-2 , suvh7-1 and peg2 mutants ( compared to triploid wild-type seeds ) . Genes marked by an asterisk are not included in the seed transcriptome dataset ( Belmonte et al . , 2013 ) and are therefore not included in panel ( C ) . ( E ) Ruthenium red staining of sections of seeds derived from crosses of Col plants pollinated with Col and osd1 at 4 DAP and 6 DAP . Red color marks the presence of demethylesterified pectin , blue color is derived from the counterstain with toluidine blue . A minimum of 100 seeds was analyzed for each cross . Scale bar , 0 . 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 01610 . 7554/eLife . 10074 . 017Figure 4—figure supplement 1 . Additional ruthenium red staining of seeds . Ruthenium red staining of sections of additional seeds derived from crosses of Col plants pollinated with Col and osd1 at 4 DAP and 6 DAP . A minimum of 100 seeds was analyzed for each cross . Scale bar , 0 . 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10074 . 017 Together , our data reveal that a subset of PEGs has a functional role in establishing interploidy hybridization barriers and that bypass of the barriers occurs by mechanisms converging on endosperm cellularization likely by suppression of pectin degradation . Thus , a subset of PEGs maintains a functional role in Arabidopsis , which is only revealed upon deregulation in triploid seeds .
A . thaliana mutants adm-2 ( Kradolfer et al . , 2013 ) , pkr2-1 ( Aichinger et al . , 2009 ) and yuc10 ( Cheng et al . , 2007 ) have been described previously . peg1 ( SAIL_659_B07 ) , peg2 ( SALK_143382 ) , peg3 ( SALK_000257 ) , peg4 ( GABI_022C09 ) , peg5 ( SALK_142234 ) , peg6 ( WiscDsLox428C06 ) , peg7 ( SALK_103601 ) , peg8 ( SALK_022399 ) , peg9-1 ( SALK_102165 ) , peg9-2 ( SALK_047888 ) , fxg1 ( SAIL_888B03 ) , hdg10 ( SALK_116071 ) , suvh7-1 ( GABI_037C06 ) , suvh7-2 ( SALK_112939 ) , suvh7-3 ( WiscDSLox297300_14H ) and vim5 ( SALK_033892 ) T-DNA insertion mutants were received from the Arabidopsis stock center ( arabidopsis . info ) and mutant plants were identified using primers listed in Supplementary file 1 . The osd1 mutant ( d'Erfurth et al . , 2009 ) was kindly provided by Raphael Mercier . Being originally identified in the Nossen background , the mutant was introgressed into Col by repeated backcrossing over five generations . Plants were grown in a growth cabinet under long day photoperiods ( 16 hr light and 8 hr dark ) at 22°C . After 10 days , seedlings were transferred to soil and plants were grown in a growth chamber at 60% humidity and daily cycles of 16 hr light at 22°C and 8 hr darkness at 18°C . For all crosses , designated female plants were emasculated and the pistils were hand-pollinated 2 days after emasculation . For RNA sequencing , seeds from 20 siliques of Ler x Col , Ler x osd1 , Ler x adm-2 osd1 , Ler x suvh7-1 osd1 , and Ler x peg2 osd1 were harvested at 6 DAP into RNA later ( Sigma–Aldrich , St Louis , Missouri ) in duplicates and homogenized ( Silamat S5; IvoclarVivadent , Germany ) using glass beads ( 1 . 25–1 . 55 mm; Carl Roth , Germany ) . RNA was extracted following a modified protocol for the RNAqueous kit ( Ambion , Life Technologies , Carlsbad , California ) . RNA was purified by Qiagen RNeasy Plant Mini Kit ( Qiagen , Germany ) after residual DNA removed by 2 μL DNaseI ( Thermo-Scientific , Waltham , Massachusetts ) . Libraries were prepared using the Truseq RNA Sample Preparation Kit ( Illumina , San Diego , California ) and sequenced at the SciLife Laboratory ( Uppsala , Sweden ) on an Illumina HiSeq2000 on two lanes in 100-bp paired-end mode . Sequencing reads have been deposited as fastq files in the Gene Expression Omnibus ( Santos-González , 2015 ) . For qPCR expression analysis three siliques of each cross were harvested , flash-frozen in liquid nitrogen and samples were disrupted using glass beads ( 1 . 25–1 . 55 mm; Carl Roth ) and a Silamat S5 machine ( IvoclarVivadent ) . RNA was extracted using the RNeasy Plant Mini Kit ( Qiagen ) according to the manufacturer's instructions and residual DNA was removed using the RNase-free DNase set ( Qiagen ) . cDNA was synthesized using the first-strand cDNA synthesis kit ( Thermo-Scientific ) according to the manufacturer's instructions . Quantitative real-time PCR was performed using an iQ5 Real-Time PCR detection system ( Bio-Rad , Hercules , California ) in triplicates using Maxima SYBR green master mix ( Thermo-Scientific ) . Results were analyzed as described by Simon ( 2003 ) using ACTIN11 as a reference gene and qPCR primers are listed in Supplementary file 1 . Gene expression data were quality trimmed and mapped to the Arabidopsis TAIR10 reference genome using TopHat v2 . 0 . 10 ( Trapnell et al . , 2009 ) . A maximum of 1 alignment to the reference was allowed for any given read , and the minimum anchor length was 10 bases . Differentially regulated genes across the two replicates were detected using the rank product method ( Breitling et al . , 2004 ) as implemented in the Bioconductor RankProd Package ( Hong et al . , 2011 ) . The test was run with 100 permutations and gene selection was corrected for multiple comparison errors using a pfp ( percentage of false prediction ) < 0 . 05 . Gene ontology ( GO ) categories were identified using AtCOECiS ( Vandepoele et al . , 2009 ) . To determine the imprinting status of selected genes , RNA was extracted from crosses Col x Col , Ler x Ler , Ler x Col and Ler x osd1 at 4 DAP . Primers used for allele specific expression analysis are specified in Supplementary file 1 and PCR products were analyzed by DNA sequencing . Seeds were fixed and embedded with Technovit 7100 ( Heraeus , Germany ) as described ( Hehenberger et al . , 2012 ) . Five-micrometer sections were prepared with an HM 355 S microtome ( Microm , Germany ) using glass knives . Sections were stained for 1 min with 0 . 1% toluidine blue and washed three times with distilled water . For pectin analysis , sections were stained for 45 min with 0 . 025% ruthenium red , counterstained for 1 min with 0 . 1% toluidine blue and washed three times with distilled water . At least ten seeds were analyzed per genotype and timepoint . Microscopy was performed using a DMI 4000B microscope with DIC optics ( Leica , Germany ) . Images were captured using a DFC360 FX camera ( Leica ) and processed using Photoshop CS5 ( Adobe , San Jose , California ) . Siliques were harvested when they turned brown and prior dehiscence . Mature seeds were separated from the siliques , spread on a document scanner with backlight unit ( Scanmaker i800; Mikrotek , Taiwan ) and analyzed as previously described ( Herridge et al . , 2011 ) . Measurements were normalized by dividing by the average seed area and plotted on histograms . To determine statistical differences between sample and control crosses , Bonferroni corrected F-Tests were performed at a level of α = 0 . 01 . Seeds were surface sterilized in a container using chlorine gas ( 10 ml hydrochloric acid plus 50 ml sodium hypochlorite ) and incubated for up to 3 hr . To determine germination frequency , seeds were plated on ½ MS media containing 1% sucrose , stratified at 4°C for 2 days in the dark and grown in a growth cabinet under long day photoperiods ( 16 hr light and 8 hr dark ) at 22°C for 10 days . For transmission analysis , seedlings were harvested after 12 days and genotyped using primers specified in Supplementary file 1 . For the generation of a Col PEG2::PEG2 construct , At1g49290 including 1 . 5 kb of upstream secuence was amplified by PCR using primers specified in Supplementary file 1 . The product was cloned into pENTR/D-TOPO ( Invitrogen , Carlsbad , California ) , followed by clonase reaction with the pB7FWG2 vector ( Karimi et al . , 2002 ) , from which the 35S promoter was removed . The Col PEG2::PEG2 construct was introduced into the peg2/− osd1/+ double mutant using Agrobacterium tumefaciens-mediated transformation ( Clough and Bent , 1998 ) and transformants were selected on ½ MS media containing 30 mg/L phosphinotricin . Independent T2 lines were selected for single locus insertions and eight independent PEG2::PEG2 lines ( in peg2/− osd1/− background ) were used for pollinations onto Col wild-type plants . | When plants and animals reproduce sexually , their offspring inherit two copies of every gene , one from each parent , which are arranged in two sets of structures called chromosomes . In some tissues , one gene copy may be switched off—through a process called ‘genomic imprinting’—while the other copy remains active . In plants , genomic imprinting is vital for seeds to develop normally . It is particularly important in the tissue that provides nutrients for the growing embryo ( the endosperm ) , in which one of the copies of many genes are switched off . Genes inherited from the male parent that have been imprinted are known as paternally expressed imprinted genes ( PEGs ) . Unlike most animals , it is common for plants to have more than two sets of chromosomes . When plants with different numbers of chromosome sets cross-fertilize each other , their offspring may have three copies of every gene instead of two . These ‘triploid’ seeds often die because their endosperm fails to develop normally . This is due to the increased activity of imprinted genes , which causes changes in the activity of many other genes in the endosperm . Although it is known that genomic imprinting in the endosperm helps to establish this reproductive barrier , it is not clear what specific roles many of the imprinted genes play . Here , Wolff et al . switched off several different PEGs in the plant Arabidopsis to investigate how they affect seed development . The experiments show that in seeds that have the normal two copies of every gene , inactivating these imprinted genes does not affect seed development . However , in triploid seeds , inactivating three of the imprinted genes rescues seeds that would normally die . These genes encode proteins that activate pathways in the endosperm that promote the formation of cell walls , which is a crucial stage in seed development . Wolff et al . 's findings reveal how imprinted genes in the endosperm establish a barrier to reproduction by preventing seeds produced from crosses between plants with different numbers of chromosome sets from being able to survive . Reproductive barriers are a major obstacle in plant breeding , so understanding how these barriers form may open new avenues for developing new plant varieties . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Material",
"and",
"methods"
] | [
"plant",
"biology",
"developmental",
"biology",
"short",
"report"
] | 2015 | Paternally expressed imprinted genes establish postzygotic hybridization barriers in Arabidopsis thaliana |
Activation of the mechanistic/mammalian target of rapamycin ( mTOR ) kinase in models of acute and chronic pain is strongly implicated in mediating enhanced translation and hyperalgesia . However , the molecular mechanisms by which mTOR regulates nociception remain unclear . Here we show that deletion of the eukaryotic initiation factor 4E-binding protein 1 ( 4E-BP1 ) , a major mTOR downstream effector , which represses eIF4E activity and cap-dependent translation , leads to mechanical , but not thermal pain hypersensitivity . Mice lacking 4E-BP1 exhibit enhanced spinal cord expression of neuroligin 1 , a cell-adhesion postsynaptic protein regulating excitatory synapse function , and show increased excitatory synaptic input into spinal neurons , and a lowered threshold for induction of synaptic potentiation . Pharmacological inhibition of eIF4E or genetic reduction of neuroligin 1 levels normalizes the increased excitatory synaptic activity and reverses mechanical hypersensitivity . Thus , translational control by 4E-BP1 downstream of mTOR effects the expression of neuroligin 1 and excitatory synaptic transmission in the spinal cord , and thereby contributes to enhanced mechanical nociception .
De novo gene expression induced by noxious stimuli markedly contributes to the development of pain hypersensitivity ( i . e . , allodynia and hyperalgesia ) . Regulation of gene expression at the level of translation enables the cell to rapidly change its proteome by modulating the rate of mRNA translation without altering mRNA levels ( Sonenberg and Hinnebusch , 2009 ) . Upregulation of mRNA translation via activation of the mechanical/mammalian target of rapamycin ( mTOR ) by noxious stimuli has been proposed to sensitize primary nociceptors and spinal circuits ( Bogen et al . , 2012; Ferrari et al . , 2013; Jimenez-Diaz et al . , 2008; Melemedjian et al . , 2010; Obara and Hunt , 2014; Price and Geranton , 2009 ) . mTOR is an evolutionarily conserved serine/threonine kinase that controls cell homeostasis through key molecular processes including translation , lipid biogenesis , autophagy , and cytoskeleton organization ( Shimobayashi and Hall , 2014 ) . mTOR is the catalytic subunit of two structurally and functionally distinct multiprotein complexes , mTORC1 and mTORC2 ( Lipton and Sahin , 2014 ) . mTORC1 is defined by the protein raptor and is sensitive to rapamycin , while mTORC2 is defined by the protein rictor and is rapamycin insensitive . mTORC1 is a key regulator of translation , whereas mTORC2 controls the actin cytoskeleton ( Jacinto et al . , 2004; Sarbassov et al . , 2004 ) . The mTORC1 pathway is activated in primary nociceptors and superficial dorsal horn neurons in rodent models of inflammatory pain ( Jiang et al . , 2013; Liang et al . , 2013; Norsted Gregory et al . , 2010; Xu et al . , 2011 ) , bone cancer-induced pain ( Shih et al . , 2012 ) , neuropathic pain ( Zhang et al . , 2013 ) , and in response to repeated morphine administration ( Xu et al . , 2015; Xu et al . , 2014 ) . The functional role of mTORC1 activation has been studied using the mTORC1 specific inhibitor , rapamycin , and its derivatives ( rapalogues ) . Systemic or intrathecal ( i . t . ) administration of rapalogues does not affect acute responses to mechanical and thermal stimuli ( Geranton et al . , 2009; Xu et al . , 2014 ) , but it reduces nocifensive behaviors , and normalizes mechanical hypersensitivity in rodent models of inflammatory pain ( Asante et al . , 2009; Jiang et al . , 2013; Price and Geranton , 2009; Price et al . , 2007 ) , bone cancer-induced pain ( Shih et al . , 2012 ) , and neuropathic pain ( Asante et al . , 2010; Cui et al . , 2014; Zhang et al . , 2013 ) . Taken together , these findings indicate that noxious stimuli-induced activation of mTORC1 plays an important role in the development of pain hypersensitivity . However , the cellular and molecular mechanisms mediating the effect of mTORC1 on nociception are not known . All nuclear transcribed eukaryotic mRNAs harbor , at their 5’ end , the structure m7GpppN ( where N is any nucleotide ) termed ‘cap’ , which facilitates ribosome recruitment to the mRNA . Recruitment of the ribosome to the mRNA is primarily regulated at the initiation step . A critical step of this process is the assembly of the eukaryotic initiation factor 4F ( eIF4F ) complex , which consists of eIF4E , the cap binding protein subunit , eIF4G , a large scaffold protein , and eIF4A , an RNA helicase that unwinds the mRNA 5’ UTR ( untranslated region ) secondary structure . Because eIF4E generally exhibits the lowest expression level of all eukaryotic initiation factors , the cap-recognition step is rate-limiting for translation and a major target for regulation ( Sonenberg and Hinnebusch , 2009 ) . mTORC1 is a key regulator of translation initiation via phosphorylation of its downstream targets , eIF4E-binding proteins ( 4E-BPs ) and p70S6 kinases ( S6Ks ) . In their hypo-phosphorylated form , 4E-BPs compete with eIF4G for binding to eIF4E to repress eIF4F complex assembly , and thereby translation initiation . Following phosphorylation by mTORC1 , 4E-BPs dissociate from eIF4E , allowing for eIF4F complex formation and activation of translation ( Gingras et al . , 1996 ) . Three 4E-BP isoforms ( 1 , 2 , and 3 ) have similar functions , differing in their tissue distribution . 4E-BP3 expression is limited to few tissues , and is excluded from the nervous system . In the forebrain , 4E-BP2 is the major isoform , and 4E-BP2 knockout mice exhibit enhanced excitation ( Gkogkas et al . , 2013; Ran et al . , 2013 ) , alterations in synaptic plasticity , memory ( Banko et al . , 2007; Banko et al . , 2005 ) , and social behavior ( Gkogkas et al . , 2013 ) . In the suprachiasmatic nucleus , 4E-BP1 controls circadian clock functions via regulation of vasoactive intestinal polypeptide ( Vip ) mRNA translation ( Cao et al . , 2013 ) . In the pain pathway , 4E-BP1 is expressed in neurons in the dorsal horn of the spinal cord , and in peripheral nerves ( Jimenez-Diaz et al . , 2008; Melemedjian et al . , 2011 ) , whereas in the dorsal root ganglion ( DRG ) it was detected only in satellite glial cells ( Xu et al . , 2010 ) . Decreasing S6Ks activity by either pharmacological inhibitors or S6K1/2 null mutation unexpectedly led to mechanical allodynia via activation of ERK , a well-known pain sensitizing molecule ( Melemedjian et al . , 2013 ) . Activation of ERK is caused by disinhibition of the S6K1/IRS-1/ERK negative feedback loop ( Carracedo et al . , 2008 ) . Here , we investigated the role of the other major mTORC1 downstream effectors , 4E-BPs , in nociception using Eif4ebp1 and 2 knockout ( Eif4epb-/- ) mice . Removal of 4E-BP1 and 2 relieves eIF4E suppression and provides an opportunity to study cellular and molecular mechanisms underlying the impact of mTORC1 , via 4E-BPs , on nociception . Previous work showed that intraplantar administration of two well-known sensitizers of nociceptors , IL-6 and NGF , upregulates eIF4F complex formation and cap-dependent translation in primary afferent neurons ( Melemedjian et al . , 2010 ) . Peripheral inhibition of eIF4F with a specific inhibitor , 4EGI-1 , completely blocked IL-6- and NGF-induced mechanical allodynia , supporting the idea that enhanced eIF4F complex formation and cap-dependent translation contribute to sensitization of primary afferents . However , the roles of eIF4F and 4E-BPs in the regulation of spinal circuits are unknown . Here we show that Eif4ebp1-/- , but not Eif4ebp2-/- mice exhibit mechanical ( but not thermal ) hypersensitivity , which can be rescued by intrathecal administration of 4EGI-1 . Mechanical hypersensitivity is also elicited by downregulation of 4E-BP1 in the spinal cord dorsal horn . In Eif4ebp1-/- mice , the excitatory synaptic input into spinal neurons is enhanced and the threshold for potentiation of electrically-evoked field post-synaptic potentials ( fPSPs ) in the superficial dorsal horn is lowered . The enhancement of excitatory synaptic transmission is mediated via increased synthesis of neuroligin 1 , which promotes the excitatory synapse function . Reduction of neuroligin 1 levels in Eif4ebp1-/- mice normalizes excitatory synaptic transmission , and partially reverses mechanical allodynia . Collectively , our data demonstrate that de-repression of eIF4E , via removal of 4E-BP1 , contributes to mechanical allodynia through a neuroligin 1-mediated increase in the excitatory synaptic drive to the spinal circuit , and provides a new paradigm to explain the mechanisms by which mTORC1 controls nociception .
To study the role of 4E-BPs and eIF4F in nociception , we assessed mechanical and thermal sensation in Eif4ebp1-/- , Eif4ebp2-/- , and Eif4ebp1-/-/2-/- mice and wild-type controls . Eif4ebp1-/- mice exhibited a marked increase in mechanical sensation in von Frey and tail clip tests ( 45% decrease in withdrawal threshold and 52% reduction in latency to attack tail clip , respectively , Figure 1A ) , whereas their thermal sensitivity was unaffected in both the radiant heat paw-withdrawal and hot-plate tests ( Figure 1B ) . Additionally , in the chemical/inflammatory formalin test ( 20 μl , 0 . 5% formalin , intraplantar ) Eif4ebp1-/- mice showed higher levels of nocifensive behavior ( Figure 1C ) and increased c-Fos expression in the dorsal horn of the lumbar spinal cord ( Figure 1D ) . In contrast , Eif4ebp2-/- mice showed no mechanical ( von Frey and tail clip tests , Figure 1—figure supplement 1A ) or thermal pain phenotype ( radiant heat paw-withdrawal and hot-plate tests , Figure 1—figure supplement 1B ) , and no alterations in formalin test behavior ( Figure 1—figure supplement 1C ) . Mice lacking both 4E-BP1 and 4E-BP2 ( Eif4ebp1-/-/2-/- ) showed mechanical hypersensitivity and increased pain behavior in formalin test ( Figure 1—figure supplement 1D–F ) , a phenotype identical to Eif4ebp1-/- mice . We did not detect any reorganization of the dorsal horn of Eif4ebp1-/- mice as assessed by neuronal marker NeuN ( Figure 1—figure supplement 2A ) , projection of peptidergic ( expressing substance P ( SP ) and calcitonin gene-related peptide ( CGRP ) , Figure 1—figure supplement 2B , C ) or non-peptidergic ( expressing isolectin B4 ( IB4 ) , Figure 1—figure supplement 2D ) C fibers , distribution of glial marker glial fibrillary acidic protein ( GFAP , Figure 1—figure supplement 2E ) , and 5HT3A receptor ( HTR3A , Figure 1—figure supplement 2F ) . To exclude the possibility that the phenotypes observed in Eif4ebp1-/- animals were the result of long-term developmental compensations and to test the role of 4E-BP1 selectively in the spinal cord , we knocked down Eif4ebp1 in the dorsal horn of adult wild-type mice using an intraparenchymal injection of lentivirus expressing shRNA against Eif4ebp1 . The virus-driven eGFP expression was restricted to the dorsal horn and showed mostly neuronal localization ( Figure 1—figure supplement 3 ) . Seven days after injection of Eif4ebp1 shRNA into the dorsal horn of the lumbar spinal cord , mice showed reduced levels of 4E-BP1 protein in the lumbar spinal cord ( 53 ± 9% reduction , Figure 1E ) , but not in DRGs , and exhibited mechanical , but not thermal hypersensitivity ( Figure 1F–G ) . To study the distribution of 4E-BP1 and 4E-BP2 , we performed western blot analysis of proteins prepared from DRG and superficial spinal cord , which showed the presence of 4E-BP1 , but much less of the 4E-BP2 isoform ( Figure 1—figure supplement 2G ) . Taken together , our data demonstrate that 4E-BP1 is the predominant functional isoform in the pain pathway and its downregulation in the spinal cord elicits mechanical hypersensitivity . 10 . 7554/eLife . 12002 . 003Figure 1 . Eif4ebp1-/- mice exhibit mechanical hypersensitivity and increased formalin response . Mechanical pain sensitivity is increased in Eif4ebp1-/- mice as evident by decreased von Frey thresholds ( A , n=8/genotype ) and shortened latencies to attack tail clip ( A , n=8/genotype ) , whereas thermal sensitivity is not altered ( B , n=8/genotype ) . ( C ) Eif4ebp1-/- mice show more nocifensive ( licking/shaking ) behavior of the injected hind paw than their wild-type ( WT ) littermates in the formalin test ( 0 . 5% , 20 μl , n=8/genotype ) . Changes in paw weight , indicative of formalin-induced inflammation , were not different in Eif4ebp1-/- mice ( not shown ) . ( D ) Intraplantar injection of formalin ( 0 . 5% , 20 μl ) induced an enhanced upregulation of c-Fos ( 2 h post injection ) in superficial layers ( I-II ) of the lumbar spinal cord of Eif4ebp1-/- as compared to WT mice ( n=3 mice/group , 10 sections/mouse ) . ( E ) Western blot analysis of lysates prepared from the dorsal horn of the lumbar spinal cord and L3/L4 DRGs seven days post intraparenchymal dorsal horn injection of lentiviruses expressing shRNA against Eif4ebp1 as compared to non-targeting ( scrambled ) shRNA ( n=3/condition ) . Mice injected with shRNA against Eif4ebp1 exhibit a reduction in von Frey threshold 7 days post injection ( F , n=9/condition ) , whereas thermal sensitivity is not altered ( G , n=9/condition ) . Data are presented as mean ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ns–not significant by Student’s t-test or t-test following repeated measures ANOVA . Scale bar: 100 μm . See also figure supplement 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 00310 . 7554/eLife . 12002 . 004Figure 1—figure supplement 1 . Eif4ebp1/2 DKO mice , but not Eif4ebp2-/- , mice exhibit mechanical hypersensitivity and enhanced formalin responding . Mechanical and thermal pain sensitivities are not altered in Eif4ebp2-/- mice as evident by data from von Frey and tail clip assays ( A , n=9/genotype ) , and radiant heat paw-withdrawal and hot-plate tests ( B , n=9/genotype ) . ( C ) Eif4ebp2-/- mice show no difference in nocifensive ( licking/shaking ) behavior as compared to wild-type littermates in the formalin test ( 0 . 5% , 20 μm , n=8/genotype ) . ( D-F ) Eif4ebp1/2 DKO mice show mechanical hypersensitivity ( D , n=7/genotype ) and enhanced pain behavior in the formalin test ( F , n=7/genotype ) , but no alterations in thermal sensitivity ( E , n=8/genotype ) . Data are presented as mean ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ns–not significant by Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 00410 . 7554/eLife . 12002 . 005Figure 1—figure supplement 2 . No gross alterations in the dorsal horn of the Eif4ebp1-/- mice . To assess the possibility of spinal cord reorganization in Eif4ebp1-/- mice , we performed immunostaining of NeuN ( A ) , Substance P ( SP , B ) , calcitonin gene-related peptide ( CGRP , C ) , isolectin B4 ( IB4 , D ) , glial fibrillary acidic protein ( GFAP , E ) , and 5HT3A receptor ( HTR3A , F ) in lumbar spinal cord sections . No gross alterations were identified in Eif4ebp1-/- mice with any marker ( n=3 mice per genotype , 12 sections/marker ) . ( G ) Western blot analysis of lysates prepare d from DRG and dorsal horn for 4E-BP1 and 4E-BP2 . Scale bar 150 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 00510 . 7554/eLife . 12002 . 006Figure 1—figure supplement 3 . Distribution of lentivirus-driven eGFP expression . Lentiviruses expressing eGFP were microinjected ( intraparenchymal injection ( IPI ) ) directly into the dorsal horn of the lumbar spinal cord . Seven days later the spinal cord was extracted and immunostained for a neuronal marker NeuN . A representative lumbar spinal cord low magnification micrographs ( upper panels ) and confocal high magnification images ( lower panels , NeuN red , eGFP green ) . Scale bar 150 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 006 4E-BP1 suppresses eIF4F complex formation by competing with eIF4G for binding to eIF4E . Thus , removal of 4E-BP1 leads to increased levels of the eIF4F complex ( Gkogkas et al . , 2013; Pause et al . , 1994; Tahmasebi et al . , 2014 ) . Therefore , we measured eIF4F complex levels in Eif4ebp1-/- mice using a cap-binding assay ( see experimental procedures ) . As expected , in the lumbar spinal cord lysates , the amount of eIF4G associated with the cap-bound eIF4E was higher in Eif4ebp1-/- mice as compared to wild-type mice ( Figure 2A ) . Next , we used an inhibitor of the eIF4F complex , 4EGI-1 , which binds to eIF4E and inhibits its interaction with eIF4G . Daily i . t . administration of low doses of 4EGI-1 ( 10 μg ) over three days normalized the enhanced binding of eIF4G to eIF4E in Eif4ebp1-/- mice ( Figure 2A ) , and rescued the mechanical hypersensitivity and enhanced formalin-induced pain behavior in these mice ( Figure 2B–C ) . These data indicate that the mechanical sensitivity and increased responses in the formalin test in Eif4ebp1-/- mice are a consequence of augmented eIF4F complex formation . 10 . 7554/eLife . 12002 . 007Figure 2 . Increased levels of eIF4F complex in Eif4ebp1-/- mice cause mechanical hypersensitivity and increased formalin response . ( A ) Left , Immunoblot analysis of cap column pull-down proteins prepared from the spinal cord of Eif4ebp1-/- and wild-type mice ( WT ) treated with 4EGI-1 ( 10 μg , i . t . once a day for 3 days ) or vehicle . Right , quantification of eIF4G1 in cap column pull-down material ( n=4/genotype/drug ) . ( B ) Chronic 4EGI-1 treatment rescues mechanical hypersensitivity of Eif4ebp1-/- mice . von Frey threshold was measured prior to ( baseline ) and after 4EGI-1 chronic treatment ( 10 μg , i . t . , once a day for 3 days , n=8/genotype/drug ) . ( C ) Chronic 4EGI-1 treatment normalizes the enhanced nocifensive ( licking/shaking ) behavior of Eif4ebp1-/- mice during the early ( 0–10 min post-formalin , 0 . 5% ) and late phases ( 10–60 min post-formalin ) of the formalin test ( n=8/genotype/drug ) . Data are presented as mean ± SEM . *p<0 . 05 , **p< 0 . 01 by Student’s t-test following two-way ( genotype x drug ) ANOVA ( in A , C ) or following two-between ( genotype , drug ) , one-within ( repeated measures ) ANOVA ( in B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 007 Next , we investigated the molecular and cellular mechanisms underlying the mechanical hypersensitivity of Eif4ebp1-/- mice . We focused on spinal mechanisms by which 4E-BP1 controls mechanical sensitivity . mTOR and 4E-BPs are strongly implicated in the regulation of synaptic transmission and synaptic plasticity in pyramidal hippocampal neurons ( Gkogkas et al . , 2013; Ran et al . , 2013; Richter and Klann , 2009; Santini et al . , 2013; Weston et al . , 2012 ) . Thus , we measured spontaneous miniature excitatory and inhibitory postsynaptic inputs into lamina II spinal neurons in spinal cord slices prepared from Eif4ebp1-/- and wild-type mice . The amplitude of miniature excitatory postsynaptic currents ( mEPSCs ) was increased in spinal neurons in Eif4ebp1-/- as compared to wild-type slices , whereas there was no statistically significant change in mEPSC frequency ( Figure 3A ) . The amplitude and frequency of inhibitory postsynaptic currents ( mIPSC ) were increased in Eif4ebp1-/- mice ( Figure 3B ) . In the spinal cord , fast inhibitory synaptic transmission is mediated via GABA and glycine receptors . Interestingly , while glycinergic mIPSC amplitude and frequency were enhanced , no alterations were measured in GABAA-mediated synaptic transmission ( Figure 3C ) . The increase in the mEPSC amplitude in lamina II spinal neurons of Eif4ebp1-/- was reversed by i . t . administration of 4EGI-1 ( 10 μg , daily for 3 days , Figure 3D ) , however , 4EGI-1 had no effect on the increased amplitude and frequency of mIPSC ( Figure 3E ) . The reasons for these differences will be addressed in the Discussion . 10 . 7554/eLife . 12002 . 008Figure 3 . Excitatory and inhibitory synaptic transmissions are increased in the spinal cord of Eif4ebp1-/- mice . Representative traces ( top ) of mEPSCs ( A ) and mIPSCs ( B ) from lamina II neurons in acute lumbar spinal cord slices from Eif4ebp1-/- and wild-type mice ( WT ) . Bar graphs ( bottom ) show mEPSC ( A ) and mIPSC ( B ) amplitude , frequency and synaptic total charge transfer ( A: n=8 WT; n=7 Eif4ebp1-/-; B: n=9 WT; n=7 Eif4ebp1-/- ) . ( C ) To isolate GABAA or glycinergic mIPSCs , strychnine ( STRY , glycine receptor antagonist , 1 μM ) or bicuculline ( BIC , GABAA receptor antagonist , 10 μM ) were used , respectively . Left: representative traces of mIPSC in the presence of strychnine or bicuculline from Eif4ebp1-/- and WT slices . Right: bar graphs showing mIPSC amplitude and frequency in the presence of strychnine ( n=7 WT; n=8 Eif4ebp1-/- ) or bicuculline ( n=9 WT; n=8 Eif4ebp1-/- ) . Eif4ebp1-/- mice were treated with 4EGI-1 ( 10 μg , daily for 3 days , i . t . ) or vehicle , and the mEPSCs ( D ) and mIPSCs ( E ) were recorded from lamina II neurons ( n=7 vehicle; n=11 4EGI-1 ) . ( F ) Synaptic potentiation was elicited by stimulation of the dorsal root ( 2 Hz ) for 20 , 40 and 60 s and recording fEPSPs 125 μm from the dorsal surface of the spinal cord . Synaptic potentiation is induced in Eif4ebp1-/- , but not WT spinal cord preparation , by 40 s stimulation . Left: representative traces of fEPSPs , evoked by stimulation for the indicated length of time . Right: bar graph showing the summary of synaptic potentiation during the last 5 min prior to stimulation ( n=7 WT; n=8 Eif4ebp1-/- ) . Data are presented as mean ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 by Student’s t-test ( A , B , D , E ) , or by Student’s t-test following two-way ANOVA ( genotype x drug in C; genotype x duration in F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 008 Additionally , we measured potentiation of fPSPs in Eif4ebp1-/-and wild-type lumbar spinal cords by stimulating dorsal roots ( 2 Hz ) and recording extracellular fPSPs in the superficial layer of the spinal cord . The threshold for induction of potentiation was lowered in Eif4ebp1-/- as compared to wild-type mice , since potentiation was elicited with 40 s of stimulation in Eif4ebp1-/- spinal cords , whereas in wild-type spinal cords 60 s of stimulation were required to potentiate EPSPs ( Figure 3F ) . The extent of potentiation achieved with 60 s of stimulation was 37 . 5% higher in Eif4ebp1-/- as compared to wild-type spinal cords . Together , the results demonstrate that basal synaptic transmission is enhanced , and the threshold for induction of synaptic potentiation is lowered in Eif4ebp1-/- mice . Based on the effects of translational de-repression on synaptic physiology , as a consequence of 4E-BP1 depletion , we reasoned that translation of specific mRNAs , which regulate synaptic transmission , is affected . Consistent with previous reports , e . g . Tahmasebi et al . , 2014 , we found no differences in rates of general translation between wild-type and Eif4ebp1-/- mice , using the puromycin incorporation assay ( Figure 4A ) , supporting the idea that alterations in translation of a small subset of mRNAs engender nociceptive phenotypes in Eif4ebp1-/- mice . We previsouly investigated the mechanisms underlying the enhanced excitatory and inhibitory synaptic transmission in hippocampal neurons of Eif4ebp2-/- mice ( Gkogkas et al . , 2013 ) . 4E-BP2 is the major 4E-BP isoform in the hippocampus ( Banko et al . , 2005 ) . Members of the family of post-synaptic adhesion proteins , neuroligins , were upregulated in the hippocampus of 4E-BP2 depleted mice , leading to changes in synaptic transmission . Specifically , translation of neuroligin ( Nlgn ) 1 , 2 and 3 mRNAs was increased in the brain of Eif4ebp2-/- mice . Neuroligin 1 controls excitatory synapse formation , maturation , and function , whereas neuroligin 2 regulates inhibitory synaptic transmission ( Krueger et al . , 2012 ) . Neuroligin 3 controls both excitatory and inhibitory synapses . An additional study has revealed a translational upregulation of α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor ( AMPA ) receptor subunits , Glua1 and Glua2 in response to 4E-BP2 ablation ( Ran et al . , 2013 ) . Protein levels of neuroligins ( 1–3 ) were increased in synaptosomes prepared from Eif4ebp1-/- spinal cords ( Figure 4B ) , whereas their mRNA levels were unaltered ( Figure 4C ) , suggesting that translation of neuroligin 1–3 mRNAs is increased in Eif4ebp1-/- mice . No differences in Glua1 and Glua2 protein expression were found in Eif4ebp1-/- spinal cords ( Figure 4B ) . 10 . 7554/eLife . 12002 . 009Figure 4 . Enhanced expression of neuroligin 1 contributes to mechanical hypersensitivity of Eif4ebp1-/- mice . ( A ) General translation is not altered in Eif4ebp1-/- mice as assessed by puromycin incorporation ( n=3/genotype ) . ( B ) Protein expression of neuroligin ( NLGN ) 1 , 2 and 3 is increased in synaptosomes prepared from the dorsal horn of the lumbar spinal cord of Eif4ebp1-/- mice ( n=5/genotype ) , whereas their mRNA levels are not changed ( C ) . ( D ) Wild-type ( WT ) and Eif4ebp1-/- mice were treated with 4EGI-1 ( 10 μg , daily for 3 days , i . t . ) or vehicle , and the protein levels of neuroligin 1 were measured in lysates prepared from the dorsal horn of the lumbar spinal cord using western blot analysis ( n=5/condition ) . ( E ) Nlgn1-/- mice exhibit elevated von Frey thresholds , whereas Nlgn1+/- mice show no alteration in mechanical sensitivity ( n=10 WT and Nlgn1-/-; n=15 Nlgn1+/- ) . ( F ) Deletion of one allele of Nlgn1 in Eif4ebp1 knockout mice ( Eif4ebp1-/- / Nlgn1+/- ) normalizes neuroligin 1 protein levels in the dorsal horn of the lumbar spinal cord of Eif4ebp1-/- mice ( left ) and partially attenuates mechanical allodynia of Eif4ebp1-/- mice ( right , n=10 wild-type , n=7 Eif4ebp1-/- , n=14 Eif4ebp1-/- / Nlgn1+/- ) . Data are presented as mean ± SEM . *p<0 . 05 , **p<0 . 01 , ns–not significant , by Student’s t-test ( A–C ) , Bonferroni post-hoc test following two-between ( genotype , condition ) one-within ( repeated measures ) ANOVA ( E ) , or Bonferroni post-hoc test following one-way ANOVA ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 009 In light of these results we hypothesized that the increased excitatory synaptic input into the spinal neurons and mechanical hypersensitivity in Eif4ebp1-/- mice are the result of increased neuroligin 1 expression , which is known to strongly promote excitatory synaptic transmission . The enhanced expression of neuroligin 1 in Eif4ebp1-/- spinal cord was normalized by 4EGI-1 treatment ( 10 μg , daily for 3 days , i . t . Figure 4D ) , consistent with our finding that 4EGI-1 normalizes the increased mEPSC amplitude in Eif4ebp1-/- spinal neurons ( Figure 3D ) . To study the role of neuroligin 1 in nociception , we examined mice with a null mutation of neuroligin 1 . Nlgn1-/- mice showed elevated von Frey thresholds ( Figure 4F ) , indicating reduced mechanical sensation , whereas heterozygous mice ( Nlgn1+/- ) exhibited no change in their mechanical thresholds . Next , we normalized the enhanced expression of neuroligin 1 in Eif4ebp1-/- mice by removing one allele of Nlgn1 . To this end , we generated Eif4ebp1-/-/ Nlgn1+/- double knockout mice and compared their mechanical sensitivity to Eif4ebp1-/- littermates . Deletion of one allele of Nlgn1 in Eif4ebp1-/- mice normalized the increased neuroligin 1 expression ( Figure 4G ) and reversed mechanical sensitivity ( Figure 4G ) . Taken together , our data demonstrate that elevated levels of neuroligin 1 contribute to the mechanical hypersensitivity of Eif4ebp1-/- mice . Finally , we explored whether the increase in excitatory synaptic transmission in Eif4ebp1-/- mice is mediated via enhanced expression of neuroligin 1 . Strikingly , deletion of one allele of Nlgn1 in Eif4ebp1-/- mice ( Eif4ebp1-/-/Nlgn1+/- ) rescued the enhanced excitatory synaptic transmission ( mEPSC amplitude and total charge transfer ) in spinal neurons ( Figure 5A ) , but had no effect on the inhibitory synapses ( Figure 5B ) . No differences were found in input resistance and membrane capacitance of spinal neuron between the four genotypes examined ( Figure 5—figure supplement 1 ) . Thus , our data demonstrate that the enhanced expression of neuroligin 1 in mice lacking 4E-BP1 strengthens excitatory spinal synapses , and thereby contributes to mechanical hypersensitivity . 10 . 7554/eLife . 12002 . 010Figure 5 . Reduction of neuroligin 1 levels in Eif4ebp1-/- mice normalizes the enhanced excitatory synaptic transmission . mEPSCs ( A ) and mIPSCs ( B ) were recorded from lamina II spinal neurons of wild-type ( WT ) , Eif4ebp1-/- , Eif4ebp1-/-/Nlgn1+/- and Nlgn1+/- mice . Top: representative traces of recordings from the four genotypes . Bottom: bar graph showing mEPSC ( A ) and mIPSC ( B ) amplitude , frequency and synaptic total charge transfer ( n=11 WT; n=9 Eif4ebp1-/-; n=10 Eif4ebp1-/-/ Nlgn1+/-; n=7 Nlgn1+/- ) . The increase in mEPSC amplitude and total charge in Eif4ebp1-/- mice is rescued following deletion of one Nlgn1 allele , whereas mIPSCs are not affected . Data are presented as mean ± SEM . *p<0 . 05 by Bonferroni post-hoc test following one-way ANOVA . See also figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 01010 . 7554/eLife . 12002 . 011Figure 5—figure supplement 1 . Input resistance and membrane capacitance are not altered in lamina II spinal neurons from Eif4ebp1-/- , Eif4ebp1-/-/Nlgn1+/- and Nlgn1+/- mice . Input resistance ( A ) and membrane capacitance ( B ) of lamina II spinal neurons from wild-type ( WT ) ( n=11 ) , Eif4ebp1-/- ( n=9 ) , Eif4ebp1-/-/Nlgn1+/- ( n=10 ) and Nlgn1+/- mice ( n=7 ) . No differences were found in either input resistance ( p>0 . 05 , Bonferroni’s post-hoc test following one-way ANOVA ) or membrane capacitance ( p>0 . 05 , Bonferroni’s post-hoc test following one-way ANOVA ) . Data are presented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 12002 . 011
mTORC1 stimulates translation by relieving the 4E-BP-mediated block on eIF4F complex formation and cap-dependent translation . IL-6 and NGF-mediated enhancement of eIF4F complex formation play an important role in mediating mechanical nociceptive plasticity in primary afferent neurons ( Melemedjian et al . , 2010 ) . However , the roles of eIF4F and 4E-BPs in regulation of spinal circuits remain unknown . Here , we show that de-repression of translation by removal of 4E-BP1 leads to enhanced eIF4F complex formation in the spinal cord , and to mechanical hypersensitivity and increased response to noxious chemical/inflammatory stimuli . This effect was observed in both Eif4ebp1-/- mice and in mice with selective downregulation of 4E-BP1 in the spinal cord dorsal horn . Normalization of eIF4F complex levels in Eif4ebp1-/- mice by intrathecal administration of 4EGI-1 reversed the exaggerated mechanical sensation , thus demonstrating that the enhanced eIF4F complex is the underlying cause for mechanical hypersensitivity in Eif4ebp1-/- mice . eIF4F controls translation of a specific subset of mRNAs . Consistent with this and with previous reports ( Gkogkas et al . , 2013; Tahmasebi et al . , 2014 ) , we found that general translation is not altered in Eif4ebp1-/- mice ( Figure 4A ) , indicating that upregulated translation of certain mRNAs contributes to the nociceptive phenotypes of these mice . Previous studies have documented that mRNAs in the family of neuroligin proteins are being translationally activated by eIF4F in the nervous system , such as in mice overexpressing eIF4E or lacking 4E-BP2 ( Gkogkas et al . , 2013 ) . We therefore reasoned that the enhanced mechanical nociception in Eif4ebp1-/- mice could be mediated via enhanced translation of Nlgn mRNAs . Indeed , we found that the expression of neuroligins was increased in the spinal cord of Eif4ebp1-/- mice without concomitant changes in mRNA levels , indicating enhanced translation . The role of neuroligins in the spinal cord is not well studied . A recent study showed that knock-down of neuroligin 2 , which promotes inhibitory synaptic transmission , in the spinal cord of naïve rats increases mechanical sensitivity ( Dolique et al . , 2013 ) . The role of neuroligin 1 , which promotes excitatory synaptic transmission , in pain processing has not been investigated . Consistent with the higher levels of neuroligin 1 , 2 and 3 in mice lacking 4E-BP1 , we found that excitatory and inhibitory spinal synaptic transmission are increased in these animals . Chronic pain is associated with an increase in the excitatory synaptic drive to dorsal horn neurons ( Sandkuhler , 2009 ) . Therefore , in the current study we focused on neuroligin 1 and its role in promoting excitatory synaptic transmission in Eif4ebp1-/- mice . Interestingly , the enhancement of mEPSC amplitude and neuroligin 1 spinal expression in Eif4ebp1-/- mice was rescued by intrathecal 4EGI-1 . In agreement with these results , we also found that mechanical sensation is reduced in Nlgn1-/- mice , indicating that neuroligin 1 has a pronociceptive function . Strikingly , normalization of neuroligin 1 levels rescued the increased mEPSC amplitudes in Eif4ebp1-/- mice , and partially reversed the mechanical hypersensitivity in these mice . These data strongly indicate that augmented spinal expression of neuroligin 1 in Eif4ebp1-/- mice contributes to mechanical hypersensitivity via neuroligin 1-induced upregulation of excitatory synaptic transmission in the spinal cord . The finding that normalization of neuroligin 1 levels only partially reverses mechanical hypersensitivity in Eif4ebp1-/- mice indicates that additional mechanisms , peripheral or central , downstream of 4E-BP1/eIF4e contribute to the pain phenotype . We observed that the increased inhibition in Eif4ebp1-/- mice is solely mediated via enhanced glycinergic , but not GABAergic synaptic transmission ( Figure 3C ) . Similarly to GABAA , glycine receptors mediate fast synaptic inhibition in the spinal cord , where they regulate nociceptive and tactile sensory processing . Little is known about how synaptic strength of glycinergic synapses is regulated in vivo . Surprisingly , we found that the inhibitory synaptic transmission was not affected by intrathecal administration of 4EGI-1 , possibly because of reduced sensitivity to 4EGI-1 or because inhibitory synaptic transmission is regulated by 4EGI-1 insensitive processes downstream of 4E-BP1 . The differential regulation of glycinergic synaptic transmission by mTOR/4E-BP1 pathway in the spinal cord is intriguing , and further studies will be necessary to better understand the precise molecular mechanism of this regulation . However , the increased inhibition in Eif4ebp1-/- mice was not reversed by normalization of neuroligin 1 expression ( Figure 5B ) , and is thus unlikely to play major role in the reduced nociceptive thresholds of Eif4ebp1-/- mice , which were reversed by normalization of neuroligin 1 expression ( Figure 4E , G ) . The threshold for enhancement of electrically-evoked field potentials is lowered in Eif4ebp1-/- mice . This finding is consistent with studies in the hippocampus showing that 4E-BP2 ablation engenders the conversion of early-phase long-term potentiation ( LTP ) into late-phase LTP in the Schaffer collateral pathway ( Banko et al . , 2005 ) . Lowered threshold for LTP induction is thought to be a mechanism contributing to increased responsiveness of spinal circuits under conditions of pathological pain ( Ji et al . , 2003; Sandkuhler , 2009 ) . An mTORC1-mediated increase in eIF4F complex formation , which leads to the enhancement of synaptic transmission and lowering of the threshold for the induction of synaptic potentiation , is likely to contribute to abnormal pain processing in models of chronic pain . Though our study demonstrates an important role of enhanced expression of neuroligin 1 in the nociceptive phenotype of Eif4ebp1-/- mice , it is conceivable that translational upregulation of other mRNAs contributes to the regulation of nociception downstream of mTORC1 and ribosome-profiling studies are required to identify these transcripts . Our data demonstrate that the mTOR/4E-BP/eIF4E pathway controls mechanical , but not thermal sensitivity . A recent study showed that somatostatin-positive cells , which are enriched in lamina II of the dorsal horn where they comprise 53% of all glutamatergic excitatory neurons , are required for sensation of mechanical but not thermal pain ( Duan et al . , 2014 ) . Interestingly , we found that lamina II neurons receive stronger excitatory input in Eif4ebp1-/- mice ( Figure 3 ) , raising the intriguing possibility that the effect of 4E-BP1 depletion on mechanical , but not thermal sensation , could be mediated via increased excitatory drive to somatostatin-positive neurons . Conditional deletion of 4E-BP1 in somatostatin positive cells will be required to address this idea . mTORC1 controls protein synthesis as well as lipid and ribosome biogenesis , autophagy and mitochondria function ( Costa-Mattioli and Monteggia , 2013; Shimobayashi and Hall , 2014 ) . Studies with rapalogues have shown that mTORC1 is strongly implicated in the development of hypersensitivity states in a variety of pain models ( Obara and Hunt , 2014; Price and Geranton , 2009 ) . However , whether the effects of mTORC1 on nociception are mediated via translation control mechanisms or via protein synthesis-independent functions of mTORC1 have remained largely unknown . Our study clearly shows that translational regulators ( 4E-BP1 and eIF4F ) , acting downstream of mTOR , control nociception via their effect on mechanical sensation and synaptic responses . In summary , our results shed important light on the mechanisms by which activation of the mTOR/4E-BP/eIF4E pathway enhances mechanical pain sensation . We show that 4E-BP1 is a major functional isoform in the pain pathway , where it represses translation of neuroligins in the spinal cord . Removal of 4E-BP1 increases neuroligin 1 expression , which in turn promotes excitatory synaptic transmission in the spinal cord and leads to mechanical hypersensitivity .
Eif4ebp1 , Eif4ebp2 and Eif4ebp1/2 null mutant mice were backcrossed for more than 10 generations into a C57BL/6J background ( Le Bacquer et al . , 2007 ) . Nlgn1 knockout mice were kindly provided by Craig M . Powell ( The University of Texas Southwestern Medical Center , Dallas ) . Wild-type mice were littermates of the corresponding knockout mice . All behavioral experiments were performed on 8–12-week-old mice of both sexes by male experimenters blinded to genotype and drug . Food and water were provided ad libitum , and mice were kept on a 12:12 hr light/dark cycle ( lights on at 08:00 hr ) . All procedures complied with the Canadian Council on Animal Care guidelines and International Association for the Study of Pain , and were approved by McGill University's Downtown Animal Care Committee . von Frey Testing: Mice were placed individually in transparent Plexiglas cubicles ( 5 × 8 . 5 × 6 cm ) set on a perforated steel floor and habituated for 1 hr prior to testing . Nylon monofilaments ( Stoelting #2-#9 ) were firmly applied to the plantar surface of each hind paw for 0 . 5 s . The up-down method of Dixon ( Chaplan et al . , 1994 ) was used to estimate the 50% withdrawal threshold ( average of two measurements for both hind paws separated by at least 30 min ) . Tail Clip test: A small alligator clip ( 700 g force ) was applied at 1 cm from the base of the tail . The latency to attack/bite the clip was measured with a stopwatch to the nearest 0 . 1 sec . Upon attack , the clip was removed and the animals were returned to their cages . Radiant Heat Paw-Withdrawal: Mice were placed in cubicles ( described above ) on a glass floor and a focused beam of high-intensity light was aimed at the plantar surface of the hind paw . The intensity of the light was set to 15% or 20% of maximum ( IITC Model 390 ) with a cut-off value of 40 s . The latency to withdraw the hindpaw was measured to the nearest 0 . 1 s . Measurements consisted of testing both hind paws twice on two separate occasions separated by at least 30 min . Hot-plate test: Mice were placed into a clear Plexiglas cylinder atop a metal surface ( Columbus Instruments ) maintained at 50°C . The latency to lick or shake either hind paw was measured with a stopwatch to the nearest 0 . 1 sec . Formalin test: Mice were placed into Plexiglas cylinders on a glass floor and habituated for at least 30 min . Following habituation , all mice were given intraplantar injections of formalin ( 20 μl , 0 . 5% ) into the left hind paw and placed back into the cylinders . Cameras , placed under the glass floor , recorded the licking behavior over 60 min . Video files were sampled at 1-min intervals for the presence or absence of licking behavior in the first 10 sec of each interval . Data are expressed as the percent of positive ( licking ) samples . The early phase of the formalin test was defined as 0–10 min . and the late phase as 10–60 min . post-injection . Electrophysiological recordings of miniature excitatory and inhibitory post-synaptic currents ( mEPSCs and mIPSCS ) and dorsal root-evoked field post-synaptic potentials ( fPSPs ) were made using lumbar spinal cord tissue . Adult ( 3−6 months old ) male mice were anesthetized with urethane ( 2 g/kg ) and perfused with ice-cold sucrose-substituted artificial cerebral spinal fluid ( sucrose artificial cerebro-spinal fluid [aCSF]; contains in mM: 252 sucrose , 2 . 5 KCl , 1 . 5 CaCl2 , 6 MgCl2 , 10 d-glucose , bubbled with 95%:5% oxygen:CO2 ) . The lumbar spinal column was removed and immersed in ice-cold sucrose aCSF after which the spinal cord was quickly removed via laminectomy . For patch-clamp experiments , the dorsal and ventral roots were removed from the spinal cord and 300 μm-thick parasagittal slices were cut from the lumbar portion in ice-cold sucrose aCSF . After cutting , the spinal cord slices were kept in oxygenated ( 95% O2 , 5% CO2 ) aCSF ( in mM: 126 NaCl , 2 . 5 KCl , 2 MgCl2 , 2 CaCl2 , 1 . 25 NaH2PO4 , 26 NaHCO3 , 10 d-glucose ) at room temperature until recording . For LTP experiments , the ventral roots and connective tissue were removed from the spinal cord after laminectomy , and the tissue explant was placed in room temperature aCSF for 1 hr before experimentation . In most patch-clamp experiments , slices were continuously perfused ( 8 ml/min ) with oxygenated aCSF supplemented with tetrodotoxin ( 1 μM ) . Lamina II cells were visually identified and patched with borosilicate pipettes containing ( in mM ) : 110 CsMeSO3 , 11 Cs-EGTA , 10 CsCl , 20 HEPES , 2 MgCl2 , 1 CaCl2 , 4 Mg-ATP , 0 . 4 Tris-GTP , 0 . 1% ( w/v ) Lucifer Yellow; pH was adjusted to 7 . 2 with CsOH . Synaptic events were low-pass filtered at 2 kHz and recorded at 10 kHz using a Multiclamp 700B amplifier ( Molecular Devices , Sunnyvale , CA ) , Digidata 1322A digitizer ( Molecular Devices ) , and pClamp 10 software ( Molecular Devices ) . DL-2-amino-5-phosphonovaleric acid ( AP5; 40 µM ) was added to the aCSF to block NMDA receptors . mEPSCs and mIPSCs were recorded at holding potentials of -70 mV and 0 mV , respectively . For experiments designed to isolate GABAA or glycinergic mIPSCs , strychnine ( 1 μM ) or bicuculline ( 10 μM ) were added to the aCSF , respectively , and visually identified lamina II cells were patched with pipettes containing ( in mM ) : 120 CsCl , 5 KCl , 11 EGTA , 1 CaCl2 , 4 Na-ATP , 0 . 4 Na-GTP , 10 HEPES and 0 . 1% Lucifer Yellow ( wt/vol ) , pH 7 . 2 . Stable 3 min recordings of mEPSCs and mIPSCs were selected for analysis in Clampfit 10 ( Molecular Devices ) . Electrically-evoked field potentials in the superficial dorsal horn were recorded as previously described ( Bonin and De Koninck , 2014 ) . fPSPs were recorded via a borosilicate glass electrode inserted into the dorsal side of the spinal cord explant in the dorsal root entry zone . Electrodes were inserted superficially to a depth of no more than 125 μm from the dorsal surface of the spinal cord measured with an MPC-200 manipulator ( Sutter Instrument Company , Novato , CA , USA ) . Electrodes had a tip resistance of 3−4 MΩ when filled with aCSF . fPSPs were evoked by electrical stimulation of the dorsal root using a suction electrode that is pulled from borosilicate glass and filled with aCSF , and placed near the cut end of the dorsal root . Field potentials were amplified with a Multiclamp 700B amplifier ( Molecular Devices , Sunnyvale , CA , USA ) , digitized with a Digidata 1322A ( Molecular Devices ) , and recorded using pClamp 10 software ( Molecular Devices ) . Data were filtered during acquisition with a low pass filter set at 1 . 6 kHz and sampled at 10 kHz . Test stimuli were presented every 60 s to evoke fPSPs and LTP was evoked by stimulation at 2 Hz for variable lengths of time . The stimulus intensity was sufficient to activate C-fibers , as indicated by the appearance of a third distinct fiber volley after the stimulus artifact ( Bonin and De Koninck , 2014 ) , while a slightly ( 10−20% ) higher intensity was used to induce synaptic potentiation . Data were analyzed using ClampFit 10 software ( Molecular Devices ) . The area of fPSPs relative to baseline was measured from 0−800 ms after the onset of the fPSP . 4EGI-1 was purchased from Calbiochem ( San Diego , CA ) and dissolved first in DMSO ( stock solution ) and then in 30% polyethylene glycol . All drugs and chemicals for electrophysiology experiments were purchased from Sigma-Aldrich ( Oakville , ON , Canada ) . Tissue extracts for western blotting were prepared in ice-cold homogenization buffer containing ( in mM ) : 20 Tris-HCl , pH 7 . 4; 150 NaCl; 1 EDTA; 1% Triton , 5 NaF; 1 . 5 Na3VO4 , and protease inhibitor cocktail ( complete , EDTA-free , Roche Applied Science ) . Synaptosomes were prepared using Syn-PER Synaptic Protein Extraction Reagent ( Thermo Scientific CAT#8779 . Following centrifugation at 12 , 000 × g for 10 min , the supernatant protein concentration was measured and samples containing equal protein amounts were boiled for 5 min in Laemmli sample buffer and separated by SDS-PAGE . Following electrophoresis , proteins were transferred to 0 . 2 μm nitrocellulose membranes . Membranes were blocked in 5% dry milk powder in Tris-buffered saline containing 0 . 1% Tween-20 ( TBS-T ) for 1 hr prior to overnight incubation with the primary antibody . The membranes were then washed , incubated for 1 h with HRP-conjugated secondary antibody , washed again , treated with Enhanced Chemiluminescence reagent ( Perkin Elmer ) , and exposed to an autoradiography film ( Denville Scientific Inc . ) . All signals were obtained in the linear range for each antibody , quantified using ImageJ ( NIH ) , and in some experiments normalized to β-actin . The western blot experiments were performed in triplicate . The antibodies and the dilutions for the western blots used are as follows: 4E-BP1 ( 1:1000 , Cat#9644 , Cell Signaling Technology ) , 4E-BP2 ( 1:1000 , Cat#2845 , Cell Signaling Technology ) , eIF4E ( 1:1000 , Cat#610269 , BD Transduction Laboratories ) , eIF4G1 ( 1:1000 , Cat#ab2609 Abcam ) , neuroligin 1–3 ( 1:1000 , Synaptic Systems ) , GluA1-2 ( 1:1000 , Alomone labs ) and β-actin ( 1:5000 , Cat#A5441 , Sigma ) . Eight-week-old wild-type and Eif4ebp1-/- mice were injected with puromycin ( 10 mg/kg , intraperitoneal , i . p . ) , and 45 min later DRGs were collected and processed for Western blotting , using anti-puromycin monoclonal antibody ( 3RH11 , KeraFast ) . DRGs from mice injected with anisomycin ( 100 mg/kg i . p , injected 20 min before puromycin ) or vehicle were processed in parallel and used as controls . Protein synthesis was determined by measuring total lane signal from 250–15 KDa and subtracting unlabelled protein control . Signals were quantified using ImageJ . Mice were deeply anaesthetized and transcardially perfused with 30 ml vascular rinse ( 0 . 1% w/v sodium nitrite in 0 . 01M perfusion buffer ) ( for composition see Côté et al , 1993 ) and 200 ml fixative solution ( 4% paraformaldehyde in 0 . 1 M PB , pH 7 . 4 ) at room temperature ( RT ) . Spinal cords were extracted trough laminectomy and post-fixed for 2 h at 4°C . Lumbar levels ( L3-L5 ) of spinal cords were sectioned using a vibratome and collected in PB saline ( PBS ) ( 50-µm thick free floating transverse sections ) . Sections were washed in PBS containing 0 . 2% Triton X-100 ( PBS-T; Sigma Aldrich ) for 30 min at RT . To block unspecific staining , sections were treated with PBS-T containing 10% normal goat serum ( NGS ) for 1 hr at RT . Sections were then incubated with primary antibodies diluted in PBS-T containing 5% NGS , overnight at 4°C . Sections were washed in PBS-T for 30 min and incubated with secondary antibodies diluted 1:800 in PBS-T , at RT for 2 hr , protected from light . For isolectin B4 ( IB4 ) staining , sections were incubated with IB4 conjugated to AlexaFluor 546 , 1:200 ( Molecular Probes ) for 1 hr at RT . All steps were carried out under constant shaking . After 30 min of final washes in PBS-T and 10 min in PBS , sections were mounted with SlowFade Gold ( Invitrogen , Burlington , ON , Canada ) . The slides were stored protected from light at 4°C until examined by use of the Olympus confocal microscope using a 20x oil-immersion objective . Antibodies used for immunohistochemistry were: glial fibrillary acidic protein ( GFAP , 1:2000 , GFAP Mouse mAb , Cell Signalling , #3670 ) , NEUronal Nuclei ( NeuN , 1:5000 , Mouse Anti-NeuN Antibody , clone A60 , MAB377 ) , calcitonin gene-related peptide ( CGRP , 1:1000 , Sigma-Aldrich , C8198 ) , substance P ( 1:400 , Neuromics , MO15094 ) , HTR3A ( 1:200 , Sigma-Aldrich , AV13046 ) , and c-Fos ( 1:10000 , Calbiochem , PC38 ) . For detection , species-specific secondary antibodies were used: goat anti-rabbit Alexa-Fluor 488 ( 1:800 , Invitrogen Molecular probes ) and goat anti-mouse Alexa-Fluor 546 ( 1:800 , Invitrogen Molecular Probes ) . Lentiviral vectors for shRNA silencing of 4E-BP1 and a scrambled sequence were obtained from Sigma . The Sigma MISSION shRNA vectors accession numbers were: mouse 4E-BP1 ( TRCN0000075612 ) , and the Non-Target shRNA Control ( SHC002 ) . Each shRNA vector was co-transfected into HEK293T cells with the lentivirus packaging plasmids PLP1 , PLP2 , and PLP-VSVG ( Invitrogen ) using Lipofectamine 2000 ( Invitrogen ) . Viral supernatants were collected 48 and 72 h post-transfection . Supernatants were filtered through a 0 . 45 µm nitrocellulose filter either before being applied to target cells in the presence of polybrene ( 5 µg/ml ) , or processed for concentration by ultracentrifugation at 26 , 000 rpm/2 h/4°C with 20% sucrose cushion . Virus concentrates were resuspended in serum free media , and aliquots were stored at -80°C until use . Viral titre ( TU/ml ) was calculated using puromycin selection of MEFs infected with different viral particle dilutions and stained with 0 . 2% crystal violet and 20% methanol solution . Colonies were measured using the CellCount plugin of ImageJ ( NIH ) . Viral titres were adjusted to 1 . 5 107 TU/ml for in vivo injections . For intraparenchymal spinal cord dorsal horn ( SCDH ) injection , mice were anaesthetized with an i . p . injection of a cocktail containing xylazine ( 3 . 33 mg/ml ) , ketamine ( 55 . 55 mg/ml ) , and Domitor ( 0 . 27 mg/ml ) . The mouse was placed in a spinal frame and a laminectomy was used to remove spinous process VL3 and VL4 . Six unilateral injections of 0 . 25 μl ( 1 × 107 viral particles/ml ) were administesred 0 . 7 mm apart , at a depth of 0 . 2 mm ( SCDH ) , using a glass pipette attached to a Hamilton syringe with plastic tubing at a rate of 0 . 1 μl/min . To allow for the solution containing viral particles to diffuse from the tip of the glass pipette into the tissue , the pipette stayed in the tissue for one additional minute . The syringe was mounted on a microinjector ( Nanomite , Harvard Apparatus ) attached to a stereotaxic unit ( David Kopf Instruments ) . The mice were allowed to recover for 1 week before the experiments . Lumbar spinal cord was homogenized in lysis buffer containing 40 mM HEPES-KOH ( pH 7 . 5 ) , 120 mM NaCl , 1 mM EDTA , 0 . 1 mM GDP , 10 mM pyrophosphate , 10 mM β-glycerophosphate , and 50 mM NaF . Extracts were mixed with 30 μl of m7 GDP-agarose for 1 . 5 h at 4°C . The resins were washed four times and proteins were eluted with SDS–PAGE sample buffer . All results are expressed as mean ± SEM . All statistical comparisons were made with either Student’s t-test or one-way or two-way ANOVA , followed by between-group comparisons using t-test or Bonferroni’s post-hoc test , unless otherwise indicated , with p<0 . 05 as the significance criteria . | Despite the unpleasant feeling it causes , pain is necessary for survival as it helps individuals to avoid objects , environments and situations that cause damage to their body . However , millions of people experience long-lasting “chronic” pain , or are hypersensitive to pain . There are few treatments available for these conditions , but these treatments do not work well for the majority of patients , and can have serious side effects . To develop new treatments , researchers must first better understand how chronic pain develops . Pain is transmitted to the brain in the form of electrical signals “fired” along nerve fibers . Different nerves transmit information about different types of pain: for example , pain caused by a sharp object pressed against the skin activates a different set of neurons to those activated when touching something dangerously hot . Studies in mice have suggested that a protein called mTOR that is found inside neurons is important for them to fire pain signals . However , it is not clear exactly how mTOR contributes to pain signaling , although it is known to affect the activities of several other proteins in neurons . One protein that mTOR affects the activity of is called 4E-BP1 . Now , Khoutorsky , Bonin , Sorge et al . show that mice that lack 4E-BP1 behave in ways that suggest they are hypersensitive to poking or pinching sensations . However , the mice did not show hypersensitivity when they touched a hot surface . Further investigation revealed that the neurons in the spinal cord of mice that lack 4E-BP1 produce abnormally high amounts of a molecule called neuroligin 1 , which makes the neurons more likely to fire and thus signal pain . Khoutorsky , Bonin , Sorge et al . found that treating mice that lack 4E-BP1 with a compound that reduces neuroligin 1 production causes their neurons to fire more normally . This also reduces the animals’ apparent signs of hypersensitivity to pressure on their skin . It will be important in future studies to identify additional targets of 4E-BP1 in the spinal cord that could contribute to increased mechanical sensation , and also to study the role of 4E-BP1 in peripheral nerves . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"neuroscience"
] | 2015 | Translational control of nociception via 4E-binding protein 1 |
Insulin-like growth factor-I receptor ( IGF-IR ) preferentially regulates the long-term IGF activities including growth and metabolism . Kinetics of ligand-dependent IGF-IR endocytosis determines how IGF induces such downstream signaling outputs . Here , we find that the insulin receptor substrate ( IRS ) −1 modulates how long ligand-activated IGF-IR remains at the cell surface before undergoing endocytosis in mammalian cells . IRS-1 interacts with the clathrin adaptor complex AP2 . IRS-1 , but not an AP2-binding-deficient mutant , delays AP2-mediated IGF-IR endocytosis after the ligand stimulation . Mechanistically , IRS-1 inhibits the recruitment of IGF-IR into clathrin-coated structures; for this reason , IGF-IR avoids rapid endocytosis and prolongs its activity on the cell surface . Accelerating IGF-IR endocytosis via IRS-1 depletion induces the shift from sustained to transient Akt activation and augments FoxO-mediated transcription . Our study establishes a new role for IRS-1 as an endocytic regulator of IGF-IR that ensures sustained IGF bioactivity , independent of its classic role as an adaptor in IGF-IR signaling .
Insulin-like growth factor ( IGF ) -I receptor ( IGF-IR ) is an important receptor tyrosine kinase ( RTK ) that regulates a variety of biological processes including proliferation , cell survival , and control of metabolism in a wide range of mammalian tissues by binding the ligands IGF-I and IGF-II ( Nakae et al . , 2001 ) . Ligand binding to the IGF-IR extracellular domain causes conformational changes of the intracellular region , inducing the tyrosine kinase domain to autophosphorylate multiple Tyr residues and activate intrinsic RTK activity ( Kavran et al . , 2014; Favelyukis et al . , 2001 ) . IGF-IR then initiates downstream signaling through tyrosine phosphorylation of insulin receptor substrate ( IRS ) adaptor proteins to activate the phosphatidylinositol 3-kinase ( PI3K ) -Akt pathway and its various biological responses ( Myers et al . , 1996; Sun et al . , 1993; White , 2002 ) . IGF/IGF-IR stimulates the PI3K-Akt pathway in a stereotypical way – sustained tonal induction . Sustained induction is thought to define the specific biological outcomes of IGF signaling , and distinguish the function of the IGF ligand from other RTKs/ligands that access the Akt cascade ( Gross and Rotwein , 2016; Kubota et al . , 2012 ) . In particular , sustained activation of the PI3K-Akt pathway , mediated by IGF-IR , induces cell proliferation in multiple types of cells , cell survival in neural cells , and protein homeostasis in skeletal muscle cells ( Fernandez and Torres-Alemán , 2012; Fukushima et al . , 2012; Ness and Wood , 2002; Sacheck et al . , 2004; Stewart and Rotwein , 1996 ) . To date , the mechanism by which IGF-IR produces sustained signaling remains poorly understood . Clathrin-mediated endocytosis ( CME ) is a major regulator of RTKs ( Goh and Sorkin , 2013 ) involving the heterotetrameric AP2 complex composed of large α and β2 , medium μ2 , and small σ2 subunits ( Collins et al . , 2002 ) . AP2 binds to transmembrane cargo proteins that contain specific motifs such as YxxΦ ( Y denotes Tyr; x , any amino acid; and Φ , bulky hydrophobic residue ) serving as μ2 binding sites ( Owen and Evans , 1998; Traub and Bonifacino , 2013 ) . In addition , AP2 associates with clathrin and with endocytic accessory proteins at the plasma membrane to coordinate clathrin-coated pit ( CCP ) formation ( Schmid and McMahon , 2007 ) . Ligand-bound RTKs enter the endocytic process through CME , but perhaps with different signaling consequences . If endocytosed RTKs are sorted to lysosomes for degradation , this process down-regulates signaling as exemplified by the model RTKs including epidermal growth factor receptor ( EGFR ) and platelet-derived growth factor receptor ( Goh and Sorkin , 2013 ) . On the other hand , some RTKs continue to signal locally across the endosome membrane even after endocytosis ( Schenck et al . , 2008; Villaseñor et al . , 2015; Lin et al . , 2006 ) . In either case , RTK internalization strongly impacts its signaling outputs . Thus , the duration at the cell surface of ligand-bound RTKs , which is tightly regulated by CME , critically fine-tunes their signaling and biological functions . Accordingly , we hypothesize that ligand-bound IGF-IR , which exhibits sustained activation and slow degradation ( Fukushima et al . , 2012; Mao et al . , 2011; Zheng et al . , 2012 ) , undergoes slow or delayed CME ( Martins et al . , 2011; Monami et al . , 2008 ) . To evaluate this idea , here we study the molecular components regulating perdurance of IGF-IR at the cell surface through its interactions with CME , and elucidate how this dictates IGF signaling and outputs . Among IRS family proteins IRS-1 and IRS-2 are well known as major substrates of IGF-IR ( Taniguchi et al . , 2006; White , 2002 ) . We and others have shown that IRS-associated proteins contribute to the regulation of IRS-1/IRS-2 function through distinct mechanisms ( Ando et al . , 2015; Hakuno et al . , 2015 , 2007; Lee et al . , 2013; Shi et al . , 2011; Fukushima et al . , 2015; Yoneyama et al . , 2013 ) . In this study , we discovered AP2 is also an IRS-1-associated protein . Unexpectedly IRS-1 promotes the surface retention of activated IGF-IR through inhibiting AP2-dependent internalization of IGF-IR , and this is independent of IRS’s classic role as an adaptor protein in IGF-IR and insulin receptor signaling . The ability of IRS-1 to prolong surface retention of IGF-IR is essential for long-term PI3K-Akt signaling . Our results establish a novel role of IRS-1 in ensuring the sustained effects of IGFs via its direct control of IGF-IR internalization .
To identify the IRS-1-interacting proteins that potentially regulate insulin/IGF signaling , we searched the candidates in our previous yeast two-hybrid screening ( Hakuno et al . , 2007 ) . We found the μ2 subunit of clathrin adaptor AP2 complex among the frequently obtained clones ( Figure 1A ) . Co-immunoprecipitation assay using HEK293T cells expressing FLAG-tagged IRS-1 or IRS-2 revealed that endogenous AP2 subunits ( α-adaptin and μ2 ) were detected in a complex with IRS-1 , but not with IRS-2 ( Figure 1B ) . In addition , a portion of AP2 was co-immunoprecipitated with endogenous IRS-1 as well as ectopically expressed FLAG-IRS-1 in L6 myoblasts , and this interaction was not affected by IGF-I stimulation ( Figure 1C , D ) . Using IRS-1 truncated mutants , we mapped the central region ( amino acid residues 543–865 ) which is necessary for the binding to AP2 ( Figure 1E ) . This region is almost identical to that for the clathrin adaptor AP1 complex found in our previous study , which binds to YxxΦ motifs of IRS-1 including Tyr 608 , Tyr 628 , and Tyr 658 via its μ1 subunit ( Yoneyama et al . , 2013 ) . Indeed , the Ala mutation of all these Tyr residues in IRS-1 , but not a single substitution , completely abolished the binding to μ2 in vitro ( IRS-1 3YA mutant; Figure 1F , G ) . We also analyzed the crystal structures of μ2 C-terminal subdomain ( C-μ2 ) bound to IRS-1 YxxΦ motifs ( Figure 1—figure supplement 1A , B and Table 1 ) . Importantly , the side chains of Tyr and Met residues of IRS-1 YxxΦ motifs are inserted into the binding pockets of μ2 , which are shared by the AP2 cargo proteins ( Owen and Evans , 1998 ) ( Figure 1—figure supplement 1C ) . Collectively , these results indicate that IRS-1 is recognized by the AP2 complex via the μ2 subunit in the very similar manner to conventional endocytic cargos . The μ2 subunit of AP2 cannot recognize phosphorylated YxxΦ sequence due to its limited capacity ( Kittler et al . , 2008; Owen and Evans , 1998 ) . However , IGF-I stimulation did not inhibit the co-immunoprecipitation of IRS-1 with AP2 ( Figure 1C , D ) . To evaluate the stoichiometry of IRS-1 Tyr phosphorylation in IGF-I-stimulated cells , we analyzed the amount of IRS-1 capable of binding to GST-C-μ2 in lysates of cells treated with or without IGF-I ( Figure 1—figure supplement 1D , E ) . Although the amounts of both pulled-down and immunoprecipitated IRS-1 were comparable , Tyr-phosphorylated IRS-1 was hardly pulled down by μ2 ( Figure 1—figure supplement 1F ) , indicating low stoichiometry of IRS-1 Tyr phosphorylation after IGF-I stimulation and existence of a non-phosphorylated IRS-1 pool which interacts with AP2 . Since AP2 plays a central role in the CME of RTKs , we reasoned that the interaction of IRS-1 with AP2 affects the internalization of IGF-IR . Using the surface biotinylation assay , we first analyzed the changes in cell surface IGF-IR in L6 cells . Long-term stimulation with IGF-I ( 3 to 12 hr ) induced the significant reduction of phosphorylated IGF-IR ( phospho-IGF-IR ) , which was assessed by Tyr 1131 phosphorylation in the activation loop ( Favelyukis et al . , 2001 ) , at the cell surface ( Figure 2A , B ) . Similar results were obtained in the analyses of other phosphorylation sites in IGF-IR ( Figure 2—figure supplement 1A ) . No reduction of phospho-IGF-IR or total IGF-IR at the cell surface was observed during short-term stimulation with IGF-I ( 5 to 60 min ) ( Figure 2A and B ) . Ubiquitination of IGF-IR has been proposed as an important event inducing its internalization and down-regulation ( Monami et al . , 2008; Mao et al . , 2011 ) . We observed that IGF-I-induced ubiquitination of IGF-IR reached the maximum 60 min after IGF-I stimulation in L6 cells ( Figure 2—figure supplement 1B ) . We next generated L6 cell lines stably expressing IRS-1 fused with green fluorescent protein ( GFP-IRS-1 ) ( Figure 2C ) . Strikingly , phospho-IGF-IR at the cell surface was sustained even after prolonged IGF-I stimulation in GFP-IRS-1-expressing cells while the reduction was observed in the control cells expressing GFP only ( Figure 2D , E ) . In contrast , GFP-IRS-2 expression did not affect the reduction in phospho-IGF-IR ( Figure 2—figure supplement 1C , D ) . To investigate the requirement of IRS-1 interaction with AP2 for the surface retention of phospho-IGF-IR , we analyzed the cells expressing the GFP-IRS-1 3YA mutant , which lacks the binding motifs for the μ2 subunit of AP2 complex . In contrast to GFP-IRS-1 wild-type ( WT ) -expressing cells , surface phospho-IGF-IR was reduced by prolonged IGF-I stimulation in GFP-IRS-1 3YA-expressing cells ( Figure 2D , E ) . These data strongly suggest that IRS-1 can promote cell surface retention of activated IGF-IR via its YxxΦ motifs . The Tyr residues of the YxxΦ motifs of IRS-1 for binding to AP2 ( Tyr 608 , Tyr 628 , and Tyr 658 ) are known to be phosphorylated by IR/IGF-IR and in turn serve as putative binding sites of PI3K ( Sun et al . , 1993; Myers et al . , 1996 ) . We next asked whether their Tyr phosphorylation of IRS-1 is involved in the surface retention of IGF-IR . Here , we used the IRS-1 ΔPTB mutant which lacks the phosphotyrosine binding domain ( PTB ) and therefore cannot be phosphorylated due to the inability to interact with IGF-IR ( Figure 2—figure supplement 1E ) . As with GFP-IRS-1 WT , expression of GFP-IRS-1 ΔPTB resulted in the surface retention of phospho-IGF-IR after prolonged IGF-I stimulation ( Figure 2F , G ) , indicating that the IRS-1-induced surface retention of activated IGF-IR is independent on the Tyr phosphorylation of IRS-1 . We investigated whether long-term IGF-I-induced reduction in activated IGF-IR depends on CME . In clathrin-depleted cells , the reduction in phospho-IGF-IR observed after long-term IGF-I stimulation was completely blocked ( Figure 3A ) . Similarly , the knockdown of AP2 ( μ2 ) , but not of another clathrin adaptor AP1 ( μ1 ) , inhibited the reduction of phospho-IGF-IR ( Figure 3B and Figure 3—figure supplement 1A ) . The canonical CME model of RTKs involves their rapid depletion from the cell surface in response to the ligands ( Goh and Sorkin , 2013 ) . Surface biotinylation analysis in Figure 2A revealed that the total amount of IGF-IR at the cell surface is not changed by IGF-I . Surface IGF-IR level reflects the balance between endocytosis , recycling , and the transport of newly synthesized receptor to the plasma membrane . When the recycling was inhibited by primaquine ( van Weert et al . , 2000 ) , surface IGF-IR levels were reduced by IGF-I treatment within 1 hr , and phospho-IGF-IR levels followed this time-dependent changes ( Figure 3C ) , indicating that IGF-I indeed triggers IGF-IR endocytosis from cell surface and that the recycling contributes to the apparent surface maintenance of IGF-IR . We also assessed the contribution of newly synthesized IGF-IR by using cycloheximide which could inhibit the increase in precursor IGF-IR observed in long-term IGF-I-stimulated cells . IGF-I reduced surface IGF-IR in the presence of cycloheximide ( Figure 3—figure supplement 2A ) . These observations support the notion that transport mechanisms other than endocytosis contribute to the maintenance of surface IGF-IR level . Protein tyrosine phosphatase 1B ( PTP1B ) , an endoplasmic reticulum-resident phosphatase , has been reported to down-regulate IGF-IR by dephosphorylation ( Buckley et al . , 2002 ) . We tested the possible involvement of PTP1B in long-term IGF-I-induced reduction in activated IGF-IR by using the substrate-trapping mutant ( PTP1B D181A ) . Phosphorylation levels of IGF-IR observed 1 hr after IGF-I treatment and the subsequent reduction at the later period ( 6 hr ) were comparable for both PTP1B D181A-expressing and non-expressing cells as revealed by immunofluorescence ( Figure 3—figure supplement 2B ) , indicating a negligible role of PTP1B in the down-regulation of phospho-IGF-IR in our observation . To directly monitor the internalized IGF-IR , we stimulated surface-biotinylated cells with IGF-I and then analyzed the internalized IGF-IR fraction ( see Materials and methods ) . It revealed that internalized IGF-IR was detected within 15 min after surface biotinylation ( Figure 3—figure supplement 3A ) . Similar results also came from the immunofluorescence analysis of a double-tagged IGF-IR-transfected cells . The IGF-IR-HA-EGFP construct that we developed contains an extracellular HA-tag and intracellular EGFP and can be utilized to directly monitor the internalization by following uptake of anti-HA antibody added to the media prior to ligand treatment ( Figure 3—figure supplement 3B , C ) . Internalized fraction of the double-tagged IGF-IR was detected within 15 to 60 min in both IGF-I-stimulated and non-stimulated conditions ( Figure 3—figure supplement 3D , E ) . The internalization of IGF-IR observed in the non-stimulated state was not affected by knockdown of AP2 ( Figure 3—figure supplement 3F ) , indicating that the basal endocytosis of IGF-IR is not dependent on AP2 . In contrast , phospho-IGF-IR was predominantly localized to the cell surface and did not overlap with internalized IGF-IR ( HA-positive ) within 1 hr in the ligand-stimulated cells ( Figure 3—figure supplement 3D ) . At the later period ( 6 hr ) , phospho-IGF-IR was detected in LysoTracker-positive compartments ( Figure 3—figure supplement 1B , left ) . More importantly , the phospho-IGF-IR targeting to lysosomes was abolished by knockdown of AP2 ( Figure 3—figure supplement 1B , right; Figure 3—figure supplement 1C ) , suggesting that ligand-activated IGF-IR undergoes AP2-dependent endocytosis . Using live cell total internal fluorescence microscopy ( TIRF-M ) , we investigated the detailed onset of IGF-IR internalization . The assembly of AP2 into clathrin-coated structures can be monitored by the expression of AP2 σ2 subunit fused with monomeric red fluorescent protein ( mRFP ) ( Ehrlich et al . , 2004 ) . IGF-IR-EGFP was uniformly distributed within the plasma membrane , and then gradually colocalized with σ2-mRFP 30 min after IGF-I stimulation ( Figure 3D , left ) . We also observed similar results in the fixed cells where phospho-IGF-IR was overlapped with AP2 and clathrin ( Figure 3—figure supplement 4A–C ) . In more detail , IGF-IR clustered after IGF-I stimulation , and then accumulated in pre-existing AP2-positive spots ( Figure 3—figure supplement 4D ) . EGFP-fused EGFR , which is a representative RTK showing rapid CME , was rapidly re-distributed into AP2-positive spots after EGF stimulation ( Figure 3C , right ) . Intriguingly , quantitative analyses revealed that IGF-I-induced increase in the colocalization rate of IGF-IR with AP2 was significantly slower than EGFR ( Figure 3D ) . Expression of IRS-1 WT , but not 3YA mutant , induced surface retention of activated IGF-IR ( Figure 2D , E ) , which phenocopies that of AP2 knockdown ( Figure 3B ) . We next asked whether IRS-1 could disrupt IGF-IR internalization . To clearly evaluate ligand-dependent receptor internalization , we performed surface biotinylation assay of IGF-I-stimulated cells when the recycling was inhibited by primaquine . While surface IGF-IR levels were gradually reduced after the ligand stimulation in cells expressing GFP and GFP-IRS-1 3YA , such reduction turned to be slower in cells expressing GFP-IRS-1 WT ( Figure 4A ) . In addition , the TIRF-M revealed that expression of GFP-IRS-1 WT , but not of 3YA mutant , significantly inhibited the targeting of phospho-IGF-IR in AP2-positive spots with diffused localization of phospho-IGF-IR ( Figure 4C , D ) , indicating that the IRS-1 binding to AP2 inhibits the ligand-induced association of IGF-IR with AP2-positive spots . Since AP2 regulates CME of various membrane cargoes , we next asked if ectopic expression of IRS-1 affects endocytosis of other cargoes . The internalization of transferrin receptor ( TfR ) , integrin , and EGFR , which are endocytosed through CME , was evaluated . We analyzed the endocytosis of TfR , which has no physical interaction with IGF-IR ( Figure 4—figure supplement 1A ) , by measuring uptake of fluorescent-labeled transferrin . Overexpression of IRS-1 did not affect the uptake of transferrin ( Figure 4—figure supplement 1B ) . Integrins including β1 are involved in the crosstalk with IGF-IR signaling ( Kiely et al . , 2005 ) . Surface level and internalization of integrin β1 were assessed by labeling cell surface with anti-integrin β1 antibody and chasing its uptake ( see Materials and methods ) . In L6 cells stably expressing integrin β1 which modestly interacts with IGF-IR ( Figure 4—figure supplement 2A ) , surface expression of integrin β1 was not statistically different between IRS-1-expressing and control cells ( Figure 4—figure supplement 2B; p=0 . 188 ) . The incorporated amount of anti-integrin β1 antibody was partially reduced in IRS-1-expressing cells ( Figure 4—figure supplement 2C , D ) . We also examined the endocytosis of EGFR induced by low-dose EGF , which is dependent on CME ( Sigismund et al . , 2008 ) , by observing localization of the transfected EGFR-GFP . Modest delay of EGFR endocytosis was observed at the early period of EGF stimulation in IRS-1-expressing cells ( Figure 4—figure supplement 2E–F ) . These observations indicate that IRS-1 can influence endocytosis of receptors other than IGF-IR . We also confirmed that the number of AP2 spots at TIRF field was not affected by the expression of IRS-1 ( Figure 4—figure supplement 1A ) . By using TIRF-M , we noticed that GFP-IRS-1 colocalizing with AP2 is localized to submembraneous actin fibers , which possess critical roles in CME ( Kaksonen et al . , 2006 ) ( Figure 4—figure supplement 1D ) . If endogenous IRS-1 inhibits IGF-IR internalization , knockdown of IRS-1 would accelerate the process of active IGF-IR reduction triggered by long-term IGF-I stimulation . IRS-1 knockdown in L6 cells resulted in a faster reduction of phospho-IGF-IR ( ~2 fold ) with a partial decrease in IGF-IR level ( Figure 5A , B; reduction rate of p-IGF-IR from 1 to 3 hr of IGF-I treatment ( value ± SEM ( /hr ) ) , siCtrl , 7 . 8 ± 2 . 2; siIRS1_1 , 15 . 1 ± 1 . 7; siIRS1_2 , 17 . 2 ± 2 . 4; p<0 . 05 versus siCtrl ) . Furthermore , phospho-IGF-IR accumulated in lysosomes in IRS-1-depleted cells 1 hr after IGF-I stimulation when phospho-IGF-IR is predominantly localized to the plasma membrane in control cells ( Figure 5—figure supplement 1A , B ) . Notably , the partial reduction of total IGF-IR levels observed in IRS-1-depleted cells was rescued by the combined knockdown of AP2 ( Figure 5C , D ) . The accelerated reduction of phospho-IGF-IR after IGF-I stimulation in IRS-1-depleted cells was also attenuated by the combined knockdown of AP2 ( Figure 5E , F ) , indicating that knockdown of IRS-1 accelerates IGF-I-induced IGF-IR internalization as well as reducing IGF-IR levels in an AP2-dependent manner . These results further support the notion that IRS-1 inhibits AP2-mediated internalization of IGF-IR and its long-term attenuation . Previous studies have demonstrated a negative feedback loop in which long-term IGF/insulin stimulation induces the degradation of IRS-1 in a PI3K/mTOR complex 1 ( mTORC1 ) -sensitive and proteasome-dependent fashion ( Harrington et al . , 2004; Haruta et al . , 2000 ) . In L6 cells , the amount of IRS-1 was significantly reduced 3 to 6 hr after IGF-I stimulation with a concomitant increase in its phosphorylation ( Figure 6A ) . Pharmacological inhibition of mTORC1 with rapamycin or Torin1 blunted the IRS-1 degradation ( Figure 6B ) . Simultaneously , the reduction of phospho-IGF-IR after IGF-I stimulation was also blocked by mTORC1 inhibition ( Figure 6B , C ) . TIRF-M analysis revealed that phospho-IGF-IR was less clustered , and overlapped very little with AP2 in Torin1-treated cells ( Figure 6D , E ) . In IRS-1-depleted cells , phospho-IGF-IR levels were decreased after long-term IGF-I stimulation even in the presence of Torin1 ( Figure 6F , G ) . Collectively , these results suggest that the degradation of IRS-1 via mTORC1-mediated feedback loop is required for the internalization of activated IGF-IR . Given that CME affects signaling duration , we tested the role of IRS-1 in the temporal changes in downstream pathways of IGF-IR . Like phospho-IGF-IR , IGF-I-induced phosphorylation of Akt was sustained within 1 hr with a gradual decrease afterwards in L6 cells ( Figure 6A ) . Ectopic expression of IRS-1 WT , however , significantly prolonged the phosphorylation of Akt in response to IGF-I ( Figure 7A , B ) . Phosphorylation of FoxO1 , a transcription factor targeted by Akt ( Calnan and Brunet , 2008 ) , was also prolonged in IRS-1 WT-overexpressing cells . These described effects on Akt and FoxO1 were not observed in cells overexpressing IRS-1 3YA mutant ( Figure 7A , B ) . In addition , overexpression of IRS-2 did not prolong the IGF-I-dependent Akt phosphorylation with a slight increase in its maximum response ( Figure 7—figure supplement 1A ) . We next assessed the role of endogenous IRS-1 in the Akt-FoxO signaling duration by using siRNA-mediated knockdown of IRS-1 . In IRS-1-depleted cells , the phosphorylation of Akt showed a very transient pattern with the acute decrease in the later period of IGF-I stimulation ( Figure 7C , D ) . During the shorter stimulation , IRS-1 depletion had a minimal effect on the Akt phosphorylation , which may be explained by the compensatory increase in IRS-2 protein ( Figure 7—figure supplement 1B ) . The phosphorylation of FoxO1 was transient in IRS-1-depleted cells while it was stable ( phospho-S256 in FoxO1 ) or accumulated ( phospho-T24 in FoxO1 or T32 in FoxO3 ) in control cells ( Figure 7C , D ) . The shift from sustained to transient phosphorylation of Akt in IRS-1-depleted cells was completely recovered by the rescue expression of IRS-1 ( Figure 7—figure supplement 1C ) . These results indicate a role of IRS-1 in sustaining the Akt-FoxO signaling as well as prolonged surface retention of active IGF-IR . Since Akt inhibits the transcriptional activity of FoxOs via their phosphorylation ( Calnan and Brunet , 2008 ) , we reasoned that sustained activation of Akt in response to IGF could efficiently suppress FoxO-targeting gene expression . Here , we measured the mRNA expression levels of a series of FoxO-regulated genes related to muscle atrophy in which ubiquitin-proteasomal and autophagic protein degradation is enhanced ( Milan et al . , 2015; Mammucari et al . , 2007; Moses et al . , 2014; Stitt et al . , 2004; Zhao et al . , 2007 ) . In L6 myotubes long-term IGF-I stimulation significantly reduced the mRNA expression level of the two muscle-specific E3 ubiquitin ligases ( Atrogin1 and Murf1 ) and recently reported E3 ligases ( Smart and Musa1 ) as well as autophagy-related genes ( Lc3b and Gabarapl1 ) ( Figure 8A and Figure 8—figure supplement 1A ) . These genes were also down-regulated by IGF-I in L6 myoblasts ( Figure 8—figure supplement 1B ) . To reveal the contribution of IRS-1 to their expression , we analyzed their mRNA levels in IRS-1-depleted L6 myoblasts . In these cells , IGF-I-induced decrease in the atrophy-related genes was markedly attenuated ( Figure 8B ) . We also tested whether IRS-1 knockdown would affect the myotube morphology ( Figure 8—figure supplement 1C ) . We confirmed that lentiviral IRS-1 knockdown did not affect the fusion rate ( the number of nuclei in myotube fiber ) ( Figure 8—figure supplement 1D ) . IRS-1-depleted myotubes showed a significant reduction in their diameter ( Figure 8C , D ) . These data indicate that IRS-1 depletion leads to insufficient suppression of the FoxO-targeting genes in response to IGF even when Akt is being activated , but in a transient fashion .
The canonical function of IRS proteins is to mediate signaling of IGF-IR to the PI3K-Akt pathway through Tyr phosphorylation ( Figure 9A ) ( White , 2002 ) . The present results reveal a new role of IRS-1 independent of its Tyr phosphorylation: IRS-1 regulates IGF-IR internalization to produce sustained activation of IGF signaling ( Figure 9B ) . IRS-1 binds with AP2 to prevent IGF-IR recruitment into clathrin-coated structures and thus enhance surface retention of activated IGF-IR . This function of IRS-1 in prolonging IGF-IR activity is critical for sustained activation of the PI3K-Akt pathway , and provides a key mechanism for how IGF-IR signaling induces specific biological actions of IGF ( Sacheck et al . , 2004; Ness and Wood , 2002; Bailey et al . , 2006; Stewart and Rotwein , 1996 ) . Thus , IRS-1 plays a dual role as a signaling adaptor of IGF-IR and an endocytic regulator of IGF-IR . The first key finding of the present study is that IRS-1 interacts with AP2 thereby regulating the rate of ligand-dependent internalization of IGF-IR . AP2-mediated recognition of YxxΦ motif in cargos is a critical step for CCP formation ( Traub and Bonifacino , 2013; Kadlecova et al . , 2017 ) . Our results indicate that IRS-1 inhibits the recruitment of IGF-IR to CCPs through YxxΦ motifs in IRS-1 . We have previously reported that another clathrin adaptor complex AP1 also binds to the same sites of IRS-1 as AP2 ( Yoneyama et al . , 2013 ) . Since AP1 depletion did not prevent the down-regulation of activated IGF-IR ( Figure 3—figure supplement 1A ) , the inhibitory effect of IRS-1 on IGF-IR internalization is based on its interaction with AP2 . In addition , IRS-1-depleted cells show the fast onset of IGF-IR internalization in response to IGF-I and the partial decrease in IGF-IR levels , both of which are presumably caused by the promotion of AP2-dependent IGF-IR internalization and subsequent degradation ( Figure 5 ) . These results suggest that IRS-1 is an inhibitory upstream regulator for AP2-dependent internalization of IGF-IR ( Figure 9B ) . EGFR and some G-protein coupled receptor/β-arrestin complexes are known to be recruited into pre-existing CCPs after the ligand stimulation ( Rappoport and Simon , 2009; Scott et al . , 2002 ) . We observed the similar behavior of IGF-IR in live-cell TIRF-M ( Figure 3—figure supplement 4D ) . Notably , less IGF-IR was recruited to AP2-positive spots in the cells ectopically expressing IRS-1 WT , but not 3YA mutant , suggesting that IRS-1 interferes with the recruitment step of IGF-IR to clathrin-coated structures through competing out AP2 from IGF-IR . This will need to be tested by more detailed observation at higher resolution . The second key finding of this study is that the ability of IRS-1 to promote surface retention of IGF-IR can be separable from Tyr phosphorylation-mediated signaling function of IRS-1 . The Tyr residues of the YxxΦ motifs of IRS-1 ( Tyr608 , 628 , and 658 ) critical for the binding to AP2 are part of phosphorylation sites among multiple Tyr residues in the C-terminus of IRS-1 that mediate the interaction of IRS-1 with PI3K and subsequent activation of PI3K ( Myers et al . , 1996; Sun et al . , 1993 ) . We showed that ectopic expression of the IRS-1 mutant ΔPTB led to the accumulation of active IGF-IR at cell surface to the same degree as that of IRS-1 WT ( Figure 2 ) , indicating that IRS-1 inhibits the internalization of IGF-IR in a manner independent of its Tyr phosphorylation . In addition , AP2 would preferentially bind non-phosphorylated IRS-1 since AP2 cannot recognize phosphorylated YxxΦ sequence due to its limited capacity ( Kittler et al . , 2008; Owen and Evans , 1998 ) . In line with this , our biochemical analyses support the notion that non-phosphorylated IRS-1 acts as an inhibitory factor for IGF-IR internalization via its interaction with AP2 ( Figure 9B ) . Our observation indicates that ectopic expression of IRS-1 affects endocytosis of receptors other than IGF-IR . As long as we tested , endocytosis of integrin β1 and EGFR , which could interact with IGF-IR , but not of TfR , was inhibited by IRS-1 , raising the possibility that IRS-1 influences endocytosis of cargoes in the close proximity of IGF-IR . As observed in our TIRF-M observation ( Figure 4—figure supplement 1D ) , a fraction of IRS-1 has been demonstrated to localize to membrane-associated cytoskeleton ( Clark et al . , 1998 ) . IRS-1 may locally regulate the specific cargo recruitment to CCPs through association with a portion of AP2 at the actin cytoskeleton . Indeed , preferred sites of endocytosis have been observed in some cargo proteins ( Grossier et al . , 2014; Weng et al . , 2014 ) , although the molecular mechanisms of such spatial regulation for IGF-IR and other cargos remain unknown . In addition to the role of IRS-1 in controlling the rate of IGF-IR internalization , we found that this ability of IRS-1 is negatively regulated by mTORC1 ( Figure 6 ) . mTORC1 has been reported to suppress IGF-IR activity via its direct substrate Grb10 ( Yu et al . , 2011; Hsu et al . , 2011 ) . Our findings propose another mode of IGF-IR regulation by mTORC1: mTORC1 feedback signaling leads to the degradation of IRS-1 , which functions as a brake release to trigger IGF-IR internalization ( Figure 9B ) . Hence , the time length needed for IRS-1 degradation , which is critically regulated by mTORC1 , should determine the initiation timing of IGF-IR internalization . Receptor endocytosis is now considered to play both negative and positive roles in the downstream signaling ( Goh and Sorkin , 2013 ) . Our data demonstrated that CME is required for long-term attenuation of activated IGF-IR ( Figure 3 ) . Previous studies have demonstrated that ligand-activated IGF-IR is ubiquitinated and subsequently undergoes CME for its down-regulation ( Monami et al . , 2008; Zheng et al . , 2012 ) . In addition , the recycling of IGF-IR has been shown to in part contribute to sustained activation of Akt in response to IGF-I ( Romanelli et al . , 2007 ) . In this study we showed that stable expression of IRS-1 inhibits ligand-dependent internalization of IGF-IR , leading to sustained activation of IGF-IR kinase and the downstream Akt signaling . This effect of IRS-1 on prolonging the Akt signaling is likely based on two independent functions of IRS-1 . First , the interaction of IRS-1 with AP2 is required since expression of the IRS-1 mutant 3YA could prolong neither IGF-IR phosphorylation nor Akt phosphorylation in IGF-I-stimulated cells . Second , the ability of IRS-1 to engage PI3K is also necessary because expression of the IRS-1 mutant ΔPTB could prolong phosphorylation of IGF-IR but failed to sustain Akt phosphorylation ( Figure 7—figure supplement 1F ) . Similar signaling events were also observed in AP2-depleted cells where IRS-1 degradation , a consequence of negative feedback , was normally induced by long-term IGF-I stimulation ( Figure 7—figure supplement 1D , E ) . Notably , the ability to interact with AP2 , enhance the surface retention of IGF-IR , and prolong the Akt signaling is specific for IRS-1 , but not for IRS-2 . Thus , IRS-1 can act as a pivotal modulator for IGF signaling duration via its control of IGF-IR internalization while the downstream signaling activation can be mediated by either IRS-1 or IRS-2 ( Figure 9B ) . It is generally recognized that IGF-IR preferentially mediates growth whereas insulin receptor ( IR ) functions in glucose homeostasis in spite of the fact that both receptors share common signaling pathways mediated by the IRS proteins ( Accili et al . , 1996; Liu et al . , 1993; Nakae et al . , 2001 ) . However , these functional differences between IR and IGF-IR cannot be attributed to characteristics of the receptors themselves , such as their kinetics of ligand binding or their tissue/cellular distribution ( Siddle , 2012 ) . Moreover , insulin levels fluctuate in response to the nutrients while IGF levels are constantly maintained by circulating IGF binding proteins and by paracrine/autocrine production ( Jones and Clemmons , 1995 ) . Yet , despite these differences in temporal pattern , this is unlikely to explain the specificity of IGF-IR and IR because even in cell culture these receptors mediate different bioactivities as well as gene expression profiles ( Lammers et al . , 1989; Palsgaard et al . , 2009 ) , including in a recent study using reconstituted model cell lines solely expressing either receptor ( Cai et al . , 2017 ) . While differential substrate preference for each receptors has been proposed to explain this specificity ( Cai et al . , 2017 ) , both receptors still induce signaling through the PI3K-Akt cascade and involve many IRS proteins ( White , 2002; Taniguchi et al . , 2006 ) . In addition , the Akt signaling cascade itself can produce different temporal dynamics in response to specific stimuli and to induce different cellular outcomes ( Gross and Rotwein , 2016; Kubota et al . , 2012 ) . Our study demonstrates that the IGF-IR pathway encodes prolonged Akt activation via IRS-1-mediated delay of IGF-IR internalization ( Figure 9B ) . In contrast , IR has been shown to undergo rapid CME in response to insulin ( Choi et al . , 2016; Morcavallo et al . , 2012 ) . These observations raise the possibility that the bioactive difference between IGF-IR and IR arises in part through their differential temporal activation of the PI3K-Akt pathway governed by CME kinetics unique to each receptor . In this context , future studies could productively address whether and how the CME of IGF-IR and IR are selectively regulated , which is also a general issue in the context of CME selectivity for multiple cargos ( Grossier et al . , 2014; Weng et al . , 2014 ) . Notably , Choi et al . ( 2016 ) revealed that IR , but not IGF-IR , uses the receptor-associated adaptor BUBR1/MAD2 to facilitate rapid CME by recruiting AP2 to IR . We are likely to better understand the role of differential endocytic regulation of IGF-IR and IR in temporal dynamics of the PI3K-Akt pathway when we identify the specific adaptors for IGF-IR and IR that engage their CME , and determine their relationship with IRS-1 . Our results demonstrate that the prolonged Akt signaling elicited by IRS-1-mediated surface retention of IGF-IR affects the FoxO-targeting gene expression . Long-term action of IGF is fundamental for various physiological aspects including growth control and neural cell survival ( Ness and Wood , 2002; Gross and Rotwein , 2016; Stewart and Rotwein , 1996 ) . Thus , IRS-1-mediated delay of IGF-IR internalization is likely to be a common mechanism for long-term IGF actions .
Anti-phospho-IGF-IRβ ( Tyr1131 ) antibody ( 3021 ) , anti-phospho-IGF-IRβ ( Tyr980 ) antibody ( 4568 ) , anti-phospho-IGF-IRβ ( Tyr1316 ) antibody ( 6113 ) , anti-IGF-IRβ antibody ( 9750; for immunofluorescence staining ) , anti-Akt antibody ( 9272 ) , anti-phospho-Akt ( Thr308 ) antibody ( 9275 ) , anti-phospho-Akt ( Ser473 ) antibody ( 9271 ) , anti-phospho-p70 S6K ( Thr389 ) antibody ( 9234 ) , anti-phospho-FoxO1 ( Thr24 ) /FoxO3a ( Thr32 ) antibody ( 9464 ) , anti-phospho-FoxO1 ( Ser256 ) antibody ( 9461 ) , and anti-FoxO1 antibody ( 2880 ) were purchased from Cell Signaling Technology ( Tokyo , Japan ) . Anti-IGF-IRα antibody ( sc-712 ) , anti-IGF-IRβ antibody ( sc-713; for immunoblotting and immunoprecipitation ) , anti-IRS-2 antibody ( sc-8299 ) , anti-clathrin HC antibody ( sc-12734; for immunoblotting ) , anti-α-adaptin antibody ( sc-17771 ) , anti-γ-adaptin antibody ( sc-10763 ) , anti-p70 S6K antibody ( sc-230 ) , anti-HSP90 antibody ( sc-7947 ) , anti-ubiquitin antibody ( sc-8017 ) and anti-GFP antibody ( sc-9996 ) were purchased from Santa Cruz Biotechnology ( Santa Cruz , CA ) . Anti-FLAG M2 antibody , anti-α-tubulin antibody ( DM1A ) , and anti-phospho-Tyr antibody ( 4G10 ) were purchased from Sigma-Aldrich ( Tokyo , Japan ) . Anti-IRS-1 antibody ( 06–248 ) , anti-myosin heavy chain ( 05–716 ) antibody , anti-Myc antibody ( 05–419 ) , and anti-p85 PI3-kinase antibody ( 06–195 ) were purchased from Upstate ( Lake Placid , NY ) . Anti-μ2 antibody ( 611350 ) was purchased from BD Biosciences ( Tokyo , Japan ) . Anti-clathrin antibody ( ab2731; for immunofluorescence staining ) , and anti-integrin β1 antibody ( ab52971 ) were purchased from abcam ( Tokyo , Japan ) . Anti-transferrin receptor antibody ( H68 . 4 ) and anti-integrin β1 antibody ( TS2/16 ) were purchased from Invitrogen ( Tokyo , Japan ) . Anti-HA antibody ( 3F10 ) was purchased from Roche ( Tokyo , Japan ) . IRS-1 polyclonal antibody for immunoprecipitation was raised in rabbit as previously described ( Yoneyama et al . , 2013 ) . L6 and HEK293T cells were cultured as previously described ( Yoneyama et al . , 2013 ) . The differentiation of L6 cells was induced as previously described ( Hakuno et al . , 2011 ) . PLAT-E cells ( provided by T . Kitamura , The University of Tokyo , Tokyo , JAPAN ) were cultured for retrovirus packaging as previously described ( Yoneyama et al . , 2013 ) . We tested each cell line for mycoplasma contamination and confirmed its absence using PCR Mycoplasma Test Kit I/C ( PromoKine , Heidelberg , Germany ) before experiments . The transfection of expression plasmids was performed by using polyethylenimine ( PEI ) for HEK293T cells as previously described ( Lanzerstorfer et al . , 2015 ) , or by using Lipofectamine LTX ( Invitrogen ) for L6 cells . For RNA interference ( RNAi ) , the cells were transfected with the following siRNAs ( RNAi Corp . , Tokyo , Japan ) by using Lipofectamine RNAiMAX ( Invitrogen ) according to the manufacturer’s instructions: clathrin ( #1 ) , 5’-GUAUGCCUCUGAAUCGAAAGA-3’; clathrin ( #2 ) , 5’-CAGAAGAAUCGACGUUAUUUU-3’; μ2 ( #1 ) , 5’-CGAAGUGGCAUUUACGAAACC-3’; μ2 ( #2 ) , 5’-CUGCUUUGGGAUAGUAUGAGC-3’; IRS-1 ( #1 ) , 5’-CAAUGAGUGUGCAUAAACUUC-3’; IRS-1 ( #2 ) , 5’-GCCUCGAAAGGUAGACACAGC-3’; μ1 , 5’-CAGACGGAGAAUUCGAACUCA-3’; non-targeting control ( Ctrl , 5’-GUACCGCACGUCAUUCGUAUC-3’ . A series of IRS-1 deletion mutants ( amino acid residues 1–865 , 1–542 , 1–259 and full-length of rat IRS-1 ) were cloned into pFLAG-CMV vector . The full-length of IRS-1 was also cloned into pmRFP-C1 vector . EGFP-fused IRS-1 and 3YA ( Y608A/Y628A/Y658A ) ( Yoneyama et al . , 2013 ) were cloned into pMXs-Puro vector ( provided by T . Kitamura , The University of Tokyo , Tokyo , JAPAN ) . FLAG-fused IRS-1 was also cloned from pFLAG-CMV-IRS-1 into pMXs-Puro . The construction of pFLAG-CMV-IRS-2 was described previously ( Fukushima et al . , 2011 ) . EGFP-fused IRS-2 was also cloned from pEGFP-IRS-2 ( Lanzerstorfer et al . , 2015 ) into pMXs-Puro . Full-length IGF-IR was cloned into pEGFP-N1 to generate the construct of IGF-IR fused with EGFP at its C-terminus . IGF-IR-EGFP and IGF-IR-FLAG ( Fukushima et al . , 2012 ) were then cloned into pMXs-Puro . To generate the double-tagged IGF-IR construct ( IGF-IR-HA-EGFP ) , the fragment encoding the α subunit attached to the HA epitope ( α + HA ) and the fragment encoding the β subunit attached to the HA epitope ( β + HA ) were prepared by PCR with independent primer sets as follows: for α + HA , 5’-CTCAAGCTTCGAATTCATGAAGTCTGGCTCCGGA-3’ and 5’-TGGAACATCGTATGGGTACATGGTggccacttgcatgacatctctc-3’; for β + HA , 5’-CCATACGATGTTCCAGATTACGCTaacaccaccatgtccagccgaa-3’ and 5’-GGCGACCGGTGGATCCGCGCAGGTCGAAGACTGGGGCA-3’ . The two fragments were cloned into pEGFP-N1 by using In-Fusion Cloning HD Kit ( TAKARA ) . The IGF-IR-HA-EGFP was then cloned into pMXs-Puro . The cDNA of human integrin β1 was cloned into pMXs-Puro . The expression plasmid of EGFR fused with EGFP was purchased from Addgene ( #32751 ) . The cDNA encoding rat σ2 subunit of the AP2 complex was obtained from pACT2-σ2 ( provided by H . Ohno , RIKEN , Kanagawa , Japan ) , and cloned into pCS2-mRFP4 ( provided by M . Taira , The University of Tokyo , Tokyo , Japan ) . The cDNA encoding human PTP1B was cloned into pCMV5-Myc vector , and the D181A mutation was introduced by site-directed mutagenesis . Construction of pGEX-μ1 was described previously ( Yoneyama et al . , 2013 ) . The full-length cDNA of mouse μ2 was obtained from pcDNA-μ2 ( provided by H . Ohno , RIKEN , Kanagawa , Japan ) and cloned into pGEX-5X-3 . To generate the construct for the recombinant C-terminal region of rat μ2 fused with His-tag , the region corresponding to amino acid residues 158–435 was cloned by RT-PCR using total RNA isolated from L6 cells and subcloned into pET15b . Retrovirus production and retrovirus transduction in L6 cells were performed as described previously ( Yoneyama et al . , 2013 ) . Briefly , PLAT-E cells were transiently transfected with pMXs-Puro vectors by using PEI reagent , and the medium containing retrovirus was collected . L6 cells were incubated with the virus-containing medium supplemented with 2 μg/ml of polybrene . Uninfected cells were removed by puromycin selection . L6 cells expressing EGFP-fused constructs were further isolated using a FACSAria II cell sorter ( BD Biosciences ) as EGFP-positive cells . For lentiviral RNAi , shRNA sequences against IRS-1 were cloned into pLV-hU6-EF1a-green ( Biosettia , San Diego , CA ) according to the manufacturer’s instructions . The shRNAs used in this study comprised the following sequences: shLacZ , 5’-GCTACACAAATCAGCGATTT-3’; shIRS-1_5 , 5’-GCAGGCACCATCTCAACAATCC-3’; shIRS-1_6 , 5’-GAGAATATGTGAATATTGAATC-3’ . HEK293T cells were transiently transfected with pLV-hU6-EF1a-green vectors together with pCAG-HIVgp and pCMV-VSV-G-RSV-Rev ( provided by RIKEN BRC , Ibaraki , Japan ) by using PEI reagent , and the medium containing lentivirus was collected followed by concentration with Lenti-X Concentrator ( Clontech , Fremont , CA ) to achieve high titer virus . The virus titer was evaluated by GFP fluorescence expressed from pLV-hU6-EF1a-green vector in L6 myoblasts infected with serially diluted virus-containing medium . Lentiviral infection was conducted on the second day of differentiation . The virus-containing medium supplemented with 8 μg/ml of polybrene was added into L6 myotube culture , and the culture plates were spun at 1200 g for 1 hr at room temperature to increase the infection efficiency . After incubation for 1 day , differentiation medium was replaced , and the myotubes were cultured for additional 5 days . Purification of GST-fused proteins from E . coli BL21 and pull-down assays were performed as described previously ( Yoneyama et al . , 2013 ) . Briefly , lysates of L6 cells or HEK293T cells expressing GFP-IRS-1 mutants were incubated with purified GST-fused proteins bound to Glutathione Sepharose 4B ( GE Healthcare , Tokyo , Japan ) . Bound proteins were analyzed by immunoblotting with the indicated antibody . Recombinant human IGF-I was kindly donated by T Ohkuma ( Astellas Pharma Inc . , Tokyo , Japan ) . Recombinant human EGF was purchased from Thermo Fisher . Prior to ligand stimulation , the cells were serum-starved for 12 hr in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 0 . 1% bovine serum albumin ( BSA ) , and then treated with the ligand ( 100 nM IGF-I or 100 nM EGF ) for the indicated time . When needed , cells were preincubated for 30 min with chemical inhibitors at the following concentrations: 250 μg/ml leupeptin ( PEPTIDE INSTITUTE , INC . , Osaka , Japan ) , 10 μg/ml pepstatin A ( Sigma-Aldrich ) , 100 nM Torin1 ( Cayman Chemical ) , 100 nM rapamycin ( Sigma-Aldrich ) , 0 . 1 mM primaquine ( Sigma-Aldrich ) , and 10 μg/ml cycloheximide ( Nacalai Tesque , Inc . , Kyoto , Japan ) . After the treatment , the extraction of cell lysate and immunoblotting were performed as described previously ( Yoneyama et al . , 2013 ) . Densitometry was performed in the linear phase of the exposure by using ImageJ software . The results were expressed as the percent of max , which corresponds to the highest value of phosphorylation among the time course experiments of control cells . Values represent means ±SEM from at least three independent experiments . After the treatment of inhibitors and ligands , cells were rinsed once with ice-cold PBS and then lysed in lysis buffer ( 25 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 1 mM EDTA , 10% glycerol , 1% Triton X-100 , 100 Kallikrein inhibitor units [KIU]/ml aprotinin , 20 μg/ml phenylmethylsulfonyl fluoride [PMSF] , 10 μg/ml leupeptin , 5 μg/ml pepstatin A , 500 μM Na3VO4 , and 10 mg/ml p-nitrophenyl phosphate [PNPP] ) . After brief sonication , the clear supernatant was obtained by centrifugation at 15 , 000 g for 15 min at 4°C . For immunoprecipitation of IRS-1 or IGF-IR , the lysates were incubated with anti-IRS-1 antibody or anti-IGF-IRβ antibody ( Santa Cruz ) overnight at 4°C , and further incubated in the presence of Protein G Sepharose beads ( GE healthcare ) . For immunoprecipitation of FLAG fusion proteins , the lysates were incubated with anti-FLAG M2 affinity gel beads ( Sigma-Aldrich ) for 2 hr . Immunoprecipitates were collected by centrifugation and washed three times with lysis buffer , and then proteins were eluted with Laemmli’s sample buffer . Samples were analyzed by immunoblotting with the indicated antibodies . Surface IGF-IR levels were measured as follows . L6 cells were treated with IGF-I for the indicated time , then placed on ice , washed three times with ice-cold PBS , and labeled for 30 min with Sulfo-NHS-LC-biotin ( 0 . 5 mg/ml; Pierce ) in PBS at 4°C . Biotinylation was then quenched with 15 mM glycine in PBS . After washing the cells with PBS once , they were lysed in lysis buffer . After brief sonication , the supernatant was obtained by centrifugation at 15 , 000 g for 15 min at 4°C . The cleared lysates were then incubated with Streptavidin agarose beads ( Pierce , Tokyo , Japan ) overnight at 4°C . The beads were washed three times with lysis buffer , and bound proteins were eluted with Laemmli’s sample buffer . Samples were analyzed by immunoblotting with the indicated antibodies . Internalization of IGF-IR was measured as follows . Serum-starved L6 cells were washed three times with cold PBS before incubation with 0 . 2 mg/ml Biotin-SS-Sulfo-OSu , a nonpermeable and reversible biotinylation reagent ( Dojindo , Kumamoto , Japan ) , in PBS for 30 min at 4°C . After surface labeling , cells were washed twice with 15 mM glycine in PBS on ice , and transferred to 0 . 1% BSA in DMEM with or without the ligand to allow internalization . At the indicated times , cells were washed once with cold PBS and treated twice with 100 mM MesNa ( 50 mM Tris-HCl , pH 8 . 6 , 100 mM NaCl , and 0 . 1% BSA ) , a nonpermeable reducing regent , for 15 min at 4°C to remove biotin . MesNa was quenched with 5 mg/ml iodoacetamide in PBS for 10 min at 4°C . After two cold PBS washes , cells were lysed followed by streptavidin pull-down as described above . Cells were washed with ice-cold PBS and then lysed with lysis buffer supplemented with 100 mM N-ethylmaleimide ( NEM ) . The cleared lysates were subjected to immunoprecipitation with anti-FLAG M2 affinity gel beads . The immunoprecipitates were then washed three times with lysis buffer supplemented with 100 mM NEM , and heated in 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , and 1% SDS at 98°C for 5 min to disrupt non-covalent protein-protein interactions . The supernatants diluted with lysis buffer ( 1:10 ) were re-immunoprecipitated with anti-FLAG M2 affinity gel beads , and then subjected to SDS-PAGE . After transfer to PVDF membranes , the membranes were subjected to a denaturing treatment prior to blocking the primary antibody by incubation for 30 min at 4°C in 50 mM Tris-HCl , pH 7 . 5 , 6M guanidine-HCl , and 5 mM 2-mercaptoethanol . For confocal microscopy L6 cells were grown on coverslips . For TIRF microscopy the cells were grown on Glass Bottom Dish Hydro ( MATSUNAMI , Osaka , Japan ) . In both cases , the cells were fixed for 20 min at room temperature in prewarmed 4% paraformaldehyde in PBS . The fixed cells were then washed three times with PBS and subsequently incubated for 5 min in 50 mM ammonium chloride in PBS . After washing three times with PBS , cells were permeabilized with 0 . 25% Triton X-100 in PBS at room temperature for 5 min . The cells were washed three times with PBS and then blocked for 1 hr at room temperature with BSA blocking buffer ( 3% BSA and 0 . 025% NaN3 in PBS ) . Primary antibodies diluted in BSA blocking buffer were added overnight at 4°C . The samples were washed three times with PBS and incubated for 1 hr at room temperature in the solution of Alexa Fluor-conjugated secondary antibodies diluted in BSA blocking buffer . For LysoTracker experiments , LysoTracker Red DND-99 ( Life Technologies , Tokyo , Japan ) was added to cells at the concentration of 50 nM 30 min prior to fixation . Fixed cells were stained with Hoechst 33342 ( Molecular Probes , Tokyo , Japan ) to visualize nuclei . Coverslips were mounted in Vectashield ( Vector Laboratories , Burlingame , CA ) for confocal microscopy . Fixed cells in glass bottom dishes were imaged in PBS for TIRF microscopy . To chase surface IGF-IR , L6 cells stably expressing IGF-IR-HA-EGFP were serum-starved , washed three times with ice-cold Hank’s Balanced Salt Solution ( HBSS ) , and then incubated on ice for 1 hr with 2 μg/ml anti-HA antibody diluted in HBSS . After removing the excess antibody , cells were incubated in 0 . 1% BSA in DMEM with or without IGF-I at 37°C for different time periods . At each time point , non-permeabilized cells were either fixed to visualize the surface receptor or acid washed in an ice-cold buffer ( 100 mM glycine , 20 mM Mg ( OAc ) 2 , and 50 mM KCl , pH 2 . 2 ) to strip surface-bound antibody . Cells were fixed and permeabilized to visualize the internalized receptor . To examine endocytosis of transferrin , L6 cells were serum-starved for 30 min , and incubated with 25 μg/ml Alexa Fluor 546-conjugated transferrin ( Invitrogen ) for the indicated time . Surface-bound fraction was evaluated from the cells labeled with Alexa Fluor 546-conjugated transferrin at 4°C . The rate of uptake is expressed as internalized/surface-bound fluorescent intensity . To examine EGF-dependent internalization of EGFR , L6 cells transfected with pEGFR-EGFP plasmid were treated with 2 nM EGF for indicated time . To examine internalization of integrin β1 , L6 cells stably expressing human integrin β1 were serum-starved and then labeled with anti-integrin integrin β1 antibody ( TS2/16 ) , which recognizes human integrin β1 , for 30 min on ice . After removing the excess antibody , cells were incubated in 0 . 1% BSA in DMEM at 37°C . At each time point , cells were washed in ice-cold acid buffer to strip surface-bound antibody . Fixed cells were observed by confocal microscopy . Confocal imaging of fixed and fluorescently stained samples was performed on an inverted Olympus FV1200 microscope . Appropriate excitation and emission wavelengths were configured by the instrument running FV10-ASW software , and emission signals in the different channels were collected in the sequential scan mode . TIRF imaging of fixed and fluorescently stained samples was performed on Leica AF6000LX total internal reflection ( TIRF ) microscopy equipped with a 100 × 1 . 46 NA oil-immersion objective and a Cascade II EMCCD camera ( Roper , Tucson , AZ ) . Images were analyzed with Adobe Photoshop CC2017 and Fiji software . Live cell dual-color TIRF microscopy was carried out as described previously ( Lanzerstorfer et al . , 2015 ) . Quantifications were performed with Fiji software . Mean fluorescence levels in individual cells minus the background fluorescence were calculated and averaged . For colocalization analysis , background intensity was subtracted by median subtraction , the value of Mander’s colocalization coefficient ( MCC ) , which is one of the most widely accepted methods to measure colocalization of different markers ( Dunn et al . , 2011 ) , was calculated by Fiji plugin in individual cells . The number of AP2-positive spots was determined as follows . Punctate structures were extracted using median subtraction , and binary images were created . Small punctae less than 5 pixel2 were removed , and the number of spots was counted using the morpheme analysis program . Images of differentiated myotubes were obtained by BZ-9000 microscope ( Keyence , Osaka , Japan ) . Myotube diameter was quantified by measuring a total of over 100 tube diameters from ten random fields using Fiji software . Total RNA from L6 cells was extracted with TRIzol reagent ( Invitrogen ) from three independently collected cells . First-strand cDNA was synthesized with ReverTra Ace qPCR Master Mix ( TOYOBO , Osaka , Japan ) . Quantitative PCR was performed with THUNDERBIRD SYBR qPCR Mix ( TOYOBO ) on an ABI StepOnePlus Real Time PCR System ( Applied Biosystems ) . To normalize the relative expression , a standard curve was prepared for each gene for relative quantification , and the expression level of each gene was normalized to the Rn18s gene . Specific primers for atrophy-related genes were used: Fbxo32 F: ACTTCTCGACTGCCATCCTG; Fbxo32 R: TCTTTTGGGCGATGCCACTC; Trim63 F: GGGAACGACCGAGTTCAGAC; Trim63 R: GCGTCAAACTTGTGGCTCAG; Fbxo30 F: TGCAGTGGGGGAAAAAGAAGT; Fbxo30 R: TGCAGTACTGAATCGCCACA; Fbxo21 F: ACTCCATCGGGCTCGTTATG; Fbxo21 R: TGTTTCGGATCCACTCGTGC; Map1lc3b F: GCCGGAGCTTCGAACAAAGA; Map1lc3b R: GCTTCTCACCCTTGTATCGC; Gabarapl1 F: ACAACACTATCCCTCCCACC; Gabarapl1 R: GCTTCTGCCTCATTTCCCGTA; Rn18s F: TCCCAGTAAGTGCGGGTCATA; Rn18s R: CGAGGGCCTCACTAAACCATC . Yeast two-hybrid assay using pAS-IRS-1 and pACT2-μ2 to assess the interaction between IRS-1 and μ2 was performed as described previously ( Hakuno et al . , 2007 ) . Construct of 6 × His tagged C-μ2 ( rat μ2 amino acid residues 158–435 ) cloned into pET15b was transformed into an E . coli strain BL21-CodonPlus ( DE3 ) -RIL ( Agilent Technologies , Santa Clara , CA ) . Bacteria were grown in LB supplemented with ampicillin and chloramphenicol at 37°C to OD600 of 0 . 7 . Expression was induced with 0 . 1 mM isopropyl β-D-thiogalactopyranoside ( IPTG ) at 17°C overnight . The cells were harvested by centrifugation and homogenized with a sonicator in a buffer of 50 mM Tris-HCl ( pH 8 . 0 ) , 500 mM NaCl , 20 mM imidazole , 5% glycerol , and 0 . 1% Triton X-100 supplemented with cOmplete EDTA-free protease inhibitor cocktail ( Roche ) . Insoluble material was removed by centrifugation . The protein was affinity-purified on HisTrap HP column ( GE Healthcare ) . The His-tag was removed by cleavage of thrombin at room temperature for 4 hr . Thrombin-cleaved C-μ2 was further purified with HiTrap SP HP column ( GE Healthcare ) , and uncleaved fusion protein was removed by passage through HisTrap HP column . The C-μ2 was finally purified by gel filtration on HiLoad 16/60 Superdex200 column equilibrated in a buffer of 10 mM HEPES-KOH ( pH 7 . 5 ) , 150 mM NaCl , and 2 mM dithiothreitol ( DTT ) for crystallization . Three eight-residue peptides of IRS-1 were chemically synthesized with their sequences GY ( 608 ) MPMSPG , DY ( 628 ) MPMSPK and GY ( 658 ) MMMSPS , where the tyrosine residue in a YxxΦ motif is indicated with its residue number in parentheses ( Toray Research Center , Inc . , Tokyo , Japan ) . Hereafter , they are referred to as Y608 peptide , Y628 peptide , and Y658 peptide , respectively . The peptides were dissolved in 10 mM HEPES buffer ( pH 7 . 5 ) containing 150 mM NaCl and 2 mM DTT . C-μ2 was mixed with each peptide in the molecular ratio of 1:10 . Crystals of the Y608 peptide were grown by the sitting drop method at 293 K with the reservoir solution containing 1 . 4 M sodium formate , 50 mM nickel chloride and 100 mM sodium acetate ( pH 6 . 0 ) . Crystals of the Y628 and Y658 peptides were grown by the hanging drop method at 291 K with the reservoir solution containing 2 . 2–2 . 3 M sodium chloride , 400 mM sodium potassium phosphate , 10 mM DTT , 15% ( v/v ) glycerol and 100 mM MES ( pH 6 . 5 ) . Crystals were briefly soaked in well solution containing 20% ( v/v ) glycerol before flash-cooled in liquid nitrogen . Diffraction data were collected on BL26B2 at SPring-8 , Harima , Japan , and processed using HKL2000 ( Otwinowski and Minor , 1997 ) and the CCP4 suite ( Winn et al . , 2011 ) . Molecular replacement was carried out with CCP4 program MOLREP ( Vagin and Teplyakov , 1997 ) using the μ2 subunit in the complex with EGFR internalization signal peptide ( Owen and Evans , 1998 ) ( PDB 1BW8 ) as the search model . Refinement was performed with REFMAC5 ( Murshudov et al . , 1997 ) and PHENIX ( Adams et al . , 2010 ) , while model building was performed with COOT ( Emsley and Cowtan , 2004 ) . The N-terminal residue and residues 220–237 of C-μ2 were not modeled for the complexes of the Y628 and Y658 peptides . As for the complex with the Y608 peptide , it appeared that the region encompassing residues 219–260 underwent a conformational change where the electron density was not enough to precisely trace the structure . Residues 224–260 were not modeled except for a five-alanine strand which was placed as unconfirmed residues in a patch of visible electron density . Structural models in the figures were drawn using PyMOL ( The PyMOL Molecular Graphics System , Schrödinger , LLC ) . Coordinates and structure factors of the three complexes have been deposited in the Protein Data Bank ( PDB ) with accession codes indicated in Table 1 . Comparisons between two groups were performed using two-tailed , unpaired Student’s t test , whereas comparisons among more than two groups were analyzed by analysis of variance ( ANOVA ) and the Tukey post hoc test . p Values of < 0 . 05 were considered statistically significant . | Mammals , including humans , use signaling molecules called hormones to carry information from one cell to another . Insulin-like growth factor ( or IGF for short ) is a hormone that is essential throughout an animal’s lifetime . It is needed for growth and for many of the chemical processes that must occur to maintain life ( which are collectively referred to as an animal’s metabolism ) . IGF binds to and activates a protein found on the surface of cells , which then transmits the signal inside the cells . This surface protein is known as the IGF-I receptor , and once it is activated by IGF binding , it is removed from the cell surface and then incorporated inside the cell to switch off the signal . The IGF signal in cells needs to be properly balanced to prevent disorders of growth and metabolism . How long the activated IGF-I receptor remains at the cell surface and when the IGF-I receptor starts to enter inside the cells after cells receive IGF influence the signals within the cell . Often IGF signaling must be activated for long periods , for example when cells maintain their balance between making and breaking proteins . However , it remains poorly understood how the IGF-I receptor produces a sustained signal . Yoneyama et al . have now focused on a protein called IRS-1 , which was known to act downstream of the receptor . The experiments revealed that this protein determines how long activated IGF-IR remains at the cell surface before it enters inside cells . It achieves this by binding to a complex of proteins , known as AP2 , which normally internalizes the IGF-I receptor . However , when IRS-1 binds , it inhibits AP2 . This means that the receptor is no longer rapidly removed from the cell surface and can continue signaling for long periods of time . The findings of Yoneyama et al . help to explain how long-term IGF signaling is regulated . Further work that builds on these findings could help scientists to understand how uncontrolled IGF signals cause the development of diseases including cancer and metabolic disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2018 | IRS-1 acts as an endocytic regulator of IGF-I receptor to facilitate sustained IGF signaling |
Detrimental microbes caused the evolution of a great diversity of antimicrobial defenses in plants and animals . Insects developing underground seem particularly threatened . Here we show that the eggs of a solitary digger wasp , the European beewolf Philanthus triangulum , emit large amounts of gaseous nitric oxide ( NO⋅ ) to protect themselves and their provisions , paralyzed honeybees , against mold fungi . We provide evidence that a NO-synthase ( NOS ) is involved in the generation of the extraordinary concentrations of nitrogen radicals in brood cells ( ~1500 ppm NO⋅ and its oxidation product NO2⋅ ) . Sequencing of the beewolf NOS gene revealed no conspicuous differences to related species . However , due to alternative splicing , the NOS-mRNA in beewolf eggs lacks an exon near the regulatory domain . This preventive external application of high doses of NO⋅ by wasp eggs represents an evolutionary key innovation that adds a remarkable novel facet to the array of functions of the important biological effector NO⋅ .
Microbes pose a major threat to the health of all animals and plants . These have responded by evolving a great diversity of defenses including hygienic behaviors ( Gilliam et al . , 1983 ) , antimicrobial chemicals ( Herzner et al . , 2013; Vilcinskas et al . , 2013; Gross et al . , 2008 ) , complex immune systems ( Hooper et al . , 2012; Iwasaki and Medzhitov , 2010 ) , and defensive symbioses ( Kaltenpoth et al . , 2005; Flórez et al . , 2017 ) . Besides such pathogenic effects , many bacteria and fungi are severe , but often neglected , competitors of animals for nutrients , thus prompting the evolution of mechanisms to preserve food sources ( Janzen , 1977; Rozen et al . , 2008 ) . Some animals are particularly prone to suffer from microbial attack due to ( 1 ) high abundance of potentially harmful microbes in their environment , ( 2 ) a microbe-friendly microclimate and/or ( 3 ) limited defense mechanisms . The progeny of many insect species develop under warm and humid conditions in the soil , where they are exposed to a high diversity of bacteria and fungi . Moreover , compared to adult insects , immature stages , in particular eggs , have usually reduced abilities to prevent microbial infestation due to , for example , a thin cuticle or an inability to groom ( Wilson and Cotter , 2013; Tranter et al . , 2014 ) . The situation is even aggravated when eggs and larvae have to develop on limited amounts of provisions that are susceptible to attack by ubiquitous and fast growing putrefactive bacteria and mold fungi ( Janzen , 1977; Arce et al . , 2013 ) . Such hostile conditions prevail in nests of subsocial Hymenoptera like the European beewolf Philanthus triangulum ( Hymenoptera , Crabronidae ) . The offspring of these solitary digger wasps develop in subterranean brood cells provisioned by the female wasps with paralyzed honeybee workers ( Apis mellifera , Apidae , Hymenoptera ) ( Strohm and Linsenmair , 2000 ) ( Figure 1A ) . The beewolf egg is laid on one of the bees , the larva hatches after three days , feeds on the bees for six to eight days , then spins a cocoon and either emerges the same summer or hibernates . The warm and humid microclimate in the brood cell promotes larval development but also favors fast growing , highly detrimental fungi ( Engl et al . , 2016 ) . Without any countermeasures the provisions will be completely overgrown by mold fungi within three days ( Figure 1B ) , and the beewolf larva becomes infested by fungi or starves to death ( Strohm and Linsenmair , 2001; Herzner et al . , 2011a ) . We have previously documented two adaptations that beewolves have evolved to counter the detrimental effects of fungi on their brood . First , beewolf females reduce molding of the larval provisions by coating the paralyzed bees with ample amounts of unsaturated hydrocarbons ( Herzner et al . , 2007 ) . This embalming changes the physicochemical properties of the preys' surface causing reduced water condensation on the bees ( Herzner and Strohm , 2007 ) . Due to the deprivation of water , germination and growth of fungi is delayed by two to three days ( Herzner et al . , 2011b ) . Second , during the long period of overwintering in their cocoons beewolf larvae are protected by antibiotics on their cocoons ( Kaltenpoth et al . , 2005; Kroiss et al . , 2010 ) . Prior to oviposition , beewolf females apply a secretion containing symbiotic Streptomyces bacteria to the ceiling of the brood cell . The secretion is taken up by the larvae and incorporated into the silk threads of their cocoons . There , the bacteria produce several antibiotics that effectively protect the cocoon and , thus , the larvae against fungus infestation ( Kroiss et al . , 2010; Engl et al . , 2018 ) . Despite the considerable effect of prey embalming , when removed from brood cells at least 50% of embalmed bees showed fungus infestation within six days after oviposition ( Strohm and Linsenmair , 2001 ) . Since in natural brood cells only around 5% of the progeny succumb to mold fungi ( Strohm and Linsenmair , 2001 ) , we searched for an additional antimicrobial defense mechanism that takes effect during the early stages of beewolf development . Here we report on a unique antifungal strategy that is employed by beewolf eggs to defend themselves and their provisions against mold fungi . Employing bioassays we discovered that beewolf eggs emit a strong antifungal agent that we identified as the gaseous radical nitric oxide ( NO⋅ ) . We characterize the amount , time course and temperature dependence of emission and show that synthetic NO⋅ exerts a similar effect as the gas emitted by beewolf eggs . Furthermore , we tested whether there was an interaction of the gas emitted by the eggs and the embalming of the prey by beewolf females . Using histological methods , inhibition assays , and gene expression analysis , we elucidate a biosynthetic pathway for NO⋅ synthesis in beewolf eggs . To explore the evolutionary background of this remarkable antimicrobial strategy , we sequenced the relevant gene and mRNA . Our findings reveal a novel function of the eminent and widespread biological effector NO⋅ in providing an extended immune defense to the producer by sanitizing its developmental microenvironment .
Thorough examination of beewolf nests in observation cages ( Strohm and Linsenmair , 1994 ) revealed that within 24 hr after oviposition , a conspicuous pungent smell occurred that was clearly emanating from the eggs and disappeared by the time the larvae hatched . We hypothesized that this smell was due to an antifungal agent . When paralyzed honeybees from completed beewolf brood cells were incubated individually , bees carrying an egg showed significantly delayed fungus growth compared to bees without egg over the period from oviposition to cocoon spinning ( Kaplan Meier survival analysis , Breslow test , day 0–11: Chi square = 12 , df = 1 , p=0 . 001; Figure 2A ) . This difference was also significant for the period from oviposition to the hatching of the larvae ( day 0–3: Chi square = 9 . 5 , df = 1 , p=0 . 002 ) , suggesting that this effect is not due to possible antifungal mechanisms of the larvae but that it is mediated by the egg . Considering the distinctive odor that emanated from the eggs , we tested whether the antifungal effect is caused by a volatile agent . Two experiments supported this assumption . First , provisioned bees without wasp eggs that were kept in artificial brood cells together with bees carrying an egg ( but without physical contact ) showed significantly delayed fungal growth compared to control bees that were kept alone ( Breslow test , day 0–11: Chi square = 7 . 6 df=1 , p=0 . 006; day 0–3: Chi square = 9 . 1 , df = 1 , p=0 . 003; Figure 2B ) . Second , when one of the most abundant mold species from infested beewolf brood cells , the fast growing Aspergillus flavus ( Engl et al . , 2016 ) , was exposed to the volatiles presumably emanating from beewolf eggs on nutrient agar for three days , its growth was entirely inhibited , whereas it thrived in controls ( for all observation times 24 hr , 48 hr , 72 hr: binomial test: N = 20 , p<0 . 001 , Figure 3 ) . Notably , when the beewolf larvae were removed from the assays shortly after hatching ( three days after oviposition ) , no fungal growth occurred in the exposed areas during another three days . A similar experiment showed that paralyzed honeybees alone did not show any antifungal activity ( Appendix 1: Additional data 1 ) . Analogous bioassays with five other fungal strains ( A . flavus strain B , Mucor circinelloides , Penicillium roqueforti , Candida albicans and Trichophyton rubrum ) revealed that in all cases fungus growth was likewise completely inhibited in the area that was exposed to volatiles from beewolf eggs , whereas the fungi thrived in the respective control areas ( for each strain: N = 8 , binomial test: p<0 . 01 ) . We conclude that beewolf eggs release a volatile compound with broad spectrum fungicidal properties . The odor emanating from the eggs was similar to that of strong oxidants like chlorine , ozone and nitrogen dioxide ( subjective evaluation of several observers , [Mücke and Lemmen , 2010] ) . In fact , the generation of a strong blue coloration when placing a beewolf egg into the lid of a reaction tube filled with a iodide/starch solution ( [Jander and Blasius , 1971] , see iodometry in Materials and methods ) revealed the existence of an oxidant in the headspace of beewolf eggs . There are few gaseous oxidants that might be considered , in particular chlorine , ozone and nitrogen dioxide . Ozone has , to our knowledge , not been described to be synthesized by organisms . Molecular chlorine has been reported as an intermediate in some organisms but its occurrence seems to be restricted to phagocytosis ( Hazen et al . , 1996 ) . The most likely candidate was the radical nitrogen dioxide ( NO2⋅ ) , because there is a plausible way for it being generated by wasp eggs: Insect embryos synthesize small amounts of nitric oxide ( NO⋅ ) as signaling effectors for developmental processes ( Andersen et al . , 2013 ) . If such odorless NO⋅ was emitted from the egg , it would spontaneously react with oxygen ( Soegiarto et al . , 2003; Mur et al . , 2011 ) to yield the strong-smelling NO2⋅ . Moreover , belonging to the reactive nitrogen species ( RNS ) , NO⋅ and NO2⋅ show considerable antimycotic activity ( Fang , 1997; Lai et al . , 2011 ) that would explain the observed fungicidal effect of beewolf eggs . Hence , we hypothesized that eggs synthesize and emit NO⋅ that reacts with the oxygen in brood cells to NO2⋅ thus generating the pungent smell and the antimycotic activity . We tested whether beewolf eggs produce and emit NO⋅ and/or NO2⋅ by conducting a series of experiments . First , headspace samples of confined beewolf eggs were subjected to the Griess assay , the standard procedure for the specific detection of NO⋅ and NO2⋅ ( Tsikas , 2007 ) . The emerging red color of the resulting azo dye ( Figure 4—figure supplement 1 ) clearly indicated the presence of NO⋅/NO2⋅ . To visualize the emission of NO⋅ from beewolf eggs , we sprayed a solution of an NO⋅ specific fluorescent probe , Diaminorhodamin-4M AM ( DAR4M-AM ) , onto prey bees carrying freshly laid eggs . The small droplets of the DAR4M-AM solution on the bees showed a clear fluorescence around the egg that increased over several hours ( Figure 4B ) . No such effect was seen on control bees without eggs ( Figure 4A ) . Moreover , beewolf eggs injected with the DAR4M-AM solution showed a strong fluorescence that peaked about one day after oviposition ( N = 45 , Figure 5A ) . The same treatment yielded only weak fluorescence in the eggs of two other Hymenoptera ( the Emerald cockroach wasp , Ampulex compressa , N = 9 , and the Red mason bee , Osmia bicornis , N = 12; Figure 5C and D ) and in newly hatched beewolf larvae ( N = 4 , Figure 5—figure supplement 1 ) . Autofluorescence of beewolf eggs injected with buffer only ( N = 10 ) was negligible ( Figure 5B ) . These findings strongly imply that beewolf eggs produce and release NO⋅ . Using iodometry , we determined that a beewolf egg ( volume: 4 . 1 ± 0 . 5 mm3 , N = 16 ) emits on average 0 . 25 ± 0 . 09 µmol NO⋅ ( N = 233 ) . The rate of NO⋅ production was initially very low , but increased to a distinct peak 14–15 hr ( at 28°C ) after oviposition ( Figure 6 ) ; around 90% of NO⋅ emission occurred within a two-hour period . Assuming no loss due to reactions or leaking out of the confined space of brood cells ( volume 3 . 2 ± 0 . 9 cm3 , N = 250 ) , the nitrogen oxides would accumulate to average concentrations of 1690 ± 680 ppm . The timing of the onset of NO⋅ emission was strongly temperature dependent ( Figure 6—figure supplement 1 ) , with higher temperatures resulting in an earlier NO⋅production ( temperature coefficient Q10 = 2 . 74 ) . To test whether the observed antifungal effect of the gas that is emitted by beewolf eggs was due to NO⋅ , we conducted an experiment using synthetic NO⋅ but otherwise emulated natural conditions as closely as possible . Artificial brood cells containing a bee without egg were injected with either synthetic NO⋅ to generate a peak concentration of 1500ppm or with nitrogen as controls . There was a significant delay in the onset of fungus growth on the bees exposed to synthetic NO⋅ as compared to controls ( Figure 7 , Breslow test: Chi square = 13 . 3 , df = 1 , p<0 . 0001 ) . Since both , NO⋅ emission and embalming of the prey bees by beewolf females take effect during the first days after oviposition , we assessed the antifungal effects of these defense mechanisms alone and in combination . Bioassays with bees in artificial brood cells that were either embalmed or not and carried an egg or not revealed that the onset of fungal growth differed significantly among treatment groups ( Figure 8: Kaplan Meier survival analysis , Breslow test: Chi square = 69 . 6 , df = 3 , p<0 . 001 ) . Pairwise comparisons showed that , on average , fungus growth was first detected on honeybees that had not been embalmed and did not carry an egg , second were prey items that were embalmed but did not carry and egg , third were not embalmed honeybees carrying an egg and least susceptible were embalmed honeybees with an egg ( Breslow test for all pairwise comparisons: Chi square ≥8 . 6 , df = 1 , p≤0 . 003 ) . The timing of conidia formation followed the same pattern ( Breslow test: Chi square = 67 . 4 , df = 3 , p<0 . 001; for all pairwise comparisons: Chi square ≥4 . 5 , df = 1 , p≤0 . 034; Figure 8 ) . Eukaryotes synthesize NO⋅ from the amino acid L-arginine by the enzyme nitric oxide synthase ( NOS ) ( Röszer , 2012 ) which is highly conserved also in insects ( Regulski and Tully , 1995 ) . The exceptional level of NO⋅ emission of beewolf eggs raised the question of whether they employ the same pathway or have evolved a different mechanism . First , using histological staining , we assessed evidence for and site of NOS activity in beewolf eggs during the time of peak NO⋅ emission . The fixation insensitive nicotinamide-adenine-dinucleotide phosphate ( NADPH ) -diaphorase assay resulted in strong blue staining only in embryonic tissue ( Figure 9 ) , thus indicating NOS activity . Second , by employing reverse transcription and real time quantitative PCR , we revealed that the temporal expression pattern of NOS-mRNA showed a clear peak around 19–20 hr after oviposition ( Figure 10 ) . For this experiment , the eggs were kept at 25°C , so the timing of peak NOS expression corresponds to the timing of peak NO⋅ emission at this temperature ( Figure 6—figure supplement 1 ) . Third , to directly test for the involvement of NOS we injected beewolf eggs with Nω-nitro-L-arginine methylester ( L-NAME ) , a NOS-inhibiting analog of L-arginine ( Willmot et al . , 2005 ) . This treatment caused a significant decrease in NO⋅ emission , whereas the non-inhibiting enantiomer D-NAME had no such effect ( Figure 11 ) . In contrast to vertebrates , most invertebrates appear to have only one type of NOS ( Rivero , 2006; Whitten et al . , 2007 ) . Considering the high level of NO⋅ production in beewolf eggs , we hypothesized that beewolves have more than one NOS gene or that the NOS responsible for the NO⋅ synthesis in beewolf eggs might exhibit considerable changes in enzyme structure compared to the NOS of related species . Sequencing of the NOS-gene ( s ) of P . triangulum ( Pt-NOS ) revealed only one Pt-NOS copy in the beewolf genome comprising 9 . 36 kbp with 25 exons ( Figure 11—figure supplement 1 ) . A phylogenetic analysis of the resulting amino acid sequence revealed a high similarity to the NOS of the closely related bees ( Apidae , Figure 11—figure supplement 2 ) . However , mRNA sequencing showed that , in contrast to adult beewolves and honeybees , the NOS-mRNA of beewolf eggs ( 3 . 72 kbp ) lacks exon 14 comprising 144 bp . In the NOS-mRNA of adult beewolves this exon is located between the binding domains for calmodulin and flavin mononucleotide ( FMN ) ( Figure 11—figure supplement 1 ) .
Fighting pathogens is of outstanding importance for any organism and has driven the evolution of a great diversity of antimicrobial defenses . Internal immune systems have been extensively documented especially in vertebrates ( Akira et al . , 2006; Hirano et al . , 2011 ) but also in insects ( Lemaitre and Hoffmann , 2007; Siva-Jothy et al . , 2005 ) , including insect eggs ( Gorman et al . , 2004 ) . However , comparatively little is known about external antimicrobial strategies that provide protection for the own body , for the progeny , or for food . Mechanical grooming is an important mechanism to remove microbes ( Zhukovskaya et al . , 2013 ) . There are some reports on the application of antimicrobial secretions on the body surface by adult insects ( Wilson and Cotter , 2013; Otti et al . , 2014 ) or inside a host by larvae of a parasitoid wasp ( Herzner et al . , 2013 ) . Carrion beetles preserve the larval food , buried carcasses , by application of antimicrobials ( Degenkolb et al . , 2011 ) and by controlling the microbiome on the carcasses ( Shukla et al . , 2018 ) . Females of some insect species deposit antimicrobial chemicals ( Vander Meer and Morel , 1995; Marchini et al . , 1997 ) or antibiotics producing symbiotic bacteria ( Flórez et al . , 2017; Flórez et al . , 2015 ) onto their eggs and ant workers can counter microbial infestation of the brood by applying venom ( Tragust et al . , 2013 ) . Recently , the employment of volatile antimicrobials by insects as a means of external defense has gathered some interest ( Gross et al . , 2008; Gross and Schmidtberg , 2009; Weiss et al . , 2014; Lopes et al . , 2015 ) . Like other insects that develop in the soil , beewolves are particularly menaced by a diverse and unpredictable range of detrimental microbes . In fact , beewolf progeny and their provisions are under severe threat from fast growing mold fungi ( Strohm and Linsenmair , 2001 ) . The development of beewolf progeny from oviposition to cocoon spinning lasts about 11 days and is , thus , rather fast . So even a few days delay in fungus growth provides a considerable benefit for the larvae . Beewolves have evolved at least three very different antimicrobial defenses that provide an effective , coordinated , and long-term protection against a broad spectrum of microbes during the whole development . First , throughout the long period of winter diapause prior to emergence progeny are protected by antibiotics on their cocoons that are produced by symbiotic Streptomyces bacteria ( Kaltenpoth et al . , 2005; Kroiss et al . , 2010; Kaltenpoth et al . , 2014 ) . Second , during the early egg and larval stages , molding of the provisions is retarded by an embalming of the honeybees with lipids by the mother wasp ( Strohm and Linsenmair , 2001 ) . Third , as shown here , the emission of gaseous nitrogen oxide radicals by the beewolf egg results not only in delay of molding but in killing of detrimental fungi in their immediate environment thus , at least partly , eliminating this major threat . The emission of a gaseous agent by beewolf eggs to their confined brood cells is an ideal way to sanitize such intricately structured surfaces as the bodies of honeybees and the rough walls of the brood cell . NO⋅ seems to be a most suitable gaseous agent because it can obviously be produced by beewolf eggs in amounts that effectively kill mold fungi in their brood cell . Such volatile sanitation mechanisms that provide a front-line defense against microbes ( Gross et al . , 2008; Weiss et al . , 2014; Lopes et al . , 2015 ) will mostly be inconspicuous and might turn out to be a wider theme in nature . Exact quantification of nitrogen oxides ( NO⋅ and NO2⋅ ) in beewolf brood cells on a micro scale or with time has not yet been accomplished . Brood cells are located in rather compact fine grained sandy soil with some moisture . Moreover , the walls of the nest burrows and the brood cells are covered with a layer of hydrocarbons ( Kroiss et al . , 2009 ) that might provide an additional barrier . Thus , brood cell walls are neither very porous nor are they sealed . Accordingly , the concentration of NO⋅ and NO2⋅ in the brood cell will decrease but at a slow rate . By the time the larvae hatch ( three days after oviposition ) the smell of NO⋅ has vanished , indicating that the nitrogen oxides have disappeared or at least decreased considerably , explaining why the larvae remain unaffected without actually being resistant to NO⋅ . However , even assuming some loss during the two hour period of peak NO⋅ production the estimated maximum concentration of nitrogen oxides ( NO⋅ and NO2⋅ ) in beewolf brood cells ( probably around 1500 ppm or 60 µmol/l ) considerably exceeds the concentrations observed in animal tissues ( mostly lower than 0 . 1 µmol/l [Wink et al . , 2011] , 0 . 85–1 . 3µmol/l in muscle tissue [Vaughn et al . , 1998] ) . The maximum concentration in beewolf brood cells might be even higher than what is used in medical applications against multiple drug resistant bacteria ( 200 ppm NO⋅[Ghaffari et al . , 2006] ) or in antifungal treatment of fruit ( 50–500 ppm NO⋅[Lazar et al . , 2008] ) and is far beyond permissible exposure limits for humans ( e . g . for the USA: 25 ppm for NO⋅ , 5 ppm for NO2⋅[Administration USOSaH , 2014] ) . Synthetic NO⋅ applied to artificial brood cells at a concentration of 1500ppm , the estimated concentration of nitrogen oxides in natural brood cells , significantly delayed fungus growth on bees . Since there was oxygen available in the brood cell , NO⋅ was oxidized to NO2⋅ similarly to natural brood cells . The effect size ( NO⋅ treatment vs control , Figure 7: hazard ratio = 0 . 41 , 95% confidence interval 0 . 198–0 . 845 ) was slightly lower than in a comparable experiment with eggs as the source of the antifungal gas ( data for unembalmed bees from the experiment of a combined effect of embalming and emissions from the egg: Appendix 1—table 2: hazard ratio = 0 . 22 , 95% confidence interval 0 . 1–0 . 47 ) . However , since the confidence intervals of the hazard ratios are mutually overlapping , there is no evidence for a significant difference in the antifungal effect of 1500ppm synthetic NO⋅ and the gas emitted by beewolf eggs . Although we cannot exclude that small amounts of other active volatiles are released by beewolf eggs , we thus conclude that the antifungal effect of brood cell fumigation by beewolf eggs is predominantly or exclusively due to NO⋅ and its oxidation product NO2⋅ . Notably , the combination of prey embalming with unsaturated hydrocarbons that reduces condensation of water on the bees ( Herzner and Strohm , 2007 ) and brood cell fumigation with NO⋅/NO2⋅ seems to affect fungal growth beyond either of these antimicrobial measures alone . One possible explanation is based on the fact that NO⋅ and NO2⋅ dissolve in water ( confirmed by the spraying of a bee with a fluorescent dye , Figure 4 ) to yield nitric acid and nitrous acid , with the latter being a potent antibacterial agent ( Gao et al . , 2015 ) . Although embalming reduces the amount of water condensation on the prey bees , some very small droplets occur . The concentration of nitrous acid in these droplets will be considerably higher than in the larger droplets that would occur without embalming . Thus , fungal germs that might have survived the NO⋅/NO2⋅ atmosphere are not only impaired by limited availability of water , but the accessible water might be toxic for them . Moreover , due to the solubility of NO⋅ and NO2⋅ , abundant water droplets on the bees could reduce the concentration of these radicals in the brood cell , thus lessening their antimicrobial effect . The reduction of water on the bees due to prey embalming could thus help to keep fumigation effective . The combination of prey embalming and fumigation , thus , has a twofold effect . Many fungi will be killed by the NO⋅/NO2⋅ . The remaining spores will encounter unfavorable conditions that slow-down germination and growth so that the majority of larvae are able to consume most of their provisions without severe competition by mold fungi . Once larvae have spun their cocoon the antibiotics that are produced by the symbiotic bacteria take over protection until emergence ( Kaltenpoth et al . , 2005; Kroiss et al . , 2010; Engl et al . , 2018 ) . Within this multifaceted antimicrobial strategy of beewolves , brood cell fumigation might be the most important component since it takes effect at a very early developmental stage and , thus , provides beewolf offspring with a decisive head start over the fast growing mold fungi . NO⋅ is an ancient biological effector of immense importance for all kinds of organisms ranging from prokaryotes to higher plants and animals ( Röszer , 2012; Moroz and Kohn , 2007 ) . Owing to its high diffusibility across biomembranes and specific chemical properties , this gaseous radical plays a crucial role in a multitude of biological processes ( Röszer , 2012; Moroz and Kohn , 2007 ) . In vertebrates , NO⋅ is synthesized from L-arginine by three different isoforms of NOS that are encoded by different genes ( Röszer , 2012; Moroz and Kohn , 2007 ) . Low levels of NO ( <1µmol/l ) are produced by constitutive NOS ( cNOS ) isoforms ( endothelial eNOS , neuronal nNOS ) and have signaling functions , for example in neuronal development and in the regulation of vascular tone in vertebrates . Higher NO⋅ concentrations ( 1–10 µmol/l , [Thomas et al . , 2003] ) are generated by an inducible NOS ( iNOS ) . At such levels NO⋅ is highly cytotoxic ( Thomas et al . , 2003 ) , making it a powerful antimicrobial ( Fang , 1997; Lai et al . , 2011 ) , for example in macrophages ( Röszer , 2012 ) . However , overproduction of NO⋅ due to inflammatory processes ( Filipović et al . , 2010 ) or certain diseases ( e . g . Alzheimer's disease , [Lüth et al . , 2002] ) may cause harmful side-effects ( Pacher et al . , 2007 ) and even septic shock ( Titheradge , 1999 ) . Moreover , NO⋅ might affect carcinogenesis and tumor progression in a positive as well as in a negative way ( Burke et al . , 2013 ) . In living tissues , NO⋅ is usually removed within seconds by reacting with the heme group of molecules such as oxyhemoglobin ( Beckman and Koppenol , 1996; Wink et al . , 2011 ) ( very low concentrations may still persist for hours [Moroz and Kohn , 2007] ) . In brood cells , there is enough oxygen ( 670 µl ) to support the metabolism of the egg and of the paralyzed bee as well as the oxidation of NO⋅ ( for more details see Appendix 1: Additional discussion 1 ) . In air , the autooxidation to NO2⋅ is comparatively slow so that NO⋅ may persist ( depending on its concentration ) for several seconds to minutes ( Mur et al . , 2011; Wink et al . , 2011 ) or even hours ( Soegiarto et al . , 2003 ) . Thus , the NO⋅ emitted by beewolf eggs might directly affect fungi , for exapmle by damaging DNA ( Lai et al . , 2011; Jones et al . , 2010 ) or by reacting with the heme group of enzymes like cytochrome P450 and cytochrome c oxidase , thus inhibiting these crucial components of the mitochondrial respiratory chain ( Thomas et al . , 2003; Feelisch , 2008; Canessa and Larrondo , 2013 ) . Yet , most of the antimicrobial activity of NO⋅ is attributed to indirect effects via reactive nitrogen species ( RNS ) , in particular nitrogen oxides ( NO2⋅ , N2O3 ) and peroxynitrite ( ONOO- , upon reaction with superoxide ) ( Thomas et al . , 2003 ) . NO2⋅ , has been reported to be severely cytotoxic , for example by nitration of tyrosine residues and oxidation of proteins and lipids ( Fang , 1997; Bogdan , 2001 ) . A beewolf egg of approximately 5 mg emits 0 . 25 µmol NO⋅ within a period of about 2 . 5 hr , or 20 . 000 µmol/kg*h , a value that is about four orders of magnitude higher than reported baseline levels of NO⋅ synthesis in humans ( 0 . 15 - ~ 4 . 5 µmol/kg*h [Castillo et al . , 1996] , rats ( 0 . 6–9 µmol/kg*h [Wu et al . , 1999] ) and plants ( Arabidopsis thaliana , 0 . 36–3 µmol/kg*h [Zeidler et al . , 2004] ) , and even considerably higher than in lipopolysaccharide ( LPS ) -activated macrophages ( ~800 µmol/kg*h , estimated from Wu et al . , 1999 ) . To investigate whether a NOS was involved in NO⋅ production in beewolf eggs we conducted three experiments . First , a specific histochemical assay indicated NOS activity in embryonic tissue but not in other parts of the egg . Second , quantitative PCR revealed elevated expression of the NOS gene at the time of peak NO⋅ production . Finally , competitive inhibition of NOS by L-NAME caused a significant reduction in NO⋅ production . While each of the three results might have alternative explanations ( e . g . L-NAME might not be a perfectly specific NOS inhibitor [Peterson et al . , 1992] ) , taken together these findings provide strong evidence that a NOS , located in the embryonic tissue , is involved in NO⋅ production of beewolf eggs . Searching for possible adaptations that might accomplish this extremely high rate of NO⋅ production the beewolf NOS gene was sequenced . Only one beewolf NOS ( Pt-NOS ) gene was found . The derived amino acid sequence did not reveal considerable differences compared to the NOS of the closely related bees ( Figure 11—figure supplement 2 ) . Thus , there is no evidence for extensive evolutionary changes with regard to the gene itself . Moreover , the structure of the Pt-NOS gene is largely homologous to other insects , for example Anopheles stephensi mosquitoes ( Luckhart et al . , 1998 ) . However , in contrast to adult beewolves , the NOS-mRNA in beewolf eggs lacks exon 14 ( 144 bp , Figure 11—figure supplement 1 ) . Such alternative splicing that results in different NOS-mRNAs , including the deletion of exons ( but others than in beewolves ) , has been documented in A . stephensi in response to Plasmodium infection ( Luckhart and Li , 2001 ) . Moreover , NOS splice variants may result in organ-specific enzymes in other organisms ( Röszer , 2012 ) . Presumably , beewolf eggs produce smaller amounts of another NOS splice variant to support signaling functions in the developing embryo . In adult beewolves , the exon missing in the NOS-mRNA of eggs is located between the binding domains for calmodulin and FMN . Since calmodulin is believed to be responsible for NOS regulation ( Smith et al . , 2013 ) the deletion of an adjacent part might affect the control of NOS activity in beewolf eggs . Thus , the alternative splicing might enable the production of such large amounts of NO⋅ . Notably , compared to the cNOS ( comprising eNOS , and nNOS ) the inducible NOS isoform of vertebrates ( iNOS ) that generates higher concentrations of NO⋅ to combat microbes lacks a section of about 40 amino acids ( 120 bp ) near the FMN domain . Interestingly , this section is thought to be responsible for autoinhibition of the cNOS ( Salerno et al . , 1997 ) and its lack enhances NO⋅ production by the iNOS . The conspicuous similarity between vertebrate iNOS and the NOS in beewolf eggs with regard to the length of the missing section and its position might suggest a convergent modification to achieve a NOS with high synthetic capacity . Whereas vertebrates have evolved another gene , beewolf eggs might accomplish a similar effect by alternative splicing of the mRNA . The possible loss of regulation of the NOS and the pattern of Pt-NOS expression in the eggs suggest that in beewolf eggs the activity of the enzyme is regulated by gene expression like the NOS in Plasmodium infested A . stephensi ( Luckhart et al . , 1998 ) and the iNOS of vertebrates ( Wong et al . , 1996; Morris Jr , 1999 ) . However , in contrast to these caes , in beewolf eggs expression of the Pt-NOS seems not to be induced by immunostimulants but to occur obligatorily at a certain stage in the development of the beewolf embryo . While we cannot exclude that there is an additional , yet unknown , pathway of NO⋅ production in beewolf eggs , we hypothesize that the NOS and in particular its alternative splice variant plays a significant role in brood cell fumigation by beewolf eggs . However , even the combined effect of prey embalming and brood cell fumigation does not provide perfect protection as fungus infestation still causes larval mortality in 5% of the brood cells in the field ( Strohm and Linsenmair , 2001 ) . Some fungal spores might survive under the bees because they were screened against the gas . Another possibility , namely that strains of the ubiquitous mold fungi that are the main causes of molding in beewolf brood cells ( Engl et al . , 2016 ) , have evolved resistance against the toxic effects of NO⋅/NO2⋅ seems rather unlikely . Ultimately , there will be only weak selection for resistance at all since beewolf brood cells are certainly a rare habitat for the ubiquitous mold fungi ( Engl et al . , 2018 ) . Moreover , there will be no repeated exposure of the same fungal strains to fumigation that would be required to favor the evolution of resistance . While there are examples for detoxification of lower concentrations of NO⋅ ( mainly by scavengers like flavohemoglobins ) in different fungi , including species of Aspergillus ( Martins et al . , 2011; Zhou et al . , 2009 ) , the NO⋅/NO2⋅ levels emitted by beewolf eggs are very high and likely affect several very basic biochemical processes , thus making the evolution of an effective resistance unlikely . While brood cell fumigation clearly retards molding of larval provisions , the antimicrobial effect of NO⋅ and NO2⋅ might harm the symbiotic Streptomyces bacteria that beewolf females apply to the brood cell prior to egg laying ( Kaltenpoth et al . , 2005; Kroiss et al . , 2010 ) . Since the symbiotic bacteria are important for the survival of larvae in the cocoon and are vertically transmitted from beewolf mothers to their daughters ( Kaltenpoth et al . , 2014 ) , a considerable number of symbionts have to survive the brood cell fumigation . At the moment we can only speculate how the bacteria can survive . Conceivably , because of strong selection due to specialization and repeated exposition , the symbiotic bacteria have evolved mechanisms to cope with the high concentrations of NO⋅/NO2⋅ ( Poole , 2005; Wareham et al . , 2018 ) . Possibly , the fumigation slowly evolved after the establishment of the symbiosis; thus bacteria might have been able to gradually evolve resistance . Moreover , the bacteria are applied to the ceiling of the brood cell , which might reduce negative effects of the nitrogen oxides since these are heavier than air ( Lide , 1995 ) and will accumulate in the lower part of the brood cell . Additionally , the bacteria are embedded in copious amounts of a secretion consisting of mostly unsaturated hydrocarbons ( Kaltenpoth et al . , 2009 ) that might shield the bacteria from the fumigants . Finally , host- and/or symbiont derived antioxidants in the hydrocarbon matrix could detoxify NO⋅ and NO2⋅ and protect the symbiotic Streptomyces bacteria . How could brood cell fumigation with high concentrations of NO⋅/NO2⋅ have evolved ? Generally , it has been assumed that the primary purpose of NO⋅ was signaling at low concentrations and that the antimicrobial functions of higher concentrations are derived ( Fang , 2004 ) . Assuming a similar scenario for beewolves , small amounts of NO⋅ that were originally produced for developmental processes ( Andersen et al . , 2013 ) might have accidentally been released into the confines of the subterranean brood cell and slightly affected the germination or growth of fungi by interfering with regulatory processes ( Röszer , 2012; Wang and Higgins , 2005 ) . Given the severe threat posed by microbes , such initial benefits would have caused strong selection for elevated NO⋅ emission by the eggs . This would have considerably increased progeny survival and might have allowed ancestral beewolves to nest in an expanded range of habitat types , including nesting sites with high risk of microbial infestation , or to exploit highly susceptible but readily available prey species . Brood cell fumigation with large doses of NO⋅ thus represents a key evolutionary innovation . Since NO⋅ is used as an antimicrobial in the immune systems of many animals ( Bogdan et al . , 2000 ) , its deployment as an antifungal gas can be viewed as an innate , externalized immune defense of beewolf eggs . Such externalized components of the immune system have recently been recognized as important and possibly widespread antimicrobial measures ( Otti et al . , 2014 ) . The clear benefit of brood cell fumigation , however , is probably accompanied by substantial costs in terms of energy and biochemical resources ( Rivero , 2006 ) . NO⋅ is synthesized from L-arginine , an amino acid that is an important constituent of many proteins and biochemical pathways ( Morris Jr , 2000 ) and it is an essential amino acid for most insects ( Barbehenn et al . , 1999; Payne and Loomis , 2006 ) ( e . g . phytophagous insects [Berenbaum , 1995] , mosquitos [Uchida , 1993] , aphids [Sasaki and Ishikawa , 1995; Akman Gündüz and Douglas , 2009] , butterflies [Erhardt and Rusterholz , 1998; O'Brien et al . , 2003] , true bugs [Mesquita et al . , 2015] , parasitoid wasps [Thompson , 1976; Barrett and Schmidt , 1991] , bees [de Groot , 1952; Weiner et al . , 2010] ) . Thus , beewolves have either evolved the capacity to synthesize L-arginine or female beewolves have to provide each egg with sufficient L-arginine for both brood cell fumigation and embryogenesis . Moreover , NO⋅ synthesis by NOS requires the cofactors flavin adenine dinucleotide ( FAD ) , FMN , ( 6R- ) 5 , 6 , 7 , 8-tetrahydrobiopterin ( BH4 ) and NADPH ( Förstermann and Sessa , 2012 ) , thus competing with other metabolic pathways in the developing beewolf embryo . One of the most remarkable aspects of our study is that the embryos inside the egg survive the high concentrations of toxic nitrogen oxides during synthesis and emission as well as after its release to the brood cell . This is all the more surprising since beewolf larvae that were accidentally exposed to the gas emitted by eggs died ( Strohm , unpublished observations ) . The synthesis and emission of such high amounts of NO⋅ likely requires a number of concomitant adaptations that protect beewolf embryos against the cytotoxic effects of high concentrations of NO⋅ and NO2⋅ . One possibility is the employment of carrier molecules to transfer NO⋅ to the egg shell . In blood sucking hemipterans , for example , nitrophorins carry NO⋅ to its release site to dilate blood vessels ( Davies , 2000 ) . The mechanistic basis of NO⋅ tolerance of beewolf eggs is of particular interest , since excessive production of NO⋅ due to inflammatory processes ( Guzik et al . , 2003 ) or certain diseases ( e . g . Alzheimer's disease , [Lüth et al . , 2002; Pacher et al . , 2007; Calabrese et al . , 2007; Pautz et al . , 2010] ) might cause severe pathological complications in humans . Thus , understanding how beewolf eggs avoid the toxic effects of NO⋅ might inspire the development of novel medical applications . Our findings reveal a surprising adaptation in a mass-provisioning digger wasp to cope with the threat of pathogen infestation in the vulnerable egg and larval stages . Sanitizing the brood cell environment by producing high amounts of NO⋅ significantly enhances the survival of immatures by reducing fungal growth on their provisions . Given that mass-provisioning and development underground are widespread ecological features among digger wasps and bees and considering the difficulties of detecting volatiles in subterranean nests , such gaseous defenses might be more widespread and as yet underappreciated . In addition to revealing new perspectives on antimicrobial strategies in nature and amplifying the biological significance of NO⋅ , beewolves offer unique opportunities to elucidate general questions on the evolution and regulation of NOS as well as the production of and resistance to high concentrations of NO⋅ .
Beewolf females , Philanthus triangulum F . ( Apoidea , Crabronidae ) , were either caught in the field from populations in Franconia ( Germany ) or were the F1 progeny of such females kept in the laboratory . They were housed in observation cages ( Strohm and Linsenmair , 1994 ) that provided access to newly completed brood cells . The cages were placed in a room with temperature control ( 20–22° at night , 25–28°C in the daytime ) and were lit for 14 hr per day by neon lamps . Honeybees , Apis mellifera L . ( Apoidea , Apidae ) , the females' prey , were caught from hive entrances or from flowers and provided ad libitum . Honey was provided ad libitum in the flight cage for the nutrition of both honeybees and beewolf females . To obtain freshly laid eggs , observation cages were checked hourly . Completed brood cells were opened , their length and width was measured using calipers and the egg and/or honeybees were removed and used for the experiments . Brood cell volume was estimated as a prolate spheroid with brood cell length as the major and width as the minor axis . The bees in brood cells had been paralyzed , embalmed with lipids ( Herzner et al . , 2007 ) , and provisioned by beewolf females . Egg volume was estimated by calculating the volume of a cylinder with the respective length and width of an egg ( both determined using a stereomicroscope with eyepiece micrometer ) . The temperatures at which eggs were kept reflect natural conditions ( Herzner and Strohm , 2008 ) and allow for an optimal development ( Strohm , 2000 ) . For all experiments beewolf eggs were harvested from brood cells of various females . Eggs were randomly allocated to different treatment groups . Sample sizes refer to independent biological replicates , that is each replicate represents a different egg or brood cell – with the exception of quantitative PCR , where several eggs were pooled for one sample ( see below ) . As it is very demanding to obtain beewolf eggs , the availability of eggs of a certain developmental stage was limited . Generally , we used as many eggs as feasible ( e . g . for quantitative PCR ) . For some experiments we decided on a meaningful sample size based on experience from preliminary experiments ( e . g . we already knew that inhibition assays with beewolf eggs in Petri dishes were really clear-cut and required only few replicates ) . Moreover , due to the limited availability of beewolf eggs on a given day , replicates were conducted consecutively over several days . To test whether the time course of fungus growth on bees differed between those carrying an egg and those without egg , we used brood cells ( N = 22 ) that had been provisioned with two bees . We placed each bee individually into an artificial brood cell of natural shape and volume in sand-filled Petri dishes ( diameter 10 cm ) and with moisture levels similar to natural conditions . Petri dishes were placed in a climate chamber at 25°C in the dark . Bees were carefully checked visually every 24 hr for fungus growth without opening the Petri dishes . First signs of fungus infestation ( hyphae ) were recorded . The experiment was terminated after eleven days since all larvae had finished feeding and spun a cocoon by then . Since these are time event data , we used survival analysis ( Kaplan Meier , Breslowe test; hazard ratios and their 95% confidence intervals are presented as estimates of effect sizes , SPSS Statistics 24 ) to compare the timing of fungus infestation of the bees with and without an egg . Larvae hatched on the third day after oviposition and started to feed on the bee . There was no evidence that hatched larvae were able to prevent fungus growth on the bee they occupied or others in the brood cell . However , to take a possible effect of the larva on the experimental bee into account , we carried out the analysis not only over the whole period from oviposition until the larvae spun into a cocoon ( 11 days ) but also for the period from oviposition to the hatching of larvae ( 3 days ) . A significant difference already until the third day indicates that this effect was associated with the egg . We examined whether beewolf eggs emit a volatile antimicrobial by conducting two experiments . For the first test , we used brood cells ( N = 16 ) that contained three bees . The bees were transferred to artificial brood cells in sand-filled Petri dishes as described above . The bee with the egg and one of the bees without egg ( the experimental bee ) were placed together in the same artificial brood cell but without physical contact . The other bee without egg ( the control ) was kept alone in another artificial brood cell ( in another Petri dish ) . We monitored the timing of fungus infestation as described above . We used survival analysis as described above . Again , to take an ( unlikely ) effect of the larva into account , we also carried out the analysis for the period from oviposition until the larvae hatched ( day 3 after oviposition ) . A significant difference already until day three could only be caused by volatiles emanating from the egg . For the second assay , we exposed conidiospores of a diverse spectrum of fungi to the volatiles emanating from beewolf eggs . Petri dishes ( 10 cm ) containing culture medium ( malt extract agar or Sabouraud-agar [Atlas , 2004] ) were inoculated with conidia from different fungal strains ( Aspergillus flavus strain A , Trichocomaceae , that was isolated from infested beewolf brood cells , [Engl et al . , 2016] , N = 20; A . flavus strain B , Mucor circinelloides , Mucoraceae; Penicillium roquefortii , Trichocomaceae; Candida albicans , Saccharomycetaceae; Trichophyton rubrum , Arthrodermataceae; N = 8 for all the latter strains and species; these were kindly provided by the Department of Hygiene and Microbiology of the Würzburg University Hospital ) . Conidiospores were harvested by sampling mature fungus colonies that were reared from stock cultures . A suspension of the conidia in sterile water was evenly distributed on the Petri dishes to obtain uniform growth of fungi . To recreate the concentrations of potential antibiotic volatiles in the brood cell , we used small plastic caps ( 3 ml , about the size of a brood cell ) to confine test areas on the agar . Freshly laid eggs were placed singly on the bottom of a cap where they readily attached due to their natural stickiness . Each cap was then placed on a freshly inoculated Petri dish so that the agar under the cap was not in contact with the egg but was exposed to volatiles that emanated from the egg . An empty cap was placed on the same Petri dish as a control . The Petri dishes were incubated in a dark climate chamber at 25°C . Fungus growth under the experimental and control caps was recorded after 24 , 48 and 72 hr . After 72 hr the caps with the hatched larvae were removed , and fungal growth was further recorded after another 24 , 48 and 72 hr . Since the results were clear-cut with either no fungal growth or substantial growth ( Figure 3 ) and no intermediate cases , the experimental and control areas were compared using binomial tests ( software PAST [Hammer et al . , 2001] ) . We hypothesized that nitric oxide ( NO⋅ ) and its main reaction product with oxygen , nitrogen dioxide ( NO2⋅ ) , were the most likely compounds emanating from beewolf eggs . The standard test for the detection of NO⋅ and NO2⋅ employs the Griess reaction . We used a solution of sulfanilic acid and N- ( 1-naphthyl ) -ethylenediamine ( Spectroquant Nitrite Test , Merck , Germany , according to the manufacturer’s instructions ) . The Griess reagent specifically reacts with the nitrite anion ( NO2- ) to form a distinctive red azo dye ( Guevara et al . , 1998 ) . NO⋅ reacts with water to form nitrous acid ( HNO2 ) and can thus be directly verified by the Griess reaction . NO2⋅ , however , disproportionates in water into nitrous acid and nitric acid ( HNO3 ) and the latter must be reduced to nitrous acid to react with the Griess reagent . Freshly laid beewolf eggs ( collected within 2 hr after oviposition , N = 11 ) were placed in the lid of a 1 . 5 ml reaction tube where they readily attached due to their natural stickiness . Tubes without eggs ( N = 11 ) were used as controls . Then 1 mL of the Griess test solution was added to the tube . For another sample ( N = 15 ) the nitrate , which might be present in the solution , was reduced to nitrite by placing a glass fiber filter disc with small amounts of zinc powder ( Jander and Blasius , 1971 ) on the surface of the solution . The same setting without an egg was used as control ( N = 15 ) . The tubes were incubated at 25°C for 24 hr , and the occurrence of the red coloration was examined visually and with a photometer ( at 520 nm , Nanophotometer , Implen , Germany , quantitative measurements were not meaningful with this set-up since the azo dye is not perfectly stable over time , according to the manufacturer’s instructions ) . The samples with and without nitrate reduction showed qualitatively the same results . NO⋅ can also be detected by specific fluorescent probes . In particular , diaminorhodamin-4M AM ( DAR4M-AM ) , a cell permeable , photostable fluorescent dye , has a high sensitivity and specificity for NO⋅[Kojima et al . , 2000] ) . A DAR4M-AM ( Alexis Biochemicals , USA ) solution was prepared according to the supplier's instructions ( 10µmol/l in 0 . 1 mol/l phosphate buffer , pH 7 . 4 ) . To verify and to visualize the emission of NO⋅ from the egg , paralyzed honeybees either with freshly laid eggs ( N = 8 ) or controls without eggs ( N = 8 ) were sprayed with the DAR4M-AM solution using a nebulizer ( the egg itself was screened from droplets during spraying ) and kept in the dark ( at 25°C in artificial brood cells as described above ) . After 20 hr , the bees were examined under a fluorescence microscope ( Axiophot II , Zeiss , Germany , filter set 43: excitation 520–570 nm , emission 535–675 nm ) and digital photos were taken ( Nikon DS-2 Mv , Nikon Japan ) at constant exposure times , to allow comparison of fluorescence intensity . Due to the size of the bees , several pictures had to be taken in the X , Y plane as well as along the Z axis . Pictures along the z-axis were stacked using the software Combine-ZP ( www . hadleyweb . pwp . blueyonder . co . uk ) . Then these stacks were stitched using Photoshop Elements 5 ( PSE5 , Adobe Systems Inc USA ) . Since small peripheral background parts within the frame of the stacked and stitched picture were ‘empty’ these parts were filled with other background parts by using the clone stamp tool . Images were corrected for contrast and sharpness using PSE5 with identical settings for experimental and control specimens . DAR4M-AM can also be used to detect NO⋅ in tissues . Aliquots of 0 . 1–0 . 5 µl of the DAR4M-AM solution ( see above ) were injected into beewolf eggs ( within 1 hr after oviposition , N = 64 , in N = 45 eggs the embryo survived and developed ) with a custom made microinjector equipped with glass capillaries ( Eppendorf Femtotips II , Eppendorf , Germany ) under microscopic control . Control eggs injected with buffer only ( N = 10 ) were monitored in the same way to assess autofluorescence . For comparison , eggs of two other Hymenoptera ( Osmia bicornis , Apoidea , Megachilidae , N = 12 , and Ampulex compressa , Apoidea , Ampulicidae , N = 9; eggs from both species were obtained from our own laboratory populations ) as well as freshly hatched beewolf larvae ( N = 4 ) were injected with the DAR4M-AM solution . All eggs were kept in dark chambers at 25°C ( a temperature within the optimal range for development for all these species ) , and fluorescence was observed directly after injection and 1 , 3 , 5 , 24 , sometimes 48 and 72 hr later . For some eggs not all time points were available . Fluorescence was examined under a fluorescence microscope and documented with a digital camera as described above . Contrast and sharpness of the images were optimized using Photoshop Elements 5 ( Adobe , USA ) with identical settings for all specimens . Iodometry provides a simple but sensitive , reliable and precise method to quantify strong oxidants . To assess the amount of emitted nitrogen oxides , we placed freshly laid eggs ( N = 233 ) individually into the lid of 1 . 5 ml reaction tubes where they readily attached due to their natural stickiness . Then 1 ml of a potassium iodide-starch solution ( containing 1% KI and 1% soluble starch in distilled water ) was added , the reaction tube was closed and kept for 24 hr at 28°C in a dark climate chamber . Oxidation of iodide results in iodine that forms a blue complex with starch ( Jander and Blasius , 1971 ) . The degree of coloration was quantified by measuring the absorbance at 590 nm in a spectrophotometer ( Uvikon 860 , Kontron , Germany ) . To assess the absolute amount of the oxidant , the solutions were subsequently calibrated by titration with a reference solution of sodium thiosulfate ( concentration: 0 . 001 M; Merck , Germany ) until the blue color of the iodine-starch complex disappeared . To establish the time course of gas production , individual beewolf eggs ( N = 4 ) were transferred within 1 hr after oviposition into the lid of reaction tubes and kept in a dark climate chamber at 28°C . Every hour , the cap with the egg was transferred to another reaction tube with fresh iodide-starch solution . Immediately after removal of the egg from a reaction tube , absorbance of the solution was measured at 590 nm as described above . To investigate the temperature dependence of gas production , tubes with a newly laid egg and iodide-starch solution ( as described above , N = 33 in total ) were placed in a rack ( with white background ) inside a climate chamber and incubated at seven different constant temperatures ( 20 , 22 . 5 , 24 , 25 . 5 , 27 , 28 . 5°C and 30°C ) . The time course of coloration of the iodide-starch solution was recorded using a digital camera ( Canon EOS 20D , Canon , Japan ) programmed to take pictures at 30 min intervals . The onset of gas production could be easily determined since the color of the solution turned from clear to dark blue from one picture to the next , that is within a 30 min interval . A quadratic regression curve was fitted to the data ( SPSS Statistics 24 ) and the Q10 value for the temperature dependence was estimated . We assessed the effect of synthetic NO⋅ on the beginning of fungus growth on honeybees in artificial brood cells . Sand filled Petri dishes with artificial brood cells ( volume 3 ml ) were prepared as described above and a honeybee ( collected at the entrance of a bee hive and killed by freezing ) was placed into the brood cell . Nitric oxide was generated by the oxidation of zinc powder with nitric acid ( HNO3 ) ( Jander and Blasius , 1971 ) . In order not to affect the composition of the gases in the artificial brood cell , we adjusted the concentration of the generated NO⋅ so that an addition of 10% of the volume of the brood cell resulted in an initial concentration of 1500ppm NO⋅ . Employing iodometry as described above , we adjusted the amounts of reactants so that the addition of 300 µl to the brood cell volume of 3 ml resulted in a NO⋅ concentration of 1500ppm ( the presumable peak concentration in natural brood cells ) . Zinc powder ( 0 . 5 g ) was placed in a vial ( 20 ml ) and the vial was closed with a plastic lid with two small holes ( ~0 . 5 mm , one for pressure compensation ) . Then the vial was extensively flushed with pure nitrogen to remove oxygen that would otherwise oxidize NO⋅ to NO2⋅ . Immediately , 150 μL of 20% HNO3 were added with an insulin syringe so that the resulting gas mixture in the vial was composed of 15000ppm NO⋅ and 98 . 5% N2 . Using a gastight syringe ( Hamilton , Reno , NV , USA ) an aliquot of 300 µl of this gas mixture was injected into the artificial brood cell with a bee through a small hole ( ~0 . 5 mm ) in the lid of the Petri dish and the hole was immediately closed with adhesive tape ( N = 20 ) . Thus the concentration of NO⋅ in the artificial brood cell was 1500ppm . As controls , otherwise identically prepared Petri dishes with bees in artificial brood cells were injected with 300 µl of pure nitrogen ( N = 20 ) . The Petri dishes were incubated in a dark climate chamber at 25°C . All bees were carefully checked for fungus growth under a stereomicroscope for three days ( twice per day ) . As a consequence , first signs of fungus growth were detected earlier than in other experiments of this study , where we used the unaided eye . Data were analysed using survival analysis as described above . To assess whether synthetic NO⋅ has a similar antifungal effect as the gas emitted by beewolf eggs we compared the effect size ( hazard ratio ) of this experiment with the data testing for the effect of the egg produced gas on unembalmed bees ( as part of the experiment on the ‘combined effects’ , see below ) . If the 95% confidence intervals of the hazard ratios overlap , there is no evidence for a difference between the effects . Fungus growth and conidia formation on bees of four different experimental groups were recorded for eleven days ( until the larvae had spun their cocoons ) . The groups consisted of: ( 1 ) paralyzed honeybees that were not embalmed and did not carry an egg ( n = 25 ) , ( 2 ) paralyzed and embalmed honeybees without egg ( n = 68 ) , ( 3 ) paralyzed honeybees that were not embalmed but an egg was carefully transferred onto them from another bee ( n = 21 ) , and ( 4 ) paralyzed and embalmed honeybees with egg ( n = 21 ) . To control for effects of the transfer of eggs in group ( 3 ) , each egg in group ( 4 ) was sham treated by using tweezers to lift it up from the bee and putting them back onto the same bee . Non-embalmed bees were removed from beewolf females immediately after paralysation . Embalmed bees were removed from brood cells in observation cages within 12 hr after oviposition . All bees were transferred to artificial brood cells ( one bee per brood cell ) in Petri dishes filled with moist sand . The Petri dishes were incubated in a dark climate chamber at 25°C . All bees were checked daily for both fungus growth and formation of conidia under a stereomicroscope . As above , first signs of fungus growth were detected earlier than in the experiments of this study , where we used the unaided eye . Data were analysed using survival analysis as described above with pairwise comparisons of treatment groups ( SPSS 24 ) . To assess whether there was NOS activity in the egg tissue and where it was located , we used fixation-insensitive NADPH diaphorase staining with nitroblue tetrazolium ( Virgili et al . , 2001; Müller , 1994 ) . Eggs were fixed in PBS containing 4% paraformaldehyde for 2 hr at 4°C , followed by cryoprotection in PBS with 12% sucrose for 20 hr . The tissue was soaked in Tissue Tec ( Sakura Finetek , Netherlands ) for 30 min , frozen , and 10 µm sections were cut on a cryostat microtome ( CM3000 , Leica , Germany ) . The sections were incubated for 60 min at 30°C with 50 mmol/l Tris-HCI , pH 7 . 8 , 0 . 1% Triton X-100 , and 0 . 2 mmol/l nitroblue tetrazolium chloride in the presence or absence ( each N = 5 ) of 0 . 2 mmol/l β-NADPH to demonstrate fixation-insensitive NADPH diaphorase activity . The sections were dehydrated , mounted with Depex ( Serva , Germany ) and observed under a compound microscope ( Zeiss Axiophot II ) . Photos were taken with a digital camera ( Nikon DS-2 Mv ) . Since the egg was larger than the field of view of the camera , two pictures had to be taken and were stitched ( Photoshop Elements 5 , Adobe USA ) . Contrast and sharpness were optimized . If NOS is responsible for NO⋅ production in beewolf eggs , the time pattern of NOS gene expression should largely resemble the time course of NO⋅ production by showing a pronounced peak several hours after egg laying ( the timing of the peak depending on temperature ) . We used reverse transcription and real time quantitative PCR to quantify the NOS mRNA in beewolf eggs at different times after oviposition . Since the amount of mRNA that could be obtained from single eggs was insufficient to get reproducible results , we conducted two trials for each of four different time intervals after oviposition ( 4–5 , 9–10 , 14–15 and 19–20 hr after oviposition ) . For each trial and per each time interval we pooled 25 eggs ( all kept at 25°C ) , as well as 25 freshly hatched larvae repsectively . The eggs and larvae were removed from the brood cells at the specified times , shock frozen with liquid nitrogen and stored at −80°C . The RNA of each sample was extracted using the peqGOLD total RNA Kit ( Peqlab , Germany ) according to the supplier's instructions and eluted with 20 µL RNase free water . An aliquot of 3 µL of the RNA was digested with DNaseI ( Fermentas , Lithuania ) and transcribed into cDNA with BioScript ( Bioline , Germany ) using an Oligo-dT primer ( Fermentas , Lithuania ) in a final volume of 20 µL . As a reference for basic levels of gene expression during the experimental period , mRNA of the housekeeping gene β-actin was quantified and the ratio of NOS/β-actin mRNA was calculated for each sample . For quantitative PCR , we established new primers for both the NOS and β-actin genes of P . triangulum ( based on the complete NOS sequences , see below ) ( NOS_qPCR_F1 and R4; Actin_qPCR_F1 and R1 , Supplementary file 1 ) . All primers were intron-overlapping to avoid the measurement of contaminating genomic DNA . The NOS and actin primers amplified fragments of 312 bp and 321 bp , respectively . The specificity of both primer sets was confirmed by sequencing purified PCR products . The qPCRs were performed on an Eppendorf Realplex cycler ( Eppendorf , Germany ) in a final volume of 25 µL , containing 1 µL of template cDNA ( 1 µL of the 20 µL RT reaction mix ) , 2 . 5 µL of each primer ( 10 pmol/l ) and 12 . 5 µL of SYBR Green Mix ( SensiMixPlus SYBR Mit , Quantace , UK ) . Standard curves were established by using 10−9 – 10−3 ng of PCR products as template . A NanoDrop TM1000 spectrophotometer ( Peqlab , Germany ) was used to measure DNA concentrations of the templates for the standard curves . PCR conditions were as follows: 95°C for 5 min , followed by 50 cycles of 56°C ( β-actin ) or 65°C ( NOS ) for 60 s , 72°C for 60 s and 95°C for 60 s . Then a melting curve analysis was performed by increasing the temperature from 60°C to 95°C within 20 min . Based on the standard curves , the amount of NOS and β-actin template and their ratio was calculated . To verify the role of NOS in NO⋅ production by beewolf eggs , we used an inhibition assay ( Willmot et al . , 2005 ) . Since L-arginine is the substrate for NO⋅ production by NOS , we injected either an inhibiting L-arginine analog or , for controls , a non-inhibiting enantiomer into freshly laid beewolf eggs . Chemicals were dissolved in 0 . 1 mol/l phosphate buffer pH 7 . 4 . Using a microinjector ( see above ) eggs were injected with about 0 . 2 µl of 1 . 5 mol/l solutions of ( 1 ) the competitive inhibitor Nω-nitro-L-arginine methylester ( L-NAME , Sigma-Aldrich , USA ) ( experimental group , N = 14 ) , or ( 2 ) the non-inhibiting Nω-nitro-D-arginine methylester ( D-NAME , Sigma-Aldrich , USA ) ( control group 1 , N = 9 ) or ( 3 ) not injected at all ( N = 14 , control group 2 ) . Each egg of the three groups was placed individually in the lid of a reaction tube with an iodide-starch solution as described above and incubated for 24 hr at 28°C . Then NO⋅ production was assessed by measuring absorbance of the solution with a photometer ( Implen Nanophotometer ) at 590 nm . Statistical comparison of the groups was conducted using Mann-Whitney U-tests with correction after Holm ( Holm , 1979 ) ( SPSS Statistics 24 ) . DNA was extracted from female beewolf heads with the Epicentre MasterPure Complete DNA and RNA Purification kit ( Epicentre , USA ) according to the manufacturer's guidelines for tissue extraction . Eggs for RNA extraction were kept at a temperature of 27 . 5°C ( range 26–29°C ) , collected 14–15 hr after oviposition , immediately frozen in liquid nitrogen and stored at −70°C until RNA extraction . Twenty eggs were pooled for extraction and homogenized by repeatedly pipetting in lysis buffer of the PeqGOLD Total RNA kit ( Peqlab , Germany ) . Samples were processed according to the kit manual and frozen at −70°C . For the full transcriptome sequencing ( to obtain the 5’ terminal region ) RNA was extracted from the antennae of eight frozen female beewolves according to manufacturer's protocol 1 of the innuPrep RNA Mini Kit ( Analytik Jena , Germany ) . Most of the beewolf NOS gene was amplified and sequenced by primer walking . Sequencing reactions were performed by a commercial service ( Seqlab , Germany ) . Four degenerate primers ( NOS860fwd2 , NOS1571rev1 , NOS_seq_F1_deg , and NOS_seq_R1_deg ) were designed ( Supplementary file 1 ) based on published NOS sequences of Drosophila melanogaster ( U25117 . 1 ) , Apis mellifera ( AB204558 . 1 ) , Anopheles stephensi ( AH007775 . 1 ) , Rhodnius prolixus ( U59389 . 1 ) , Manduca sexta ( AF062749 . 1 ) and Nasonia vitripennis ( NM_001168232 . 1 ) . First , the central region ( ~700 bp , between NOS860fwd2 and NOS1571rev1 ) was amplified and sequenced . Based on this sequence , we designed a pair of P . triangulum specific primers ( NOS_qPCR_F2 and NOS_qPCR_R2 , Supplementary file 1 ) . Using one specific central and one degenerate terminal primer ( NOS_seq_F1_deg and NOS_seq_R1_deg , Supplementary file 1 ) , respectively , fragments of 4–5 kb were amplified and sequenced by primer walking , which yielded the central 9 . 5 kb of the NOS gene . Fragments larger than 2 kb were amplified with the PeqGOLD Mid-Range PCR System on a thermocycler ( TGradient , Biometra , Germany ) . Reaction volumes of 12 . 5 µL contained 1 µL DNA template , 50 mmol/l Tris-HCl ( pH 9 . 1 ) , 14 mmol/l ( NH4 ) 2SO4 , 1 . 75 mmol/l MgCl2 , 350 mmol/l of each dNTP , 400 mmol/l of each primer and 0 . 5 U 'MidRange PCR' enzyme mix . An initial 3 min melting step at 94°C was followed by 35 cycles of 0 . 5 min at 94°C , 0 . 5 min at 58°C and 3 min +20 s per cycle at 68°C and a final extension time of 20 min at 68°C . Fragments up to 2 kb were amplified using the PeqGOLD Taq . Reaction volumes of 12 . 5 µL contained 1 µL of DNA template , 50 mmol/l Tris-HCl pH 9 . 1 , 14 mmol/l ( NH4 ) 2SO4 , 3 mmol/l MgCl2 , 240 µmol of each dNTP , 800 nmol/l of each primer and 0 . 5 U Taq . An initial 3 min melting step at 95°C was followed by 35 cycles of 1 min at 95°C , 1 min at 60°C and 2 min at 72°C and a final extension time of 3 min at 72°C . The 3’ terminus was sequenced following the 3’ RACE protocol ( Sambrook and Russell , 2001a ) . Briefly , cDNA was generated by reverse transcription with a poly-T primer . Before reverse transcription , co-extracted DNA was digested using DNaseI ( New England Biolabs , UK ) . The DNA digestion mix contained 1 mmol/l Tris-HCl , 0 . 25 mmol/l MgCl2 and 1 mmol/l CaCl2 and 0 . 4 U DNaseI . DNA was digested for 10 min at 37°C , followed by DNase inactivation for 10 min at 75°C . The final reverse transcription mix contained 25 mmol/l KCL , 10 mmol/l Tris-HCl , 0 , 6 mmol/l MgCl2 , 2 mmol/l DTT , 4 µmol poly-T or gene specific primer , 0 . 5 mmol/l of each dNTP and 200 U of BioSkript Moloney Murine Leukemia Virus reverse transcriptase ( Bioline , Germany ) . The entire digestion mixture was incubated with the primer for 5 min at 70°C to enable primer annealing , then cooled on ice . Reverse transcription was carried out for 1 hr at 42°C and the enzyme was subsequently inactivated for 10 min at 70°C . The cDNA including the 3’ terminal region was amplified with the specific primer NOS_seq_3-F3 and a 'poly-T adapter primer’ , that is a polyT primer to which a specific adapter sequence was added ( Sambrook and Russell , 2001b ) ( Supplementary file 1 ) . Subsequently , a nested PCR was performed using a second specific primer ( NOS_seq_3-F6 ) and a primer that contained only the specific adapter sequence of the ‘polyT adapter primer’ to increase PCR specificity ( Supplementary file 1 , same PCR conditions as above ) . The 5’ terminal region of 200 bp was obtained from a full transcriptome sequencing approach of female antennae , which covered the full-length NOS mRNA sequence . RNA sequencing was performed by a commercial service provider ( Fasteris , Switzerland ) , using the HiSeq TM2000 Sequencing System ( Illumina , USA ) with 100 bp single reads , on 5 μg total RNA isolated from female P . triangulum antennae . CLC Genomics Workbench was used for sequence assembly of the resulting 75 million reads . Reads were quality-trimmed with standard settings and subsequently assembled using the following CLC parameters: nucleotide mismatch cost = 2; insertion cost = 2; deletion cost = 2; length fraction = 0 . 3; similarity = 0 . 9 . Conflicts among the individual bases were resolved by voting for the base with highest frequency . Contigs shorter than 250 bp were discarded . To sequence the entire NOS transcript from eggs , cDNA was generated by reverse transcription with a poly-T primer and additionally a specific , central NOS_RT_R1 primer , followed by PCR amplification using various primer combinations to cover the whole transcript sequence ( Supplementary file 1 ) . Additionally , the sequence of the 5’ terminal region was confirmed by RT-PCR of mRNA from P . triangulum eggs , using primers NOS_seq_5-F6 and NOS_seq_5-R3 ( Supplementary file 1 ) and subsequent sequencing . Even though we used a large number of primers to cover the gene , we did not find sections with signals for two different bases at the same site . Thus we infer that there is only one NOS gene in the P . triangulum genome , as in most invertebrates ( Labbé et al . , 2009 ) . In addition , the transcriptome dataset did not reveal any other transcript that was annotated as nitric oxide synthase . The GenBank accession numbers for the P . triangulum NOS ( Pt-NOS ) gene sequence is: KJ425525 , for the NOS mRNA of P . triangulum eggs: KJ425526 , and for the NOS mRNA in P . triangulum female antennae: KJ425527 . NOS coding sequences of 23 insect species from five orders were acquired from the NCBI database . Along with the P . triangulum NOS sequence , these were translated and aligned using Geneious ( Version 6 . 0 . 5 , created by Biomatters , Geneious , New Zealand ) . The highly variable 5’ end was trimmed . An approximately-maximum-likelihood tree was created with FastTree ( Price et al . , 2010; Price et al . , 2009 ) . Local support values were estimated with the Shimodaira-Hasegawa test based on 1000 samples without re-optimizing the branch lengths for the resampled alignments ( Price et al . , 2010 ) . Bayesian estimates were made with the program MrBayes 3 . 1 . 2 ( Huelsenbeck et al . , 2001; Huelsenbeck and Ronquist , 2001; Ronquist and Huelsenbeck , 2003 ) . The MCMC analysis was conducted under a mixed amino acid rate model ( prset aamodelpr = mixed ) . After 1 , 000 , 000 generations , with trees sampled every 1000 generations , the standard deviation of split frequencies was consistently lower than 0 . 01 . We discarded the first 100 of the sampled trees ( 10% burn-in ) and computed a 50% majority rule consensus tree with posterior probability values for every node . The trees estimated by both methods were nearly identical , so they were combined into a single figure . | Humans use heat , cooling , and freezing to protect their foods from mold and bacteria . Many animals , including a wasp called the European beewolf , have also developed ways to store and preserve food . Female beewolves hunt honeybees . After paralyzing a bee , the beewolf takes the body into an underground chamber and lays an egg on it . When a larva hatches from the egg , it feeds on the bee . The warm , humid conditions in the chamber provide ideal conditions in which larvae can develop , but also encourage mold and bacteria to grow . Previous research has uncovered two methods used by beewolves to fight off mold . In 2007 , researchers discovered that the female beewolf coats her bee prey with a layer of fats . This prevents water loss and keeps the outside of the bee dry so that mold spores cannot grow . In 2010 , a further study showed that the female beewolf grows helpful bacteria inside her antennae and transfers some to her young . The bacteria produce antibiotics that protect the larvae and their cocoons from mold . But these two strategies alone cannot explain the high survival rate of beewolf young . This suggests that the beewolves have at least one more strategy to prevent mold from growing . Now , Strohm et al . – including some of the researchers involved in the 2007 and 2010 studies – show that beewolf eggs emit high levels of a gas called nitric oxide , which reacts with oxygen to form nitrogen dioxide . Nitrogen dioxide is part of the air pollution generated by cars and is harmful to many species in high concentrations . Nitric oxide also plays an important role for many biochemical processes in virtually all organisms , albeit in very low concentrations . The beewolf eggs produce comparatively huge amounts of this gas to fumigate their brood chambers and protect themselves and their food from mold . Strom et al . then investigated how the eggs produce nitric oxide . The eggs appear to use the same enzyme that some other organisms use to produce nitric oxide . However , the wasp version of the enzyme contains a small modification that might explain why the eggs can produce the gas in such large amounts . Learning more about how beewolves evolved different anti-mold strategies could help researchers to develop new antimicrobial treatments for medical applications . In addition , it is not yet clear how the wasp eggs survive in high concentrations of nitric oxide and nitric dioxide . Inflammation and some human diseases produce nitric oxide , killing nearby cells . Understanding how the beewolf eggs survive could therefore help to treat these cells or protect them from damage . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2019 | Nitric oxide radicals are emitted by wasp eggs to kill mold fungi |
Disturbance from whale-watching can cause significant behavioural changes with fitness consequences for targeted whale populations . However , the sensory stimuli triggering these responses are unknown , preventing effective mitigation . Here , we test the hypothesis that vessel noise level is a driver of disturbance , using humpback whales ( Megaptera novaeangliae ) as a model species . We conducted controlled exposure experiments ( n = 42 ) on resting mother-calf pairs on a resting ground off Australia , by simulating whale-watch scenarios with a research vessel ( range 100 m , speed 1 . 5 knts ) playing back vessel noise at control/low ( 124/148 dB ) , medium ( 160 dB ) or high ( 172 dB ) low frequency-weighted source levels ( re 1 μPa RMS@1 m ) . Compared to control/low treatments , during high noise playbacks the mother’s proportion of time resting decreased by 30% , respiration rate doubled and swim speed increased by 37% . We therefore conclude that vessel noise is an adequate driver of behavioural disturbance in whales and that regulations to mitigate the impact of whale-watching should include noise emission standards .
Whale-watching comprises the largest component of marine mammal-based tourism , and the multi-billion-dollar industry is increasing globally ( Hoyt , 2018 ) . The most common form of whale-watching is boat-based , where tours often repeatedly target specific cetacean populations in easily accessible coastal waters . With an increase in the number of people whale-watching , there is consequently a rise in the number and/or the size of vessels to accommodate for the expansion of the industry . Commercial whale-watching began in 1955 and has previously been viewed as a non-invasive activity that can generate a revenue for local economies as an alternative to whaling whilst allowing depleted whale stocks to recover . It is , however , increasingly clear that boat-based whale-watching can have short-term behavioural impacts on individuals ( New et al . , 2015 ) . Short-term behavioural impacts include alterations of dive patterns , swim speeds , swim orientation , group cohesiveness , behavioural state and changes in acoustic behaviour ( for reviews see Senigaglia et al . , 2016; Machernis et al . , 2018 ) . Repeated behavioural disruptions on individuals can lead to long-term negative effects on health ( i . e . body condition ) , survival and reproduction , which in turn , if a sufficient number of individuals are affected , can negatively influence population dynamics ( Bejder et al . , 2006; Lusseau et al . , 2006 ) . However , while whale-watching has been shown to have both short- and long-term negative effects on cetaceans , the sensory drivers that elicit these behavioural responses remain unclear , preventing informed mitigation . To mitigate such negative effects and to facilitate the sustainability of the whale-watch industry , there is an increasing push by regulators for best-practice regulations , guidelines or codes of conduct . By stipulating approach distance ( typically ~100 m ) , angle ( typically from the rear and side ) , speed ( typically below wake speed ) , current guidelines rest on the premise that physical proximity is the primary vehicle of disturbance ( Higham et al . , 2014 ) . As such , a very quiet whale-watch vessel is considered to have the same impact on the target animals as a very noisy vessel at the same distance , angle of approach and speed . Given that whales in most waters of the world are offered no underwater visual cues at a range of 100 m , it therefore seems plausible that hearing rather than vision is the sensory modality that serves to mediate behavioural responses to approaching whale-watch vessels . Thus , a fundamental knowledge gap remains , as to whether behavioural reactions of cetaceans to whale-watch activities are attributable to noise level , visual cues , or a combination of both . To alleviate that pertinent data gap , we test the hypothesis that underwater noise level from whale-watch vessels is an adequate stimulus that elicits short-term behavioural responses in whales . To do this , we measured behavioural responses of 42 humpback whale ( Megaptera novaeangliae ) mother-calf pairs to different levels of vessel noise during controlled exposure experiments on a resting ground with murky waters excluding visual cues . The humpback whale was used as a model species as it is the most targeted species for whale-watching and swim-with-whale activities globally , mainly due to its cosmopolitan distribution and highly acrobatic visual displays ( Hendrix and Rose , 2014; Hoyt , 2018 ) . In keeping with the hypothesis , we predicted that high vessel noise levels would elicit greater short-term behavioural responses than medium and control levels .
Fieldwork was conducted in Exmouth Gulf , Western Australia ( Figure 1—figure supplement 1; between 21°45′–22°33′ S and 114°08′–114°40′ E ) , from August 1 to October 31 2018 . Humpback whales enter the Gulf during their southern migration , between late August and early November . The Gulf is an important resting ground for mother and calves , that may rest and nurse for a few weeks before continuing their southern migration to their high latitude feeding grounds ( Jenner et al . , 2001; Bestley et al . , 2019 ) . The Western Australian whale population has increased substantially since the cessation of whaling and is estimated to comprise 20–30 , 000 whales , increasing 9–12 . 7% yr−1 ( Bejder et al . , 2016 ) . The Gulf is shallow , with a mean water depth of 9 m and maximum of 20 m . The soundscape in the Gulf is dominated by biological sounds , such as the continuous melody of male humpback whale song and omnipresent snapping shrimp , with minimal noise from anthropogenic activities , such as vessels ( Bejder et al . , 2019 ) . Controlled exposure experiments ( CEEs ) ( Tyack et al . , 2003 ) were conducted from a small research vessel ( Quintrex , 6 m rigid hull , 4-stroke Yamaha 100hp outboard engine ) which simulated whale-watch vessel approaches ( Figure 1 ) . Vessel approaches consisted of a typical whale-watch approach; transiting past a logging mother-calf pair at 100 m distance at slow speed ( first gear , 800 rpm/1 . 5 knots; Figure 1A ) . To test the hypothesis that underwater noise level from whale-watch vessels is an adequate stimulus that elicits short-term behavioural responses in whales , the only variable that we changed during replicates of CEEs was the noise level of the same vessel signature to maximise statistical power to uncover effects of level and avoid the confounds of small differences between vessel signatures . Vessel noise was played through a laptop ( Acer Aspire ES 15 ) , amplifier ( Boss PM2500 monoblock ) , bridging transformer box ( AC1424 HP ) and emitted through an underwater acoustic transducer ( Lubell LL1424 , flat [±3 dB] frequency range 200 Hz-9 kHz ) . The transducer was suspended from the side of the vessel to 1 . 5 m below the surface to mimic typical depth of propellers/shaft/exhaust of whale-watching vessels ( Figure 1B ) . Vessel noise was emitted with a 60 s ramp up/down to avoid an acoustic startle response from the whales . Vessel noise was set to different LF-weighted broadband source levels ( SLs ) of control ( 124 dB re 1 μPa ) , low ( 148 dB re 1 μPa ) , medium ( 160 dB re 1 μPa ) and high noise ( 172 dB re 1 μPa RMS ) . Noise files for playback were generated by recording the research vessel noise . To do this , the vessel was tied to a stationary mooring whilst the vessel was in gear at 1300–1400 rpm ( equivalent of ~5 knts which is a typical speed of whale-watching vessels ) . Vessel noise was recorded for 60 min with an autonomous sound recorder ( SoundTrap , www . oceaninstruments . co . nz ) . The SoundTrap was abeam at 6 m distance to engine , at 2 m depth ( sampling rate of 48 kHz , 16 bit , rendering a flat ( ±2 dB ) frequency response from 0 . 02 to 20 kHz , clip level 174 dB re 1 μPa ( high gain ) ) . Vessel noise was not recorded whilst in transit to eliminate flow noise from the passing water . The vessel noise recording was processed in Adobe Audition ( version 3 . 0 ) using tube-modelled compression , to remove extreme values . The modified noise file was then normalised to 0 . 9 to amplify the sound without clipping . This modified vessel noise was used during CEEs to generate the desired SLs by changing the appropriate gain in the playback wav . files . From the 60 min noise recording , we extracted 15 different 15 min sound files at random times in the recording , for each low , medium and high noise files ( i . e . 45 noise files , and a control noise ) . To avoid pseudoreplication , each recording was selected at random for consecutive CEEs using the function randperm in MATLAB ( The MathWorks , Inc , Natick , MA ) . We deliberately chose to use a single vessel signature for noise playback to maximise the statistical power to uncover effects of noise level rather having such power diluted by the confounding effects of differences inherent in the spectral signatures of different whale-watch vessels . However , the broad scale applicability of our findings is supported by the fact that the spectral features of the playback noise is similar to a range of whale-watch vessels in the same frequency range ( see Figure 1—figure supplement 2 , Arranz , P . et al . unpublished data ) ( Erbe , 2002; Jensen et al . , 2009; Wladichuk et al . , 2019 ) . To achieve the high vessel noise level , maximum output noise level was tested at the Exmouth marina , where a maximum undistorted output level of 173 dB re 1μPa RMS was achieved ( laptop volume 100% and amplifier input level 50% ) . To determine the low vessel noise level , we first measured the SL of the research vessel at a transit speed of 800 rpm ( ~1 . 5 knts ) . This was the selected transit speed as it was suitable to manoeuvre the vessel with the transducer in the water and was slow enough to replicate a whale-watch scenario ( i . e . slow speed and longer duration in the presence of the whales ) . Source level of the research vessel was recorded by driving past a SoundTrap suspended on a weighted buoy at 5 m depth in 10 m of water at 17–18 m distance ( replicated three times ) . The research vessel noise SL was calculated at 140 ± 2 dB re 1 μPa low frequency ( LF ) -weighted ( Southall et al . , 2019 ) . We therefore had a range from 140 to 173 dB re 1 μPa to select for low , medium and high levels for CEEs , and a 12 dB difference was chosen ( 148 , 160 and 172 dB re 1 μPa ) . These selected noise levels represent a range of slow moving ( <10 knts ) motorised whale-watch vessels that a whale would experience in the wild , having SLs ranging from 138 to 169 dB re 1 μPa @ 1 m ( Jensen et al . , 2009; Wladichuk et al . , 2019 ) . The control was set to 124 dB re 1 μPa to ensure it was ~16 dB quieter than the vessel noise at 800 rpm . Prior to conducting CEEs , the accuracy of the transducer was tested at Exmouth marina , by playing the high vessel noise and recording the SL at 2 dB increments from 140 dB re 1 μPa until 173 dB re 1 μPa at 3 m depth ( SoundTrap at 2 m depth , sampling rate of 48 kHz , clip level 184 dB re 1 μPa ( low gain ) ) . The SLs were linear with expected playback levels from the gain-adjusted audio files and no clipping was observed . Received levels ( RLs ) of playback noise were then measured in calm waters ( Beaufort 0–1 ) in 14 m water depth using a SoundTrap attached to a vertical weighted buoy ( sampling rate 48 kHz , 16 bit , clip level: 174 dB re 1 µPa ( high gain ) ) . The SoundTrap was located at 1 . 5 m depth from the surface as this is approximately where the lower jaw/inner ear of an adult humpback whale would be when resting at the surface . We conducted noise recordings whilst the vessel was stationary ( anchored ) at ~100 m distance from the SoundTrap . Ten vessel noise files were played for each low , medium and high noise treatment to confirm the LF-weighted SLs of 148 , 160 and 172 dB re 1µPa ( rms ) ( Southall et al . , 2019; Tougaard and Beedholm , 2019 ) . These LF-weighted SLs were then used to predict LF-weighted RLs by subtracting the transmission loss for each closest point of approach from the SL of the given treatment assuming spherical spreading . For verification , the RLs of vessel noise as both LF-weighted ( Southall et al . , 2019 ) and third-octave level ( TOLs ) from the moving vessel were recorded whilst the vessel was transiting past the SoundTrap at ~100 m distance as per CEEs ( Figure 1A ) . We played four files for each noise treatment . Segments of 5 s were extracted during the closest point of approach and the different noise levels were quantified as TOLs with a 0 . 125 s averaging window ( RMSfast ) and graphed as 50th percentiles ( median ) . Figure 1—figure supplement 3 shows the spectral components of the playback sounds compared to the original moored vessel recording ( 6 m , 5 knts ) . It is shown that the playback noise is slightly more tonal than the recorded noise , but well within the spectral realm of whale-watching vessel spectral signatures ( Figure 1—figure supplement 2 ) . There is less energy at frequencies below 200 Hz in the playback noise compared to the recorded noise; however , because the noise was played to whales logging at the surface , energy at frequencies below 200–400 Hz cancel out via interference with the surface and therefore does not reach the whale ( Figure 1—figure supplement 3 ) . The echosounder was switched off during all recordings and experiments . To compare the RLs of playback noise to the surrounding underwater noise in the Gulf , ambient noise was recorded . Ambient noise was recorded continuously by a SoundTrap ( sample rate of 48 kHz , clip level 174 dB re 1 µPa attached to a weighted buoy in 15 m water depth ( the SoundTrap was ~4 m below the surface ) at 22°00’19’ S 114°08’25’ E ( Figure 1—figure supplement 1 ) from 31 August to 23 September . Ambient noise was also recorded for 24 hr periods ( 18 September and 18 October ) . For analyses , ambient noise was quantified as TOLs in dB re 1 μPa RMS ( 0 . 125 s averaging windows ) . Five different relevant TOLs ( nominal center frequencies of 250 Hz , 400 Hz , 1250 Hz , 2500 Hz , 4000 Hz ) were extracted following Bejder et al . , 2019 . Self-noise of SoundTraps were recorded in a silent anechoic chamber at Aarhus University , Denmark . Controlled exposure experiments were conducted on humpback whale mother-calf pairs , where the mother was predominantly logging on the surface . Mother-calf pairs were selected as i ) many whale-watch and swim-with-whale operators target them during tours due to their slow , calm behaviour ( Sprogis et al . , 2020 ) , ii ) they are likely the most sensitive to anthropogenic disturbance ( Lundquist et al . , 2013; Argüelles et al . , 2016 ) , and iii ) they offer a standardised behaviour that facilitates detection of noise induced disturbance . Controlled exposure experiments consisted of a before , during and after experimental design ( before phase = absence of vessel , stationary >300–400 m from whales with engine in neutral; during phase = vessel approach ( Figure 1 ) ; after phase = departure of vessel , stationary >300 m from whales with engine in neutral ) . In the during phase , the driver of the vessel aimed for a tangential to parallel approach , and pass at the same distance ( ~100 m ) distance and speed ( ~1 . 5 knts ) . To ensure the distance to the whale , a laser range finder ( Bushnell 10 × 42 Fusion 1 mile laser ) was used . We aimed to conduct ten replicates of each treatment to ensure sufficient power for analyses . Replicates of CEEs were conducted >3 km from any previous CEE on the same day . To ensure samples were independent ( same mother-calf pair never sampled twice ) , photo-identification of the dorsal fin using a DSLR ( Canon 50D 400 mm lens ) and aerial photographs of the dorsal side were taken using an unmanned aerial vehicle ( UAV ) . Focal follows were conducted during good weather conditions ( <15 knot winds , Beaufort sea state <3 ) and ceased if the weather deteriorated . Controlled exposure experiments were included in analyses if before , during and after data was recorded on the same mother-calf pair , if the research vessel approached ~100 m , if no other vessels passed <500 m of the focal pair , and if no conspecifics or other species approached and/or interacted <100 m of the focal pair . Throughout CEEs , focal follows of humpback mother-calf pairs were conducted using UAVs to video-record continuously for all occurrence sampling ( Altmann , 1974 ) . A quadcopter UAV ( DJI Phantom 4 Advanced , diameter = 350 mm , weight = 1368 g , video = 2 . 7K , 2720 × 1530 , 48fps ) was flown , which had a maximum flight time of ~25 min . The UAV provided a live video feed to the remote controller connected to an iPad ( Figure 1C ) . The UAV was launched and retrieved by hand from the front of the vessel . Two UAVs were flown consecutively to ensure a near-continuous video recording . UAVs were flown at 25–30 m altitude and positioned above the mother , facing north and the camera vertically down ( Nielsen et al . , 2019 ) . At these altitudes , the presence of the UAV has no negative noise effects due to the low RLs underwater ( Christiansen et al . , 2016b ) , and cause no apparent behavioural changes on baleen whales ( Christiansen et al . , 2020 ) , thus this technology is non-invasive and appropriate to record undisturbed ( control ) behavioural responses . The UAV logged UTC time , GPS position ( WGS84 ellipsoid ) and altitude ( barometric and GPS ) every 100 ms . Behavioural responses of interest were common short-term responses that are altered during whale-watch activities ( Senigaglia et al . , 2016 ) , namely i ) behavioural events , ii ) respiration rate , iii ) heading and iv ) swim speed . Data were filtered , and CEEs that ceased early and/or were not used in analyses were due to i ) identifying repeat whales ( n = 2 ) , ii ) if a boat approached close by ( n = 5 ) , iii ) if the sun was setting and did not allow sufficient time to complete the CEE ( n = 1 ) , iv ) if conspecifics arrived <100 m to the focal pair ( n = 12 ) , v ) if the vessel ended up approaching the whale too close ( e . g . 70 m , n = 1 ) , vi ) if the whales possibly reacted to a loud gear-shift in the during phase ( n = 2 ) , vii ) if the whales were predominantly slow travelling in the before phase ( n = 15 ) , viii ) due to technical issues ( n = 4 ) , ix ) if the weather deteriorated ( n = 1 ) , and x ) if other species interacted with the whales appearing to cause behavioural changes ( e . g . a school of fish , n = 1; silver gulls pecking the skin of the whale , n = 6 ) . The remaining CEEs that had before , during and after data on the same mother-calf pair were used in analyses . Mixed effect models were constructed to investigate the effects of underwater vessel noise on mother-calf pairs , and were developed in R v3 . 5 . 2 ( R Development Core Team , 2014 ) . Prior to modelling , data exploration was conducted following Zuur et al . , 2010 . We examined i ) within treatments to determine if there was an effect of treatment , and ii ) among treatments to determine the severity of the treatment . Treatment and phase were the fixed effects of interest , thus an interaction term between treatment and phase was added as treatment was dependent on phase . A CEE was composed of three phases ( before , during , after ) on the same focal whale ( i . e . repeated measures ) , thus to account for any effect of individual , mother-calf identity was added as a random effect . Five response variables were tested for within and among treatment effects: 1 ) the proportion of time resting , 2 ) presence of instantaneous events , 3 ) respiration rate , 4 ) heading change , and 5 ) swim speed ( Supplementary file 2 for model summary ) . We developed linear mixed effects models ( LMMs: models 3 , 4 and 5 ) in the nlme package ( Pinheiro et al . , 2019 ) , and when data did not conform to normality we used generalised linear mixed models ( GLMM: models 1 and 2 ) in the MASS package ( Venables and Ripley , 2002 ) using penalized quasi-likelihood ( GLMM-PQL ) to account for overdispersion ( Bolker et al . , 2009 ) following Zuur et al . , 2009 . To validate models , normalised residuals versus fitted values were calculated to identify potential violations of model assumptions . We explored scatterplots for homogeneity , histograms for normality , Cook’s distance for influential points and auto-correlation functions ( ACF plots ) for temporal dependence . The goodness-of-fit for each model was assessed using a coefficient of determination ( R2 , ranging from 0 to 1 ) . The marginal ( R2 ( m ) ) and conditional ( R2 ( c ) ) values were calculated in the MuMIn package ( Nakagawa and Schielzeth , 2013 ) . R2 ( m ) explains the variance explained by the fixed effects , while R2 ( c ) explains the variance in the full model ( including the random effects ) .
Controlled exposure experiments were conducted across 53 days ( 290 hr on the water ) from 25 August to 28 October , 2018 ( Figure 2 ) . Vessel track lines searching for resting mother-calf pairs covered 2337 km ( Figure 2—figure supplement 1 ) . Over 60 CEEs were conducted , including 273 UAV flights ( 78 hr of flight time ) . The length of mothers ranged from 9 . 7 m to 16 . 7 m ( mean = 13 . 2 m; n = 105 ) , and for calves from 4 . 5 m to 8 . 4 m ( mean = 6 . 3 m; n = 104; Figure 2—figure supplement 1 ) . After data filtering , 13 control/low ( including 4 low ) , 14 medium and 15 high noise CEEs were used in analyses ( 42 mother-calf pairs; Figure 2; Video 1 for examples of reactions ) . Control and low noise treatments were pooled as the RLs of vessel noise were around ambient noise ( Figure 3 ) . For these filtered CEEs , there was 29 . 4 hr of data conducted in daylight hours ( 7:20 to 18:20 ) . The mean duration for before flights was 13:09 mins ( 0 . 003 SD ) , during was 15:04 min ( 0 . 002 SD ) and after was 14:48 min ( 0 . 004 SD ) . The average closest point of approach was 135 m ( 56 . 0 SD ) . Filtered CEEs were conducted in water temperatures between 19°C and 24°C ( mean = 22°C ) and in water depths between 13 m and 21 m ( mean = 17 m ) . From the transiting calibration , all treatments had a peak level in the third octave band around 4 kHz , with mean LF-weighted RLs @100 m of: control at 104 dB , low noise at 112 dB ( ±1 ) ( average distance: 103 m , range: 102–103 m ) , medium noise at 122 ± 3 dB ( average distance: 102 m , range: 94–108 m ) , and high noise at 133 ± 2 dB ( average distance: 101 m , range: 85–113 m ) re 1 μPa RMS ( 0 . 125 s ) ( Figure 3; Figure 3—figure supplement 1 for shallower depth ) . When we correct for spherical spreading , the back calculated SLs are: medium noise at 162 ± 2 dB , and high noise at 173 ± 2 dB re 1 μPa RMS ( 0 . 125 s ) , in keeping with the predictions from the stationary calibrations ( Supplementary file 3 ) . Control/low noise was equal to the ambient noise statistic , so it was not meaningful to compute SLs from these RLs . Ambient noise was mostly dominated by humpback whale song and snapping shrimp ( Figure 3—figure supplement 2 ) . Median LF-weighted levels increased by a moderate 5 dB from 103 to 108 dB re 1uPa RMS ( 0 . 125 s ) . However , the TOLs between 400 Hz and 2500 Hz rose by 10–15 dB as singing humpback whale males arrived through the breeding season ( Figure 3 ) , whereas high frequency TOLs changed little due to the ever-present snapping shrimp noise .
Humpback whale mothers rest on average for 35% of their time on the breeding ground as the early phase of lactation is the most energetically demanding phase in their reproductive cycle , with females sometimes loosing >25% of their body condition in only 3–4 months ( Christiansen et al . , 2016a; Bejder et al . , 2019 ) . Thus , an increase in respiration rate and movements due to anthropogenic disturbance ( e . g . whale-watching with a loud vessel ) will increase maternal energy expenditure , especially if loud noise exposures are cumulative ( e . g . repeated throughout the day , from many sources , or prolonged exposure ) . Such noise-induced disturbances inevitably leads to a negative offset in the energy available for nursing , fending off males/predators and migrating back to their polar feeding ground ( Braithwaite et al . , 2015 ) . For calves , there was no significant effect of playbacks on instantaneous events or respiration rate; however , the proportion of time resting decreased with increased vessel noise level . As the mothers were significantly disturbed during high noise playbacks , and as the calf ( <3–4 months old ) is dependent on its mother , the energetic consequences are also likely to increase for the calf . A calf is required to nurse substantially ( ~20% of their time; Videsen et al . , 2017 ) to grow in strength and size ( by ~3 cm in length a day; Christiansen et al . , 2016a ) within a short period of time to reduce predation ( killer whales and sharks; Pitman et al . , 2015 ) and endure the long migration to high latitude feeding grounds ( Bestley et al . , 2019 ) . Thus , disturbing mother-calf pairs on a resting ground through loud vessel noise should be minimised and mitigated appropriately , especially as the whale-watch industry is expanding globally . We show that with a quiet vessel , mothers continued to rest , thus in a whale-watch scenario mothers and calves are more likely to remain at the surface which is beneficial for tourist viewing compared to if the mother was to dive and swim away as documented during medium and high vessel noise playbacks . The demonstration that noise exposure level from a vessel is a driver of short-term behavioural disturbance in whales is perhaps not surprising , as for cetaceans hearing is the primary sensory modality which is more efficient than sight at longer ranges ( Richardson et al . , 1995 ) . Cetacean sight underwater is highly limited and hence even under the best conditions they are unlikely to be able to see a vessel underwater at 100 m distance . Thus , to facilitate sustainable whale-watching operations , guidelines and legislation should be based on approach distance , angle , speed and as highlighted here noise level . For humpback whales , we provide an evidence-based noise threshold for recommended application into global whale-watch guidelines: for vessels with spectral signatures similar to the playback used here , we suggest that emission standards be implemented so that vessels operating around whales as close as 100 m do not have LF-weighted SLs of more than 150 dB re 1 μPa RMS @ 1 m . Such low vessel SLs will lead to RLs for a logging whale at 100 m distance that are close to the ambient noise ( depending on habitat , sea state and timing in the whale season ) , offering noise levels that are perhaps audible to the whales but with a low perceived loudness . Clearly , it remains an open question of how this noise limit may apply to other cetacean species with different predator avoidance strategies , behavioural states , exposure history , habitats and hearing capabilities . Nevertheless , as noise from non-cavitating vessels has most of the energy at low frequencies ( Jensen et al . , 2009 ) , it seems parsimonious to hypothesise that smaller toothed whales would similarly not respond to whale-watching vessels with broadband SLs <150 dB re 1µPa RMS at 100 m given their poor low frequency hearing ( Au et al . , 1997 ) . To further examine the impacts of vessel noise , we suggest that future research should address the noise effects from the presence of multiple vessels , different vessel engine types , vessel proximity and different vessel approach types ( e . g . in-path vs . parallel approaches during whale-watching/swim-with-cetacean tourism ) . There is currently limited information on the range of SLs produced by commercial whale-watch vessels around whales . Available data suggests a SL range from around 138 to 169 dB re 1 μPa @ 1 m for slow moving ( <10 knts ) whale-watch vessels ( Au and Green , 2000; Jensen et al . , 2009; Wladichuk et al . , 2019 ) , largely covering the low to high range of playback SLs used here . Similarly , the chosen playback noise has spectral features representative of signatures of whale-watch vessels ( Figure 1—figure supplement 2 ) and other motorised vessels ( Erbe , 2002; Jensen et al . , 2009; Wladichuk et al . , 2019 ) . Thus , if these spectral signatures and SLs are representative , some vessels are already at the low output levels recommended here when moving slowly , while others produce higher SLs that are expected to evoke clear responses at 100 m from humpback whales similar to those documented here . For loud vessels , perhaps most importantly , operators can reduce the SL of their vessel by driving slowly ( Erbe , 2002; Jensen et al . , 2009; Wladichuk et al . , 2019 ) , avoid gear-shifts which generate high-level transient sounds ( Jensen et al . , 2009 ) , and/or increase the distance to the focal whales , although just a 6 dB increase in SL would require a doubling in range to 200 m , which tour operators are unlikely to do . To permanently reduce SLs , operators can apply a range of techniques , including using larger , slower moving propellers to minimise cavitation , quieter engines/electric engines and installing noise absorption gear ( International Maritime Organisation , 2014 ) . This study demonstrates that noise level from a vessel drives the short-term behavioural response of humpback whales to disturbance . Thus , in the commercial whale-watch industry , if two operators adhere to distance guidelines ( e . g . 100 m distance to the whale ) with two different vessels with a similar spectral signature as our playback ( one quiet and one loud ) , the vessels would evoke vastly different responses from the whales , resulting in differences in energetic and fitness consequences . To reduce or avoid disturbance , we propose that noise emission standards be incorporated into whale-watch and swim-with-cetacean guidelines along with the already stipulated speed , distance and angle of approach regulations . We recommend that whale-watch vessels employ broadband SL <150 dB re 1 μPa RMS when operating around whales at guideline distances of 100 m . Given that some vessels already comply with this , these recommendations are feasible to implement into existing whale-watch guidelines . These recommendations will allow operators to approach cetaceans in a responsible , sustainable manner in a way that also offers eco-tourists a view of undisturbed wildlife . | Whale-watching is a multi-billion-dollar industry that is growing around the world . Typically , tour operators use boats to transport tourists into coastal waters to see groups of whales , dolphins or porpoises . There is , however , accumulating evidence that boat-based whale-watching negatively affects the way these animals behave and so many countries have put guidelines in place to mitigate activities that may disturb the animals . These guidelines generally stipulate the boat’s angle of approach , how close the boat can get and the speed at which it can pass by the animals . In general , these guidelines are based on the assumption that the animals are disturbed by the closeness of the whale-watching boats . However , whales , dolphins and porpoises have very sensitive hearing , and only have a short range of vision underwater . Therefore , it seems plausible that the animals hear whale-watching boats long before they see them and so the loudness of underwater noise from the boats may be enough to disturb these animals' behaviour . To test this hypothesis , Sprogis et al . performed experiments where they simulated a whale-watching vessel approaching humpback whale mothers and calves who were resting off the northwest coast of Australia . A small motorised research boat travelling at a low speed passed different mother-calf pairs at a target distance of 100 meters , which is a common whale-watching distance guideline in many countries . The boat had an underwater speaker that played recordings of the boat noise at different volumes , while a drone with a video camera flew overhead to record the whales’ behaviours in detail and to identify individual animals . These “controlled exposure experiments” showed that the quiet boat noise did not appear to disturb the mothers and calves . However , compared to when the quiet boat passed the animals the louder boat noise decreased how long the mother whale rested on the surface by 30% , made her swim 37% faster , and doubled the number of breaths she took per minute . If there are many disturbances from humans , then it can negatively impact the energy the mother and calf have available for nursing , fending off males and predators , and migrating back to their feeding ground nearer the Earth’s poles . Based on these findings , it is shown that the loudness of the underwater noise from boats can explain why whales may be disturbed during whale-watching activities . To help reduce this disturbance , Sprogis et al . recommend that noise emission standards should be added to the current whale-watching regulations such that boats should be as quiet as possible and ideally around the volume of the ambient background noise . This would allow operators to approach the animals in a responsible , sustainable manner and offer tourists a view of undisturbed wildlife . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"ecology"
] | 2020 | Vessel noise levels drive behavioural responses of humpback whales with implications for whale-watching |
The circumventricular organs ( CVOs ) in the central nervous system ( CNS ) lack a vascular blood-brain barrier ( BBB ) , creating communication sites for sensory or secretory neurons , involved in body homeostasis . Wnt/β-catenin signaling is essential for BBB development and maintenance in endothelial cells ( ECs ) in most CNS vessels . Here we show that in mouse development , as well as in adult mouse and zebrafish , CVO ECs rendered Wnt-reporter negative , suggesting low level pathway activity . Characterization of the subfornical organ ( SFO ) vasculature revealed heterogenous claudin-5 ( Cldn5 ) and Plvap/Meca32 expression indicative for tight and leaky vessels , respectively . Dominant , EC-specific β-catenin transcription in mice , converted phenotypically leaky into BBB-like vessels , by augmenting Cldn5+vessels , stabilizing junctions and by reducing Plvap/Meca32+ and fenestrated vessels , resulting in decreased tracer permeability . Endothelial tightening augmented neuronal activity in the SFO of water restricted mice . Hence , regulating the SFO vessel barrier may influence neuronal function in the context of water homeostasis .
In vertebrates , the endothelial blood-brain barrier ( BBB ) is crucial for providing a permissive microenvironment for neuronal function . During developmental brain vascularization , blood vessels undergo Wnt/β-catenin signaling , driven by Wnt7a/7b that is required for angiogenesis as well as for BBB formation ( Daneman et al . , 2009; Stenman et al . , 2008; Liebner et al . , 2008 ) . In the adult , the Wnt pathway remains instrumental to maintain BBB function in endothelial cells ( ECs ) of the central nervous system ( CNS ) ( Zhou et al . , 2014 ) . Herein activation of β-catenin/TCF signaling can be induced by two flavors of the canonical Wnt pathway mediated by the ligands Wnt7a/7b and the non-Wnt-related norrin disease protein ( Ndp ) , binding to the receptor complexes frizzled-4/Lrp5/6/Gpr124/Reck and frizzled-4/Lrp5/6/Tspan12 , respectively ( Junge et al . , 2009; Chang et al . , 2017; Cullen et al . , 2011; Posokhova et al . , 2015; Kuhnert et al . , 2010; Wang et al . , 2014; Cho et al . , 2017; Vanhollebeke et al . , 2015; Eubelen et al . , 2018 ) . Although a strict control of the exchange between blood and the CNS tissue by the endothelial BBB is realized in most parts of the CNS , some areas of the brain and the ciliary body of the eye are exceptions to this rule , providing a physiologically highly relevant door to the CNS . The circumventricular organs ( CVOs ) are a number of small midline structures found in all vertebrate brains , located around the third and fourth ventricle . CVOs have a rich capillary plexus , which physiologically lacks BBB properties ( Ganong , 2000; Ufnal and Skrzypecki , 2014; Langlet et al . , 2013; Benarroch , 2011 ) . These characteristics are regarded as important sites of communication between brain and blood . Based on their function , CVOs are commonly classified into secretory and sensory organs . The median eminence , the neurohypophysis ( posterior pituitary , PP ) , the pineal gland ( PI ) and the subcommissural organ ( SCO ) belong to the secretory group . The vascular organ of the lamina terminalis ( organum vasculosum of the lamina terminalis , OVLT ) , the subfornical organ ( SFO ) and the area postrema ( AP ) are considered as sensory organs ( Ufnal and Skrzypecki , 2014 ) . The leaky vessels of the CVOs evidently have different morphological and structural characteristics from those of typical BBB vessels , lacking a cellular organization as neuro-vascular unit ( NVU ) , without distinctive astrocytic endfeet and ECs with numerous fenestrations and vesicles ( Morita et al . , 2016 ) . Interestingly , in previous reports neither the tight junction proteins claudin-5 ( Cldn5 ) , occludin ( Ocln ) and zonula occludens 1 ( Tjp1/ZO-1 ) ( Mullier et al . , 2010; Maolood and Meister , 2010; Norsted et al . , 2008; Sisó et al . , 2010 ) , nor the transporter proteins glucose transporter 1 ( Glut-1 ) and transferrin receptor ( Norsted et al . , 2008; Maolood and Meister , 2009 ) , showed a BBB-like staining in leaky CVO ECs . In line with these vascular features , Morita et al . showed that 10 kDa dextran accumulates in the perivascular space between the inner and outer basement membranes , whereas smaller tracers up to 3 kDa dextran diffuse into the parenchyma ( Morita et al . , 2016 ) . Not only the endothelium , but also the perivascular space in CVOs has specific properties , being enlarged and filled by collagen fibers , fibroblasts , astrocytic processes and axons ( Morita and Miyata , 2012 ) . Beside the leaky vasculature of the CVOs , free diffusion of substances into the brain parenchyma is prohibited by tanycytes , specialized cells of the ependymal lining . Tanycytes contain long processes , unlike typical ependymal cells , which project to the parenchyma of the CVOs making contact to the fenestrated vascular wall of the CVOs ( Langlet et al . , 2013; Benarroch , 2011; Mullier et al . , 2010 ) . The detailed function of all CVOs has not intensely been explored in the past , but more recently , the SFO together with the OVLT , the median preoptic nucleus ( MnPO ) and the PP were identified as a functional circuit , regulating drinking behavior and water homeostasis of the organism ( Zimmerman et al . , 2016; Oka et al . , 2015; Augustine et al . , 2018 ) . The SFO is a tiny organ located underneath the fornix at the foramen of Monro , protruding into the third ventricle at the meeting point with the lateral ventricles . The dense vascular network in the SFO is similarly organized in different vertebrate species . It presents with heterogeneous vessel phenotypes and can therefore be divided into two zones . Whereas the outer shell contains more BBB expressing vessels , the majority within the ventromedial core is fenestrated with a wide perivascular space . In general , the vascular density is four to five times higher than in other brain regions with tortuous vessels , exhibiting a high blood volume and slow perfusion rate , thereby contributing to high permeability rates ( Sisó et al . , 2010; Duvernoy and Risold , 2007; Fry et al . , 2007; Bouchaud et al . , 1989 ) . Although a cell type-specific expression analysis by single cell sequencing , as it has been performed for the brain parenchyma ( Vanlandewijck et al . , 2018 ) , has not been conducted yet , neurons and astrocytes of sensory CVOs ( SFO , OVLT and AP ) were shown to express specific receptors and ion channels . Those permit them to detect several blood–derived molecules such as salts , hormones , lipids and toxins and convey this information to other parts of the brain , involved in controlling autonomic and peripheral functions ( Benarroch , 2011; Sisó et al . , 2010 ) . Evidence supports that 25–60% of CVO neurons respond to signals in the circulatory system and a single neuron may respond to multiple signals such as osmolarity and angiotensin II . Most sensory CVOs play a role in the control of blood pressure , fluid and sodium balance , cardiovascular regulation , feeding and energy homeostasis and immunomodulation ( Ufnal and Skrzypecki , 2014; Benarroch , 2011; Sisó et al . , 2010; Morita and Miyata , 2012; Smith and Ferguson , 2012 ) . Given the fact that endothelial Wnt/β-catenin signaling is necessary for BBB development and maintenance , the question remains if the pathway is instrumental in the establishment of vascular heterogeneity in the CNS . Hence , we addressed the question if Wnt/β-catenin signaling is operational in CVO vessels during development and if local regulation of β-catenin signaling is involved in establishing a leaky vascular phenotype in CVOs . Finally , we asked if dominant activation of β-catenin in ECs can overwrite the leaky vessel fate thereby affecting CVO function . Here we show by investigating CVOs during BAT-gal reporter mouse development that at any embryonic stage analyzed , starting from E13 . 5 , when the first SFO primordium could be identified , to P21 , no activation of β-catenin signaling in CVO vessels could be detected . Focusing on the SFO as a crucial CVO in the regulation of water homeostasis , we show that blood vessels in the caudal portion were mainly leaky evidenced by Plvap/Meca32 staining , whereas capillaries in the rostral portion of the organ were tighter . Interestingly , within the vessel continuity , individual ECs might be Plvap/Meca32+//Cldn5- followed by a Plvap/Meca32-//Cldn5+ EC , suggesting a locally confined regulation of barrier properties . Dominant , genetic activation of β-catenin signaling ( gain-of-function , GOF ) in ECs resulted in tightening of CVO blood vessels , evidenced by the switch from a Plvap/Meca32+//Cldn5- to a Plvap/Meca32-//Cldn5+ vascular phenotype . Vessel tightening was accompanied by a significant reduction in endothelial fenestrations , that likely contributed to the reduction of transcellular permeability evidenced by decreased tracer leakage . Interestingly , endothelial tightening did not coincide with the formation of astrocytic endfeet towards a BBB-like NVU . Finally , we observed augmented neuronal activity in the SFO under thirst conditions after sealing CVO vessels , supported by significantly increased neuronal c-fos staining in the SFO of GOF mice .
As previously shown , BAT-gal mice report active Wnt/β-catenin in brain parenchymal vessels during embryonic and early postnatal brain vascularization ( Figure 1A ) ( Liebner et al . , 2008 ) . So far , the developmental formation of CVOs and of the SFO in particular has not been investigated in mice . We made use of BAT-gal mice to monitor Wnt/β-catenin activity in CVO vessels at different time points of embryonic and postnatal mouse development , starting from E13 . 5 which was the first timepoint we could identify the primordial SFO , to P21 ( Figure 1B–E , Figure 1—figure supplements 1 and 2 ) . In the SFO ( Figure 1B–E ) we could not detect an overlap of reporter gene-expression and CD31+ cells at any developmental stage analyzed . Instead , adjacent , non-endothelial cells in the ependymal lining as well as cells in the stroma of the organ showed active Wnt/β-catenin signaling , evidenced by nuclear β-galactosidase staining ( Figure 1B–E ) . The latter observation suggested that Wnt growth factors are indeed available in the CVO region , but the canonical pathway was not activated in ECs . Similarly , we did also not observe reporter gene-expressing ECs in the OVLT and in the PI ( Figure 1—figure supplements 1 and 2 ) , providing evidence for the interpretation that CVO endothelia generally show low or no Wnt/β-catenin activity . We further wanted to address the question , whether the lack of Wnt pathway activation in CVO vessels is evolutionary conserved and analyzed the OVLT of adult Wnt pathway reporter zebrafish ( Jeong et al . , 2008 ) . In all fish analyzed , OVLT vessels were largely devoid of GFP reporter gene expression , suggesting that Wnt/β-catenin activation is strongly reduced or absent in this CVO of the fish ( Figure 2 ) . In order to further characterize the vascular organization of the SFO , we analyzed adult wild type ( WT ) mice by confocal and light sheet microscopy . As it has previously been proposed by Pócsai et al . that the SFO can be divided into a shell and a core region with different properties of astrocytes and extracellular matix ( ECM ) , we intended to analyze the distribution of leaky and tight vessels within the SFO by staining for Plvap/Meca32 and Cldn5 , respectively ( Pócsai and Kálmán , 2015 ) . In order to visualize the organs , relevant for water homeostasis , we initially applied fluorescent microscopy on sagittal sections , showing that indeed , vessels in the SFO , OVLT and PP were Plvap/Meca32+ , but also showed a considerable degree of intermingling with Cldn5+ ECs ( Figure 3—figure supplement 1 ) . In order to have a global view on vessel heterogeneity within the SFO , we prepared brains of WT C57Bl6 mice for whole mount staining ( Figure 3 ) . Light sheet microscopy analysis of whole mount preparations revealed that the majority of SFO vessels in the rostral portion , as well as of the shell were Plvap/Meca32-//Cldn5+ , suggesting that these vessels possess BBB properties . Instead , vessels of the caudal SFO region were mainly Plvap/Meca32+//Cldn5- , providing evidence for their leaky phenotype ( Figure 3D; Video 1 ) . Higher magnification of the rostral part and the outer shell of the organ showed that some vessels exhibit a mosaic-like staining for Plvap/Meca32 and Cldn5 along their longitudinal extension ( Figure 3D; Video 1 ) . In order to visualize the alternating expression of leaky and tight vessel markers in more detail , we applied confocal microscopy on sagittal sections , revealing that neighboring cells may be positive either for Plvap/Meca32 or for Cldn5 ( Figure 3C ) . However , some cells also showed a mixed identity , allowing the interpretation that there is a continuous transition from leaky-to-tight-to-leaky vessels in the SFO . As we observed an alternating barrier phenotype in the SFO vasculature , we addressed the questions if dominant activation of β-catenin signaling in ECs may lead to vessel tightening of CVO vessels , particularly in the SFO . To dominantly activate the Wnt/β-catenin pathway in ECs , Cdh5 ( PAC ) -CreERT2:Ctnnb1E×3fl/fl ( GOF ) double-transgenic mice were induced with tamoxifen ( TAM ) either directly after birth for three ( 50 µg/day , P1-P3 , Figure 4—figure supplement 1 ) or in the adult for five ( 500 µg/day , 8–10 week-old mice , Figure 4 ) consecutive days . We initially determined recombination efficiency in the brain by analyzing Cdh5 ( PAC ) -CreERT2:mTmG reporter mice ( Wang et al . , 2010Muzumdar et al . , 2007 ) , suggesting that VE-cadherin efficiently drives endothelial recombination in brain vessels ( Figure 4—figure supplement 2 ) . β-Catenin GOF pups were analyzed at P6 and P14 and compared to respective controls . At both timepoints analyzed ( P6 and P14 ) , control vessels exhibited high levels of Plvap/Meca32 and low levels of Cldn5 immunolabeling ( Figure 4—figure supplement 1A , B ) . Interestingly , endothelial-specific β-catenin GOF reverted this phenotype , resulting in significantly decreased Plvap/Meca32 immunoreactivity , whereas Cldn5 expression was markedly increased in these vessels without any changes in VE-cadherin immunolabeling at P6 ( Figure 4—figure supplement 1C ) . In the adult , the same antagonistic regulation of Plvap/Meca32 and Cldn5 by β-catenin GOF was observed as in postnatal stages ( Figure 4B–D; quantification 4E , F ) . When analyzing different timepoints after TAM induction of adult GOF mice for the expression of Plvap/Meca32 and Cldn5 , we observed that activation of β-catenin signaling significantly suppressed Plvap/Meca32 and induced Cldn5 already by day 16 ( Figure 4E , F ) . Maximal Cldn5 induction was observed after 26 days , being significantly higher than at day 19 after the first TAM injection ( Figure 4E ) . Analysis of pooled mRNA from 17 whole SFOs from GOF or control mice , revealed that Plvap/Meca32 was indeed down-regulated in the GOF condition on the mRNA level , whereas Cldn5 did not show an obvious regulation . This suggested that transformation of the leaky into a tight vessel phenotype in the SFO by β-catenin GOF requires around 26 days after induction of recombination . To understand if beside Cldn5 also other tight junction components are regulated upon β-catenin activation in SFO endothelial cells , we stained for occludin ( Ocln ) and zonula occludens 1 ( ZO-1 ) . Analyzing the junctional localization of Ocln normalized to the vessel length in the SFO , we observed only a punctuated staining of Ocln at endothelial cell-cell junctions of controls as previously described by Morita et al . ( Morita and Miyata , 2012 ) . In vessels of GOF mice we noted a significant increase in line-like junctional Ocln staining compared to controls ( Figure 5A , B; quantification Figure 5C ) . Further analysis of mRNA of pooled SFO samples from GOF or control mice , revealed that Ocln , like Cldn5 , did not show an obvious regulation in the GOF condition ( Figure 5D ) . As opposed to Cldn5 and Ocln , ZO-1 showed only a moderately increased localization at cell-cell junctions in SFO vessels of GOF mice ( Figure 5—figure supplement 1 ) . Specifically , ZO-1 was consistently present at inter-endothelial junctions of the SFO in the control condition . As the upregulation of the junctional proteins Cldn5 and Ocln support the interpretation of an SFO vessel tightening in GOF mice , it remained to be clarified if vessel permeability is indeed affected by dominant β-catenin activation in ECs . To this end , GOF and control mice were intravenously injected with FITC-bovine serum albumin ( FITC-BSA ) ( ~68 kDa ) , and examined after 1 . 5 hrs of circulation . Analysis of FITC-BSA leakage normalized to vessel area revealed a significant reduction of extravasation in SFO vessels of the GOF versus control mice ( Figure 6 ) . Specifically , the leaky vessels in the controls showed pronounced FITC-BSA distribution in the circumference of vessels indicated by a prominent cloudy FITC signal in the entire SFO , whereas in the GOF condition the tracer remained confined to the vessel lumen ( Figure 6C ) . To strengthen the observation of SFO vessel tightening , we performed β-catenin GOF experiments also with the Pdgfb-iCreERT2 mouse driver line ( Claxton et al . , 2008 ) , resulting in comparable regulation of Plvap/Meca32 and Cldn5 ( Figure 4—figure supplement 3A , B ) . Activation of the Wnt/β-catenin pathway was supported by significantly increased nuclear Sox17 localization that was reported to be a downstream target of Wnt/β-catenin and to be upstream of Notch ( Corada et al . , 2013; Zhou et al . , 2015 ) ( Figure 4—figure supplement 3C , D ) . Given the increase in Cldn5 expression in SFO vessels , we addressed if the endothelial tightening may also have an effect on the organization of the NVU within the core region of the SFO , in which no astrocytic endfeet are formed around vessels . Therefore , we stained GOF and control SFOs for the astrocytic endfeet markers aquaporin-4 ( Aqp4 ) , α-dystroglycan ( αDag ) and Kir4 . 1 ( Figure 7—figure supplements 1 and 2 ) , as well as the ECM markers laminin α 2 ( Lama2 ) and collagen IV ( ColIV ) ( Figure 7—figure supplement 2 ) . All markers revealed the expected polarized distribution around BBB vessels in the striatum , nicely confirming staining specificity ( Figure 7—figure supplements 1A and 2A , D ) . As previously shown for astrocytic endfeet proteins ( Pócsai and Kálmán , 2015 ) , leaky SFO vessels did not exhibit pronounced staining of the polarity markers αDag and Kir4 . 1 ( Figure 7—figure supplements 1B , C and and 2B , C ) . Moreover , none of these stainings were found to be affected by the GOF conditions , meaning that no distinct staining of vascular endfeet could be observed . Specifically , Aqp4 and αDag showed only a weak , unpolarized localization around vessels in GOF and controls , whereas the sodium channel Kir4 . 1 was mainly expressed by cells morphologically resembling tanycytes in the SFO ( Figure 7—figure supplements 1B , C and and 2B , C ) . The ECM components Lama2 and ColIV , revealed that in GOF and in control vessels of the SFO a vascular and an astrocytic basal lamina was present with no obvious differences in structure and distribution between conditions ( Figure 7—figure supplement 2 ) . In order to further characterize the blood vessels in the SFO of β-catenin GOF mice , we employed electron microscopy to visualize their subcellular phenotype . As expected , the vessels in control SFOs showed the typical large and lacuna-like structure with an extensive ECM circumference ( Figure 7A–C ) . Moreover , fenestrations were frequently observed in the control condition , a morphological feature that is consistent with a high Plvap/Meca32 expression and a permeable phenotype ( Figure 7C; quantification Figure 7D ) . Although the vessel morphology did not show major differences regarding vessel perimeter and structure , the vessels of the GOF mice appeared to have a more compact ECM deposition in their circumference ( Figure 7A–C ) . Endothelial vesicles did not show obvious alterations between GOF and controls ( data not shown ) . Instead , the junctional area of GOF vessels was considerably more elaborate compared to the controls , which exhibited typical blunt-ending connections ( Figure 7C ) . Additionally , fenestrations were significantly reduced in GOF mice , being in line with the reduction of Plvap/Meca32 immunostaining ( Figure 4E; quantification Figure 4G ) and further suggesting that β-catenin GOF in ECs is crucial for the suppression of a leaky vessel phenotype . Upon the observation that dominant endothelial activation of Wnt/β-catenin signaling established barrier properties in SFO vessel , we wanted to elucidate if the tightening of SFO vessels affects neuronal function in this organ . In order to understand if dominant activation of β-catenin signaling in ECs of the SFO may influence neuronal activity in the context of water homeostasis and drinking behavior , we induced thirst in adult mice and analyzed neuronal activity . To this end , WT mice were either kept for 72 hrs under water restriction ( Figure 8A ) or were intraperitoneally injected with a hyperosmolar NaCl ( 3 M ) solution 50 min prior to sample collection ( Figure 8B ) and subsequent assessment of neuronal activation by c-fos staining in the SFO ( Figure 8C , D ) . The Nissl staining of the so-called Nissl flounders nicely documents the neuronal identity of the c-fos+ cells ( Figure 8 ) . As opposed to the general nuclear staining by the fluorescent Nissl stain , the flounders are specific for neurons only . Both thirst-inducing paradigms lead to a significant increase in c-fos+ neurons in the SFO of WT mice ( Figure 8 ) . We could also show a dose-dependent c-fos activation in thirst induction by hyperosmolar NaCl , comparing 2 M and 3 M solutions ( Figure 8—figure supplement 1 ) . Given that water restriction is a more physiological setting which reflects the restricted availability of resources in nature , we made use of this paradigm to investigate the influence of β-catenin GOF on neuronal activity in the SFO . Under control conditions , in which mice received water ad libitum , we could not detect any genotype-specific differences in c-fos+ nuclei in the SFO between control and GOF animals ( data not shown ) . β-Catenin GOF and control mice were subjected to water restriction 26 days after induction by TAM ( Figure 9A ) . In case of water restriction for 72 hrs ( Figure 9A ) a slight , but stable weight loss was induced in GOF and control mice in the same manner ( Figure 9B ) . Analysis of c-fos activation revealed a significantly higher neuronal activity in the SFO of GOF mice ( Figure 9D ) . This suggests that tightening SFO blood vessels may have physiological consequences for the water homeostasis in mice .
The present study deals with the regulation of the leaky vascular phenotype in the CVOs and in the SFO in particular . Specifically , we addressed the questions , a ) if the Wnt/β-catenin pathway is operational in ECs of CVOs during murine development and in the adult mouse and zebrafish , b ) if endothelial-specific , dominant activation of β-catenin transcription could convert the leaky vascular phenotype in CVOs and c ) if the latter may have an effect on CVO function . The principle findings of this study are: 1 ) Wnt/β-catenin signaling is undetectable in CVO vessels during BAT-gal reporter mouse development; 2 ) similarly , β-catenin-mediated transcription is strongly reduced in the adult zebrafish OVLT; 3 ) SFO vessels are heterogenous regarding the expression of Plvap/Meca32 and Cldn5; 4 ) upon genetic β-catenin GOF in ECs , leaky SFO vessels are partially converted into tight vessels; 5 ) functional conversion of SFO vessel towards a BBB-like identity affects neuronal activity in the SFO . Wnt/β-catenin is crucial for brain vascularization and BBB development , by regulating endothelial sprouting as well as by promoting a BBB expression profile in ECs , respectively ( Vanhollebeke et al . , 2015; Liebner et al . , 2008; Daneman et al . , 2009; Stenman et al . , 2008; Zhou et al . , 2014 ) . CVOs are well known , but poorly investigated , structures in the midline of vertebrate brains , conferring neurosensory and/or neurosecretory function . Because of this physiological function , CVO blood vessels were described for a long time to lack BBB characteristics , a feature that is considered to be important for allowing neurons to ‘sense’ salts , hormones , lipids and toxic compounds in the blood ( Sisó et al . , 2010; Kiecker , 2018 ) . Indeed , it has been shown that neurons send axons into the extended perivascular space , which is in line with their sensory function ( Morita and Miyata , 2012 ) . The peculiar , leaky specialization of the CVO vascular system is well documented , showing tortuous and fenestrated vessels with poorly developed inter-endothelial junctions ( McKinley et al . , 2003 ) . However , how this specialization is induced on a molecular level during development and how it is maintained is currently not well understood . Vascular endothelial growth factor ( VEGF ) is the best described inducing factor for endothelial fenestrations and is reported to be expressed in sensory CVOs as well as in other tissues that physiologically require endothelial fenestrations , such as the choroid and the ciliary body of the eye ( Furube et al . , 2014; Ford et al . , 2012; Kinnunen and Ylä-Herttuala , 2012 ) . As the Wnt/β-catenin pathway is considered a master switch for barriergenesis , we hypothesized that β-catenin transcription is not operational during CVO vascularization . The data provided here support this interpretation , as in BAT-gal reporter mice , from the initial identification of the SFO primordium at E13 . 5 , none of the investigated developmental stages revealed a single β-galactosidase-positive vessel within the CVOs ( Figure 1; Figure 1—figure supplements 1 and 2 ) . Although this finding may formally not exclude low level activation of the pathway in ECs , the observation that neighboring , non-endothelial cells in the CVOs , do show Wnt pathway activation , supports the interpretation of low or absent Wnt/β-catenin signaling in CVO ECs . Specifically , we observed that the ependymal cells covering the SFO as well as stromal cells in the core of the organ show Wnt/β-catenin pathway activation ( Figure 1 ) . Quantitative RT-PCR revealed also expression of Wnt3a , Wnt7 as well as Fzd4 in the SFO , suggesting that at least the BBB-inducing machinery is expressed ( data not shown ) . As also in the adult , β-catenin-mediated transcription in ECs is required to maintain BBB function , the absence of reporter activity in mouse ( data not shown ) or zebrafish models presented in this study , further underlines that Wnt is evidently not operational in CVO vessels . This may support the hypothesis that Wnt/β-catenin signaling is actively suppressed in the SFO and likely also in other CVOs that have fenestrated vessels . So far , no conclusive data are available demonstrating Wnt pathway inhibitors in the CVOs , however , it has been shown that expression of the soluble frizzled receptor protein 1 ( Sfrp1 ) is about 30 times higher in the rat choroid plexus ( CP ) , that also lacks BBB vessels , compared to the striatum and parietal cortex ( Bowyer et al . , 2013 ) . Nevertheless , a detailed analysis of the CVOs regarding cell type-specific expression profiles has not been published yet . Still , it has to be noted that vessels in the SFO , OVLT and PP are heterogenous regarding the expression of Plvap/Meca32 and Cldn5 ( Figure 3 , Figure 3—figure supplement 1 ) . Similarly , differentially tight vessels were also shown in other CVOs ( Morita and Miyata , 2012 ) . This raises the question if in the CVOs , unlike in the brain parenchyma a ‘…gradual phenotypic change ( zonation ) along the arteriovenous axis…’ is realized ( Vanlandewijck et al . , 2018 ) , or if alternating endothelial differentiation might be established by factors yet to be discovered . The present findings may suggest that , at least to some degree , vascular phenotypes in the SFO are locally regulated , which would be in line with their role providing local access for neurons to the blood milieu . If vessel differentiation might also be dynamically regulated to control water homeostasis in a circadian rhythm ( Gizowski et al . , 2016 ) , is currently unknown and subject to ongoing investigation . In this regard it is interesting to note however , that Cldn5 and Ocln mRNA were not significantly upregulated when analyzed in whole mount dissected SFOs from GOF mice ( Figures 4H and 5D ) . This might be due to several reasons , such as signal masking by other vessels in the whole mount preparations . Alternatively , this finding might support the interpretation that Cldn5 and Ocln are not transcriptionally regulated by β-catenin , but rather regulated on a post-transcriptional level . Interestingly , there is still some controversy about Cldn5 regulation by Wnt/β-catenin , as it has been shown by Taddei et al . that β-catenin cooperates with FOXO1 to suppress Cldn5 at the promotor level under pro-angiogenic conditions ( Taddei et al . , 2008 ) . On the other hand , it has been shown that Sox18 , a member of the SOX family of high-mobility group box transcription factors , is instrumental in activating Cldn5 transcription , contributing to endothelial barrier formation ( Fontijn et al . , 2008 ) . Given the high redundancy of SoxF genes ( Sox7 , 17 , 18 ) ( Zhou et al . , 2015 ) , it might be feasible that Sox17 , that we report here to be upregulated in SFO vessels of β-catenin GOF mice ( Figure 5—figure supplement 1 ) , mediates Cldn5 regulation . Although the regulation of Cldn5 on the promotor level and the role of β-catenin herein requires additional investigation , the tightening of SFO vessels by Cldn5 protein upregulation in β-catenin GOF mice is consistent with previous reports in other regions of the brain ( Zhou et al . , 2014 ) . Interestingly , in the SFO of GOF mice , we also observed a significantly augmented junctional localization of Ocln , which is in line with the endothelial tightening , but , like for Cldn5 , at which molecular level the Ocln regulation occurs remains to be clarified . Moreover , the adherens and tight junction-associated protein ZO-1 qualitatively showed a slight increase in junctional continuity in the GOF condition , which fits with the overall formation of more elaborate junctional complexes between ECs . The fact that ZO-1 exhibits also junctional staining in the controls ( Figure 5—figure supplement 1 ) , is consistent with its role in VE-cadherin-based adherens junctions , which are also formed by SFO vessels ( Figure 4—figure supplement 1 ) ( Tornavaca et al . , 2015 ) . Beside the mere upregulation of Cldn5 and Ocln , endothelial β-catenin GOF resulted in the abolishment of fenestrations and strengthened inter-endothelial junctions . These findings are well in line with a reduction in VEGF signaling in glioma ECs upon Wnt/β-catenin activation via the downregulation of VEGF receptor 2 ( VEGFR2 , flk-1 ) and upregulation of VEGFR1 ( Reis et al . , 2012 ) . This suggests that also upon β-catenin GOF in CVO vessels the responsiveness of ECs for VEGF could be reduced , leading to regression of fenestrations . Interestingly , the structural components of the NVU such as astrocytic endfeet and ECM , additional crucial BBB features , were not observed to be changed by the GOF condition ( Figure 7—figure supplements 1 and 2 ) . Also , the vessel coverage by pericytes showed no major changes in the SFO comparing GOF and controls ( data not shown ) . If the perivascular fibroblasts , recently described by Vanlandewijck et al . ( Vanlandewijck et al . , 2018 ) , are present at SFO vessels and if yes , whether they are affected by β-catenin GOF in ECs has to be determined in future investigations . Hence , to form the NVU structure might require additional cues and/or prolonged time to form , although the latter explanation might not be as likely as the first , given that even after sixty days after TAM injection the control-like phenotype persisted ( data not shown ) . These findings support the conclusion that tightening the ECs in the SFO via β-catenin GOF does not lead to pronounced structural alterations at the NVU . As the leaky vessels of the SFO core are surrounded by a prominent perivascular space , which is considered to be important for the communication of neuronal axons with the blood milieu , it might be therapeutically beneficial that the NVU is not affected by the dominant activation of endothelial β-catenin . One of the main questions investigated in the present work is how the vasculature in the sensory CVOs like the SFO functionally cooperates with the neurons and other stromal cells to achieve proper physiological regulation of fundamental body parameters like water homeostasis . So far , the vasculature has drawn little attention in this respect , even though considerable progress has recently been made to unravel the regulation of drinking behavior by the SFO , OVLT and the PP ( Gizowski et al . , 2016; Matsuda et al . , 2017; Oka et al . , 2015; Zimmerman et al . , 2016; Augustine et al . , 2018 ) . Specifically , it was shown for the SFO that two distinct populations of neurons expressing ETV-1 and Vgat mediate thirst-ON and thirst-OFF signals , respectively ( Oka et al . , 2015 ) . Here we provide evidence for an essential role of endothelial barrier function in neuronal activation in water restricted mice , as neuronal c-fos reactivity was increased in water-deprived GOF animals ( Figure 9 ) . How this finding relates to the drinking behavior and to the activity of excitatory and inhibitory neuronal signals to and from the median preoptic nucleus ( MnPO ) , which was shown to host the behavioral output neurons ( Augustine et al . , 2018 ) , is beyond the scope of this study and is subject to future work . Moreover , it remains to be clarified if the increased c-fos signal in the SFO of GOF mice is directly caused by the tightened vessel phenotype , or indirectly affected by an altered angiocrine profile of the tightened endothelium , potentially leading to altered drinking behavior . Preliminary analysis of primary mouse brain microvascular ECs ( MBMECs ) treated with Wnt3a revealed no regulation of VEGF that was previously described to be neuroactive ( data not shown ) ( Mackenzie and Ruhrberg , 2012 ) . Although own preliminary experiments aiming to pharmacologically tighten CVO vessels with a systemically administered Wnt/β-catenin activator did not result in SFO vessel tightening ( data not shown ) , this might be a potential way to therapeutically modulate water intake . Interestingly , patients that chronically receive LiCl , an FDA-approved drug for bipolar disorders and a potent Wnt/β-catenin activator , frequently develop polyuria that is linked to altered anti-diuretic hormone ( ADH; vasopressin ) function , which is released by the PP . Moreover , many patients develop polydipsia and urinate more frequently ( Malhi , 2015 ) . Hence in-depth investigation of the pharmacologic modulation of SFO vessel permeability is required . Although the detailed mechanisms underlying the neuro-vascular coupling in the CVOs have to be investigated in more detail , in light of the present work however , the CVO vasculature likely participates actively in controlling water homeostasis .
Mice were housed under standard conditions with 12 hrs light dark cycle and water and mouse chow available ad libitum if not declared otherwise . All experimental protocols , handling and use of mice were approved by the Regional Council Darmstadt , Germany ( V54-19c20/15-FK/1052 and V54-19c20/15-FK/1108 ) . Wildtype ( WT ) C57BL6/J as well as transgenic animals were used . The following mouse strains were included Cdh5 ( PAC ) -CreERT2 ( Wang et al . , 2010 ) , PDGFB-iCreERT2 ( Claxton et al . , 2008 ) , Ctnnb1Ex3fl/fl ( Harada et al . , 1999 ) , BAT-gal+/wt Wnt/β-catenin reporter ( Maretto et al . , 2003 ) and mT/mG ( Muzumdar et al . , 2007 ) . Zebrafish ( Danio rerio ) were maintained under standard conditions at 28°C and a 14 hr light/10 hr dark cycle , in accordance with European and national animal welfare and ethical guidelines ( protocol approval number: CEBEA-IBMM2017-22:65 ) . Transgenic lines used in this study were Tg ( kdrl:Hsa . HRAS-mCherry ) s896 ( Chi et al . , 2008 ) and Tg ( 7xTCF-Xla . Siam:GFP ) ia4 ( Moro et al . , 2012 ) . After euthanasia with 0 . 3 mg . ml-1 Tricaine methanesulfonate ( MS-222 ) for 10 min , adult brains aged 6 to 12 months were dissected and fixed overnight in sweet fixative ( 4% PFA , 4% sucrose in PBS ) . Brains were washed in PBS and embedded in 4% low-melting agarose . 300 μm sections were obtained using a LeicaVT1200s automated vibratome ( Leica Biosystems ) . Sections were imaged on a Zeiss LSM710 confocal microscope using separate channels . To investigate β-catenin activity in ECs , Wnt/β-catenin reporter mice ( BAT-gal+/wt ) were bred with C57BL6/J mice to generate either heterozygous positive pups for β-galactosidase or homozygous negative control littermates . At different developmental stages ( embryonic days E13 . 5 , E17 . 5 ) embryos were harvested . For postnatal day 0 ( P0 ) pups were sacrificed by decapitation , for postnatal day 21 ( P21 ) pups were sacrificed by cervical dislocation . Brain preparation was performed in ice cold PBS and followed by overnight fixation in 4% PFA in PBS . For cryo-sectioning the whole brain was embedded in Tissue-Tek O . C . T . after incubation in 12/15/18% sucrose . β-Catenin endothelial specific gain of function system was kept by the use of Ctnnb1Ex3fl/fl ( Harada et al . , 1999 ) mice crossed with the Cdh5 ( PAC ) -CreERT2 ( Wang et al . , 2010 ) . To activate the Cre-recombinase , tamoxifen ( TAM , 500 µg/day in corn oil; central pharmacy , Steinbach , Germany ) was i . p . injected on five consecutive days . Brains were harvested and embedded for cryo-sectioning at 16 , 19 and 26 days after the first TAM injection . To investigate the tightening effect at different postnatal stages , pups were i . p . injected with TAM ( 50 µg/day in corn oil ) at P0-P3 and analyzed on P6 and P14 . Animals were injected with TAM and kept for 26 days to assure SFO vessel tightening as described above . Mice were anesthetized and intravenously injected with 50 µl FITC-albumin ( #A9771; Sigma-Aldrich ) . After 1 . 5 hr mice were sacrificed by cervical dislocation . The embedded brain was cryo sectioned ( 20 µm , counterstained for Podxl and analyzed after confocal imaging . For analysis the FITC covered area as well as the Podxl+ vessel area within the SFO was measured for each optical section of at least one stack . To quantify the tracer leakage , the FITC covered SFO area was normalized to the vessel area , indicated by Podxl staining . Mice were i . p . injected with either 3 M or an isotonic ( 0 . 15 M ) NaCl solution ( 150 µl/20 g mouse ) as described in Zimmerman et al . ( Zimmerman et al . , 2016 ) . After an incubation time of 50 min without any access to drinking water animals were sacrificed and SFOs were analyzed in cryo sections for c-fos activation . At first the tissue was embedded in low melt agarose . The following dehydration and delipidation protocol was adapted from Orlich et al . and Renier et al . ( Orlich and Kiefer , 2018; Renier et al . , 2014 ) . In brief , MetOH ( 50/70/100% ) in PBS was used for 1 hrs for each step in dark-brown glass vials slightly shaking at RT , followed by an overnight incubation in 100% MetOH . To remove lipids an incubation with dichlormethane followed until the tissue sank down . Afterwards ethylcinnamate ( ECi ) ( #112372 , Sigma-aldrich ) clearing was performed as described in Klingberg et al . ( Klingberg et al . , 2017 ) . Samples were stored in ECi solution that was renewed one day before the acquisition . Samples were imaged in ECi solution with an UltraMicroscope II ( LaVision , Germany ) and stacks with 1 µm step size were further processed for visualization either by Imaris 9 ( BitPlane , Switzerland ) or the volume visualization framework Voreen ( volume rendering engine ) ( Meyer-Spradow et al . , 2009 ) . Either native frozen tissue or sucrose embedded samples were cryo-sectioned coronal or sagittal in 10 µm thickness and then fixed with 4% PFA for 10 min at room temperature or with ice cold MetOH for 3 min . To block/permeabilize tissue slides were incubated for 1 hr ( overnight for vibratome section ) ( 10% NDS , 0 . 1% Triton-X100 in PBS ) . Primary antibodies ( Key resource table ) were incubated for 2 hr ( 24 hr for vibratome section ) and secondary for 1 hr ( 4 hr for vibratome section ) in antibody incubation buffer ( 1% BSA , 0 . 1% Triton-X100 in PBS ) . If required , sections of PFA-fixed samples were subjected to antigen-retrieval ( * ) by boiling slides for 45 min in AR6 Buffer ( #AR600250ML; Perkin Elmer ) . After cooling them down and an additional washing step , slides were stained as described above . Images were acquired using either a Nikon 80i wide field fluorescent microscope , or a Nikon C1si Confocal Laser Scanning Microscope , together with NIS-Elements Microscope Imaging Software for image analysis ( Nikon Instruments , Inc . , Düsseldorf , Germany ) . SFO vessels were defined as regions of interest ( ROI ) for area measurements . Staining was evaluated as a ratio of Cldn5 or Plvap/Meca32 to vessel area , evidenced by Podxl or Cdh5 labeling . The number of c-fos+ neurons as well as the total number of nuclei within the SFO , defined as ROI , were counted and the ratio of c-fos + to total nuclei was calculated . Animals were anesthetized and transcardially perfused with PBS/heparin for 1 min followed by 4 min with 4% PFA in cacodylate buffer ( CB , pH 7 . 4 ) . The SFO was whole mount prepared in ice cold PBS directly after brain isolation . Afterwards the tiny SFO whole mount tissue pieces were post-fixed with 4% PFA and 2% glutaraldehyde/CB overnight at 4°C . Prior to embedding the tissue was incubated in 1% Os for 2 hr at RT followed by dehydration in graded acetone including contrast enhancement with uranyl acetate solution at 4°C o/n . Samples were embedded in Epon finally polymerized at 60°C for 24 hr . Ultra-thin sections ( 50 nm ) were cut with Leica Ultracut UCT and analyzed using a Tecnai Spirit BioTWIN FEI electron microscope at 120kV . Images were taken with an Eagle 4K CCD bottom-mount camera . For the quantification of fenestrations , 5 SFO vessels were analyzed for each animal ( n = 4 per genotype ) . RNA isolation was done using the RNeasy plus Microkit ( Qiagen ) according to the manufacturer recommendations with DNase on-column digestion ( Qiagen ) like suggested in the RNeasy Minikit ( Quiagen ) . For cDNA synthesis ( RevertAidTM H minus first strand cDNA synthesis kit , #K1632 , Thermo Fisher Scientific ) 57 ng RNA were used from SFO tissue of β-catenin GOF ( Cre+ ) and control ( Cre- ) mice . Quantitative real time RT-PCR ( qRT-PCR ) was performed in technical triplicates for each sample using the Absolute qPCR SYBR Green Fluorescein Mix ( AB-1219 , Thermo Fisher Scientific ) according to the manufacturer's protocol . Rplp0 was used as a housekeeping gene for normalization . Expression data were analyzed with ∆∆ct method . Primer sequences used for cDNA amplification by qRT-PCR are listed in Table 2 . No statistical tests were used to predetermine sample size . Several independent experiments were performed to ensure reproducibility . The investigators were blinded by the experimental design during the analysis of the experiments shown in Figures 4E , 5F–H , 6D , 7E–F and 8C as well as in Figure 4—figure supplement 1C , Figure 5—figure supplement 1D , Figure 7—figure supplement 2B . Raw data are presented in the additional source data files . The number of biological replicates is provided as ‘n’ in the legend of each figure . Technical replicates , such as the number of sections analyzed or replicates for qRT-PCR analyses are indicated in the figure legend and the respective material and methods section , respectively . Results are shown as mean ±SEM . Statistical significance was assessed by an unpaired t-test using GraphPad Prism version 6 . 0 ( GraphPad Software Inc . , USA ) . p-Values were considered significant at p<0 . 05 and individual p-values are provided in each figure . | Infections and diseases in the brain and spine can be very damaging and debilitating . Indeed , the central nervous system also needs a carefully controlled biochemical environment to survive . As such , all animals with a backbone have barriers and defenses to protect and preserve this key system . One of these is the blood-brain barrier , a physical barrier between the brain and the outside world . Where most blood vessels allow relatively free exchange of chemicals between the blood and surrounding cells , the blood-brain barrier controls what can move between the bloodstream and the brain . Yet , there are gaps in the blood-brain barrier , specifically within structures in the brain called the circumventricular organs . These leaky vessels allow the brain cells in these regions to monitor the blood and respond to changes , for example , by triggering sensations such as hunger , thirst or nausea . It is not clear what stops the blood-brain barrier from forming in these regions and what effect the presence of a barrier would have on the brains activity , or the health and behavior of the animal . Benz et al . have now used mice and zebrafish to examine the development and structure of the blood-brain barrier . The investigation revealed that the signals that induce the blood-brain barrier throughout the brain are absent in the circumventricular organs of both species . Next , by artificially activating a protein involved in cell-cell interactions in mice , Benz et al . created blood-brain barrier-like structures in circumventricular organs by converting the leaky vessels into tight ones . This change meant that the brain cells in these regions did not respond properly to water deprivation , which potentially may have affected the regulation of thirst in these mice . Understanding the blood-brain barrier could have a variety of impacts on how we treat diseases in the central nervous system . This includes stroke , brain tumors and Alzheimers disease . These findings could particularly help scientists to better understand conditions that affect basic needs like thirst and hunger . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2019 | Low wnt/β-catenin signaling determines leaky vessels in the subfornical organ and affects water homeostasis in mice |
When human fibroblasts take up plasma low density lipoprotein ( LDL ) , its cholesterol is liberated in lysosomes and eventually reaches the endoplasmic reticulum ( ER ) where it inhibits cholesterol synthesis by blocking activation of SREBPs . This feedback protects against cholesterol overaccumulation in the plasma membrane ( PM ) . But how does ER know whether PM is saturated with cholesterol ? In this study , we define three pools of PM cholesterol: ( 1 ) a pool accessible to bind 125I-PFO* , a mutant form of bacterial Perfringolysin O , which binds cholesterol in membranes; ( 2 ) a sphingomyelin ( SM ) -sequestered pool that binds 125I-PFO* only after SM is destroyed by sphingomyelinase; and ( 3 ) a residual pool that does not bind 125I-PFO* even after sphingomyelinase treatment . When LDL-derived cholesterol leaves lysosomes , it expands PM's PFO-accessible pool and , after a short lag , it also increases the ER's PFO-accessible regulatory pool . This regulatory mechanism allows cells to ensure optimal cholesterol levels in PM while avoiding cholesterol overaccumulation .
Animal cells tightly control the level of cholesterol in their plasma membranes ( PMs ) . Control is mediated by SREBP-2 , a membrane bound protein that activates transcription of genes encoding most , if not all , enzymes of cholesterol biosynthesis ( Brown and Goldstein , 1997; Horton et al . , 2002; Espenshade and Hughes , 2007 ) . SREBP-2 is synthesized on endoplasmic reticulum ( ER ) membranes where it binds to an escort protein called Scap . When ER cholesterol is less than ∼5 mole% of total ER lipids ( Radhakrishnan et al . , 2008 ) , the Scap/SREBP-2 complex enters COPII-coated vesicles that move to the Golgi , where two proteases liberate the active fragment of SREBP-2 ( Sun et al . , 2007 ) . The active fragment enters the nucleus where it activates transcription of the cholesterol-synthesizing genes and also the gene for the low density lipoprotein ( LDL ) receptor , which supplies the cell with exogenous cholesterol ( Brown and Goldstein , 1997 ) . When cholesterol-depleted cells are incubated with cholesterol-carrying LDL , the lipoprotein binds to its receptor , enters the cell through endocytosis in clathrin-coated vesicles and reaches lysosomes where its cholesterol is liberated ( Brown and Goldstein , 1986 ) . The lysosome-derived cholesterol is delivered to the PM and the ER membrane . When the ER cholesterol rises above a sharp threshold of 5 mole% of total ER lipids ( Radhakrishnan et al . , 2008 ) , the Scap/SREBP complex binds to an ER anchor protein called Insig , and this prevents its transport to the Golgi ( Goldstein et al . , 2006 ) . As a result , cholesterol synthesis and uptake from LDL are reduced . If excess cholesterol accumulates , it is removed from the PM and delivered to the ER where it is esterified by acyl-CoA:cholesterol acyltransferase ( ACAT ) for storage as cytoplasmic cholesteryl ester droplets ( Brown and Goldstein , 1986 ) . The above studies expose a potential paradox . The Scap/SREBP system functions to regulate the concentration of cholesterol in the PM where the vast majority ( 60–90% ) of a cell's total cholesterol is located ( De Duve , 1971; Lange et al . , 1989 ) . Within PMs , cholesterol accounts for 40–50% of lipids ( Ray et al . , 1969; Lange et al . , 1989; van Meer et al . , 2008 ) . Yet , the cholesterol sensor for the Scap/SREBP system is located not in the PM , but in the ER , which contains only ∼1% of a cell's total cholesterol ( Lange and Steck , 1997 ) . Within the ER , cholesterol accounts for no more than 5% of total membrane lipids . How does the small ER cholesterol pool monitor the cholesterol concentration in the large PM pool ? The current paper suggests a possible resolution of this paradox . The key to the proposed solution lies in the paths by which cholesterol moves from lysosome to the PM and to the ER . Before the current understanding of SREBP was obtained , it was suggested that LDL-derived cholesterol moves from lysosomes to the PM , and after the capacity of the PM is saturated , then secondarily to the ER ( Xu and Tabas , 1991; Lange et al . , 1997 ) . Others proposed a different route whereby cholesterol moves directly from lysosomes to the ER without involving the PM ( Neufeld et al . , 1996; Underwood et al . , 1998 ) . Studies supporting either route rely on indirect approaches to monitor the arrival of LDL-derived cholesterol at the PM and at the ER . In all studies , arrival of cholesterol at the ER was monitored by measuring the formation of cholesteryl esters mediated by ACAT , a resident ER membrane protein ( Chang et al . , 1995 ) . On the other hand , arrival of LDL-derived cholesterol at the PM was monitored using four different methods: ( 1 ) treatment of intact cells with cholesterol oxidase , a soluble cholesterol-modifying enzyme that converts cholesterol to cholestenone ( Lange et al . , 1997; Underwood et al . , 1998 ) ; ( 2 ) treatment of intact cells with 2-hydroxypropyl-β-cyclodextrin ( HPCD ) , a soluble acceptor that extracts cholesterol from membranes ( Neufeld et al . , 1996 ) ; ( 3 ) treatment of intact cells with amphotericin B , a polyene antibiotic that forms pores in cholesterol-rich membranes ( Underwood et al . , 1998 ) ; and ( 4 ) lipid extraction of unfractionated whole cells , followed by thin layer chromatography ( Xu and Tabas , 1991 ) . None of the aforementioned approaches directly measured the concentration of cholesterol in purified PMs . Once LDL-derived cholesterol arrives at the PM or ER , the manner in which it is organized within each membrane also remains controversial . In particular , the distribution of cholesterol in the PM has been studied intensely . The idea that cholesterol forms complexes with PM phospholipids such as sphingomyelin ( SM ) in membranes has a long history ( Finean , 1953; Radhakrishnan and McConnell , 2005 ) . It is possible that these complexes exist in the PM of living cells . Under the right conditions , they could coalese to form higher-order structures , such as the proposed lipid raft domains that are enriched in cholesterol and SM and are involved in cell signaling ( Brown and Rose , 1992; Simons and Ikonen , 1997 ) . However , the size and lifetime of these lipid rafts remain hotly debated ( Edidin , 2001; Munro , 2003; Gowrishankar et al . , 2012 ) . Whatever the underlying organization of cholesterol in the PM may be , reducing the SM content of PM by treatment with sphingomyelinase ( SMase ) causes a portion of PM cholesterol to move to the ER ( Slotte and Bierman , 1988; Scheek et al . , 1997; Abi-Mosleh et al . , 2009 ) . We recently described a probe that measures an accessible pool of cholesterol in the PM ( Das et al . , 2013 ) . This probe , designated 125I-PFO* , is an 125I-labeled , mutant version of Perfringolysin O , a bacterial protein toxin that binds to cholesterol-rich membranes ( Flanagan et al . , 2009; Das et al . , 2013 ) . 125I-PFO* retains the ability to bind cholesterol , but ( unlike the native version , PFO ) it no longer forms pores at 4°C . ( Das et al . , 2013 ) . We showed that cholesterol in PMs is not accessible to 125I-PFO* until the cholesterol concentration exceeds a threshold of ∼35 mole% ( Das et al . , 2013 ) , which is considerably higher than the 5 mole% threshold for PFO binding to purified ER ( Radhakrishnan et al . , 2008; Sokolov and Radhakrishnan , 2010 ) . Below these thresholds , PM and ER cholesterol are inaccessible to PFO . In the current paper , we use 125I-PFO* binding to define three pools of cholesterol in the PM . The first pool is the ‘PFO-accessible pool’ that binds 125I-PFO* when the membrane is in its native cholesterol-replete state . This pool is labile , and it is depleted selectively when cells are deprived of cholesterol . The second pool binds 125I-PFO* only after SM in the PM is destroyed by SMase . This pool , which is not depleted by cholesterol deprivation , is referred to as the ‘SM-sequestered pool’ . The remaining cholesterol exists as a third pool and does not bind 125I-PFO* even after SMase treatment . We name this third pool the ‘essential pool’ because depletion of this pool causes cells to round up and dissociate from the petri dish . After LDL-cholesterol is liberated in lysosomes , it is transported out of lysosomes and expands the PM's PFO-accessible pool . After a short lag , it also increases the ER's PFO-accessible regulatory pool . These data indicate how one pool of cholesterol in the PM can regulate overall cholesterol homeostasis in animal cells .
The experiments in Figures 1 and 2 were designed to determine the time course of delivery of LDL-derived cholesterol to the ER as compared with the PM . In Figure 1 , we used 125I-PFO* binding to monitor delivery to the PM , and we used ACAT-mediated cholesterol esterification as an indirect monitor of cholesterol delivery to the ER . Human SV-589 fibroblasts were depleted of cholesterol by incubation in lipoprotein-deficient serum plus the HMG CoA reductase inhibitor compactin . We then added varying amounts of LDL together with [14C]oleate . After the incubation , one set of cells was harvested for measurement of cholesteryl [14C]oleate . The other set was chilled to 4°C and used to measure 125I-PFO* binding . As shown in Figure 1A , when LDL was added at 50 μg protein/ml , the amount of 125I-PFO* binding increased within 1 hr and continued to increase for 6 hr . In contrast , we detected no synthesis of cholesteryl [14C]oleate until after 2 hr . In Figure 1B , we fixed the incubation time at 4 hr and varied the LDL concentration . LDL at 3 μg protein/ml clearly increased the binding of 125I-PFO* without significantly increasing cholesteryl [14C]oleate formation , which increased only when the LDL concentration was at least 10 μg protein/ml . Figure 1C plots cholesteryl [14C]oleate formation vs 125I-PFO* binding for the data in Figures 1A , B . Irrespective of whether the incubation time or LDL concentration was varied , cholesteryl [14C]oleate formation did not increase until PM 125I-PFO* binding rose above 3 μg/mg protein . These data raise the possibility that LDL-derived cholesterol must first expand a pool of cholesterol in the PM before it is delivered to the ER in sufficient amounts to undergo esterification by ACAT . 10 . 7554/eLife . 02882 . 003Figure 1 . Movement of LDL-derived cholesterol from lysosomes to cell surface and to ER in human fibroblasts . On day 0 , SV-589 cells were set up in medium A at 1 × 105 cells per 60-mm dish . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were refed with medium C containing 50 μM compactin and 50 μM sodium mevalonate and incubated for 16 hr at 37°C . On day 4 , cells received 2 ml of fresh medium D containing 50 μM compactin and 50 μM mevalonate together with either 50 μg protein/ml LDL ( A ) or the indicated concentration of LDL ( B ) . The cells were incubated for either the indicated time ( A ) or 4 hr ( B ) in the presence of either 0 . 2 mM unlabeled sodium oleate-albumin ( ) or 0 . 2 mM sodium [14C]oleate-albumin ( 7780 dpm/nmol ) ( ) . For 125I-PFO* binding ( ) , after the indicated incubation , the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 11 × 103 cpm/μg ) . After 2 hr at 4°C , the total cell surface binding of 125I-PFO* was determined , and the amount bound after subtraction of the zero-time value ( 1 . 6 μg/mg protein ) is plotted as ‘Increase in 125I-PFO* Bound’ . For measurement of cholesteryl [14C]oleate formation ( ) , after the indicated incubation , the cells were harvested and the increase in content of cholesteryl [14C]oleate was determined after subtraction of the zero-time value ( 0 . 09 nmol/mg protein ) . All values shown are the average of duplicate incubations . ( C ) Graph showing relation between the increase in 125I-PFO* binding and the increase in cholesteryl [14C]oleate formation . Data taken from ( A ) and ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 00310 . 7554/eLife . 02882 . 004Figure 2 . Kinetics of transport of LDL-derived cholesterol from lysosomes to PMs and ER . On day 0 , SV-589 cells were set up in medium A at 1 × 105 cells per 60-mm . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium B containing 50 μM compactin and 50 μM sodium mevalonate and then incubated for 16 hr at 37°C . On day 4 , each monolayer received 2 ml of fresh medium E containing 50 μM compactin , 50 μM mevalonate , and 50 μg protein/ml of LDL . Groups of 18 dishes were incubated at 37°C for the indicated times , after which each monolayer was washed with PBS at room temperature and then treated with 2 ml of fresh medium E containing 50 μM compactin and 50 μM mevalonate for 15 min at 37°C . Six dishes from each group were pooled for purification of PMs , and the remaining 12 dishes were pooled for purification of ER membranes , as described in ‘Materials and methods’ . Purified membranes were incubated with PFO for 1 hr at 37°C as described in ‘Materials and methods’ . ( A ) Oligomer formation of PFO was assessed by SDS-PAGE and immunoblotting with an antibody against the His tag . Exposure time for the films was 10 s . ( B ) Densitometric analysis of the scanned gel was performed to quantify the percentage of PFO in its oligomeric form relative to the total ( oligomer + monomer ) for each time point . The percentage of oligomer formed at the 2-hr time point ( 50% for PM; 80% for ER ) was designated as 100% . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 004 [14C]Oleate incorporation into cholesteryl esters and 125I-PFO* binding to cell surfaces are indirect measures of ER membrane cholesterol and PM cholesterol , respectively . In order to have a more direct measurement of the cholesterol content of ER and PMs , we isolated pure ER membranes and pure PMs using previously described techniques ( Radhakrishnan et al . , 2008; Das et al . , 2013 ) . We then estimated the accessibility of cholesterol in both the membranes by measurement of the polymerization of native PFO , which forms SDS-resistant oligomers upon binding to membrane cholesterol ( Sokolov and Radhakrishnan , 2010 ) . In the experiment of Figure 2 , we first depleted cells of cholesterol and then added LDL . After varying times , one set of cells was harvested for purification of PMs , and a duplicate set was harvested for purification of ER membranes . The purified membranes were incubated with His-tagged PFO , after which the protein was solubilized with SDS-containing buffer and subjected to SDS-PAGE . SDS-resistant PFO oligomers were visualized by immunoblotting with anti-His antibody , as shown in Figure 2A . When PMs were studied , oligomers of PFO reached a half-maximal value after 30 min ( top panel of Figure 2A , Figure 2B ) . Interaction of PFO with ER membranes was markedly different . At time periods up to 1 hr after LDL addition , PFO migrated as a monomer ( bottom panel of Figure 2A , Figure 2B ) . At 1 . 5 hr , the oligomer abruptly appeared , and it did not increase further with time . These results are similar to those obtained when ER cholesterol was monitored by [14C]oleate incorporation into cholesteryl esters ( Figure 1A ) . The time-lag of 1 hr between PFO binding to PM vs ER membranes in Figure 2 does not necessarily mean that cholesterol levels in ER do not increase during this period—only that cholesterol levels have not increased past the 5 mole% threshold level for PFO binding to ER membranes . The absence of an initial lag for the PM indicates that PM cholesterol rapidly surpassed the 35 mole% threshold level for PFO binding to PM . Despite several efforts , we have been unable to measure directly the kinetics of arrival of LDL-derived cholesterol at PM and ER , owing to the difficulty in obtaining sufficient amounts of purified ER membranes that allow accurate and reproducible measurements of the mass of newly arrived cholesterol . Nonetheless , the results of Figures 1 and 2 suggest that LDL-derived cholesterol expands a pool of PM cholesterol before raising ER cholesterol by sufficient amounts to undergo esterification by ACAT or be bound by 125I-PFO* . The estimation of membrane cholesterol by the PFO oligomerization assay is non-linear and reflects the threshold nature of PFO binding to cholesterol-containing membranes . Binding of PFO to purified cell membranes does not commence until the cholesterol concentration exceeds a threshold of 5 mole% in ER membranes ( Sokolov and Radhakrishnan , 2010 ) and 35 mole% in PMs ( Das et al . , 2013 ) , respectively . One possible explanation for this higher threshold in the PM is that some of the PM cholesterol is sequestered in complexes with SM and is not accessible to PFO . To test this hypothesis ( Figure 3 ) , we treated cells without or with compactin to create cholesterol-replete or cholesterol-depleted cells . Half of the dishes from each set were treated with SMase to hydrolyze SM , after which the cells were chilled and incubated with 125I-PFO* . In cholesterol-replete cells , 125I-PFO* binding was abundant , and SMase treatment increased the binding by 16 μg/mg protein ( Figure 3A ) . When the cells had been treated with compactin to deplete cholesterol , 125IPFO* binding was reduced by 92% . Nevertheless , SMase treatment increased 125I-PFO* binding by 14 μg/mg protein , which was nearly the same as the increase in cholesterol-replete cells . We use the term ‘PFO-accessible’ to refer to the pool of PM cholesterol that binds to 125I-PFO* before SMase treatment and the term ‘SM-sequestered’ to denote the pool of PM cholesterol that is released to bind to 125I-PFO* after SMase treatment . 10 . 7554/eLife . 02882 . 005Figure 3 . Effect of SMase treatment of human fibroblasts on amount of cell surface binding of 125I-PFO* ( A ) and PM content of SM and ceramide ( B–D ) . On day 0 , SV-589 cells were set up in medium A at 1 × 105 cells per 60-mm dish ( A–D ) . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium B containing 50 μM sodium mevalonate in the presence or absence of 50 μM compactin as indicated . On day 4 , each monolayer received fresh medium B containing 50 μM mevalonate in the absence or presence of 50 μM compactin and 100 milliunits/ml of SMase as indicated . ( A ) 125I-PFO* binding . After incubation for 15 min at 37°C , the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml of ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 11 × 103 cpm/μg ) . After 2 hr at 4°C , the total amount of cell surface binding of 125I-PFO* was determined . Each bar represents the mean of triplicate incubations with individual values shown . ( B–D ) Lipid measurements . Cells were cultured under identical condition as described above . For each treatment , six 60-mm dishes were pooled together for purification of PMs by surface biotinylation as described in ‘Materials and methods’ . Lipids were extracted from the membranes , and the content of cholesterol ( B ) , SM ( C and D ) , and ceramide ( C and D ) were measured as described in ‘Materials and methods’ . The data represent the mean ± SEM obtained from three independent experiments . Each individual data point denotes the average of duplicate measurements of each pooled sample . Bracketed numbers denote the increase in 125I-PFO* binding resulting from SMase treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 00510 . 7554/eLife . 02882 . 006Figure 3—figure supplement 1 . Movement of FCS-derived cholesterol and effect of SMase treatment in hamster cells . ( A ) Movement of FCS-derived cholesterol from lysosomes to cell surface and to ER in hamster cells . On day 0 , CHO-7 cells were set up in lipoprotein-deficient medium G at 3 × 105 cells per 60-mm dish . On day 2 , cells were switched to medium G containing 50 μM compactin and 50 μM sodium mevalonate and incubated for 16 hr at 37°C . On day 3 , cells received 2 ml of fresh medium H containing 50 μM compactin and 50 μM mevalonate . After incubation for the indicated time with 5% FCS ( containing lipoprotein-cholesterol ) in the presence of either 0 . 2 mM unlabeled sodium oleate-albumin ( ) or 0 . 2 mM sodium [14C]oleate-albumin ( 4466 dpm/nmol ) ( ) , the cells were harvested for assays . For 125IPFO* binding ( ) , after the indicated time the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml ice-cold buffer A containing 25 μg/ml 125IPFO* ( 45 × 103 cpm/μg ) . After 2 hr at 4°C , the total cell surface binding of 125I-PFO* was determined , and the amount bound after subtraction of the zero-time value ( 0 . 4 μg/mg protein ) is plotted as ‘Increase in 125I-PFO* Bound’ . For measurement of cholesteryl [14C]oleate formation ( ) , after the indicated time the cells were harvested , and the increase in content of cholesteryl [14C]oleate was determined after subtraction of the zero-time value ( 0 . 0 nmol/mg protein ) . All values shown are the average of duplicate incubations . ( B ) Effect of SMase treatment of hamster cells on amount of cell surface binding of 125I-PFO* . On day 0 , CHO-K1 cells were set up in medium F at 4 × 105 cells per 60-mm dish . On day 1 , cells were switched to lipoprotein-deficient medium G . On day 2 , cells were treated with fresh medium G containing 50 μM sodium mevalonate in the presence or absence of 10 μM compactin as indicated . On day 3 , each monolayer received fresh medium G containing 50 μM mevalonate in the absence or presence of 10 μM compactin and 100 milliunits/ml of SMase as indicated . After incubation for 30 min at 37°C , the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml of ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 10 . 5 × 103 cpm/μg ) . After 2 hr at 4°C , the total amount of cell surface binding of 125I-PFO* was determined . Each bar represents the average of duplicate incubations with individual values shown . Bracketed numbers denote the increase in 125I-PFO* binding resulting from SMase treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 006 We also purified PMs and measured their cholesterol content directly ( Figure 3B ) . These data show that compactin reduced PM cholesterol from 43% to 27 mole% of PM lipids ( blue bars in Figure 3B ) , with a corresponding decrease in 125I-PFO* binding by 12 µg/mg protein ( blue bars in Figure 3A ) . Using this result , we can derive an approximate relationship between 125I-PFO* binding and the magnitude of the PFO-accessible pool: 1 µg/mg protein of 125I-PFO* binding corresponds to 1 . 3 mole% of PM cholesterol that is PFO-accessible . We can then estimate the fraction of PM cholesterol that is in the SM-sequestered pool from the increase in 125I-PFO* binding ( 14 μg/mg ) after SMase treatment to be 18 mole% of PM lipids . Considered together , these data indicate that when PM cholesterol was reduced to 27 mole% of PM lipids , little of the residual cholesterol was accessible to PFO , but 2/3 of this residual pool ( 18 mole% ) was made accessible by SMase treatment . The remaining 1/3 of the residual pool ( 9 mole% ) was inaccessible to PFO even after SMase treatment . The ‘Discussion’ contains a further elaboration of these cholesterol pools . To verify that the SMase treatment depleted SM in PMs , we directly measured SM and ceramide levels in purified PMs by high-performance liquid chromatography ( HPLC ) -tandem mass spectrometry ( MS ) . The distribution of the major SM species that we identified is shown in Table 1 . Consistent with previous MS analysis of SM composition in human fibroblasts ( Valsecchi et al . , 2007 ) , we found that more than 90% of the SM in our purified PMs was comprised of two sphingosine species , one with a palmitoyl ( 16:0 ) acyl chain and the other with a nervonoyl ( 24:1 ) acyl chain . SMase treatment reduced the levels of all SM species and produced a reciprocal increase in the corresponding ceramide species . SMase treatment of cells incubated without or with compactin reduced total SM levels in purified PM by >90% ( Figure 3C ) and increased total ceramide levels in these same membranes by a somewhat lesser amount ( Figure 3D ) . 10 . 7554/eLife . 02882 . 007Table 1 . Major subspecies of sphingomyelin and ceramide in sterol-repleted ( −compactin ) and sterol-depleted ( +compactin ) PMs from SV-589 cells treated without and with SMaseDOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 007Lipid− Compactin+ CompactinSpecies− SMase+ SMase− SMase+ SMasemole % of total PM lipidsSphingomyelin C16:05 . 38 ± 0 . 190 . 32 ± 0 . 028 . 04 ± 0 . 480 . 58 ± 0 . 12 C18:00 . 11 ± 0 . 000 . 01 ± 0 . 000 . 16 ± 0 . 010 . 01 ± 0 . 00 C24:00 . 35 ± 0 . 010 . 01 ± 0 . 000 . 55 ± 0 . 030 . 02 ± 0 . 00 C24:12 . 59 ± 0 . 220 . 14 ± 0 . 014 . 41 ± 0 . 230 . 27 ± 0 . 05Ceramide C16:00 . 04 ± 0 . 002 . 95 ± 0 . 500 . 07 ± 0 . 013 . 93 ± 1 . 22 C18:00 . 00 ± 0 . 000 . 11 ± 0 . 010 . 01 ± 0 . 000 . 15 ± 0 . 04 C24:00 . 01 ± 0 . 000 . 25 ± 0 . 040 . 01 ± 0 . 000 . 36 ± 0 . 11 C24:10 . 02 ± 0 . 001 . 39 ± 0 . 160 . 04 ± 0 . 002 . 14 ± 0 . 60Lipids from the purified PMs isolated from the cells used in Figure 3 were extracted with 85:15 ( vol/vol ) ethyl acetate: isopropanol , and the contents of the four indicated acyl chain subspecies of SM and ceramide were quantified as described in ‘Materials and methods’ . The data are expressed as mole % of total PM lipids and represent the mean ± SEM obtained from three independent experiments with duplicate measurements of each sample . Levels of SM and ceramide with oleoyl ( 18:1 ) , arachidoyl ( 20:0 ) , and behenoyl ( 22:0 ) acyl chains are not included in this table as their levels were less than 0 . 1% of total PM lipids . The striking aspect of Figure 3A is that the SM-sequestered pool of cholesterol was nearly identical whether or not the total membrane cholesterol had been depleted with compactin . To determine whether this constancy is maintained over a broad range of PM cholesterol concentrations , we incubated cells with or without compactin and then exposed them to varying levels of LDL to increase PM cholesterol prior to measurement of 125I-PFO* binding ( Figure 4 ) . In the absence of compactin treatment , 125I-PFO* binding was relatively high in the absence of LDL , and the binding rose only slightly when increasing amounts of LDL were added ( Figure 4A ) . SMase treatment increased 125I-PFO* binding at all concentrations of LDL , and the shape of the curve was parallel to the curve without SMase . When the cells were treated with compactin in the absence of LDL , 125I-PFO binding was very low ( Figure 4B ) , and it rose to a saturating value when increasing amounts of LDL were added . SMase treatment increased 125I-PFO* binding at all LDL concentrations , and the saturation curve paralleled the curve in the absence of SMase . Figure 4C plots the SM-sequestered pool as a function of LDL concentration . This pool is defined as the difference between 125I-PFO* binding in the presence and absence of SMase . Whether or not the cells had been treated with compactin , the SM-sequestered pool remained remarkably constant at all LDL concentrations . The only deviation occurred in the compactin-treated cells in the absence of LDL , where PM cholesterol levels were very low . In this extreme example of cholesterol depletion , the SM-sequestered pool declined slightly . Using the conversion factor defined by Figure 3 and the 125I-PFO* binding value of the increase in the PFO-accessible pool ( 9 µg/mg , Figure 4C ) , we estimate the SM-sequestered cholesterol pool in this experiment to be 12 mole% of PM lipids , which is somewhat lower than the 18% value calculated for Figure 3 . 10 . 7554/eLife . 02882 . 008Figure 4 . Effect of LDL on SM-sequestered pool of PM cholesterol . ( A and B ) On day 0 , SV-589 cells were setup in medium A at 1 × 105 cells per 60-mm dish . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium C containing 50 μM mevalonate in the absence ( A ) or presence ( B ) of 50 μM compactin . On day 4 , each monolayer received fresh medium D containing 50 μM mevalonate in the absence ( A ) or presence ( B ) of 50 μM compactin and the indicated concentration of LDL . After incubation for 5 hr at 37°C , the cells were washed twice with PBS at room temperature and treated with fresh medium D containing 50 μM mevalonate in the absence ( A ) or presence ( B ) 50 μM compactin and with or without 180 milliunits/ml SMase as indicated . After incubation for 1 hr at 37°C , the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml of ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 8 × 103 cpm/μg ) . After 2 hr at 4°C , the total amount of cell surface binding of 125I-PFO* was determined . Each value represents the mean of duplicate incubations . ( C ) Graph showing the SM-sequestered pool of cholesterol in the absence and presence of compactin plotted as a function of the concentration of LDL . The SM-sequestered pool was calculated by subtracting the amount of 125I-PFO* bound in the absence of SMase from that bound after SMase treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 008 To determine whether the increase in 125I-PFO* binding after SMase treatment requires vesicular movement , we pretreated the cells with paraformaldehyde to fix the PM prior to SMase treatment ( Brown et al . , 1976 ) . First , the cells were depleted of cholesterol by incubation with compactin , and then they were treated with SMase with or without prior paraformaldehyde treatment . In the absence of paraformaldehyde , SMase caused a large increase in 125I-PFO* binding . From the increase in 125I-PFO* binding in Figure 5A , we estimate the SM-sequestered cholesterol pool to be 13 mole% of PM lipids . This increase was undiminished after paraformaldehyde treatment ( red bars in Figure 5A ) . To show that paraformaldehyde was active , we incubated the cells with LDL and measured the incorporation of [14C]oleate into cholesteryl [14C]oleate ( Figure 5B ) . Stimulation of esterification requires the receptor-mediated endocytosis of LDL ( Brown and Goldstein , 1986 ) . In the absence of paraformaldehyde , LDL caused a major increase in cholesteryl ester synthesis , and this was totally blocked by paraformaldehyde . Paraformaldehyde did not inhibit the incorporation of [14C]oleate into [14C]triglycerides ( see legend to Figure 5 ) . 10 . 7554/eLife . 02882 . 009Figure 5 . Effect of paraformaldehyde on 125I-PFO* binding to human fibroblasts after treatment with SMase . On day 0 , SV-589 cells were setup in medium A at 1 × 105 cells per 60-mm dish . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium C containing 50 μM compactin and 50 μM mevalonate and incubated for 16 hr at 37°C . On day 4 , each monolayer was washed with ice-cold PBS at 4°C and treated with 2 ml ice-cold PBS in the presence or absence of 1% ( vol/vol ) paraformaldehyde as indicated . After incubation at 4°C for 45 min , the paraformaldehyde-containing medium was removed , and each monolayer was washed four times with buffer A at room temperature . ( A ) 125I-PFO* binding . After treatment with paraformaldehyde and washing , cells were incubated with 2 ml of fresh medium D containing 50 μM mevalonate and 50 μM compactin in the presence or absence of 100 milliunits/ml SMase as indicated . After incubation for 30 min at 37°C , the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml of ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 64 × 103 cpm/μg ) . After 2 hr at 4°C , the total amount of cell surface binding of 125I-PFO* was determined . ( B ) Cholesterol esterification . After treatment with paraformaldehyde and washing , each monolayer received 2 ml of fresh medium D containing 50 μM mevalonate and 50 μM compactin in the presence or absence of 50 μg protein/ml LDL . After incubation for 4 hr , 0 . 2 mM sodium [14C]oleate-albumin ( 7733 dpm/nmol ) was added to the cell and incubated for additional 2 hr . After the desired incubation , the cells were harvested and the content of cholesteryl [14C]oleate was determined . ( A and B ) Each bar represents the mean of triplicate incubations with the individual values shown . Paraformaldehyde treatment did not significantly affect the cellular uptake of [14C]oleate as indicated by parallel measurements of the incorporation of [14C]oleate into [14C]triglycerides . In the absence of paraformaldehyde , the amount of [14C]triglycerides formed ( nmol/mg protein per hr ) was 51 ( no addition ) and 43 ( +LDL ) ; in the presence of paraformaldehyde , the values were 57 ( no addition ) , and 32 ( +LDL ) . These triglyceride values represent the mean of triplicate incubations . In a separate experiment , the aforementioned mean triglyceride values were 74 , 46 , 57 , and 66 nmol/mg per hr , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 009 SMase treatment has been reported to cause cholesterol to translocate from the PM to the ER where it is esterified ( Slotte and Bierman , 1988; Scheek et al . , 1997; Abi-Mosleh et al . , 2009 ) . This predicted cholesterol loss from the PM conflicts with the increase in PFO-accessible cholesterol that is detected with 125I-PFO* . To resolve this issue , we performed a time course experiment ( Figure 6 ) . Human fibroblasts were treated with compactin to deplete cholesterol and then incubated with [14C]oleate in the absence or presence of SMase for varying times . After the incubation , one set of cells was harvested for measurement of cholesteryl [14C]oleate , and the other was chilled and incubated with 125I-PFO* . When incubated in the absence of SMase , the cells incorporated a very small amount of [14C]oleate into cholesteryl [14C]oleate , and 125I-PFO* binding was also low ( Figure 6A ) . In the presence of SMase ( Figure 6B ) , 125I-PFO* binding rose rapidly , reaching a peak at 15 min , after which it fell by about 20% . In contrast , the content of cholesteryl [14C]oleate was low at 15 min , and it rose progressively throughout the rest of the incubation . These data indicate that SM depletion first releases cholesterol into a PFO-accessible pool , and then the excess cholesterol is transported to the ER . The absolute amount of cholesteryl [14C]oleate formed at 2 hr ( 0 . 38 nmol/mg ) was much less than we generally observed when LDL is added ( for example , 7 nmol/mg per hr in Figure 5B ) . This indicates that only a small amount of cholesterol moves to the ER after SMase treatment . From the increase in 125I-PFO* binding in Figure 6B , we estimate the SM-sequestered cholesterol pool to be 10 mole% of PM lipids in this experiment . 10 . 7554/eLife . 02882 . 010Figure 6 . Time course of 125I-PFO* binding and cholesterol esterification in human fibroblasts after treatment with SMase . On day 0 , SV-589 cells were setup in medium A at 1 × 105 cells per 60-mm dish . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium C containing 50 μM compactin and 50 μM mevalonate and incubated for 16 hr at 37°C . On day 4 , each dish received fresh medium D containing 50 μM compactin and 50 μM mevalonate . Half of the dishes received 143 milliunits/ml of SMase in the absence ( ) or presence ( ) of 0 . 2 mM sodium [14C]oleate-albumin ( 7931 dpm/nmol ) for the indicated time . For 125I-PFO* binding ( ) , after the indicated incubation the cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml ice-cold buffer A containing 25 μg/ml 125I-PFO* ( 10 × 103 cpm/μg ) . After 2 hr at 4°C , the total cell surface binding of 125I-PFO* was determined , and the amount bound after subtraction of the zero-time value ( 1 µg/mg protein ) is plotted as ‘Increase in 125I-PFO* Bound’ . For measurement of cholesteryl [14C]oleate formation ( ) , the cells were harvested and the content of cholesteryl [14C]oleate was determined . Each value represents the average of duplicate incubations . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 010 The preceding data suggest that SMase treatment increases the transfer of LDL-derived cholesterol from the PM to the ER . To study this transfer process more directly , we first depleted cells of cholesterol by treatment with compactin . Next , we treated the cells with SMase and then added cholesterol directly to the PM by delivering it in a complex with methyl-β-cyclodextrin ( MCD ) . Transfer to ER was assessed by measurement of [14C]oleate incorporation into cholesteryl [14C]oleate . After SMase treatment , the cholesterol-stimulated esterification reaction was markedly increased whether examined over time ( Figure 7A ) or as a function of the concentration of cholesterol/MCD ( Figure 7B ) . 10 . 7554/eLife . 02882 . 011Figure 7 . Prior incubation of human fibroblasts with SMase stimulates transport of cholesterol from PM to ER . On day 0 , SV-589 cells were set up in medium A at 1 × 105 cells per 60-mm dish . On day 2 , cells were switched to lipoprotein-deficient medium C . On day 3 , cells were treated with fresh medium C containing 50 μM compactin and 50 μM mevalonate and then incubated for 16 hr at 37°C . On day 4 , each monolayer received 2 ml of fresh medium C containing 50 μM mevalonate and 50 μM compactin in the absence or presence of 100 milliunits/ml of SMase as indicated . After incubation with SMase for 15 min at 37°C , the cells in ( A ) received a direct addition of 50 cholesterol/MCD together with 0 . 2 mM sodium [14C]oleate-albumin ( 8572 dpm/nmol ) and were then incubated for the indicated time , and the cells in ( B ) received a direct addition of the indicated concentrations of cholesterol/MCD together with 0 . 2 mM sodium [14C]oleate-albumin ( 6948dpm/nmol ) and were then incubated for 1 hr . After the indicated incubations , the cells were harvested and the content of cholesteryl [14C]oleate was determined . Each value represents the average of duplicate incubations . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 011 Figure 8 shows an experiment in which we compared the threshold for 125I-PFO* binding to cells treated without and with SMase . For this purpose , cholesterol was removed from cells by exposure to concentrations of 2-hydroxypropyl-β-cyclodextrin ( HPCD ) ranging from 0% to 2% . Half of the dishes were then treated with SMase , after which 125I-PFO* binding was measured . A parallel set of dishes was harvested for measurement of the cholesterol concentration in isolated PMs . In the absence of SMase treatment , 125I-PFO* binding decreased substantially when PM cholesterol was reduced below 45 mole% of total PM lipids ( black curve ) . In the absence of cholesterol depletion ( 0% HPCD ) , SMase treatment increased 125I-PFO* binding by 18 µg/mg protein . Using the conversion factor defined by Figure 3 and the 125I-PFO* binding value of the increase in the PFO-accessible pool ( 18 µg/mg ) , we estimate the SM-sequestered cholesterol pool to be 23 mole% of PM lipids . This SM-sequestered pool declined gradually as the concentration of HPCD increased ( red curve ) . When PM cholesterol fell to 25 mole% of total PM lipids , the SM-sequestered pool had declined by 80% . The residual cholesterol was in a pool that was not accessible to 125I-PFO* even after SMase treatment . We call this pool the ‘essential’ pool ( ‘Discussion’ ) . This ‘essential pool’ may be related to the fraction of PM cholesterol in fibroblasts that fails to be modified by cholesterol oxidase after prior treatment with SMase ( Pörn and Slotte , 1995 ) . 10 . 7554/eLife . 02882 . 012Figure 8 . Reduced threshold for 125I-PFO* binding after treatment of human fibroblasts with SMase . On day 0 , SV-589 cells were set up in medium A at 1 × 105 cells per 60-mm dish . On day 3 , cells were refed with medium B . On day 4 , cells were treated for 1 hr at 37°C in medium B containing 50 µM mevalonate , 50 µM compactin , and 0–2% of HPCD as indicated for each point . The cells were then washed twice with PBS at room temperature , after which each monolayer received fresh medium B containing 50 µM compactin and 50 µM mevalonate in the absence or presence of 100 miliunits/ml of SMase . After incubation for 30 min at 37°C , one set of dishes was used for 125I-PFO* binding and a parallel set was used for PMs purification . For 125I-PFO* binding , cells were washed five times as described in ‘Materials and methods’ and then incubated with 2 ml ice-cold buffer A containing 25 µg/ml 125I-PFO* ( 10 × 103 cpm/µg ) for 2 hr at 4°C . The total amount of cell surface 125I-PFO* binding was determined . Each value represents the average of duplicate incubations . For PM purification , six 60-mm dishes for each HPCD treatment were pooled together , and the PMs were purified as described in ‘Materials and methods’ . Lipids were extracted from the PM samples , and the content of unesterified cholesterol , phospholipids , and ceramide was measured . Each value represents the average of duplicate measurements of each pooled sample . The graph shows the amount of cell surface 125I-PFO* binding plotted as a function of the cholesterol content of the PM . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 012
The current studies help to resolve a conundrum in cellular cholesterol regulation that has existed ever since the discovery of SREBPs 20 years ago . How does a transcription factor residing in the cholesterol-poor ER membrane regulate the cholesterol content of the cholesterol-rich plasma membrane ( PM ) ? Here , we show how the controlled distribution of LDL-derived cholesterol to PM and ER accomplishes this regulatory task . As long as the newly expanding LDL-derived and PFO-accessible PM cholesterol pool remains low , little cholesterol moves to the ER , SREBPs are transported to the Golgi for processing , and the cell acquires cholesterol from endogenous synthesis and uptake of LDL . As this PM pool expands , more cholesterol moves to the ER . Once this ER cholesterol surpasses a 5 mole% threshold , it blocks SREBP transport , reducing the cellular accumulation of cholesterol ( Radhakrishnan et al . , 2008 ) . In the ER , excess cholesterol is esterified for storage as cholesteryl esters , further preventing the accumulation of excess cholesterol in the PM . Our analysis of the current results suggests that the PM contains three pools of cholesterol , as illustrated in Figure 9 . One pool is accessible to 125I-PFO* binding ( shown in green ) . In LDL-treated cells , the PFO-accessible pool accounts for 16% of PM lipids . The second pool , shown in red , is inaccessible to 125I-PFO , but becomes accessible when cells are treated with SMase . The size of this SM-sequestered pool , as determined from five independent experiments ( Figures 3–6 , 8 ) , ranges from 10 to 23% of total PM lipids , with a mean of 15% . The third pool , shown in purple , is termed the ‘essential pool’ . This pool was also determined from the aforementioned five experiments; it accounts for 12% of PM lipids and is not accessible to 125I-PFO* , even after SMase treatment . We call it ‘essential’ because reducing this pool with a high concentration of HPCD ( >2% ) causes cells to round up and dissociate from the petri dish , which indicates that this pool is essential for PM integrity . Although all of the results in the current study were done with human fibroblasts , we also found that the PM of CHO cells contains three pools of cholesterol ( Figure 3—figure supplement 1 ) . Whether these findings apply to other cell types grown under various conditions needs to be tested in future studies . 10 . 7554/eLife . 02882 . 013Figure 9 . Schematic diagram illustrating the three pools of cholesterol in the PM of human fibroblasts under different conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 02882 . 013 When cells are depleted of sterols by treatment with compactin in lipoprotein-deficient serum , the PFO-accessible pool becomes depleted ( Figure 3 ) . Even though the other two pools of cholesterol remain in compactin-treated cells , there is no transport to the ER and no cholesterol esterification . When sterol-depleted cells are treated with SMase , the SM-sequestered pool is released , and the liberated cholesterol becomes accessible for 125I-PFO binding ( Figure 3A ) , restoring the PFO-accessible pool . Under these conditions even without the addition of LDL , a small amount of cholesterol is transported from PM to ER where it is esterified ( Figure 6B ) . The accessibility of only a portion of PM cholesterol to PFO is consistent with earlier work by Lange et al . ( 1980 ) , who showed that only a portion of PM cholesterol is susceptible to the enzyme cholesterol oxidase . It seems likely that both PFO and cholesterol oxidase recognize the same pool of accessible cholesterol in the PM . While our work does not reveal the molecular basis of cholesterol accessibility , it is worth noting that studies on model lipid membranes have suggested a thermodynamic basis for this pool of accessible cholesterol ( Radhakrishnan and McConnell , 2000; Lange et al . , 2013 ) . According to these studies , at low concentrations of cholesterol , most of the cholesterol is tied up in complexes with phospholipids and would not be accessible to soluble proteins such as PFO or cholesterol oxidase . At higher concentrations of cholesterol , the bilayer phospholipids become limiting , and a pool of cholesterol emerges that is not tied up in complexes with phospholipids . It is this ‘free’ cholesterol that is accessible to PFO ( or cholesterol oxidase ) . The free cholesterol concentration is related to thermodynamic quantities termed chemical activity or fugacity ( Radhakrishnan and McConnell , 2000 ) . The portion of PM cholesterol that is inaccessible to extracellularly added PFO could theoretically be sequestered either in the extracellular or in the cytoplasmic PM leaflets by membrane phospholipids . This PFO-inaccessible pool is subdivided into the SM-sequestered pool and the essential pool . Loss of SM upon SMase treatment increases the chemical activity of cholesterol in the outer leaflet of the PM and expands the pool of PFO-accessible cholesterol . Some of this accessible cholesterol could be transferred to the inner leaflet of the PM by cholesterol flip-flop , but there is no mechanism in the PM to convert this excess cholesterol into an inert form as ACAT does in the ER membrane . Instead , the excess PM cholesterol is transported to the ER membrane by vesicular or non-vesicular pathways to shut down SREBP processing and to be converted into cholesteryl esters by ACAT ( Scheek et al . , 1997; Abi-Mosleh et al . , 2009 ) . The physical nature of the SM-sequestered pool of cholesterol is not known , although a model of cholesterol-SM complexes is plausible based on studies in model membranes ( Radhakrishnan and McConnell , 2005; Lönnfors et al . , 2011; Lange et al . , 2013 ) . The idea of complexes of cholesterol and phospholipids is not new: more than 60 years ago , an X-ray diffraction study of nerve fibers suggested a 1:1 complex between cholesterol and phospholipid as the structural basis of the myelin sheath ( Finean , 1953 ) . Higher-order structures such as the proposed lipid raft domains rich in SM and cholesterol provide an additional mechanism for cholesterol sequestration . The existence of an essential pool of PM cholesterol is inferred from the data of Figure 8 in which PM cholesterol was depleted by incubation of the cells with varying concentrations of HPCD . At the highest HPCD concentration tested ( 2% ) , the PFO-accessible pool was totally depleted , and the SM-sequestered pool was reduced by 80% ( red curve in Figure 8 ) . At this point , PM cholesterol was reduced from the control value of 45–22% of total PM lipids . Despite this severe reduction in PM cholesterol , the cells appeared normal by phase contrast light microscopy . When we increased the HPCD concentration above 2% , the morphology of the cells was drastically altered . The cells rounded up and many of them dissociated from the petri dish . Thus , the PM cholesterol that is retained at 2% HPCD is essential for normal PM morphology . Inasmuch as most of this residual cholesterol is neither PFO-accessible nor SM-sequestered , we term it the essential pool . The physical nature of this pool is unknown , and it may well represent multiple subpools of cholesterol in complex with PM phospholipids other than SM . The further delineation of the molecular nature of all three pools of PM cholesterol in fibroblasts and other cell types is a challenge for future membrane structural biology .
We obtained [1-14C]oleic acid ( 55 mCi/mmol ) from American Radiolabeled Chemicals , St . Louis , MO; paraformaldehyde from Electron Microscopy Sciences , Hatfield , PA; S . aureus SMase from Sigma , St . Louis , MO; and monoclonal anti-His antibody from GE Healthcare , Pittsburgh , PA . All other reagents ( tissue culture supplies , 2-hydroxypropyl-β-cyclodextrin ( HPCD ) , methyl-β-cyclodextrin ( MCD ) , 125I-NaI , LDL , lipoprotein-deficient serum , and stock solutions of sodium mevalonate and compactin were obtained from sources or prepared as previously described ( Das et al . , 2013 ) . A stock solution of cholesterol/MCD complexes was prepared at a final concentration of 2 . 5 mM and a cholesterol/MCD ratio of 1:10 ( Brown et al . , 2002 ) . Buffer A contains 25 mM Hepes-KOH ( pH 7 . 4 ) , 150 mM NaCl , and 0 . 2% ( wt/vol ) bovine serum albumin . Medium A is DMEM ( with L-glutamine ) containing 100 units/ml of penicillin , 100 μg/ml streptomycin sulfate , and 10% ( vol/vol ) FCS . Medium B is DMEM ( with L-glutamine ) containing 100 units/ml penicillin , 100 μg/ml streptomycin sulfate , and 1% ( vol/vol ) Insulin-Transferrin-Selenium . Medium C is DMEM ( with L-glutamine ) containing 100 units/ml penicillin , 100 μg/ml streptomycin sulfate , and 5% ( vol/vol ) newborn calf lipoprotein-deficient serum . Media D and E are identical to media C and B , respectively , except for the absence of L-glutamine in the DMEM in media D and E . Medium F is 1:1 mixture of Ham's F-12 medium and DMEM ( with L-glutamine ) containing 100 units/ml penicillin , 100 μg/ml streptomycin sulfate , and 5% ( vol/vol ) FCS . Medium G is 1:1 mixture of Ham's F-12 medium and DMEM ( with L-glutamine ) containing 100 units/ml penicillin , 100 μg/ml streptomycin sulfate and 5% ( vol/vol ) newborn calf lipoprotein-deficient serum . Medium H is DMEM ( without L-glutamine ) containing 100 units/ml penicillin and 100 μg/ml streptomycin sulfate . Stock cultures of human SV-589 fibroblasts ( Yamamoto et al . , 1984 ) were grown in monolayer at 37°C in a 5% CO2 incubator and maintained in medium A . Stock cultures of hamster CHO-K1 and CHO-7 ( Metherall et al . , 1989 ) were grown in monolayer culture at 37°C in a 8–9% CO2 incubator and maintained in medium F and G , respectively . PFO refers to the fully active cytolytic form of the toxin ( Flanagan et al . , 2009 ) ; PFO* refers to a mutant PFO in which tyrosine-181 was changed to alanine , yielding a version that is not cytolytic at 4°C ( Das et al . , 2013 ) . Both PFO ( Sokolov and Radhakrishnan , 2010 ) and PFO* ( Das et al . , 2013 ) contained His6 tag at the NH2-terminus . The proteins were overexpressed in Escherichia coli and purified as described in the indicated reference . PFO* was radiolabeled with 125I as previously described ( Das et al . , 2013 ) . The procedure for purification of PMs from SV-589 cells was carried out by cell surface biotinylation followed by streptavidin affinity chromatography as previously described ( Das et al . , 2013 ) . ER membranes from SV-589 cells were purified by differential gradient centrifugation as previously described ( Radhakrishnan et al . , 2008 ) . Prior to addition of 125I-PFO* , cells were washed as follows to remove surface-bound lipoproteins or HPCD: three rapid washes with buffer A at room temperature , followed by two 10-min washes with the ice-cold buffer A in a 4°C cold room . After these five washes , each 60-mm dish of cells was incubated at 4°C with 2 ml of buffer A containing 125I-PFO* as described in Legends . After the indicated time , cell monolayers were washed rapidly three times with ice-cold PBS , dissolved with 1 ml of 0 . 1 N NaOH , and shaken on a rotary shaker for 15 min at room temperature . Aliquots ( 500 μl ) of the dissolved cells were removed for scintillation counting in a gamma counter and for measurement of protein concentration ( 50 μl ) ( Lowry et al . , 1951 ) . The data are expressed as μg 125I- PFO* bound per mg cell protein . Each 100-μl reaction mixture contained 0 . 5 μg of PFO and either purified ER membranes ( 12 μg protein ) or purified PMs ( 90 μg protein ) . After incubation for 1 hr at 37°C , each mixture was combined with 5 × SDS loading buffer , incubated for 10 min at room temperature , and then subjected to 10% SDS-PAGE , followed by immunoblot analysis using an anti-His antibody ( 2 . 6 μg/ml ) . Cell lysates were divided in half , and the PMs from both halves were purified as described above . For one half , the content of unesterified cholesterol and choline-containing phospholipids was determined as previously described ( Das et al . , 2013 ) . Total phospholipids were estimated by multiplying the measured choline content by 1 . 53 to account for the proportion of phospholipids that do not contain choline ( Das et al . , 2013 ) . The second half was used to measure SM and ceramide content . The purified membranes were homogenized by sonication in 1 . 5 ml of 25 mM HEPES ( pH 6 . 8 ) buffer . Immediately afterward , 20 μl of a Ceramide/Sphingolipid Internal Standard Mixture II ( Avanti Polar Lipids , Alabaster , AL ) was added , and the mixtures were vortexed and sonicated at 40°C for 10 min . Lipids from the homogenate were extracted with 2 ml of 85:15 ( vol/vol ) ethyl acetate:isopropanol . SM and ceramide levels were resolved and detected using high-performance liquid chromatography ( HPLC ) coupled to a triple quadrupole mass spectrometer ( MS; ABSciex , Framingham , MA ) through an electrospray ionization interface ( Shaner et al . , 2009; Hammad et al . , 2010 ) . The mole% of individual lipid classes is defined as the moles of the indicated lipid divided by the sum of the moles of cholesterol , phospholipids , and ceramide . The total moles of the four most abundant species of SM and of ceramide , as determined by MS analysis , are referred to as total SM and total ceramide , respectively ( Table 1 ) . The rate of incorporation of [14C]oleate into cholesteryl [14C]esters and [14C]triglycerides by monolayers of SV-589 cells was measured as previously described ( Goldstein et al . , 1983 ) . | Cells are enclosed by a plasma membrane that is made of lipid molecules and proteins . Almost half of the lipid molecules in the plasma membranes of animal cells ( including human cells ) are cholesterol molecules . Since cholesterol helps to keep the membrane stable , its level in the plasma membrane is tightly regulated . Cholesterol is produced within animal cells , but it can also be taken up from outside the cell , primarily from low density lipoprotein ( or LDL for short ) . When ingested LDL is broken down inside a cell , most of the cholesterol molecules are added to the plasma membrane , but some end up in the membrane of an organelle inside the cell called the endoplasmic reticulum . The amount of cholesterol in the membrane of the endoplasmic reticulum regulates the activation of a protein called SREBP , a transcription factor that is attached to this membrane . If the level of cholesterol becomes too low , this transcription factor travels to the cell nucleus , where it switches on the genes that cause the cell to produce more cholesterol and also to take up more LDL from the environment . When the amount of cholesterol in the membrane is high enough , the SREBP protein remains attached to the endoplasmic reticulum , which reduces the production of cholesterol and the uptake of LDL . Here , Das et al . study the movement of cholesterol molecules between the plasma membrane and the membrane of the endoplasmic reticulum by using a toxin that binds to membranes that are rich in cholesterol . These experiments showed that the plasma membrane contains three different types or ‘pools’ of cholesterol . Das et al . found that only one of these pools is ‘labile’: it grows when there is an excess of LDL , and shrinks when cholesterol is running low in the cell . Furthermore , excess cholesterol is first added to this labile pool in the plasma membrane before it is added to the pool in the endoplasmic reticulum . This suggests that the production , uptake , and breakdown of cholesterol are all controlled by partitioning this molecule between the labile pool in the plasma membrane and the endoplasmic reticulum . The next challenge is to determine how the three pools of cholesterol in the plasma membrane are maintained , and what regulates the distribution of cholesterol between the endoplasmic reticulum and the labile pool in the plasma membrane . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2014 | Three pools of plasma membrane cholesterol and their relation to cholesterol homeostasis |
Mitochondrial fusion and fission affect the distribution and quality control of mitochondria . We show that Marf ( Mitochondrial associated regulatory factor ) , is required for mitochondrial fusion and transport in long axons . Moreover , loss of Marf leads to a severe depletion of mitochondria in neuromuscular junctions ( NMJs ) . Marf mutants also fail to maintain proper synaptic transmission at NMJs upon repetitive stimulation , similar to Drp1 fission mutants . However , unlike Drp1 , loss of Marf leads to NMJ morphology defects and extended larval lifespan . Marf is required to form contacts between the endoplasmic reticulum and/or lipid droplets ( LDs ) and for proper storage of cholesterol and ecdysone synthesis in ring glands . Interestingly , human Mitofusin-2 rescues the loss of LD but both Mitofusin-1 and Mitofusin-2 are required for steroid-hormone synthesis . Our data show that Marf and Mitofusins share an evolutionarily conserved role in mitochondrial transport , cholesterol ester storage and steroid-hormone synthesis .
Mitochondrial dynamics plays a critical role in the control of organelle shape , size , number , function and quality control of mitochondria from yeast to mammals ( Westermann , 2009; Chan , 2012 ) . It consists of fusion and fission of mitochondria , which are regulated by several GTPases ( van der Bliek et al . , 2013 ) . Mitochondrial fusion requires the fusion of the outer membrane followed by inner membrane fusion ( Chan , 2012; Mishra et al . , 2014 ) . In mammals , Mitofusin 1 ( Mfn1 ) and Mitofusin 2 ( Mfn2 ) regulate outer mitochondrial fusion whereas inner membrane fusion is controlled by Optic atrophy protein 1 ( Opa1 ) . Mitochondrial fission is regulated by Dynamin related protein 1 ( Drp1 ) ( van der Bliek et al . , 2013 ) . Decreased fusion results in fragmented round mitochondria , while defective fission leads to fused and enlarged mitochondria ( van der Bliek et al . , 2013 ) . Loss of these mitochondrial GTPases results in lethality in worms , flies and mice ( Chen et al . , 2003; Westermann , 2009; Debattisti and Scorrano , 2012 ) . Mutations in the human DRP1 gene causes a dominant fatal infantile encephalopathy associated with defective mitochondrial and peroxisomal fission ( Waterham et al . , 2007 ) . On the other hand , missense mutations in OPA1 lead to a dominant optic atrophy ( Alexander et al . , 2000; Delettre et al . , 2000 ) . Depending on the severity of the mutation , patients may also suffer from ataxia and neuropathy ( Yu-Wai-Man et al . , 2010 ) . Also , missense mutations in MFN2 cause Charcot-Marie-Tooth type 2A , a common autosomal dominant peripheral neuropathy associated with axon degeneration ( Zuchner et al . , 2004 ) . Finally , aberrant levels of mitochondrial GTPases have been associated with Parkinson's , Huntington's and Alzheimers' diseases ( Itoh et al . , 2012 ) . These observations in model organisms and human patients suggest that mitochondrial dynamics affects neuronal maintenance in many different contexts . A significant imbalance of mitochondrial fission and fusion may affect the subcellular distribution of mitochondria , especially in neurons since they need to efficiently traffic from the soma to the synapses ( Sheng , 2014 ) . Loss of Drosophila Drp1 impairs the delivery of mitochondria to neuromuscular junctions ( NMJs ) , likely because they are large and interconnected . This defect is also associated with a severe depletion of mitochondria in NMJs , which affects local ATP production . This in turn affects the trafficking of synaptic vesicles upon endocytosis during prolonged stimulation ( Verstreken et al . , 2005 ) . Similarly , in vertebrates , loss of Drp1 leads to an accumulation of mitochondria in the soma and reduced mitochondrial density in dendrites of hippocampal neurons ( Li et al . , 2004 ) . The Drp1 data in flies and vertebrates indicate that the expanded size of mitochondria affects their mobility ( Sheng , 2014 ) . Mitochondrial trafficking may also be affected by the physical interaction between the mitochondria and the transport machinery . Recent studies have documented a direct interaction between Mfn2 and a motor adaptor complex for mitochondrial transport , Miro2 ( Misko et al . , 2010 ) . Moreover , loss of MFN2 in Purkinje cells displayed reduced mitochondrial motility in cerebellar dendrites ( Chen et al . , 2007 ) and reduced mitochondrial transport in axons in cultured dorsal root ganglion neurons ( Misko et al . , 2010 ) . These data suggest that an interaction of Mfn2 with Miro2 may be important for its role in trafficking ( Misko et al . , 2010 ) . Although loss of both Drp1 and MFN2 impair mitochondrial trafficking , a careful comparison of the phenotypes associated with loss of Drosophila Drp1 , Mitofusin or Marf , would be useful as the suggested mechanisms by which they impair transport seem very different . In addition to their roles in fission and fusion , Drp1 , Mfns and Opa1 have been implicated in a variety of other processes . For example , Drp1 has been shown to facilitate the induction of apoptosis ( Frank et al . , 2001 ) whereas Opa1 was shown to affect the stability of cristae junction in inner mitochondrial membrane ( Frezza et al . , 2006 ) . Finally , Mfn2 also tethers mitochondria to the endoplasmic reticulum ( ER ) to mediate Ca2+ uptake ( de Brito and Scorrano , 2008 ) . However , the molecular mechanisms underlying these non-canonical functions are less well studied . In an unbiased screen designed to identify essential genes that affect neuronal function ( Yamamoto et al . , 2014 ) , we identified the first mutant allelic series of Marf in Drosophila . Here we exploit these mutants to determine how loss of Marf affects mitochondrial transport when compared to Drp1 loss . Surprisingly , we observe NMJ defects only in Marf mutants but not in Drp1 mutants . These defects are regulated non-cell autonomously by steroid-hormones produced in ring glands ( RG ) , a major endocrine organ in insects . Through expression of human MFN1 or MFN2 in Marf mutant RG , we show that MFN1 and MFN2 have both distinct and complementary roles .
Through a forward genetic screen on the Drosophila X-chromosome ( Yamamoto et al . , 2014 ) we isolated seven independent lethal alleles of Marf that affect electroretinogram ( ERG ) recordings in homozygous mutant clones ( Figure 1A , C , Figure 1—figure supplement 1 ) . The on- and off-transients ( Figure 1A , red arrows ) of the ERG are a read-out of synaptic transmission between photoreceptors ( PR ) and postsynaptic cells , while the amplitude of the depolarization ( Figure 1A , green bracket ) is a measure of the function of the phototransduction cascade ( Wang and Montell , 2007 ) . The Marf mutations vary in strength ( Figure 1A , E and Figure 1—figure supplement 1B ) , providing an allelic series . ERG recordings in homozygous mutant eye clones reveal a reduction in on- and off-transients as well as loss of amplitude in one day old flies ( Figure 1A ) . The ERG recordings differ from Drp1 mutants that only exhibit a loss of on- and off-transients but a normal amplitude ( Figure 1B , [Verstreken et al . , 2005] ) . In summary , loss of Marf severely impairs the phototransduction cascade as well as synaptic transmission , whereas loss of Drp1 mainly affects synaptic transmission of PRs . 10 . 7554/eLife . 03558 . 003Figure 1 . Loss of Marf impairs phototransduction and affects mitochondrial localization to photoreceptor terminals . ( A ) Electroretinograms ( ERGs ) of 1 day old ey-FLP mutant clones of 7 different Marf mutants or isogenized wild type clones ( Control ) . ERGs of Marf mutant alleles and control flies . A typical ERG trace is comprised of an on-transient ( red arrow ) , a depolarization ( green bracket ) and an off-transient ( red arrow ) . ( B ) ERGs of Drp1 mutants and control flies . ( C ) Marf protein domains and localization of EMS-induced mutations of the seven Marf mutant alleles identified by sequencing . H494fs93 = insertion of an A at nucleotide codon for amino acid H494 that generates 93 new amino acids followed by a premature stop codon . TM = transmembrane domain . HR = heptad repeat . ( D ) TEM sections of a cartridge containing fly photoreceptor terminals ( green shading ) . Marf mutant photoreceptor terminals display reduced number and size of mitochondria ( yellow arrow heads ) compared to Marf-genomic rescue controls . ( E ) Quantification of total mitochondria number per cartridge in Marf mutants and Marf-genomic rescue photoreceptor terminals ( Control ) . 50 cartridges per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 00310 . 7554/eLife . 03558 . 004Figure 1—figure supplement 1 . Mapping , lethal staging and Marf protein expression of Marf mutant alleles . ( A ) For mapping of Marf , the lethality of all Marf alleles were rescued by large duplication Dp ( 1;Y ) dx[+]5 , y[+]/C ( 1 ) M5 ( 4C11;6D8 + 1A1;1B4 ) covering the Marf locus . A 6 . 1 kb genomic rescue fragment encompassing the Marf locus was used to generate a Marf-HA tagged genomic construct ( Marf-gHA ) to rescue the Marf alleles . ( B ) Lethal staging analysis of Marf mutant alleles and lethality rescue by Marf-gHA , UAS-Marf-HA , UAS-MFN1 , UAS-MFN2 and UAS-MFN1/UAS-MFN2 cDNA constructs . ( C ) Marf Western blot ( Ziviani et al . , 2010 ) from MarfB , MarfA , MarfG , Marf-Genomic ( Marf-gHA ) and ubiquitous ( Actin-Gal4 ) Marf knockdown in third instar larvae . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 004 Lethal staging shows that most Marf mutants ( A , B , E , F and G ) die as third instars after a very extended larval stage period of 18–21 days , which typically takes 6 days in wild type animals ( Figure 1—figure supplement 1B ) . The lethality of all Marf mutants is rescued by a Marf genomic DNA construct or by a ubiquitously expressed Marf cDNA ( Figure 1—figure supplement 1A ) , showing that the Marf mutations are responsible for the lethality ( Figure 1—figure supplement 1B ) . Moreover , transheterozygous MarfB/Df ( 1 ) Exel6239 female mutants display the same lethal phase as MarfB/Y males , suggesting that MarfB is likely to be a severe loss of function allele or null allele ( Figure 1—figure supplement 1B ) . Finally , MarfB hemizygous males exhibit a severe protein loss compared to MarfG hemizygous males and controls ( Figure 1—figure supplement 1C ) , suggesting that this missense mutation in the GTPase domain ( Figure 1C ) also destabilizes the protein . Since mitochondrial transport has been shown to be affected in some neurites of MFN2-deficient vertebrate cells ( Chen et al . , 2007 ) , we performed Transmission Electron Microscopy ( TEM ) at the PR terminals . Marf mutants exhibit a very severe loss of mitochondria ( Figure 1D , yellow arrows ) in PR terminals when compared to control ( Figure 1D , E ) . The severity of the loss of mitochondria ( Figure 1E ) correlates with the loss of neuronal function gauged by ERGs ( Figure 1A ) . These data are reminiscent of the documented lack of mitochondria in PR terminals in Drp1 mutants ( Verstreken et al . , 2005 ) . However , the mitochondria in Marf mutant PRs are significantly smaller in size than controls ( Figure 1D , yellow arrows ) , suggesting that an active transport mechanism is impaired . To assess if mitochondrial size is also affected in mutant muscles , we stained Marf and Drp1 ( Figure 2—source data 1 ) mutants with an anti-mitochondrial complex V antibody ( ATP5A ) ( Baqri et al . , 2009 ) . As expected , Drp1 mutants have filamentous mitochondria whereas Marf mutants have small , rounded mitochondria ( Figure 2A and Figure 2—source data 2 ) . However , both Marf and Drp1 mutant mitochondria produce similar reduced levels of ATP when compared to controls ( Figure 2C and Figure 2—source data 2 ) . Interestingly , the mitochondrial membrane potential ( MMP ) of Drp1 mutants as measured with tetramethylrhodamine ethyl ester ( TMRE ) ( Scaduto and Grotyohann , 1999 ) is slightly elevated , as reported before ( Verstreken et al . , 2005 ) , when compared to controls whereas MMP of Marf mutants is reduced ( Figure 2B and Figure 2—source data 2 ) . Measurements of the activity of the Electron Chain Complexes ( ETC I , II , III and IV ) that pump protons across the mitochondrial inner membrane from the mitochondrial matrix to the inner membrane space to generate the MMP revealed that all ETC complex activities are similarly or more severely affected in Marf than Drp1 mutants ( Figure 2D ) . Furthermore , measurement of reactive oxygen species ( ROS ) by dihydroethidium ( DHE ) staining ( Shidara and Hollenbeck , 2010 ) and mitochondrial aconitase assay ( native activity of aconitase negatively correlates with ROS levels ) ( Yan et al . , 1997 ) shows that Marf mutants are significantly more severely affected than Drp1 mutants ( Figure 2E , F and Figure 2—source data 2 ) . The ROS data is in agreement with the ETC data as loss of function of CI and CIII are considered the major drivers of increased ROS ( Koopman et al . , 2013 ) . In summary , Marf and Drp1 mutants exhibit dysfunctional mitochondria , but loss of Marf affects their function more severely . 10 . 7554/eLife . 03558 . 005Figure 2 . Mitochondrial morphology and function in Marf and Drp1 mutants . ( A ) Mitochondrial morphology based on anti-Complex V antibody staining ( Complex V ) in larval muscles ( Zoom in view around muscle nucleus ) . ( B ) Mitochondrial membrane potential as measured by the TMRE dye in larva muscle . ( C ) Relative ATP amounts . ( D ) Measurement of the enzymatic activity of electron transport chain ( ETC ) complexes ( I–IV ) from purified mitochondria from third instar larvae . All the ETC activities were normalized to citrate synthase ( CS ) activity of controls . ( E and F ) ROS is measured by two methods: ( E ) by DHE staining in larval muscles and ( F ) by measuring aconitase activity reduction from purified mitochondria . Reducing reagents reactivate native aconitase . Aconitase activities were normalized to controls . ( C , D and F ) error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 00510 . 7554/eLife . 03558 . 006Figure 2—source data 1 . Lethal staging of Drp1 mutants . Lethal staging of Drp1 transheterozygous combinations of Drp1KG38015 , Drp1[T26] and Drp1 with Drp12 mutant alleles . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 00610 . 7554/eLife . 03558 . 007Figure 2—source data 2 . Phenotypic comparison of Marf , Drp1 and Marf and Drp1 mutants . Phenotypic comparison table of Marf , Drp1 and Marf and Drp1 mutants in mitochondria morphology , mitochondria membrane potential ( MMP ) , ATP levels , ROS ( DHE ) intensity , bouton numbers and 20-hydroxyecdysone ( 20E ) levels . Figure 2B MarfB panel has both puncta globular ( P ) and non-puncta ( NP ) staining that were both used to measure MMP . MMP , ATP levels and ROS intensity were normalized to controls and all columns are representative of three independent experiments ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 007 Loss of one copy of MFN2 in human causes a progressive and severe loss of function of neurons with long axons and affects motor neurons ( MN ) more severely than sensory neurons ( Zuchner et al . , 2004 ) . To assess if mitochondria in MN are affected in larvae we expressed MitoGFP in MN using the D42-Gal4 driver ( Pilling et al . , 2006 ) . In the ventral nerve cord ( VNC ) of control larvae , MitoGFP mostly localizes to the neuropil ( Figure 3A ) . Marf mutants show an obvious reduction in levels of mitochondria in the neuropil and the mitochondria mostly form clumps in the soma and the initial segments of axons ( Figure 3A ) . In control MN , MitoGFP also labels numerous mitochondria in axons that innervate proximal ( A3 ) and more distal ( A5 ) segments ( Figure 3B ) . In the axons of Marf mutants , fewer MitoGFP-marked mitochondria are observed in distal axons compared to controls ( Figure 3B ) . These data show that loss of Marf impairs , but does not abolish , axonal mitochondrial transport ( Figure 3B ) . 10 . 7554/eLife . 03558 . 008Figure 3 . Mitochondrial trafficking defects in distal axons and boutons . Mutations and controls were crossed to a motor neuron driver ( D42-GAL4 , UAS-MitoGFP ) to label neuronal mitochondria . ( A ) Ventral nerve cord ( VNC ) : Marf and Drp1 mutants exhibit clustered mitochondria in the soma . ( B ) Comparison of a proximal axonal segment in A3 and a distal segment in A5 . Distal segments of A5 axons in Marf mutants contain many fewer mitochondria than proximal segments . ( C ) Marf mutants contain almost no mitochondria in boutons when co-stained with post-synaptic marker Discs Large 1 ( Dlg1 ) . Percentage of boutons with no mitochondria: Genomic rescue ( 0% ) , Marf B ( 89% ) , UAS-Marf ( 0% ) , Drp12 ( 36% ) and Marf B;Drp12 ( 95% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 00810 . 7554/eLife . 03558 . 009Figure 3—figure supplement 1 . Pre-synaptic , endocytic and postsynaptic markers are present in Marf mutant boutons . A panel of different NMJ markers co-stained with Dlg1: ( A ) Bruchpilot ( Brp ) , ( B ) α-Adaptin , ( C ) Glutamate receptor IIa ( GluRIIa ) , ( D ) Dap160 , ( E ) Hrp , ( F ) Endophilin , ( G ) Synaptojanin and ( H ) Drosophila vesicular glutamate transporter ( DV-Glut ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 00910 . 7554/eLife . 03558 . 010Figure 3—figure supplement 2 . Mitochondrial trafficking defect in Marf mutants cannot be rescued by motor neuron expression of human MFN1 or MFN2 . Mutations and controls were crossed to a motor neuron ( MN ) driver ( D42-GAL4 , UAS-mitoGFP ) to label neuronal mitochondria . ( A ) Ventral nerve cord ( VNC ) , MN-knockdown of dmiro in Marf mutant exhibit more clustered mitochondria in the soma compared to Marf alone , while neither MN-expression of MFN1 or MFN2 rescued the VNC mitochondrial trafficking defect of Marf mutants . ( B ) At the proximal end of the A3 axon , MN-knockdown of dmiro in Marf mutants had severed reduction of mitochondrial trafficking compared to Marf alone . ( C ) Neither MN-expression of MFN1 or MFN2 rescued the mitochondrial trafficking defect of Marf mutants in boutons co-stained with post-synaptic marker Discs Large 1 ( Dlg1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 010 To assess the presence of mitochondria at NMJs , we counted MitoGFP positive puncta in boutons labeled by anti-Discs Large 1 ( Dlg1 [Parnas et al . , 2001] ) . While control NMJs contain numerous mitochondria per bouton , Marf boutons contain almost no mitochondria , even fewer than in Drp1 mutants ( Figure 3C , see Figure legend , [Verstreken et al . , 2005] ) . However , unlike Drp1 mutants , Marf mutant NMJs exhibit severe morphological defects ( see below ) . Interestingly , we find no obvious labeling defects with the presynaptic active zone marker Bruchpilot ( Wagh et al . , 2006 ) , endocytic markers such as α-Adaptin ( Gonzalez-Gaitan and Jackle , 1997 ) , Dap160 ( Roos and Kelly , 1998 ) , Endophilin ( Verstreken et al . , 2002 ) , and Synaptojanin ( Verstreken et al . , 2003 ) , or the postsynaptic Glutamate receptor IIA ( Qin et al . , 2005 ) in Marf mutants ( Figure 3—figure supplement 1 ) . Expression of Marf protein in MN using the D42-Gal4 driver rescues the trafficking defect and restores the presence of mitochondria at the NMJ ( Figure 3 ) . However , it does not restore the morphological defects ( Figure 3C ) , suggesting that Marf's function in mitochondrial trafficking is cell autonomous and that the defects in synapse morphology are cell non-autonomous . Recently , mammalian MFN2 was shown to physically interact with MIRO2 , an adaptor protein for motor proteins required for mitochondrial trafficking ( Misko et al . , 2010 ) . Drosophila miro ( dmiro ) mutants are severely impaired in mitochondrial trafficking in the VNC ( Guo et al . , 2005 ) . Indeed , RNAi knockdown of dmiro almost abolishes the presence of mitochondria in axons , a phenotype that is much more severe than what we observe in Marf mutants ( data not shown ) . Moreover , loss of dmiro in Marf mutant MNs largely enhances the mitochondrial trafficking defect in the VNC and proximal axons ( Figure 3—figure supplement 2A , B ) . This suggests that Marf cannot be the sole anchor that binds dMiro for mitochondrial trafficking . Loss of mitochondria at NMJs in Drp1 mutants was shown to affect synaptic transmission at high frequency stimulation ( Verstreken et al . , 2005 ) . To gauge how loss of Marf affects synaptic transmission we performed electrophysiological recordings at the NMJs , using a transheterozygous MarfB/MarfE allelic combination in order to compare larvae of the same size since MarfB mutant are small in size . When stimulated at 0 . 2 Hz , Marf mutants do not exhibit any obvious defect in transmitter release based on excitatory junction potential ( EJP ) recordings ( Figure 4A ) . Moreover , the amplitude of spontaneous release events or miniature EJPs ( mEJPs ) and quantal content are not altered in Marf mutants ( Figure 4A ) . Hence , the average number of vesicles released in response to low frequency stimulations in Marf mutants is not different from Marf genomic-rescue controls . However , Marf mutant terminals are unable to properly sustain a 10 Hz stimulus for 10 min when compared to controls ( Figure 4B ) as the EJP amplitudes progressively decrease . A rundown of synaptic transmission is often observed in endocytic mutants such as endophilin and synaptojanin ( Verstreken et al . , 2002 , 2003; Dickman et al . , 2005 ) , dap160 and eps15 ( Koh et al . , 2004 , 2007 ) , and flower ( Yao et al . , 2009 ) . We therefore assessed if endocytosis is impaired and used FM1-43 , a dye that reversibly binds membranes and is internalized into vesicles ( Verstreken et al . , 2008 ) . Unlike eps15 mutants that serve as a positive control , nerve stimulation at 60 mM K+ in the presence of FM1-43 effectively labels synaptic boutons in Marf mutants similar to controls ( Figure 4C , D ) . Hence , vesicle endocytosis or evoked responses at 0 . 2 Hz are not affected in Marf mutants . These features are similar to Drp1 mutants , suggesting that lack of mitochondria at synaptic terminals affect ATP levels required for vesicle mobilization at high frequency stimulation ( Verstreken et al . , 2005 ) . 10 . 7554/eLife . 03558 . 011Figure 4 . Marf is required to maintain synaptic transmission upon repetitive stimulation . ( A ) Excitatory Junctional Potentials ( EJP ) and miniature EJPs ( mEJP ) measured at 0 . 2 Hz in 0 . 75 mM Ca2+ are similar in Marf mutants ( day 12 or day 20 old larvae ) and controls . Hence , quantal content in Marf mutants is also similar to controls ( n = 6–11 larvae assayed ) . ( B ) Controls display facilitation whereas Marf mutants ( day 12 or day 20 old larvae ) show a rundown at 10 Hz in 0 . 75 mM Ca2+ . ( C ) Assessing endocytosis using FM-143 dye uptake at 60 mM [K+] for 1 min shows no obvious differences between wild type controls and Marf mutants . ( D ) Quantification of FM-143 uptake . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 011 A striking difference between Marf mutants and Drp1 mutants is that loss of Marf severely affects NMJ morphology whereas loss of Drp1 does not affect NMJ development ( Figure 3C , Figure 3—figure supplement 1 and Figure 2—source data 2 ) . To visualize bouton morphology , we co-stained with Eps15 , a presynaptic marker ( Koh et al . , 2007 ) and Dlg1 , a postsynaptic marker ( Parnas et al . , 2001 ) . Marf mutant displayed a severe reduction in average bouton size ( Figure 5A ) accompanied by an increase in clustering and numbers of boutons when compared to controls ( Figure 5A , C ) . This NMJ phenotype can be rescued by a Marf genomic rescue construct as well as ubiquitous expression of a Marf cDNA ( Figure 5A , C ) . An increase in bouton number and reduction in size is also observed by ubiquitous knockdown of Marf using RNAi ( Figure 5B , D and Figure 1—figure supplement 1C ) . 10 . 7554/eLife . 03558 . 012Figure 5 . Loss of mitochondrial fusion but not fission in the ring gland results in altered bouton morphology . Third instar larvae NMJs from muscles 6/7 segments A3 were stained with pre-synaptic ( EPS15 ) and post-synaptic ( Dlg1 ) markers . ( A ) Ubiquitous ( Tubulin-Gal4 ) or ring gland ( RG , Feb36-Gal4 ) expression of Marf rescue bouton morphology in Marf mutants , while motor neuron ( D42-Gal4 ) or muscle ( Mef-Gal4 ) Marf expression did not . ( B ) Ubiquitous or RG specific knockdown of Marf or Opa1 ( Poole et al . , 2010 ) phenocopy the bouton phenotype in Marf mutants while knockdown of Drp1 ( Drp1 IR knockdown of Drp1 mRNA is 82% using ubiquitous driver Actin-Gal4 ) did not . ( C and D ) Quantification of bouton numbers from three independent experiments . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01210 . 7554/eLife . 03558 . 013Figure 5—source data 1 . Tissue specific Gal4 screen to assess rescue of lethality and bouton morphology by Marf expression . Tissue specific Gal4 screen using UAS-Marf to assess rescuing ability of the Marf mutant lethal stage and bouton morphology phenotypes . Ubiquitous expression of Marf resulted in rescue of both lethality and bouton phenotype in Marf mutant , while RG specific expression of Marf rescues the Marf mutant bouton phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01310 . 7554/eLife . 03558 . 014Figure 5—source data 2 . Tissue specific Gal4 screen to assess lethality and alterations to bouton morphology by Marf knockdown . Tissue specific Gal4 screen using Marf IR for phenocopying Marf mutant lethal stage and bouton morphology phenotypes . Ubiquitous knockdown of Marf resulted in both prolonged third instar larval stage and similar Marf mutant bouton phenotype , while RG specific knockdown of Marf phenocopied the Marf mutant bouton phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01410 . 7554/eLife . 03558 . 015Figure 5—figure supplement 1 . Ring gland drivers tissues specificity . Specificity of RG driver expression used in this study: Feb36 or Phantom ( Phm ) , ( Mirth et al . , 2005 ) Gal4 expression of UAS-GFP . Third instar larval RGs were stained with anti-GFP antibody , anti-HRP ( presynaptic marker ) , and anti Dlg1 ( post synaptic marker ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 015 Since ubiquitous expression of the Marf cDNA rescues the NMJ morphology phenotype , we tested whether expression of Marf in MN , muscles or glial cells is able to rescue the phenotype . The NMJ phenotype is only partially rescued by Marf expression in MN ( Figure 5A , C ) . Moreover , muscle , glial or MN and muscle expression of Marf does not alter the Marf mutant NMJ morphology ( Figure 5A , C and Figure 5—source data 1 ) . Consistent with these observations , RNAi knock down of Marf in MN , muscles , glia and MN and muscle does not affect bouton number or size at NMJs ( Figure 5D and Figure 5—source data 2 ) . This indicates that Marf expression is required in other cells than MN , muscles or glia . To assess which other tissue/cells contribute to the NMJ defects in Marf mutants , we tested specific RNAi knockdown of Marf using Gal4 drivers that drive expression in different tissues including fat body , haemocytes , oenocytes , trachea or ring gland ( RG ) ( Figure 5—source data 2 ) . Knockdown of Marf with three independent RG-Gal4 drivers resulted in a NMJ phenotype similar to that observed in Marf mutants or ubiquitous knockdown of Marf ( Figure 5B , Figure 5—source data 2 and Figure 5—figure supplement 1 ) , clearly showing a non-cell autonomous requirement for Marf in RGs . In addition , while knockdown of Marf in neurons and RG resulted in pupal lethality , only knockdown of Marf in RG significantly lengthened the third instar larva stage ( 8–10 days ) ( Figure 5—source data 2 ) . Finally , expression of Marf in the RG using two different RG drivers rescued the bouton phenotype of Marf mutants ( Figure 5A , C , Figure 5—source data 1 and Figure 5—figure supplement 1 ) . Hence , Marf is required in RGs to regulate NMJ morphology in a cell non-autonomous manner . Given that loss of Drp1 does not cause obvious developmental defects at NMJs ( Figure 2—source data 2 , Figure 3C and Figure 5B ) ( Drp1 IR knockdown of Drp1 mRNA is 82% using a ubiquitous driver Actin-Gal4 ) , we tested whether loss of Opa1 , another fusion protein ( Cipolat et al . , 2004; Chen et al . , 2005 ) , in RGs causes a bouton phenotype . A RG specific knockdown of Opa1 ( Deng et al . , 2008; Poole et al . , 2010 ) causes a very similar alteration in synaptic morphology as Marf knockdown ( Figure 5B ) . Moreover , Opa1 knockdown in RG also lengthens the larval stages and causes pupal lethality , similar to Marf knockdown ( data not shown ) . Hence , both inner and outer mitochondrial fusion but not fission proteins alter bouton morphology and lengthen larval lifespan via RG , suggesting that the fusion proteins affect the same cell non-autonomous process . RGs are responsible for production of hormones such as ecdysone ( Huang et al . , 2008 ) and juvenile hormone ( Di Cara and King-Jones , 2013 ) . These hormones regulate growth and differentiation of numerous tissues and control the proper timing of larval molts and metamorphosis ( Yamanaka et al . , 2012; Di Cara and King-Jones , 2013 ) . Loss of production of ecdysone in RGs results in a lengthened larval stage ranging from 4 to 19 days ( McBrayer et al . , 2007 , Talamillo et al . , 2008; Rewitz et al . , 2009 ) . To determine if ecdysone production is affected we measured the levels of 20-hydroxyecdysone ( 20E ) ( Porcheron et al . , 1976 ) , in Marf mutants as well as animals with RG specific knockdown of Marf , Opa or Drp1 . Marf mutants or knockdown of Marf and Opa1 in RG exhibit severely reduced levels of 20E when compared to control or knockdown of Drp1 in the RG or Drp1 mutant alleles ( Figure 6A and Figure 2—source data 2 ) . Restoring expression of Marf in the RGs of Marf mutants partially restores the 20E levels ( Figure 6A ) . Moreover , the feeding of 20E to third instar larvae with RG specific knockdown of Marf rescued both the pupal lethality and NMJ morphology phenotype ( Data not shown and Figure 6—figure supplement 1A ) . In summary , Marf and Opa1 but not Drp1 affect ecdysone production in the RG . 10 . 7554/eLife . 03558 . 016Figure 6 . Both Marf and Opa1 regulate ecdysone synthesis in the ring gland , but only Marf promotes lipid droplet formation . ( A ) Both loss of Marf and Opa1 in the RG have reduced 20-hydroxyecdysone ( 20E ) levels when compared to loss of Drp1 and controls . 20E levels are determined and normalized by weight . ( B ) Only loss of Marf in the RG results in reduced lipid droplets ( LDs ) when stained by Nile Red compared to loss of Opa1 or Drp1 . ( C ) Quantification of LDs in the ring gland ( RG ) from three independent experiments . ( D ) TEM sections of RG were the ER is labeled in green , mitochondria in blue and lipid droplets are labeled ‘LD’ . Marf mutants display increased ER fragmentation and reduced numbers of LDs when compared to Marf-genomic rescue control animals . ( E ) Marf mutants have reduced contact length between mitochondria and ER , ER and LD , and mitochondria and LD when compared to controls . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01610 . 7554/eLife . 03558 . 017Figure 6—figure supplement 1 . Feeding of 20E rescues the NMJ morphology of RG specific knockdown of Marf . ( A ) Third instar larvae with a RG ( Feb36-Gal4 ) specific knock down of Marf were fed either 20E ( 0 . 5 mM ) or solvent ( 60% ethanol ) . NMJs from muscles 6/7 segments A3 were stained with pre-synaptic ( EPS15 ) and post-synaptic ( Dlg1 ) markers . Quantification of bouton numbers from three independent experiments . Expression of DRP1 in RGs ( Deng et al . , 2008 ) does not affect the NMJs . ( B ) Expression of DRP1 in RGs ( Deng et al . , 2008 ) does also not affect lipid droplets ( LDs ) numbers when stained by Nile Red and ( C ) 20E levels . Quantification of bouton numbers and 20E levels Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 017 The production of ecdysone ( steroid hormones ) involves many steps following uptake of cholesterol . Drosophila lacks several biosynthetic enzymes for de novo cholesterol synthesis and depends on cholesterol uptake from the food ( Clark and Block , 1959 ) . In the RG , cholesterol is processed into ‘free-cholesterol ( FC ) ’ in the ER ( Miller , 2013 ) . It is then transported into the mitochondrial inner matrix for processing by at least two cytochrome p450 enzymes ( encoded by disembodied [Chavez et al . , 2000] and shadow [Warren et al . , 2002] in Drosophila ) and finally secreted from the RG into the hemolymph ( Gilbert , 2004 ) . Because steroid hormones cannot be stored during Drosophila larva development , FC is stored in the form of cholesterol esters in lipid droplets ( LDs ) until there is a burst of ecdysone synthesis ( Talamillo et al . , 2008; Miller , 2013 ) . This process of cholesterol ester storage and steroid synthesis is highly conserved from flies to mammals . To assess cholesterol ester storage in LDs in RGs of wandering third instar larva , we first stained LDs with Nile Red , which marks neutral lipids that comprise LDs ( Greenspan et al . , 1985 ) . This larval stage precedes the large burst of ecdysone that occurs at the larval–pupal transition ( Yamanaka et al . , 2012 ) . Interestingly , the numbers of LDs are severely reduced in Marf mutants as well as in Marf knockdown in RGs ( Figure 6B , C ) . Moreover , RG expression of Marf rescues the LD phenotype and even increases the LDs numbers above control in Marf mutants , suggesting that Marf is necessary and sufficient for LD formation ( Figure 6B , C ) . Interestingly , RG knockdown of Opa1 does not affect LD number ( Figure 6B , C ) , suggesting that Marf and Opa1 have different roles in the RG . Our findings indicate that Marf plays a unique role in LD synthesis in RG and that it affects cholesterol ester storage . Loss of Opa1 on the other hand does not affect LD storage but like loss of Marf , impairs 20E production . Finally , loss of Drp1 or RG expression of Drp1 does not affect LD synthesis , nor does it affect 20E production ( Figure 6A–C , Figure 2—source data 2 and Figure 6—figure supplement 1B , C ) . Taken together , the three mitochondrial GTPases have different roles in LD dynamics and ecdysone synthesis . LD are generated from the ER through budding of the outer leaflet of the ER membrane ( Walther and Farese , 2012 ) . A physical link between the ER , LDs and mitochondria are often observed as these organelles collaborate to orchestrate numerous metabolic processes such as cholesterol transport and steroid synthesis ( Issop et al . , 2012; English and Voeltz , 2013 ) . Indeed , human MFN2 has been shown to tether the mitochondria to the ER ( de Brito and Scorrano , 2008 ) . To assess the ultrastructural features of ER , LDs , and mitochondria in RGs , we performed TEM in RG . As shown in Figure 6D , Marf mutants exhibit a fragmented ER , reduced number of LD , and morphologically altered mitochondria when compared to controls . The contacts between the mitochondria and the ER , the ER and LD , as well as mitochondria and LD , are all severely reduced in Marf mutant RG ( Figure 6D , E ) . This suggests that Marf promotes cholesterol ester storage in LDs possibly through inter-organelle connections . Human MFN2 tethers mitochondria to the ER ( de Brito and Scorrano , 2008 ) but this has not been documented for MFN1 . Similarly , loss of MFN2 leads to ER stress ( Ngoh et al . , 2012; Sebastian et al . , 2012; Munoz et al . , 2013 ) but a role for MFN1 in ER function has not been reported . If Drosophila Marf mediates connections of mitochondria to ER and if this activity is required for ecdysone synthesis , expression of human MFN2 ( Dorn et al . , 2011 ) in the RG may rescue the loss of LDs , alleviate the bouton morphology defects and restore 20E levels in Marf mutants . We find that RG specific expression of human MFN2 restores the proper number of LD levels and organelle contacts in Marf mutants whereas expression of human MFN1 ( Dorn et al . , 2011 ) does not ( Figure 7A , C and Figure 7—figure supplement 1 ) , indicating that MFN2 specifically can rescue the defect in LD synthesis . However , RG expression of human MFN2 did not rescue the bouton phenotype of Marf mutants ( Figure 7B , D ) . Moreover , ubiquitous expression of MFN1 or MFN2 alone ( Daughterless-Gal4 and Tubulin-GAL4 ) does not rescue the lethality ( Figure 1—figure supplement 1B ) , mitochondrial morphology ( Figure 7—figure supplement 2 ) , mitochondrial trafficking to synapses ( Figure 3—figure supplement 2 ) , 20E levels , and the NMJ phenotypes ( Figure 7 ) , whereas ubiquitous co-expression of both MFN1 and MFN2 rescued all phenotypes ( Figure 1—figure supplement 1B and Figure 7 ) . These data indicate that MFN1 and MFN2 play non-redundant roles and have complementary functions that are integrated into a single protein in Drosophila Marf . 10 . 7554/eLife . 03558 . 018Figure 7 . Human MFN2 restores LD numbers but both human MFN1 and MFN2 are required for steroid-hormone production in the ring glands . ( A ) Rescue of lipid droplets numbers stained by Nile Red in Marf ring glands ( RG ) by MFN2 and MFN1/MFN2 co-expression , but not MFN1 . ( B ) Rescue of Marf bouton morphology by expressing MFN1/MFN2 in RGs ( Feb36-Gal4 ) . Expression of MFN1 or MFN2 alone does not rescue the phenotype . ( C–E ) Quantification in control and Marf mutants for: ( C ) LDs ( D ) Boutons and ( E ) Ecdysone ( 20E levels ) as described in Figures 5 and 6 . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01810 . 7554/eLife . 03558 . 019Figure 7—figure supplement 1 . RG expression of human MFN2 restores organelle contact lengths in Marf mutants . TEM sections of RGs that express human MFN1 or MFN2 . The ER is labeled in green , mitochondria in blue and lipid droplets are labeled ‘LD’ . Marf mutants with RG expression of human MFN2 display increased LD droplets and organelle contact lengths when compared to Marf mutants or Marf mutants with RG expression of human MFN1 animals . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 01910 . 7554/eLife . 03558 . 020Figure 7—figure supplement 2 . Muscle expression of either human MFN2 or MFN1 does not fully restores mitochondrial morphology in Marf mutants . Mitochondrial morphology based on anti-Complex V antibody staining ( Complex V ) in larval muscles of Marf mutants with muscle expression of human MFN1 or MFN2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 020
How does loss of fission or fusion affect mitochondrial function ? In the absence of fusion mixing of mitochondrial DNA and proteins may be severely impaired . Given that mitochondrial proteins are in an environment rich in oxygen radicals , lack of fusion may cause more damage than when fission is impaired ( Chan , 2012 ) . Simply stated , loss of fusion proteins like Marf , MFN1 or MFN2 may cause more severe phenotypes than the loss of a fission protein like Drp1 . Moreover , proteins like Marf and Drp1 may perform other functions that are not directly related to fusion or fission , and hence affect other processes . Based on a careful phenotypic comparison of loss of Marf and Drp1 in Drosophila we find many similarities and differences . Marf mutants display small mitochondria whereas Drp1 mutants exhibit large fused mitochondria . Interestingly , both mutants accumulate mitochondria in the cell body of the neurons and the proximal axonal segments ( Figure 3A ) . In Drp1 mutants , the mitochondria seem to be severely elongated in axons where they fail to reach the NMJs , as previously described ( Verstreken et al . , 2005 ) . The impairment in axonal transport is thought to be due to the fact that the mitochondria are hyperfused and cannot easily be transported . Indeed , loss of Marf in Drp1 mutants can restore mitochondrial trafficking proximally but distal axonal trafficking is still impaired ( Figure 3B ) . In Marf mutants , even though mitochondria are small and can enter the axons , the numbers of mitochondria that travel distally toward the NMJs are dramatically reduced ( Figure 3B ) . Hence , loss of Marf impairs mitochondrial trafficking and longer axons are more severely affected than shorter axons . Since longer axons are more severely affected in CMT2A patients ( Scherer , 2011 ) , defects in mitochondrial trafficking may be at the root of some of the phenotypes associated with the disease . Mfn2 has been implicated in axonal transport via binding to Miro2 . Indeed , knockdown of MIRO2 in cultured vertebrate neurons affects mitochondrial transport in an identical fashion as loss of MFN2 ( Misko et al . , 2010 ) . However , the severity of mitochondrial transport that we observe in Marf mutants is much less pronounced than what has been described in dmiro mutants ( Guo et al . , 2005 ) and what we observe when dmiro is lost . Moreover , removal of dmiro in Marf mutants dramatically enhances the Marf phenotype and almost abolishes axonal localization of mitochondria ( Figure 3—figure supplement 2 ) , arguing that Marf cannot be solely responsible for mitochondrial transport in Drosophila . A comparison of the presence of mitochondria at NMJ synapses shows that Marf mutants have fewer mitochondria than Drp1 mutants ( Figure 3C ) . Moreover , Marf mutants but not Drp1 mutants display a severe increase in small clustered boutons ( Figure 2—source data 2 , Figures 3C and 5 ) . The small and clustered boutons have also been observed in other mutants like endophilin ( Dickman et al . , 2006 ) , synaptojanin ( Dickman et al . , 2006 ) , eps15 ( Koh et al . , 2007 ) , dap 160 ( Koh et al . , 2004 ) , flower ( Yao et al . , 2009 ) and dmiro ( Guo et al . , 2005 ) . However , unlike in Marf mutants , the bouton phenotypes are fully rescued by neuronal expression of the cognate protein within MN in the above mentioned mutants . Moreover , knockdown of Marf in neuron , muscle or glia does not recapitulate the bouton phenotype observe in Marf mutants ( Figure 5B and Figure 5—source data 2 ) , suggesting a unique cell non-autonomous requirement of Marf for proper NMJ morphology . Marf mutants exhibit two obvious phenotypes at NMJs: a severe depletion of mitochondria and a doubling of the number of boutons combined with a severe reduction in size whereas Drp1 mutants only exhibit a severe reduction in mitochondria . However , our electrophysiological studies show that loss of Marf does not affect basal synaptic transmission ( Figure 4 ) similar to what is observed in Drp1 mutants ( Verstreken et al . , 2005 ) . Both respond similarly to wild type NMJs when stimulated at 0 . 2 Hz and both show a progressive run down at 10 Hz when compared to controls . Moreover , endocytosis using FM1-43 and 60 mM K+ is not impaired in Marf and Drp1 mutants , suggesting a defect in reserve pool mobilization in both mutants ( Verstreken et al . , 2005 , 2008 ) . The data also show that the bouton defects observed in Marf mutants do not contribute to the run down in synaptic transmission since Drp1 boutons are normal in number and size yet also have a run down in synaptic transmission ( Figure 2—source data 2 , Figures 3 and 4; [Verstreken et al . , 2005] ) . Loss of Marf in RG recapitulates the bouton phenotype observed in Marf mutants and expression of Marf in RG fully rescues this phenotype ( Figure 5 and Figure 5—source data 1 ) . Interestingly , both Marf and Opa1 are required for steroid hormone production and both lead to extended larval lifespan when knocked down in the RG only ( 8–10 days ) , whereas Drp1 mutations do not affect steroid hormone synthesis . Reduction of ecdysone production by knockdown of the prothoracicotropic hormone receptor ( torso ) in the RG also leads to an extended larval lifespan ( 9 days ) ( Rewitz et al . , 2009 ) and an increased growth of NMJs ( Miller et al . , 2012 ) . Interestingly , knockdown of Drosophila SUMO ( dsmt3 ) in RG lead to a defect in cholesterol import in the RG , reduced 20E levels and an extended larval lifespan ( 19 days ) ( Talamillo et al . , 2008 ) . Hence , the severe reduction in ecdysone synthesis in Marf mutant RG underlies the prolonged larva stages and NMJ morphological defects . The reduction in the number of LDs in RGs when Marf is lost suggests that these RGs are unable to store cholesterol ( Figure 6B , C ) . This storage of cholesterol esters probably permits the RG to produce large amounts of ecdysone when needed , especially at the larval stage and larval to pupal transitions . Cholesterol storage and steroid hormone biosynthesis requires both the ER and mitochondria in vertebrates ( Miller , 2013 ) but loss of MFN1 or MFN2 have not been shown to affect LD synthesis . Defects of anchoring mitochondria to the ER and LDs in Marf RGs argue that these defects lead to the loss of LD and production of ecdysone ( Figure 6 ) . In agreement with this hypothesis , expression of human MFN2 , which tethers ER to mitochondria ( de Brito and Scorrano , 2008 ) , in Marf mutants restores LD synthesis and organelle contacts ( Figure 7A , Figure 7C and Figure 7—figure supplement 1 ) . Moreover , expression of human MFN2 in RNAi mediated Marf knockdown in neurons and muscles rescues ER morphology and stress ( Debattisti et al . , 2014 ) . However , MFN2 expression alone in Marf mutant RG did not restore ecydsone synthesis ( Figure 7E ) , arguing that there are other mitochondrial defects associated with the loss of Marf ( Figure 8 ) . 10 . 7554/eLife . 03558 . 021Figure 8 . Model of Marf dual function in steroid synthesis in the ring glands . ( A ) In wild type ring glands ( RG ) , cholesterol must enter the cell first . Then , cholesterol undergoes a series of modifications in endosomes and along the ER to become free-cholesterol . Then , free-cholesterol is transferred into the mitochondrial inner matrix , where it is processed from free-cholesterol to steroid hormone by p450 enzymes . The steroid hormone is then secreted . As Drosophila larva develops it stores cholesterol in the form of cholesterol ester in lipid droplets ( LDs ) in order to accumulate a reserve of substrate so it can generate bursts of steroid hormone when needed . These LDs require the ER for synthesis . ( B ) In Marf mutants , the ER is fragmented and LD formation is severely reduced . ( C ) RG-specific expression of MFN2 in Marf mutant restores LD numbers but does not rescue hormone synthesis , suggesting that Marf has a second function within the mitochondria . DOI: http://dx . doi . org/10 . 7554/eLife . 03558 . 021 Our data show that co-expression of human MFN1 and MFN2 fully rescue the observed phenotypes in Marf mutants ( Figure 7 ) . Although RG-specific expression of MFN1 in Marf mutants did not restore LD numbers or organelle contacts ( Figure 7—figure supplement 1 ) , MFN1 is still necessary for ecdysone synthesis together with MFN2 , suggesting a role downstream of cholesterol ester storage for both proteins ( Figure 8 ) . Moreover , knockdown of Opa1 in RG did not alter LD numbers but causes reduced 20E levels and aberrant NMJs ( Figure 6 ) . Opa1 resides within the inner mitochondrial membrane , suggesting its role in ecdysone synthesis is within the mitochondria . Ecdysone synthesis within the mitochondria requires two cytochrome p450 enzymes encoded by disembodied ( Chavez et al . , 2000 ) and shadow ( Warren et al . , 2002 ) . Hence , it is likely that impairment in fusion but not fission affects the function of these enzymes ( Figure 8 ) . Opa1 and MFN2 but not Drp1 have been implicated in vertebrate steroidogenesis ( Issop et al . , 2012 ) . Interestingly , in placental trophoblast cells ( BeWO ) in culture the loss of OPA-1 promotes progesterone production by 70% whereas loss of MFN2 has been reported to lead to a 20% decrease in progesterone production ( Wasilewski et al . , 2012 ) . In contrast , testosterone production in MA-10 Leydig cells was unaffected by loss of OPA1 ( Rone et al . , 2012 ) whereas loss of MFN2 did affect testosterone production by 40% in MA-10 Leydig cells ( Duarte et al . , 2012 ) . Hence , in both vertebrate endocrine cells , loss of MFN2 or OPA-1 affected steroids very differently as we observe very similar phenotypes associated with the loss of either protein . Our study also suggests that MFN2 functions upstream of cholesterol entry into the mitochondria at the cholesterol storage stage , since MFN2 restores LD synthesis in Drosophila RG . However , rescuing LD production is not sufficient to restore ecdysone synthesis , suggesting a secondary defect ( Figure 8C ) . In summary , our data indicate that MFN1 and MFN2 have separate functions in vivo that are integrated in a single protein in fly Marf .
Flies were obtained from the Bloomington Drosophila Stock Center at Indiana University ( BDSC ) unless otherwise noted . All flies were kept in standard media and stocks were maintained at room temperature ( 21–23°C ) . For all the larvae experiments described , flies were allowed to lay embryos for 48 hr on grape juice plates with yeast paste . Hemizygous mutant larvae and wild type controls were isolated via GFP selection at the first instar phase and transferred to standard fly food for the duration of their development . The following stocks were used in this study:y1 w* P{neoFRT}19Ay1 w* MarfA , B , C , E , F , G or H P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP , sn+yw eyFLP GMR-LacZ; y+; Drp12 FRT40A/CyO , Kr-Gal4 UAS-GFPcl ( 1 ) P{neoFRT}19A/Dp ( 1;Y ) y+ v+ ey-FLPy1w118 ey-FLP; Drp12 FRT40A/CyO , Kr-Gal4 UAS-GFPy1 w* MarfB or E P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; Genomic Marf-HA/TM6B , Tb+y1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; UAS-MarfHA/TM6B , Tby w;; D42-Gal4 , UAS-mito-HA-GFP , e/TM6B , Tby w; Drp12 FRT40A/CyO , Kr-Gal4 UAS-GFP; D42-Gal4 , UAS-mito-HA-GFP , e/TM6B , Tby1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP; Drp12 FRT40A/CyO , Kr-Gal4 UAS-GFPy w; Df ( 2L ) burK1 , eps15[e75]/Cyo; twi-Gal4 UAS-2xEGFPy1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; Tub-Gal4/TM6B , Tby1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP; DA-Gal4y1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; Mef-Gal4/TM6B , Tby1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP; Feb36-Gal4/CyO , Kr-Gal4 UAS-GFPy1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; Mai60-Gal4/TM6B , Tby w;; UAS-Marf IR/T ( 2;3 ) TSTL , Cyo:TM6b , Tby w;; UAS-Drp1 IR/T ( 2;3 ) TSTL , Cyo:TM6b , Tby w;; UAS-dmiro IR/T ( 2;3 ) TSTL , Cyo:TM6b , Tby1 w* MarfB P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; UAS-MFN1/TM6B , Tby1 w* Marf alleles P{neoFRT}19A/FM7c , Kr-Gal4 UAS-GFP;; UAS-MFN2/TM6B , Tbyw eyFLP GMR-LacZ; y+; Drp11 FRT40A/CyO , Kr-Gal4 UAS-GFPDrp1[T26] cn bw sp/CyO , Kr-Gal4 UAS-GFPy; Drp1[KG03815]/CyO; ryw; UAS-Drp1/TM6C , Sb TbGal4 BDSC fly lines listed on Figure 5—figure supplement 1 y , w , P{neoFRT}19Aisogenized ( iso ) male flies were treated with low concentration of ethylmethanesulfonate to induce mutations , and mutant alleles which showed ERG defects were isolated as described ( Xiong et al . , 2012; Zhang et al . , 2013; Yamamoto et al . , 2014 ) . For mapping of the Marf group , male large duplications ( ∼1–2 Mb ) covering the X chromosome ( Haelterman et al . , 2014 ) were crossed with female y , w mut* , P{neoFRT}19Aisogenized flies that were balanced with FM7c , Kr-GAL4 , UAS-GFP ( Kr > GFP ) . For the Marf group , the lethality of all alleles were rescued by Dp ( 1;Y ) dx[+]5 , y[+]/C ( 1 ) M5 ( 4C11;6D8 + 1A1;1B4 ) . Marf alleles complemented with all the available deficiencies covered by Dp ( 1;Y ) dx[+]5 , y[+]/C ( 1 ) M5 except Df ( 1 ) Exel6239 ( Parks et al . , 2004; Cook et al . , 2012 ) . We then performed Sanger sequencing for genes located to this region and identified mutations in Marf . A 6 . 1 kb genomic rescue fragment ( X: 6259600…6265700 , Drosophila melanogaster Release 5 . 7 ) was amplified using PCR from the P[acman] CH322-102K19 ( Venken et al . , 2009 ) . This DNA fragment was then subcloned into the HindIII and KpnI sites of the P element transformation vector P{CaSpeR-4-HA} ( Yao et al . , 2009 ) and sequenced . For cDNA constructs , the CDS of Marf was retrieved from cDNA clones RE04414 ( Stapleton et al . , 2002 ) , respectively , and subcloned into pUAST-HA vector ( Ohyama et al . , 2007 ) using NotI and XbaI sites . Cloning and DNA purification were performed based on standard protocols . All constructs were sequenced before injection . As previously described in Yao et al . ( 2008 ) , we chose the 22 nucleotides of the coding sequence of Marf , Drp1 , or dmiro as target sequences listed in lowercase and bold in the sequences shown below . In oligo-1 , the third nucleotide from 3ʹ end was changed to C . To synthesize essential backbone for miRNAi production , four long primers were designed . The first PCR product was generated by oligo-1 and -2 . With the first PCR template , the final construct was generated by using common oligo-3 and -4 then digested with EcoRI and NotI and cloned into the pUAST transformation vector . For ERG recording , y w *mut ( lethal ) FRT19A/FM7c , Kr-Gal4 , UAS-GFP flies were crossed to y w P{w+} cl ( 1 ) FRT19A/Dp ( 1;Y ) y+; eyFLP or y w; Drp12 FRT40A/CyO crossed to y w , eyFLP; Drp12 FRT40A/CyO to generate flies with mutant clones in the eyes and ERGs were performed as previously described ( Ly et al . , 2008 ) . Briefly , adult flies were glued to glass slides . A recording probe was placed on the surface of the eye , and a reference probe was inserted in the thorax . A 1-s flash of white light was given , and the response was recorded and analyzed by the AXON™-pCLAMP8 software . TEM of photoreceptor terminals ( Verstreken et al . , 2003 ) and ring glands ( Bellen and Budnik , 2000 ) was performed as described . TEM of photoreceptor terminals and ring glands were done using a Ted Pella Bio Wave processing microwave with vacuum attachments . Briefly , fly heads or third instar larva were dissected and fixed at 4°C in 4% paraformaldehyde , 2% glutaraldehyde , 0 . 1 M sodium cacodylate , and 0 . 005% CaCl2 ( PH 7 . 2 ) overnight , post-fixed in 1% OsO4 , dehydrated in ethanol and propylene oxide , and then embedded in Embed-812 resin ( Electron Microscopy Sciences , Hatfield , PA ) . Photoreceptors or ring glands were then sectioned and stained in 4% uranyl acetate and 2 . 5% lead nitrate . TEM images of PR sections were taken using a JEOL JEM 1010 transmission electron microscope with an AMT XR-16 mid-mount 16 mega-pixel digital camera . Staining of mitochondria membrane potential ( MMP ) by Tetramethylrhodamine ethyl ester ( TMRE; Molecular Probes , Life Technologies , Grand Island , NY ) and ROS by dihydroethidium dye ( DHE; Sigma , St . Louis , MO ) in live muscles , larvae were prepared and stained as described in Shidara and Hollenbeck ( 2010 ) . Live images were acquired using a 40× water immersion lens and a Zeiss LSM510 confocal microscope . ATP levels in larvae was determined as described ( Park et al . , 2006 ) using a kit ( Invitrogen , Life Technologies , Grand Island , NY ) . Quantification of ETC enzymatic activity assay and aconitase assay were performed on isolated mitochondria extracted as previously described ( Graham et al . , 2010; Zhang et al . , 2013 ) . Enzymatic activity assays were performed as previously described ( Emptage et al . , 1983; Das et al . , 2001; Graham et al . , 2010; Zhang et al . , 2013 ) . Aconitase activity assays were performed as previously described in Graham et al . ( 2010 ) ; Zhang et al . ( 2013 ) . For muscle or NMJ immunostaining , dissection and immunostaining of third instar larvae were performed as described in Bellen and Budnik ( 2000 ) . Briefly , third instar larvae were fixed in 3 . 7% formaldehyde for 20 min at room temperature and washed in 0 . 4% Triton X-100 . Primary antibodies were used at the following dilutions: mouse anti- ATP5A 1:500 ( Abcam , Cambridge , MA ) , chicken anti-GFP 1:1000 ( Abcam , Cambridge , MA ) , mouse anti-DLG 1:250 ( DSHB , [Parnas et al . , 2001] ) , guinea pig anti-EPS15 1:2000 ( Koh et al . , 2007 ) , mouse anti-BRP 1:1000 ( Wagh et al . , 2006 ) , rabbit anti-α-adaptin 1:500 ( Gonzalez-Gaitan and Jackle , 1997 ) , mouse anti-Glutamate receptor IIa ( DSHB , Iowa City , IA , [Schuster et al . , 1991] ) , guinea pig anti-Dap160 1:500 ( Roos and Kelly , 1998 ) , rabbit anti-HRP 1:1500 ( Jackson ImmunoResearch , West Grove , PA ) , guinea pig anti-endophilin 1:200 ( Verstreken et al . , 2002 ) , rabbit anti-synaptojanin ( Verstreken et al . , 2003 ) , and rabbit anti-Drosophila vesicular glutamate transporter ( DVGlut ) 1:2000 ( Daniels et al . , 2004 ) . Alexa 488 conjugated ( Invitrogen ) , and Cy3 or Cy5 conjugated secondary antibodies ( Jackson ImmunoResearch , West Grove , PA ) were used at 1:250 . Samples were mounted in VECTASHIELD ( Vector Labs , Burlingame , CA ) . For Lipid Droplet staining , third instar larvae were dissected in cold PBS and fixed in 4% paraformaldehyde for 30 min . Larvae were rinsed several times with 1× PBS to remove fixative and incubated for 10 min at 1:1000 dilution of PBS with 1 mg/ml Nile Red ( Sigma , St . Louis , MO ) . Subsequently the tissues were rinsed with PBS and immediately covered with VECTASHIELD ( Vector Labs , Burlingame , CA ) for same-day imaging . All confocal figures were acquired with confocal microscope ( LSM510; Zeiss ) using Plan Apochromat 40 × NA 1 . 4 and Plan Apochromat 63 × NA 1 . 4 objectives ( Zeiss ) , followed by processing in LSM software ( Zeiss ) , ImageJ , and Photoshop ( Adobe ) . Larval electrophysiological recordings were performed as described in Koh et al . ( 2004 ) . For labeling the exo-endo cycling pool ( ECP ) of vesicles , FM1-43 assays were performed as described ( Verstreken et al . , 2005 , 2008 ) . Live images were acquired using a 40× water immersion lens and a Zeiss LSM510 confocal microscope . Ecdysteroid levels were quantified by ELISA following the procedure described by Porcheron et al . ( 1976 ) , and adapted by Pascual et al . ( 1995 ) . For sample preparation , 20 to 30 staged larvae were weighed and preserved in 600 μl of methanol . Prior to the assay , samples were homogenized and centrifuged ( 10 min at 18 , 000×g ) twice and the resultant methanol supernatants were combined and dried . Samples were resuspended in 50 μl of enzyme immunoassay ( EIA ) buffer ( 0 . 4 M NaCl , 1 mM EDTA , 0 . 1% BSA in 0 . 1 M phosphate buffer ) . 20E ( Sigma , St . Louis , MO ) and 20E-acetylcholinesterase ( Cayman Chemical , Ann Arbor , MI ) were used as the standard and enzymatic tracer . Absorbance was read at 450 nm using a FLUOstar Optima Spectrophotometer ( BMG Labtech ) , results are expressed as 20E equivalents . | Mitochondria are the main source of energy for cells . These vital and highly dynamic organelles continually change shape by fusing with each other and splitting apart to create new mitochondria , repairing and replacing those damaged by cell stress . For nerve impulses to be transmitted across the gaps ( called synapses ) between nerve cells , mitochondria need to supply the very ends of the nerve fibers with energy . To do this , the mitochondria must be transported from the main body of the nerve cell to the tips of the nerve fibers . This may not happen if mitochondria are the wrong shape , size or damaged . While searching for genetic mutations that disrupt nerve function in the fruit fly Drosophila , Sandoval et al . spotted mutations in a gene called Marf . Further investigations revealed that flies with mutant versions of Marf have small , round mitochondria , and their nerves cannot transmit signals to muscles when they are highly stimulated . This is because the mutant mitochondria are not easily transported along nerve fibers , and so not enough energy is supplied to the synapses . The synapses of the Marf mutants are also abnormally shaped . Sandoval et al . found that this is not because Marf is lost in the neurons themselves , but because it is lost from a hormone-producing tissue called the ring gland . Another problem found in flies with mutated Marf genes is that they stop developing while in their larval stage . Sandoval et al . established that this could also be related to the loss of Marf from the ring gland . The Marf protein has two different functions in the ring gland: forming and storing droplets of fatty molecules used in hormone production , and synthesising a hormone that controls when a fly larva matures into the adult fly . This suggests that the lower levels of this hormone produced by Marf mutant flies underlies their prolonged larval stages and synapse defects . Vertebrates ( animals with backbones , such as humans ) have two genes that are related to the fly's Marf gene . When the human forms of these genes were introduced into mutant flies that lack a working copy of Marf , hormone production was only restored if both genes were introduced together . This indicates that these genes have separate roles in vertebrates , but that these roles are both performed by the single fly gene . The role of Marf in tethering mitochondria in the ring gland may allow us to better understand how this process affects hormone production and how the different parts of the cell communicate . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2014 | Mitochondrial fusion but not fission regulates larval growth and synaptic development through steroid hormone production |
The unfolded protein response ( UPR ) adjusts the cell’s protein folding capacity in the endoplasmic reticulum ( ER ) according to need . IRE1 is the most conserved UPR sensor in eukaryotic cells . It has remained controversial , however , whether mammalian and yeast IRE1 use a common mechanism for ER stress sensing . Here , we show that similar to yeast , human IRE1α’s ER-lumenal domain ( hIRE1α LD ) binds peptides with a characteristic amino acid bias . Peptides and unfolded proteins bind to hIRE1α LD’s MHC-like groove and induce allosteric changes that lead to its oligomerization . Mutation of a hydrophobic patch at the oligomerization interface decoupled peptide binding to hIRE1α LD from its oligomerization , yet retained peptide-induced allosteric coupling within the domain . Importantly , impairing oligomerization of hIRE1α LD abolished IRE1’s activity in living cells . Our results provide evidence for a unifying mechanism of IRE1 activation that relies on unfolded protein binding-induced oligomerization .
Protein-folding homeostasis is critical for proper cell function . Accordingly , cells evolved surveillance mechanisms to monitor protein-folding status and elicit adaptive responses to adjust protein-folding capacity according to need ( Balchin et al . , 2016; Bukau et al . , 2006; Walter and Ron , 2011 ) . In the endoplasmic reticulum ( ER ) , where the majority of transmembrane and soluble secretory proteins fold and mature , protein-folding homeostasis is ensured by a network of signaling pathways collectively known as the unfolded protein response ( UPR ) ( Walter and Ron , 2011 ) . In metazoans , perturbations leading to the accumulation of mis- or unfolded proteins in the ER are recognized as ‘ER stress’ by three unique ER-resident UPR sensors , IRE1 , PERK and ATF6 ( Cox et al . , 1993; Cox and Walter , 1996; Harding et al . , 2000; Niwa et al . , 1999; Sidrauski and Walter , 1997; Tirasophon et al . , 2000; Walter and Ron , 2011; Yoshida et al . , 1998; Yoshida et al . , 2001 ) . These sensors transmit information about the protein-folding status in the ER and drive gene expression programs that modulate both the protein-folding load and folding capacity of the ER . If ER stress remains unmitigated , the UPR induces pro-apoptotic pathways , thereby placing the network at the center life-or-death decisions that affect the progression of numerous diseases ( Bi et al . , 2005; Feldman et al . , 2005; Lin et al . , 2007; Lu et al . , 2014; Vidal et al . , 2012; Walter and Ron , 2011; Zhang and Kaufman , 2008 ) . IRE1 drives the most conserved branch of the UPR , which exhibits remarkably similar mechanistic aspects shared between yeast and mammals ( Aragón et al . , 2009; Korennykh et al . , 2009; Li et al . , 2010 ) . In mammals , IRE1 exists in two isoforms , α and β . IRE1α is ubiquitously expressed , whereas IRE1β expression is restricted to gastrointestinal and respiratory tracts ( Bertolotti et al . , 2001; Tsuru et al . , 2013 ) . Both IRE1 orthologs are trans-membrane kinase/nucleases that oligomerize in the ER-membrane in response to ER stress ( Aragón et al . , 2009; Li et al . , 2010 ) . Oligomerization is crucial for IRE1 activation as it allows for trans-autophosphorylation and allosteric activation of its endonuclease domain , which for IRE1α then initiates the unconventional splicing of the XBP1 mRNA ( Aragón et al . , 2009; Cox et al . , 1993; Cox and Walter , 1996; Korennykh et al . , 2009; Li et al . , 2010; Sidrauski and Walter , 1997; Yoshida et al . , 1998; Yoshida et al . , 2001 ) . Spliced XBP1 mRNA encodes the transcription factor XBP1s , which activates the transcription of several target genes involved in restoring ER homeostasis ( Acosta-Alvear et al . , 2007; Lee et al . , 2003 ) . While the XBP1 mRNA is the only known splicing target of IRE1 , active IRE1 can also cleave ER-localized mRNAs in a process known as regulated IRE1-dependent decay of messenger RNAs ( RIDD ) , which serves to limit the amount of client proteins entering the ER , thus helping alleviate the folding stress ( Hollien et al . , 2009; Hollien and Weissman , 2006 ) . Two alternative models are used to describe how IRE1’s lumenal domain senses ER stress: a recent model where unfolded proteins act directly as activating ligands and an earlier model where IRE1 lumenal domain is indirectly activated through dissociation of the ER-chaperone BiP . The direct activation model emerged from the crystal structure of the core lumenal domain ( cLD ) from S . cerevisiae IRE1 ( yIRE1; ‘y’ for yeast ) , where yIRE1 cLD dimers join via a 2-fold symmetric interface IF1L ( ‘L’ for lumenal ) . A putative peptide-binding groove that architecturally resembles that of the major histocompatibility complexes ( MHCs ) extends across this interface ( Credle et al . , 2005 ) . yIRE1 selectively binds a misfolded mutant of carboxypeptidase Y ( Gly255Arg , CPY* ) in vivo , and purified yIRE1 cLD directly interacts with peptides in vitro , leading to its oligomerization . Taken together , these observations support the model that direct binding of unfolded proteins in the ER lumen to IRE1 induces its oligomerization leading to IRE1 activation ( Gardner and Walter , 2011 ) . Due to structural differences between human and yeast IRE1 lumenal domains , it is not yet clear if this mechanism is also used by mammalian IRE1 . Although the crystal structure of human IRE1α ( hIRE1α ) cLD displays conserved structural elements in its core , there are several notable differences between the crystal structures of human and yeast IRE1 cLD known to date ( Figure 1 ) . First , the helices flanking the groove in yIRE1 cLD are too closely juxtaposed in the human structure to allow formation of the MHC-like groove present in the yeast ( Zhou et al . , 2006 ) . Second , the yIRE1 cLD structure displays a second interface , IF2L , which provides contacts for higher order oligomerization , which was experimentally validated to be indispensable for yIRE1 activation in vivo ( Figure 1 ) . In the yIRE1 cLD , an α-helix–turn region forms an important element in IF2L making contacts with the incomplete β-propeller in the neighboring protomer . Notably , the residues corresponding to the α-helix–turn are not resolved in the hIRE1α cLD crystal structure ( aa V307-Y358 ) . Instead , hIRE1α cLD has two other symmetry mates in addition to the dimerization interface , which appear to be crystal lattice contacts that are predicted to be too energetically unstable to form biologically important oligomerization interfaces ( Zhou et al . , 2006 ) . Indeed , the equivalent of an IF2L cannot form in the depicted hIRE1α cLD structure because of a steric hindrance from a prominent α-helix ( ‘αB helix’; aa V245-I263 ) that is absent in yIRE1 cLD ( Figure 1 ) ( Zhou et al . , 2006 ) . The structural differences between IRE1 orthologs were cast to support the indirect model of IRE1 activation in higher eukaryotes ( Zhou et al . , 2006 ) . This model poses that due to the aforementioned structural differences—rather than direct unfolded protein binding—it is the reversible dissociation of the ER-resident Hsp70-type chaperone BiP from IRE1’s lumenal domain the main driving force regulating hIRE1α activity ( Zhou et al . , 2006 ) . According to this view , titration of BiP to unfolded proteins upon ER stress licenses IRE1 activation ( Bertolotti et al . , 2000; Carrara et al . , 2015; Oikawa et al . , 2009; Zhou et al . , 2006 ) . In yeast , however , this view has been experimentally refuted ( Kimata et al . , 2004; Pincus et al . , 2010 ) . Considering the degree of conservation at various features of IRE1 mechanism of action from yeast to mammals , we favor the unifying direct activation model . Such model finds support in the notion that all structures adopted by a protein in a crystal lattice represent a singular snapshot of many possible conformational states . Therefore , it is entirely plausible that human and yeast IRE1 cLD use a common mechanism of activation and that the divergent structures aforementioned represent different states in a spectrum of possible conformational states that the IRE1 cLD from any species could assume . In this scenario , the crystal structure of hIRE1α cLD represents a ‘closed’ conformation that can shift towards an ‘open’ state to allow peptide binding in the MHC-like groove that is apparent in the structure of the yeast ortholog ( Video 1 ) ( Gardner et al . , 2013; Gardner and Walter , 2011 ) . As such , this model predicts specific outcomes that can be experimentally tested . Specifically , that ( i ) human IRE1 α cLD can bind to unfolded polypeptides , ( ii ) unfolded polypeptide binding stabilizes the open conformation of the hIRE1α cLD , and ( iii ) the open conformation of hIRE1α cLD favors its oligomerization . Here , we used complementary biochemical and structural approaches to experimentally explore the mechanism of human IRE1α activation . We show that hIRE1α cLD—just like its yeast ortholog—directly binds select peptides with a characteristic amino acid bias . State-of-the-art NMR experiments that probe dynamic conformational states further support an activation mechanism involving peptide binding to the MHC-like groove and stabilizing the open conformation of hIRE1α cLD . Moreover , we provide insights into the mechanism that couples peptide binding and oligomerization to produce active IRE1 oligomers . Importantly , we show by mutational analysis that lumenal domain driven oligomerization is crucial for IRE1 function in mammalian cells . Taken together , our results resolve the discrepancies between existing models of IRE1 activation and supports a model in which unfolded polypeptides can bind and directly activate human IRE1 .
To test whether , akin to yeast IRE1 , mammalian IRE1 also binds unfolded proteins directly , we employed peptide tiling arrays . To identify hIRE1α cLD-binding peptides , we designed tiling arrays utilizing ER-targeted model proteins known to induce the UPR either by overproduction ( proinsulin and 8ab protein from SARS-corona virus [Scheuner et al . , 2001; Sung et al . , 2009] ) or through destabilizing point mutations ( myelin protein zero ( MPZ ) ) . The peptide arrays were composed by tiling 18-mer peptides that step through the entire protein sequence , shifting by three amino acids between adjacent spots . We incubated the peptide arrays with purified hIRE1α cLD fused N-terminally to maltose-binding protein ( MBP ) and probed with an anti-MBP antibody . As shown in Figure 2A ( left panel ) , MBP-hIRE1α cLD bound a select subset of peptides on the arrays . To maximize the available sequence space , we analyzed binding of MBP-hIRE1α cLD to these peptides irrespective of their topological accessibility in the ER lumen . hIRE1α cLD recognized peptide sequences found in both the ER-lumenal and cytosolic domains of MPZ , which we considered together in our analyses to define the chemical properties of cLD peptide recognition . We found that hIRE1α cLD-binding peptides with the top 10% binding scores were enriched in cysteine , tyrosine , tryptophan , and arginine ( Figure 2B , Figure 2—figure supplement 1A , p<0 . 05 ) . By contrast , aspartate and glutamate were strongly disfavored , together with glutamine , valine , and serine . At a first glance , the amino acid preferences displayed by mammalian IRE1 cLD resemble those of the other chaperones including the ER chaperone BiP ( Blond-Elguindi et al . , 1993; Deuerling et al . , 2003; Flynn et al . , 1991 ) . Like BiP , hIRE1α cLD favored binding to aromatic and positively charged residues ( Blond-Elguindi et al . , 1993; Otero et al . , 2010 ) . BiP is a highly abundant chaperone in the ER lumen , whereas IRE1 is present at orders of magnitude lower levels ( Ghaemmaghami et al . , 2003 ) . Therefore , if IRE1 and BiP recognize the same regions of unfolded proteins , the peptide-binding activity of hIRE1α cLD would depend entirely on saturation of BiP by unfolded substrate proteins—a scenario difficult to reconcile with IRE1’s task of dynamically sensing ER stress . To address this point , we compared the binding preferences of mammalian BiP ( fused to an N-terminal 10x-histidine tag ) on the same peptide arrays . We found sequences recognized by both hIRE1α cLD and BiP ( Figure 2A , Figure 2—figure supplement 1A , B ) . Importantly , however , we also found profound differences . While IRE1 tolerated both prolines and histidines , BiP strongly disfavored these amino acids ( Figure 2B , Figure 2—figure supplement 1A , p<0 . 05 ) . Conversely , BiP tolerated serine and threonines , while IRE1 strongly disfavored them . Thus , IRE1 can recognize regions of unfolded proteins to which BiP would not readily bind and vice versa , thereby providing a plausible explanation of how IRE1 could recognize unfolded proteins despite of the vast excess of BiP over hIRE1α LD in the ER . To measure binding affinities of hIRE1α cLD’s interaction with peptides in solution , we selected the two peptides with the highest binding scores in the peptide arrays ( MPZ- and 8ab-derived peptides , henceforth referred to as ‘MPZ1’ and ‘8ab1’ , respectively ) and attached fluorophores at their N-termini . Fluorescence anisotropy revealed that hIRE1α cLD bound to MPZ1 with K1/2 = 24 ± 4 . 7 µM and to 8ab1 with K1/2 = 5 ± 1 . 7 µM ( Figure 2C ) . ( Note that we used K1/2 to denote a measure of affinity because , as we show below , hIRE1α cLD exists in solution as an ensemble of different interconverting conformational states and our measurements therefore score several superimposed equilibria . The measured affinities therefore do not reflect true Kd values ) . These affinities fall within the same order of magnitude of chaperone binding to unfolded proteins , supporting the notion that similar modes of fast transient interactions with unfolded proteins are adopted by both IRE1 and chaperones ( Karagöz et al . , 2014; Marcinowski et al . , 2011; Street et al . , 2011 ) . To identify the minimal region in MPZ1 for binding to hIRE1α cLD , we next divided MPZ1 into 12 , 11 and 9 amino acid long fragments representing its N-terminal ( MPZ1-N ) , middle ( MPZ1-M ) and C-terminal ( MPZ1-C ) regions and measured their respective affinities for hIRE1α cLD . hIRE1α cLD bound to MPZ1-N with a similar affinity as the full-length peptide ( K1/2 = 16 . 0 ± 2 . 6 µM , Figure 2D ) , whereas the other peptide fragments displayed much lower binding affinities ( K1/2 = 377 ± 54 µM and 572 ± 107 µM , respectively , assuming similar maximum anisotropy values as for the MPZ1-N peptide ) . We further truncated MPZ1-N by two residues at a time from either its N- or C-terminus . Deleting amino acids from the C-terminus gradually decreased the affinity ( Figure 2E ) . By contrast , deletion of the first two hydrophobic residues from the N-terminus ( leucine and isoleucine ) abolished its binding to hIRE1α cLD ( Figure 2F ) . These analyses revealed that the minimum peptide length with a comparable binding affinity to the full-length MPZ1 peptide is a 12-mer . This 12-mer peptide matches the chemical properties we found for hIRE1α cLD-binding peptides: it is enriched in aromatics , hydrophobic amino acids and arginines , indicating that specific binding contacts play a role in hIRE1α cLD’s interaction with unfolded proteins . To validate that peptides are valid surrogates for unfolded proteins , we next tested binding of intact but constitutively unfolded proteins to hIRE1α cLD . Immunoglobulins ( IgGs ) mature in the ER using a well-characterized folding pathway , wherein the constant region domain of the IgG heavy chain ( CH1 ) remains disordered until it binds to its cognate partner , the constant region domain of the IgG light chain CL ( Feige et al . , 2009 ) . We measured the binding affinity of CH1 to hIRE1α cLD by thermophoresis , which reports on changes in the hydration shell of a biomolecule upon interaction with a partner in solution ( Jerabek-Willemsen et al . , 2011 ) . By contrast to earlier findings that showed no measurable binding of hIRE1α cLD to CH1 under different experimental conditions ( Carrara et al . , 2015 ) , our experiments showed that hIRE1α cLD interacts with CH1 with a K1/2 = 29 . 2 ± 1 . 2 µM ( Figure 2G ) . To further validate this observation , we measured binding of hIRE1α cLD to another model unfolded protein by fluorescence anisotropy , the folding mutant of staphylococcal nuclease Δ131Δ ( Street et al . , 2011 ) . We observed a comparable binding affinity of K1/2 = 21 . 4 ± 2 . 3 µM ( Figure 2—figure supplement 2 ) . Our data thus show that hIRE1α cLD binds to full-length unfolded proteins with similar affinity as peptides , suggesting that these proteins display a distinct single binding site for hIRE1α cLD . To test whether multiple binding sites would increase the affinity for hIRE1α cLD , we synthesized a peptide consisting of two MPZ1-N tandem repeats separated by a 5-amino acid spacer ( MPZ1-N-2X ) . Intriguingly , MPZ1-N-2X bound to hIRE1α cLD with an order of magnitude higher affinity ( K1/2 = 0 . 456 ± 0 . 07 µM ) compared to MPZ1 peptide ( Figure 2H ) . As we show below , the increased apparent affinity is due to avidity of hIRE1α cLD to the peptide . To capture evidence for structural rearrangements in hIRE1α cLD predicted by a switch-mechanism that oscillates between inactive closed and active open conformations as we suggest in the Introduction , we employed nuclear magnetic resonance ( NMR ) spectroscopy . NMR spectroscopy reveals structural information at the atomic level for dynamic protein complexes and is well suited to study structural changes in hIRE1α cLD upon its interaction with peptides and unfolded proteins . The hIRE1α cLD dimer is ~80 kDa and thus is well above the size limit for conventional NMR approaches . We therefore used methyl transverse relaxation optimized spectroscopy ( methyl-TROSY ) , a specific NMR method that allows to extract structural information from large proteins after selective isotopic labeling of side chain methyl groups with carbon-13 ( 13C ) ( Tugarinov et al . , 2004; Tugarinov et al . , 2007 ) in select amino acids including isoleucines . hIRE1α cLD has 12 isoleucines per monomer , which are evenly distributed throughout the protein ( Figure 3A ) . In hIRE1α cLD’s methyl-TROSY spectra , we resolved seven peaks corresponding to isoleucines ( Figure 3B ) , which then served as sensors of peptide binding and accompanying conformational changes . All isoleucine peaks in hIRE1α cLD’s NMR spectrum displayed broad line widths ( Figure 3B ) , which is indicative of chemical exchange resulting from hIRE1α cLD sampling multiple conformational states at the conditions of the NMR experiments . These data revealed that hIRE1α cLD is dynamic in solution . To assign the resolved peaks to specific amino acids in the hIRE1α cLD sequence , we mutated each isoleucine to leucine , alanine or valine and monitored the disappearance of each resolved peak in methyl-TROSY spectra of the mutant proteins . This approach allowed us to assign six isoleucine peaks unambiguously ( Figure 3C , D , Figure 3—figure supplements 1 , 2 and 3 ) . To further increase the number of NMR visible probes in hIRE1α cLD , we mutated Leu186 and Thr159 to isoleucines ( Figure 3E , F , Figure 3—figure supplement 1C and F ) . Leu186 lies in an amphipathic unstructured loop surrounding the putative groove in hIRE1α cLD . The Leu186Ile peak displayed high signal intensity consistent with a dynamic and flexible position ( Figure 3F , Figure 3—figure supplement 1F ) . By contrast , Thr159 lies at the β-sheet floor in hIRE1α cLD structure where its side chain faces towards the MHC-like groove and , as expected , the Thr159Ile substitution resulted in a low-intensity peak ( Figure 3—figure supplement 1C ) . We further enhanced the coverage of hIRE1α cLD with NMR-visible probes in complementary experiments in which we labeled threonine side chains with 13C at their γ2 methyl groups ( Figure 3—figure supplement 4 ) . There are 33 threonine residues in hIRE1α cLD , 24 of which were detected by the NMR experiments . While we did not assign threonine peaks in hIRE1α cLD spectrum due to high spectral crowding , they provided an additional ‘fingerprint’ reporting on peptide binding-induced changes in hIRE1α cLD . Next , we used methyl-TROSY experiment to monitor changes in the environment of isoleucines and threonines in hIRE1α cLD upon peptide binding . A largely overlapping subset of isoleucine and threonine peaks shifted when hIRE1α cLD bound to the peptides MPZ1 or 8ab1 , indicating a change in a localized environment upon peptide binding ( Figure 4A , B , Figure 4—figure supplement 1A–C ) . Yet , a subset of isoleucine and threonine peaks displayed peptide specific changes . The chemical shifts displayed by the isoleucine peaks were not very large yet reproducible upon binding of different peptides allowing us to probe peptide induced changes in hIRE1α cLD . By contrast , the threonine peaks displayed larger chemical shifts , which is expected from their higher solvent exposure rendering them more sensitive to binding events ( Figure 4B , Figure 4—figure supplement 1B and C ) . Mapping the chemical shift perturbations of the isoleucine peaks upon peptide binding on the hIRE1α cLD structure ( Figure 4A–E , Figure 4—figure supplement 1B , Figure 4—figure supplement 2 ) revealed that the isoleucine resonances that shifted most significantly lie on the floor of the central β-sheet ( marked by Ile124 , Ile128 , Thr159Ile ) ( Figure 4D and E ) . Among these isoleucines , only the side chain of Thr159Ile faces towards the MHC-like groove . We noted that in comparison to the isoleucines 124 and 128 , Thr159Ile peak displayed a larger shift upon peptide binding ( Figure 4C and E , Figure 4—figure supplement 1D ) . In addition to the central β-sheet floor , the αB helix that lies at the ends of hIRE1α cLD dimer ( marked by Ile263 ) , and the β-sandwich connecting the β-sheet floor to the αB helix ( marked by Ile52 ) were affected , albeit to a lesser extent . By contrast , the unstructured loop extending from the MHC-like groove ( marked by Ile186 ) was only slightly affected and the isoleucines positioned in the flexible region that are not resolved in the crystal structure ( marked by Ile326 and Ile334 ) did not shift ( Figure 4D , E ) . Importantly , binding of the unfolded protein CH1 shifted the same peaks in the hIRE1α cLD spectra as the short peptides suggesting peptides and unfolded protein chains interact with hIRE1α cLD in a similar way ( Figure 4F , Figure 4—figure supplement 2 ) . Taken together , these results indicate that peptide as well as unfolded protein binding populate a distinct conformational state of hRE1α cLD , consistent with a peptide-induced closed-to-open conformational transition . Moreover , the results are consistent with a model in which peptide binding induces conformational changes that propagate from the MHC-like groove via the β-sandwich to affect the regions involved in oligomerization . To map the peptide-binding site in hIRE1α cLD with higher precision , we employed paramagnetic relaxation enhancement ( PRE ) experiments ( Gaponenko et al . , 2000; Gillespie and Shortle , 1997 ) using MPZ1 modified with a nitroxide spin label , 3- ( 2-Iodoacetamido ) -PROXYL , at cysteine residue , Cys5 ( Figure 5A , B ) . The unpaired electron in the spin label broadens ( in a range of 1 to 2 . 5 nm ) or entirely erases ( distances <1 nm ) NMR signals in its vicinity in a distance dependent manner ( Gottstein et al . , 2012 ) . Binding of the spin label attached peptide to hIRE1α cLD would result in a decrease in the intensity of isoleucine peaks depending on their relative distance to the peptide-binding site . Therefore , we analyzed the changes in the intensities of all isoleucine signals upon binding of MPZ1-proxyl peptide to hIRE1α cLD ( Figure 5B , Figure 5—figure supplement 1A–C ) . Binding of MPZ1-proxyl to hIRE1α cLD erased the otherwise very strong signal of Leu186Ile and broadened that of Ile124 ( Figure 5B–D and Figure 5—figure supplement 1A–C ) . Importantly , Ile128 and Ile263 signals , which shifted upon MPZ1 binding as discussed above ( Figure 4E ) , broadened to a lesser extent , suggesting that these residues lie further from the peptide-binding site ( Figure 5—figure supplement 1C and D ) . Their resonances therefore shifted due to peptide-induced distant conformational rearrangements . Displaying the normalized PRE effect on hIRE1α cLD structure revealed that MPZ1-proxyl binding mapped to the center of the MHC-like groove , suggesting that peptides bind to MHC-like groove and induce distant conformational changes in hIRE1α cLD ( Figure 5D ) . To test whether the distant conformational changes in hIRE1α cLD monitored by the NMR experiments are due to peptide binding-induced oligomerization , we employed analytical ultracentrifugation ( AUC ) sedimentation velocity experiments to assess the oligomeric status of hIRE1α cLD in the absence and presence of peptides . At the concentration range used at NMR experiments ( 25–75 µM ) , hIRE1α cLD was found as a mixture of various oligomeric states , where the main peaks corresponded to dimers and tetramers ( with higher amount of tetramers formed at higher concentrations , see Figure 6A ) . Notably , binding of MPZ1-N to hIRE1α cLD at the NMR concentrations sharpened the tetramer peak and induced formation of larger oligomeric species in these experiments ( Figure 6A ) . The peptide concentration used in these experiments does not saturate hIRE1α cLD molecules based on a determined K1/2 of 16 . 0 ± 2 . 6 µM , therefore only a small population of hIRE1α cLD formed higher oligomers ( depicted as the pink area ) ( Figure 6A ) . To assess hIRE1α cLD’s oligomeric status at varying hIRE1α cLD concentrations , we performed size exclusion chromatography and found that hIRE1α cLD eluted at earlier fractions in a concentration-dependent manner ( Figure 6—figure supplement 1A ) . AUC data confirmed these findings and showed that at concentrations close to its dimerization constant of 2 . 5 µM , hIRE1α cLD sediment as a single peak with a sedimentation coefficient corresponding to a mixture of monomers and dimers ( Figure 6B , Figure 6—figure supplement 1B ) . In this concentration regime ( from 1 to 2 . 5 µM ) , the hIRE1α cLD peak progressively shifted to higher sedimentation values with increasing hIRE1α cLD concentration ( Figure 6—figure supplement 1B ) . Peptide binding to hIRE1α cLD shifted the hIRE1α cLD population to even higher sedimentation values ( Figure 6B , blue trace ) , indicating that under these conditions peptide binding stabilized hIRE1α cLD dimers and lead to the formation of oligomers . As hIRE1α cLD populated distinct oligomeric states in a concentration-dependent manner , we next compared the conformational state of hIRE1α cLD at 5 µM ( no higher-order oligomer formation detected by AUC ) to 50 µM ( based on Figure 6C , approximately 60% higher-order oligomer formation ) by NMR spectroscopy to probe for the structural differences assumed by these two distinct states ( Figure 6C , D ) . In these experiments , we relied on the high sensitivity of selective isoleucine labeling strategy , which could readily detect hIRE1α cLD signals at concentrations as low as 5 µM ( Figure 6—figure supplement 2A , B ) . Notably and similar to effects observed upon peptide binding , oligomerization changed the environment of the αB helix ( marked by Ile263 ) and the β-sandwich connecting the β-sheet floor to the αB helix ( marked by Ile52 ) that both lie at the tips of hIRE1α cLD dimers ( Figure 6D , E , Figure 6—figure supplement 2C ) . These data suggest that these isoleucines are part of the oligomerization interface and/or that their conformational rearrangements are coupled to the formation of the interface . Moreover , NMR experiments showed chemical shifts in the isoleucines on the beta sheet floor of the groove ( marked by Ile124 and Ile128 ) upon formation of higher oligomers ( Figure 6E ) . These coupled , global conformational differences observed by NMR strongly underscore the notion that oligomeric hIRE1α cLD adopts an active conformation and displays higher affinity for unfolded protein ligands . To address this notion , we set out to experimentally determine the oligomerization interface and then impair the oligomerization of hIRE1α cLD by mutation . We employed a chemical cross-linking strategy coupled to mass spectrometry to experimentally determine residues that map to the oligomerization interface in hIRE1α cLD . To this end , we cross-linked hIRE1α cLD in the presence and absence of peptides by a homobifunctional cross-linker , BS3 ( bis ( sulfosuccinimidyl ) suberate ) , which crosslinks primary amines mainly present in lysine side chains . Denaturing SDS-PAGE analysis of hIRE1α cLD after cross-linking revealed that cross-linking captured oligomeric hIRE1α cLD ( Figure 7A , Figure 7—figure supplement 1 ) . We separately isolated the bands corresponding to hIRE1α cLD monomers , dimers and higher oligomers from the gel and analyzed peptides by mass spectrometry . We identified cross-linked peptides by accurate mass measurement of both candidate peptides and their fragment ions ( Chu et al . , 2010; Trnka et al . , 2014 ) . In comparative analyses , we separated intra- from inter-molecular cross-links by focusing on peaks that were present only in the covalent dimers and higher oligomers ( Wu et al . , 2013; Zeng-Elmore et al . , 2014 ) . These analyses revealed five abundant cross-links between lysines 120•120 , 53•347 , 53•349 , 53•351 and 265•351 ( Figure 7A , Table 1 ) . Previous studies of BS3-cross-linked proteins with known crystal structures established that the distance between the αC atoms of cross-linked lysines is less than 28 Å for most cross-links but can be up to 33 Å for a few cases due to local protein flexibility ( Leitner et al . , 2010 ) , in agreement with the additive lengths of the cross-linker itself plus twice the length of the lysine side chain . The Lys120•120 cross-link maps to hIRE1α cLD’s dimerization interface ( IF1L ) , whereas the four other cross-links are compatible with being positioned at hIRE1α cLD oligomerization interface , IF2L . The cross-links Lys53•347 , Lys53•349 , Lys53•351 and Lys265•351 each involve one lysine residue ( Lys53 and Lys263 ) that is close to the isoleucines ( Ile52 and Ile263 ) that shifted upon hIRE1α cLD oligomerization ( Figure 6D , E ) , suggesting that they report on the formation of hIRE1α cLD’s putative oligomerization interface IF2L . Lys347 , Lys349 and Lys351 are located in a region that was not resolved in hIRE1α cLD crystal structure , suggesting that these regions are contributing to the formation of the oligomerization interface in hIRE1α cLD . We next threaded the sequence of hIRE1α cLD into the yeast crystal structure of the oligomeric state , which fulfilled the distance restraints imposed by the cross-links ( Figure 7B , C ) . This structural model predicted an extensive interface formed by hIRE1α cLD oligomers that involves residues from parts of hIRE1 cLD that are not resolved in the crystal structure , as well as the incomplete β-propeller involved in the formation of the oligomerization interface in yeast Ire1 cLD ( Figure 7C ) . We used the predictive power of the structural model ( hIRE1 cLD threaded into the yeast structure ) to identify a patch of four hydrophobic residues WLLI ( aa 359–362 ) suggested to contribute to the hIRE1α cLD oligomerization interface IF2L ( Figure 7C , Figure 7—figure supplement 2 ) . Assuming that these residues would be critical for oligomerization , we mutated them ( WLLI359-362 to GSGS359-362; ‘IF2L mutant’ ) and assessed whether the hIRE1α cLD IF2L mutant formed oligomers by AUC sedimentation velocity analysis . The experiments revealed that , at a concentration ( 50 µM ) where wild type hIRE1α cLD readily forms oligomers , the hIRE1α cLD IF2L mutant sediment as a single dimeric peak , showing that the mutation prevents hIRE1α cLD oligomerization ( Figure 7D ) . To assess whether hIRE1α cLD IF2L mutant is functional , we tested peptide binding by fluorescent anisotropy experiments . The IF2L mutant bound MPZ1-N and MPZ1-N-2X peptide at similar affinities to the wild type protein ( with K1/2 = 5 . 4 ± 1 . 4 µM and K1/2 = 0 . 95 ± 0 . 4 µM , respectively ) ( Figure 7E , Figure 7—figure supplement 3A ) . These results indicated that hIRE1α cLD dimer is the functional unit for peptide binding and that hIRE1α cLD oligomers do not display a higher affinity conformation . Moreover , they also showed that the avidity effect that resulted in higher affinity binding of MPZ1-N-2X peptide to hIRE1α cLD does not require formation of higher hIRE1α cLD oligomers . AUC data confirmed these analyses and showed that binding of MPZ1-N-2X to hIRE1α cLD IF2Lmutant stabilized dimer formation but did not lead to formation of oligomers bridged by MPZ1-N-2X peptide ( Figure 7E , Figure 7—figure supplement 3B and C ) . The hIRE1α cLD IF2L mutant therefore enabled us to decouple peptide induced allosteric communication from the formation of oligomers , both of which could have contributed to the shift of the isoleucine peaks in the NMR experiments . To address this notion , we repeated the NMR experiments with the IF2L mutant ( Figure 7F , Figure 7—figure supplement 4A–C ) . Similar to WT hIRE1α cLD , MPZ1-N peptide binding to hIRE1α cLD IF2L mutant shifted isoleucines in the β-sheet floor ( marked by Ile124 and Ile128 ) ( Figure 7F and G , Figure 7—figure supplement 4B and C ) . Importantly , isoleucine peaks ( Ile52 and Ile263 ) close to the oligomerization interfaces also shifted upon peptide binding to the hIRE1α cLD IF2L mutant . Thus peptide binding-induced conformational changes in isoleucines distant to the peptide binding site persisted in the hIRE1α cLD IF2L mutant . Interestingly , MPZ1-N-2X binding shifted isoleucine peaks in the same direction and to a similar extent as binding of MPZ1-N , indicating that hIRE1α cLD IF2L binds to the same site in these peptides ( Figure 7G , Figure 7—figure supplement 4B and C ) . These data suggest that the increased affinity of MPZ1-N-2X is due to a decreased rate of dissociation of the peptide . To test the importance of lumenal domain driven oligomerization for hIRE1α function in vivo , we generated cell lines that stably express hIRE1α IF2L mutant as the only form of hIRE1α . To this end , we introduced the hIRE1α IF2L mutant into mouse embryonic fibroblasts ( MEFs ) deficient for both isoforms of IRE1 ( IRE1α−/−/IRE1β−/− ) . In addition , we attached a GFP tag to IRE1’s cytoplasmic flexible linker retaining its function as published previously for HEK293 cells ( Li et al . , 2010 ) . In parallel , we introduced hIRE1α-GFP to IRE1α−/−/IRE1β−/− MEFs to compare hIRE1α activity at similar conditions . In these cell lines , we controlled hIRE1α expression via a doxycycline-inducible promoter . In the absence of doxycycline , cells expressed low levels of hIRE1α due to the leakiness of the promoter . In those conditions , the expression level of the hIRE1α-GFP-IF2L mutant was similar to hIRE1α-GFP and to the level of endogenous IRE1α from wild-type MEFs , as assessed by Western blot analysis ( Figure 8A , B ) . We next monitored the XBP1 mRNA splicing activity of IRE1 in IRE1α−/−/IRE1β−/− MEFs harboring hIRE1α-GFP or hIRE1α-GFP-IF2L mutant ( Figure 8C ) . We found that unlike hIRE1α-GFP , hIRE1α-GFP-IF2L mutant did not splice XBP1 mRNA after induction of ER stress by tunicamycin , a chemical stressor that impairs ER-folding homeostasis by inhibiting N-linked glycosylation ( Figure 8C , Figure 8—figure supplement 1 ) ( Heifetz et al . , 1979 ) . IRE1’s RNase activity is preceded by the autophosphorylation of its kinase domain , which can be monitored by a phospho-specific antibody . Western blot analysis showed no signal corresponding to phospho-IRE1 in the IRE1α−/−/IRE1β−/− cells expressing hIRE1α--GFP-IF2L , by contrast to the same cells reconstituted with wild type hIRE1α-GFP , or in contrast to wild type MEFs , in which we detected phosphorylation of the endogenous protein ( Figure 8B , Figure 8—figure supplement 2 ) . Lastly , confocal microscopy revealed that under ER stress conditions where hIRE1α-GFP readily formed foci ( >70% , n = 88 , Figure 8D , Figure 8—figure supplement 3A ) , reflecting its assembly into active oligomers , the hIRE1α-GFP-IF2L mutant failed to do so ( Figure 8D , Figure 8—figure supplement 3B and C ) . These data confirmed that cLD-mediated oligomerization is crucial for IRE1 function in cells .
To date , the mechanism by which mammalian IRE1 senses ER stress has remained controversial . Here , we provide evidence that activation of human IRE1α occurs via direct recognition of unfolded proteins and that the mechanism of ER stress sensing is conserved from yeast to mammals . This conclusion is based on six independent lines of evidence . First , we found that hIRE1α cLD binds peptides with a characteristic amino acid bias . Second , NMR spectroscopy suggested that peptides bind to hIRE1α cLD’s MHC-like groove and induce a conformational change including the distant αB helix . In this way , occupation of the peptide-binding groove is allosterically communicated , which , we propose , culminates in the formation of a functional oligomerization interface corresponding to IF2L in yIRE1 cLD . Third , binding of minimal-length peptides induces formation of hIRE1α cLD oligomers as assessed by AUC analyses , further supporting this notion . Fourth , cross-linking experiments captured the oligomerization interfaces , which allowed identification of a functionally crucial hydrophobic patch at IF2L . Fifth , mutation of this patch uncoupled peptide binding from oligomerization but retained the allosteric coupling within the domain . Sixth , impairing the oligomerization of hIRE1α cLD abolished IRE1’s activity in living cells , attesting to the physiological relevance of the activation mechanism proposed here . Taken together , our data converge on a model ( Figure 9 ) in which unfolded protein-binding activates a switch in hIRE1α’s cLD , leading to rearrangements that render it compatible with the formation of IF2L and therefore stabilizing an active oligomeric conformation ( Video 1 ) . cLD-mediated oligomerization on the lumenal side of the ER , in turn , would juxtapose hIRE1α’s cytosolic kinase domains in the face-to-face confirmation allowing its trans-autophosphorylation , followed by stacking of its RNase domains in back-to-back orientation . These conformational rearrangements then lead to RNase activation , and thus allowing information flow across the ER membrane . Interestingly , our data show that impairment of lumenal domain oligomerization diminished IRE1’s both RNase and kinase activities in cells . Currently due to lack of biochemical and structural understanding of IRE1’s interaction with the ER-resident chaperone BiP , its role in regulating IRE1 activity remains unknown . Although it is clear that BiP is released from IRE1 upon ER stress ( Bertolotti et al . , 2000 ) , current models proposing BiP as the primary regulator of IRE1 activity do not address how active IRE1 oligomers would form ( Carrara et al . , 2015; Oikawa et al . , 2009; Zhou et al . , 2006 ) . By contrast , our data indicate that peptide-binding is important for lumenal domain-driven IRE1 oligomerization , leading to its activation . We therefore consider it most plausible that BiP binding modulates the response via tuning IRE1’s oligomerization equilibrium , similar to what was shown for the yeast counterpart ( Pincus et al . , 2010 ) . In this way , BiP binding would buffer IRE1 activity at the early stages of the ER stress when the chaperones are not overwhelmed by the unfolded protein load , and during the deactivation phase , when the protein folding homeostasis is achieved . In this scenario , unfolded protein accumulation exerts synergistic effects on IRE1 activation , simultaneously freeing more IRE1 from BiP upon ER stress and inducing IRE1’s oligomerization/activation through their direct binding to the sensor ( Pincus et al . , 2010 ) . Despite these profound similarities in the salient features of ER stress sensing and processing , yeast and human IRE1α cLD display some distinct oligomerization properties . Whereas yIRE1 cLD precipitously assembles into larger oligomers at concentrations that exceed its dimerization constant ( Gardner and Walter , 2011 ) , hIRE1α cLD forms discrete dimers , which in a concentration-dependent manner gradually assemble into tetramers . hIRE1α cLD oligomers are in a dynamic equilibrium of different states , apparent from our size exclusion chromatography and AUC analyses and hIRE1α cLD forms even larger oligomers when bound to peptides . These observations are consistent with the model that the αB helix , which may hinder formation of hIRE1α oligomers as previously suggested ( Zhou et al . , 2006 ) participates in conformational changes that release its block on oligomerization . At higher hIRE1α cLD concentrations , the conformational equilibrium of the αB helix is shifted towards the active state . Peptide binding allosterically releases this inhibition and stabilizes the active hIRE1α oligomers . We anticipate that the effect of peptide binding-induced oligomerization would be more pronounced under physiological conditions , where hIRE1α is tethered to ER-membrane with diffusion limited to two dimensions . We speculate that the conformational change in the αB helix allows the incomplete β-propeller to form contacts with the residues from the flexible region , which is not resolved in the crystal structure ( V307-Y358 ) forming the oligomerization interface in hIRE1α cLD . In this conformation , αB helix may provide additional contact sites contributing to the oligomerization interface . Interestingly , one of the symmetry mates captured by hIRE1α cLD crystal structure shows contacts of the αB helix with the hydrophobic stretch ( 359WLLI362 ) , which we show to be important for oligomerization . We anticipate that in addition to this hydrophobic stretch , additional contacts contributed by these flexible parts may further facilitate oligomer formation . hIRE1α cLD’s groove is enriched in aromatic residues and displays a negatively charged surface . In this way , the amino acids lining the groove chemically complement hIRE1α cLD binding peptides identified in our study , which are enriched in aromatics and arginines . In the crystal structure of hIRE1α cLD in the ‘closed’ conformation , the α-helices forming the MHC-like groove are close together and mask the residues on the β-sheet floor . When these helices are moved approximately 6 Å apart from one another , the groove deepens and exposes more hydrophobicity mostly contributed by newly exposed aromatic residues . Thus , opening the groove exposes surface chemistry that is conducive to IRE1 binding peptides . Our data support a model in which widening of the groove is allosterically coupled to the formation of the IF2L-like oligomerization interfaces . We showed that a 12-mer peptide is the shortest derivative of MPZ1 peptide that binds hIRE1α cLD with undiminished affinity when compared to the original 21-mer peptide , indicating that a 12-mer provides maximal contact with cognate interfaces in hIRE1α cLD groove . It is plausible that similar to MHC molecules , select amino acids in unfolded polypeptides act as ‘anchor residues’ providing contact sites for hIRE1 α cLD binding ( Fremont et al . , 1992; Matsumura et al . , 1992; Wilson and Fremont , 1993 ) . Notably , assuming an extended peptide backbone with an average length of 3 . 4 Å per peptide bond , a 12-mer peptide can fit without constraints into the 39 Å-long groove in the structural model presented here . This notion suggests that the groove ensures preferential binding of fully exposed , unfolded 39Å-stretch of a polypeptide chain . This recognition principle is therefore similar to that of Hsp70-type chaperones , where the structural constraints in the cavity of the substrate-binding domain allow interaction with the substrates only in their extended , unfolded conformation , although Hsp70 only binds a seven amino acids stretch ( Rüdiger et al . , 1997a; Rüdiger et al . , 1997b ) . Supporting the notion of mechanistic similarities in unfolded protein recognition between chaperone proteins and IRE1 , hIRE1α cLD and the ER-resident chaperone BiP bind partially overlapping as well as distinct sets of peptides tested in our peptide arrays , as previously shown for the orthologous yeast proteins ( Gardner and Walter , 2011 ) . Importantly , the presence of distinct hIRE1α cLD binding peptides liberates IRE1 from an otherwise inevitable failure to compete with highly abundant BiP for binding sites in unfolded proteins . hIRE1α cLD’s affinity for peptides measured here varied between 5 and 30 µM , which is within the same order of magnitude but at the lower range of those reported for most chaperones ( Karagöz et al . , 2014; Marcinowski et al . , 2011; Street et al . , 2011 ) . For example , hIRE1α cLD binds the IgG’s CH1 unfolded domain with ~30 µM affinity whereas BiP was shown to bind the same protein with ~7 µM ( Marcinowski et al . , 2011 ) . We surmise that this difference has been selected in evolution to set the threshold for unfolded protein recognition slightly higher for the UPR sensors when compared to that of chaperones so that the UPR is not triggered until a critical concentration of unfolded proteins accumulates . Moreover , our data with the MPZ1-N-2X peptide suggested that IRE1 could display higher affinity for select polypeptides that present more than a single IRE1 binding site . IRE1 dysfunction contributes to the development of numerous diseases , including cancer ( such as multiple myeloma [Mimura et al . , 2012] ) , metabolic disorders ( such as obesity and diabetes [Fonseca et al . , 2009; Hotamisligil , 2010] ) and neurodegenerative diseases ( such as amyotrophic lateral sclerosis and Hungtinton’s disease [Hetz et al . , 2009; Matus et al . , 2009; Vidal et al . , 2012] ) . Depending on the disease context , IRE1 makes life or death decisions in response to altered ER function manifested in these pathological conditions ( Walter P . and D . , 2011 ) . Our data showing that unfolded proteins stabilize a distinct IRE1 conformation suggest novel approaches to manipulate IRE1 pharmacologically . For example , it will be promising to design or screen for small molecule modulators that lock IRE1’s groove in the open or closed conformation based on the chemical signature of the IRE1 binding peptides identified here . Such compounds could act as agonists or antagonists of IRE1 activity . As such , it should be possible to develop new classes of pharmaceuticals to induce or inhibit the IRE1 branch of the UPR , driving the desired IRE1 output depending on the disease context .
Synthetic peptides were ordered from Elim Biosciences and GenScript at >95% purity . To express MBP-hIRE1α cLD ( aa 24–389 ) , human IRE1α cDNA sequences were cloned into a pMalC2p vector to create a hIRE1α cLD fused on its N-terminus to MBP . To express His10-hIRE1α cLD , hIRE1α cLD was cloned into pet16b ( + ) vector containing a FactorXa protease cleavage site . Additionally , His10-hIRE1α cLD and IRE1 LD coding sequences were cloned into pet47b ( + ) vector with a preScission protease cleavage site . Hamster BiP with an N-terminal His10-tag was cloned into pet16b ( + ) vector , which was modified to introduce a preScission protease site C-terminal to the His10-tag . For expression of the proteins , the plasmid of interest was transformed into Escherichia coli strain BL21DE3* RIPL ( Agilent Technologies ) or Rosetta2 cells ( Novagen ) . Cells were grown in Luria Broth at 37°C until OD600 = 0 . 6 . Protein expression was induced with 0 . 3 mM IPTG , and cells were grown at 21°C overnight . For selective labeling , cells were grown according to published protocols ( Tugarinov and Kay , 2004 ) . Briefly , cells were grown at minimal media in D2O supplemented with deuterated glucose as the primary carbon source . For purification , cells were resuspended in Lysis Buffer ( 50 mM HEPES pH 7 . 2 , 400 mM NaCl , 4 mM dithiothreitol ( DTT ) ( or 5 mM β-mercaptoethanol , if a nickel column was used ) ) and were lysed in an Avestin EmulsiFlex-C3 cell disruptor at 16 , 000 psi . The supernatant was collected after centrifugation for 40 min at 30 , 000xg . MBP-IRE1 cLD constructs were purified on an MBP-amylose resin ( New England Biolabs ) and eluted with 10 mM amylose in Elution Buffer ( 50 mM HEPES pH 7 . 2 , 150 mM NaCl , 4 mM DTT ) after washing the column with 20 column volumes of Lysis Buffer . The eluate was then diluted with 50 mM HEPES ( pH 7 . 2 ) buffer to 50 mM NaCl and applied to a MonoQ ion exchange column and eluted with a linear gradient from 50 mM to 1 M NaCl . The protein was further purified on a Superdex 200 10/300 gel filtration column equilibrated with Buffer A ( 25 mM HEPES pH 7 . 2 , 150 mM NaCl , 2 mM tris ( 2carboxyethyl ) phosphine ( TCEP ) . The initial purification of His6- and His10-hIRE1α cLD and His10-BiP constructs were performed on a His-TRAP column ( GE Healthcare ) , where the protein was eluted with gradient from 20 mM to 500 mM imidazole . The eluate was purified on a MonoQ column , before the His6-tag ( pet47b+ ) or His10-tag ( pet16b+ ) were removed by either PreScission protease ( GE Healthcare , 1 unit of enzyme for 100 µg of protein ) or FactorXa ( NEB , 1 µg of FactorXa per 100 µg of protein ) , respectively . The tag removal was performed at 4o C overnight after the protein concentration was adjusted to 1 mg/mL . CH1 domain of IgG was purified under reducing conditions as described ( Feige et al . , 2009 ) . Protein concentrations were determined using extinction coefficient at 280 nm predicted by the Expasy ProtParam tool ( http://web . expasy . org/protparam/ ) . Peptide arrays were purchased from the MIT Biopolymers Laboratory . The tiling arrays were composed of 18-mer peptides that were tiled along the CPY* , MPZ , insulin , lysozyme and PTIP sequences with a three amino acid shift between adjacent spots . In the mutational arrays , peptides were synthesized to systematically mutate each amino acid in the core region of the CPY*-derived peptide . The arrays were incubated in 100% methanol for 10 min , then in Binding Buffer ( 50 mM HEPES pH 7 . 2 , 150 mM NaCl , 0 . 02% Tween-20 , 2 mM DTT ) three times for 10 min each . For BiP experiments , ADP and MgCl2 were added to the binding buffer to final concentrations of 1 mM and 5 mM , respectively . The arrays were then incubated for 1 hr at room temperature with 500 nM MBP-hIRE1α cLD or His10-BiP and washed again three times with 10 min incubation in between the washes in Binding Buffer to remove any unbound protein . Using a semi-dry transfer apparatus , the bound protein was transferred to a nitrocellulose membrane and detected with anti-MBP antiserum ( NEB ) or anti-His6 antibody ( Abcam ) . The contribution of each amino acid to hIRE1α cLD and BiP binding was calculated as described previously ( Gardner and Walter , 2011 ) . The peptide arrays were quantified using Max Quant . The binding intensity in each spot was normalized to max signal intensity in the peptide array . The peptides with the top 10% binding scores were selected and the occurrence of each amino acid in these top-binding peptides was calculated . This value is normalized to their abundance in the arrays ( Figure 2A ) . To calculate experimental error , the amino acid occurrences of top binders were calculated for independent replicates . The statistical significance ( p<0 . 05 ) is determined using non-paired t-test by the Prism software ( Figure 2—figure supplement 1A ) . For fluorescence anisotropy measurements , MPZ1 peptide attached to 5-carboxyfluorescein ( 5-FAM ) at its C-terminus was obtained at >95% purity from ELIM Biopharmaceuticals . For the remaining peptides ( 8ab1 , MPZ1-N , MPZ1-M , MPZ1-C and MPZ1-N ) derivatives were synthesized with 5-FAM attached to their N-terminus by GenScript at >95% purity . Binding affinities of hIRE1α cLD or IRE1 mutants to FAM-labeled peptides were measured by the change in fluorescence anisotropy on a Spectramax-M5 plate reader with excitation at 485 nm and emission at 525 nm with increasing concentrations of hIRE1α cLD . Fluorescently labeled peptides were used in a concentration range of 50–100 nM . The reaction volume of each data point was 20 µL and the measurements were performed in 384-well , black flat-bottomed plates after incubation of peptide with hIRE1α cLD or its mutants for 30 min at 25o C . Binding curves were fitted using Prism Software ( GraphPad ) using the following equation: Fbound = rfree + ( rmax- rfree ) / ( 1 + 10 ( ( LogK1/2-X ) •nH ) ) , where Fbound is the fraction of peptide bound , rmax and rfree are the anisotropy values at maximum and minimum plateaus , respectively . nH is the Hill coefficient and x is the concentration of the protein in log scale . Curvefitting was performed with minimal constraints to obtain K1/2 values with high R2 values . However , as this equation does not take into account the equilibria between hIRE1α cLD dimers/oligomers , these apperant K1/2 values do not reflect the dissociation constant . MST experiments were performed with a Monolith NT . 115 instrument ( NanoTemper Technologies , Germany ) . All experiments were done with the following buffer: 25 mM HEPES pH 7 . 2 , 150 mM NaCl , 1 mM TCEP , 0 . 025% Tween-20 . hIRE1α cLD was labeled using the Monolith NT Protein labeling Kit Red-Maleimide . Labeled protein was used in the measurements at a concentration of 50 nM . It was mixed with equal volumes unlabeled interaction partner in two-fold serial dilutions . Hydrophilic-treated capillaries ( NanoTemper Technologies ) were used for all the measurements . All experiments were performed at 50% LED power and 40-60–80% IR-laser at 25°C . Sedimentation velocity experiments were carried out in a Beckman Optima XL-A analytical centrifuge at 40 , 000xg at 20°C with An-60 Ti rotor . All experiments were performed in buffer containing 25 mM HEPES pH 7 . 2 , 150 mM NaCl , 2 mM DTT . Samples ( 400 µL ) and reference buffer ( 410 µL ) were loaded into AUC cells for each experiment . Samples of hIRE1α cLD at 5 µM were scanned at 280 nm , whereas hIRE1α cLD at concentrations higher than 25 µM were scanned at 290 nm to prevent detector saturation at high protein concentrations . Data analysis was performed using the SEDFIT software employing the c ( s ) method with time invariant and radial invariant noise fitting ( Schuck , 2000 ) . Buffer viscosity was calculated by Sednterp . NMR experiments were performed on an 800 MHz Bruker AVANCE-I spectrometer with a TXI Cryoprobe equipped with an actively shielded Z-gradient at 298 . 0 K . Samples were buffer-exchanged into 25 mM phosphate buffer pH 7 . 2 , 150 mM NaCl and 2 mM DTT in 100% D2O on Vivaspin columns ( Millipore ) . The concentration of WT hIRE1α cLD and hIRE1α cLD mutants varied from 25 to 400 µM in 250 µL volume . Samples were placed in a Shigemi advanced NMR microtube . For peptide and unfolded protein binding experiments , the peptides were dissolved in the same buffer at high concentrations ( 1–2 mM ) and titrated in 1:0 . 5 , 1:1 and 1:2 molar ratios . Two-dimensional [13C , 1 H]-HMQC methyl correlation experiments on 13CH3–Ile hIRE1α cLD were acquired with 86* and 768* complex points in the 13C and 1H dimensions , respectively . All spectra were processed with TOPSPIN 3 . 2 and analyzed with Sparky . MPZ1 peptide at 200 µM was labeled with 3- ( 2-iodoacetamido ) -proxyl ( Sigma ) at the single cysteine , Cys5 in 25 mM phosphate buffer pH 7 . 2 , 150 mM NaCl in the presence of 2 mM spin-label at 4°C for 8 hr . The labeled peptide was then dialyzed in a Slide-a-Lyzer dialysis cassette ( Thermo Fisher Scientific ) with 2 kDa cut-off to remove the excess spin-label and to exchange the buffer to deuterated buffer ( 25 mM phosphate buffer pH 7 . 2 , 150 mM NaCl , 2 mM DTT ) for NMR experiments . Control samples used in the reference experiments contained ( 1-oxyl-2 , 2 , 5 , 5-tetramethylpyrroline-3-methyl ) methanethiosulfonate spin-label that was treated the same way as the proxyl-labeled peptide . Wild type hIRE1α cLD and quadruple mutant hIRE1α cLD ( Leu186Ile , Ile326/334/362Val ) and single mutant Leu186Ile were used in PRE experiments at 75 µM and 100 µM protein concentration respectively , in the presence and absence of equimolar concentrations of MPZ1-proxyl peptide . We normalized the PRE effect with the surface exposed area displayed by that isoleucine to exclude possible contributions from non-specific interactions with the spin label attached peptide ( Clore and Iwahara , 2009 ) . The normalized PRE values are calculated as follows , the solvent accessible surface area for isoleucines are calculated using the ‘GETAREA’ webserver ( http://curie . utmb . edu/getarea . html , [Fraczkiewicz and Braun , 1998] ) based on hIRE1α cLD crystal structure . The maximum solvent accessible surface by these isoleucines is normalized to one and the normalized values are multiplied with the PRE effect . The PRE effect is calculated by dividing the intensity of isoleucine signals in the control experiments with the isoleucine signals in the presence of MPZ1-proxyl peptide . 10 µM , 20 µM and 50 µM hIRE1α cLD was incubated with 500 µM and 1 mM BS3 cross-linker for 15 and 30 min at room temperature . Same reaction was performed for hIRE1 cLD pre-bound to 50 µM MPZ1-N for 30 min on ice . The reaction was stopped with the addition of 1M Tris-HCl at pH 8 . 0 at end concentration of 50 mM Tris-HCl , and incubated for 10 min at room temperature before running the SDS-PAGE gel . Cross-linked products were in-gel digested and analyzed by LC-MS and LC-MS-MS as described previously ( Wu et al . , 2013; Zeng-Elmore et al . , 2014 ) . Briefly , 1 μl aliquot of the digestion mixture was injected into an Dionex Ultimate 3000 RSLCnano UHPLC system ( Dionex Corporation , Sunnyvale , CA ) , and separated by a 75 μm × 25 cm PepMap RSLC column ( 100 Å , 2 µm ) at a flow rate of ~450 nl/min . The eluant was connected directly to a nanoelectrospray ionization source of an LTQ Orbitrap XL mass spectrometer ( Thermo Scientific , Waltham , MA ) . LC-MS data were acquired in an information-dependent acquisition mode , cycling between a MS scan ( m/z 315–2 , 000 ) acquired in the Orbitrap , followed by low-energy CID analysis on three most intense multiply charged precursors acquired in the linear ion trap . Cross-linked peptides were identified using an integrated module in Protein Prospector , based on a bioinformatic strategy described previously ( Chu et al . , 2010; Trnka et al . , 2014 ) . The score of a cross-linked peptide was based on number and types of fragment ions identified , as well as the sequence and charge state of the cross-linked peptide . Only results where the score difference is greater than 0 ( i . e . the cross-linked peptide match was better than a single peptide match alone ) are considered . Tandem MS spectra of cross-linked peptides were manually inspected to ensure data quality . With the threshold of peptide score and expectation value for oligomer-only cross-linked peptides , no decoy match was returned . The coding sequence of wild type GFP-tagged IRE1 ( Li et al . , 2010 ) was amplified by PCR with Phsuion polymerase ( NEB ) and oligonucleotides with engineered restriction sites . The PCR product was introduced into the Gateway entry vector pSHUTTLE-CMV-TO ( kind gift of A . Ashkenazi , Genentech and ( Gray et al . , 2007 ) atcognate KpnI and EcoRI sites . The hIRE1α-GFP- IF2L mutant was generated in pSHUTTLE-CMV-TO by site directed mutagenesis of the wild-type sequence . The resulting clones were recombined into pGpHUSH . puro ( kind gift of A . Ashkenazi , Genentech and [Gray et al . , 2007] ) , a single lentivirus expression vector that allows the doxycyline-regulatable ( TetON ) expression of a gene-of-interest . VSV-G pseudotyped lentiviral particles were prepared using standard protocols using 293METR packaging cells ( kind gift of Brian Ravinovich , formerly at MD Anderson Cancer Center , [Rabinovich et al . , 2006] ) . Viral supernatants were concentrated by filtration ( Amicon Ultra centrifugal filter device , 100 kDa MWCO ) and used to infect target cells by centrifugal inoculation ( spinoculation ) at 2000 rpm inn a Beckman GH3 . 8 rotor outfitted with plate carriers for 90 min in presence of 8 ug/mL polybrene . The cells were left to recover overnight following infection and were then subjected to puromycin selection as described below . IRE1 double-knockout Mouse Embryonic Fibroblasts ( MEF ) ( IRE1α−/−/IRE1β−/− ) and wild-type MEFs ( kind gift of D . Ron , University of Cambridge ) . were grown in DMEM supplemented with 10% fetal bovine serum , 2 mM L-glutamine , and penicillin/streptomycin . Cells were not tested for the mycoplasma contamination . Lentiviral-transduced cells were selected with 6 µg/mL puromycin for 72 hr based on the puromycin concentration defined by the kill curve . Subsequently , a pulse of 25 nM doxycycline was given to induce expression of the GFP-tagged IRE1 transgenes for 10–12 hr . The following day , the doxycycline was washed out and pseudoclonal cell populations were selected by fluorescent activated cell sorting based on GFP expression for both wild-type ( hIRE1α-GFP ) and IF2L mutant ( hIRE1α-GFP-IF2L mutant ) forms of IRE1 . The cells were selected in a FACS Aria instrument ( BD FACSAria3 ) , gating for a very narrow GFP expressing population . This procedure ensures selection of a pseudoclonal population where most cells have similar levels of expression of the transgene of interest while avoiding typical problems associated with monoclonal selection of IRE1-expressing cells; namely an aberrant UPR . The pseudoclonal populations were expanded and frozen as source stocks for experiments . IRE1 double-knockout MEFs ( IRE1α−/−/IRE1β−/− ) reconstituted with of hIRE1α -GFP or hIRE1α-GFP-IF2L mutant were split 2 days before imaging onto ibiTreat dishes ( ibidi ) at 5 × 104 cells/dish . 25 nM Doxyccline containing medium was added for 10–12 hr , withdrawn before imaging and replaced with imaging media consisting of Fluorobrite DMEM ( Thermo Scientific ) , 2 . 5% FBS , and 5 mM Hepes at a pH of 7 . 0 . Cells were imaged at 37oC on a spinning disk confocal with Yokogawa CSUX A1 scan head , Andor iXon EMCCD camera and 40x Plan Apo air Objective NA 0 . 95 with a 1 . 5x tube lens for additional magnification giving 60x final or 100X objective . Images were acquired using 488 nm laser at a rate of one frame per 3 min with 300 ms exposure time for each time point for an hour . Images were collected after different time points following induction of ER stress by tunicamycin ( 5 µg/mL ) or thapsigargin ( 100 nM ) . IRE1 double-knockout MEFs ( IRE1α−/−/IRE1β−/− ) reconstituted with of hIRE1α -GFP and hIRE1α-GFP-IF2L mutant were grown similar to live cell imaging experiments . After stress induction at various time points , cells were washed three times with PBS followed by 3 min fixation with 100% methanol , and a three subsequent 5 min washes with PBS . As these fixation conditions kept GFP intact , immunostaining of hIRE1α for fluorescence imaging was not required . DAPI staining is performed according to manufacturer’s instructions ( Thermo Fisher ) . Cells exposed to DMSO or thapsigargin ( 100 nM ) or tunicamycin ( 5 µg/ml ) were collected in 0 . 5 ml of TRIzol reagent ( Life Technologies ) from a six well dish and total RNA was extracted following the manufacturer’s recommendations . To generate cDNAs , 500 ng of total RNA were reverse transcribed using the SuperScript VILO system ( Life Technologies ) following the manufacturer’s recommendations . The resulting 20 µl reverse transcription reactions were diluted to 10 times to 200 µl with 10 mM Tris– HCl pH 8 . 2 , and 1% of this dilution was used for multiplex semiquantitative PCR . The multiplex PCR was set up using 1 µM of the forward reverese primers , 0 . 4 units of Taq DNA polymerase ( Thermo Scientific ) , 0 . 2 mM of each dNTP , and 1 . 5 mM MgCl2 , in a 20 µl reaction using the following buffer system: 75 mM Tris–HCl pH 8 . 8 , 20 mM ( NH4 ) SO4 , and 0 . 01% Tween-20 . The oligonucleotide sequences are the following: Hs_XBP1_Fwd: 50 -GGAGTT AAGACAGCGCTTGG-30; Hs_XBP1_Rev: 50 -ACTGGGTCCAAGTTG TCCAG-30 . The PCR products were amplified for 28 cycles and resolved on 3% agarose gels ( 1:1 mixture of regular and low-melting point agarose ) stained with ethidium bromide . Cells were lysed in SDS-PAGE loading buffer ( 1% SDS , 62 . 5 mM Tris-HCl pH 6 . 8 , 10% glycerol ) . Lysates were sonicated and equal amounts were loaded on SDS-PAGE gels ( BioRad , Hercules , CA ) . Proteins were transferred onto nitrocellulose membranes and probed with primary antibodies diluted in Phosphate-buffered saline supplemented with 0 . 1% Tween 20% and 5% bovine serum albumin at 4°C , overnight . The following antibodies were used: anti-IRE1 ( 1:1000 ) ( 14C10 , Cell Signaling Technology , Danvers , MA ) , anti-GAPDH ( 1:1000 ) ( 14C10 , Cell Signaling Technology , Danvers , MA and anti-phosho IRE1 antibody ( 1:500 ) . IRE1 anti-phospho antibody is a kind gift of Avi Ashkenazi’s group at Genentech . An HRP-conjugated secondary antibody ( Amersham , Piscataway , NJ ) was employed to detect immunereactive bands using enhanced chemiluminescence ( SuperSignal; Thermo Scientific , Waltham , MA ) detected by Li-Cor instrument ( Li-Core Biosciences ) . | Proteins are long string-like molecules that fold into specific three-dimensional shapes . Most proteins that a cell uses to communicate with its environment are folded within a part of the cell called the endoplasmic reticulum . Dedicated sensor proteins in this cellular compartment track this process to make sure that it continues to meet the cell’s demand for protein folding . If it cannot meet the demand , unfolded or poorly folded proteins build up , which stresses the cell . IRE1 is a sensor protein that detects stress in the endoplasmic reticulum . It is found in a range of organisms from yeast to humans , where it spans the membrane that encloses the endoplasmic reticulum . When unfolded proteins accumulate , IRE1 proteins come together and form so-called oligomers . The IRE1 oligomers then become active and send signals outside of the endoplasmic reticulum . These signals adjust the cell’s protein-folding capacity according to its needs at that time . The yeast version of IRE1 directly recognizes unfolded proteins in the endoplasmic reticulum . Yet , its human counterpart was found to have a different three-dimensional structure , which suggested that it might use a different mechanism to detect the stress . Now , Karagöz et al . show that , as in yeast , the sensor part of human IRE1 does indeed bind to unfolded proteins directly . This binding causes this part of the protein to engage other copies of IRE1 and form the oligomers . To understand this interaction in more detail , Karagöz et al . used a technique called nuclear magnetic resonance spectroscopy to monitor changes in the shape of proteins . These observations revealed that binding to an unfolded protein causes other parts of IRE1 protein to change shape . In turn , these shape changes act as a switch that causes the oligomers to form . Stopping the sensor domains from forming oligomers inactivated the IRE1 protein in mammalian cells; this rendered IRE1 unresponsive to stress within the endoplasmic reticulum . The regulation of IRE1 affects many health disorders , including diabetes , cancer and neurodegenerative diseases . By showing that unfolded proteins switch IRE1 into its active , oligomeric state , these findings might lead to new approaches to manipulate IRE1’s activity with small molecules to help to treat these diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | An unfolded protein-induced conformational switch activates mammalian IRE1 |
The root cap has a fundamental role in sensing environmental cues as well as regulating root growth via altered meristem activity . Despite this well-established role in the control of developmental processes in roots , the root cap’s function in nutrition remains obscure . Here , we uncover its role in phosphate nutrition by targeted cellular inactivation or phosphate transport complementation in Arabidopsis , using a transactivation strategy with an innovative high-resolution real-time 33P imaging technique . Remarkably , the diminutive size of the root cap cells at the root-to-soil exchange surface accounts for a significant amount of the total seedling phosphate uptake ( approximately 20% ) . This level of Pi absorption is sufficient for shoot biomass production ( up to a 180% gain in soil ) , as well as repression of Pi starvation-induced genes . These results extend our understanding of this important tissue from its previously described roles in environmental perception to novel functions in mineral nutrition and homeostasis control .
Mineral nutrition has an essential function in plant roots and its spatial localization has been a subject of controversy for a long time . Following the invention of the compound microscope , pioneering descriptions of plant tissues in the seventeenth century by Marcello Malpighi ( Malpighi , 1675 ) and Nehemiah Grew ( Grew , 1682 ) established the bases of plant anatomy . Although both authors drew a connection between plant nutrition and distinct anatomical location , Malpighi proposed the root hairs whereas Grew proposed the root cap , at the distal end of the root tip . Our understanding of the spatial localization of plant nutrition has undergone great advances in the past twenty-five years . Combined efforts in molecular biology , genetics , biochemistry and electrophysiology have identified several key aspects to ion uptake , with a central role for plasma membrane ( PM ) ion transporters in this process . These transporters often belong to broad multigenic families ( a likely consequence of their vital role ) with overlapping expression patterns ( Nussaume et al . , 2011 ) . However , this high redundancy has held back their genetic study to some degree , preventing analysis of ion uptake contribution at specific root localities . The uptake of phosphate ( Pi ) , an essential plant macronutrient , relies on a family of nine high-affinity transporters , identified as PHT1 in the plant model Arabidopsis ( Nussaume et al . , 2011 ) . Mineral nutrition is classically associated with the root epidermis , and the various PHT1 members were identified within this cell layer ( Mudge et al . , 2002; Karthikeyan et al . , 2002; Misson et al . , 2004; Nussaume et al . , 2011 ) . These transporters are particularly abundant where hair cells develop ( Daram et al . , 1998; Brady et al . , 2007 ) , as this increases the surface area in contact with the environment and improves nutrient uptake ( Peret et al . , 2011 ) . Transcriptomic analyses of specific root cell layers ( Birnbaum et al . , 2003 ) and analysis of the PHT1 expression pattern has revealed the accumulation of these transporters in the root tip ( Mudge et al . , 2002; Karthikeyan et al . , 2002; Misson et al . , 2004 ) . Nevertheless , determining how PHT1 proteins contribute to Pi uptake at the root tip is difficult , particularly since this area contains the primary root meristem . This tissue , which is essential for root growth , generates new cells that require Pi and its accumulation to build essential components including nucleic acids , ATP , and phospholipids . Consequently , this complicates differentiating Pi uptake from the Pi translocation ( from epidermis ) necessary to sustain active cell division in the root tip . Our approach to overcome this problem employs a recently developed high-resolution live radioisotope micro-imaging system ( Kanno et al . , 2012 ) combined with targeted cell ablation or complementation by genetic manipulation . These findings establish the importance of the root cap in phosphate uptake and homeostasis .
PHT1;1 and PHT1;4 , the two most important high-affinity phosphate transporters required for up to 75% of Pi uptake in Arabidopsis , are localized to the distal part of the root tip ( Shin et al . , 2004 ) . Both transporters are observed in the root tip ( primarily in the root cap ) , where they are expressed at the transcriptional and protein levels ( Figure 1A ) with PHF1 , a crucial component facilitating PHT1 post-translational regulation and its targeting to the PM ( Gonzalez et al . , 2005; Bayle et al . , 2011 ) ( Figure 1A ) . The root tip surrounds and protects the meristem , while also acting as the initial contact point between the roots and soil . In order to visualize Pi in plants , pulses of radioactive tracer were applied by immersing the whole root system in 33Pi solution . When fed to plantlets , 33Pi accumulation was observed at the extremity of the root tip ( Figure 1B ) , including the meristematic zone ( Figure 1C ) . 10 . 7554/eLife . 14577 . 003Figure 1 . Active Pi transporters are localized in the root cap . ( A ) Reporter lines expressing transcriptional and translational fusions for the high affinity transporters expressed in the root ( PHT1;1 and PHT1;4 ) are localized in the root cap , in addition to PHF1 , a major post-translational regulator required for PHT1 targeting to the plasma membrane . Scale bars: 50 μm . ( B ) Accumulation of 33P in the root tip of Arabidopsis plantlets . Whole roots were immersed in 33P-enriched solution for 1 day . ( C ) Pi accumulation in the root tip is abolished by targeted ablation with 5FC in the Q0171>>FCY-UPP line . The short pulse of 33Pi applied to the WT and Q0171>>FCY-UPP lines was revealed by a live radioisotope microimaging system . Light transmission and 33P images are merged . 33P content is represented as color intensity . Scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 00310 . 7554/eLife . 14577 . 004Figure 1—figure supplement 1 . Assay for Pi translocation to the root tip . ( A ) 33P ( 200 kBq ) was initially applied to the middle of the root in a zone isolated from the medium by Vaseline . ( B ) The presence of 33P in the root tip is observed 30 min later by the live radioisotope microimaging system . Presented image is a magnified view of the box in ( A ) . Scale bar: 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 00410 . 7554/eLife . 14577 . 005Figure 1—figure supplement 2 . Conditional negative marker expression in the root cap . ( A ) Enzymatic transformation of 5-Fluorocytosine ( 5-FC ) into toxic 5-Fluoro-UMP by FCY and UPP enzymes . ( B ) Transactivation system used to express a conditional negative marker in the root cap . The Q0171 transgenic line contains a GAL4 activator gene ( in T-DNA ) driven by a minimal promoter ( enhancer trap ) , which is transactivated in the root cap by a cell layer-specific gene X ( black arrow ) . After crossing with a line containing the conditional ablation marker FCY-UPP driven by UAS , GAL4 binds to the UAS sequence , activating transcription of the GFP reporter gene as well as FCY-UPP ( red arrows ) . ( C ) Effect of 5FC ( 3 . 8 mM ) on GFP expression in the transgenic Q0171>>FCY-UPP line . The line contains a transactivating GFP marker ( green ) and an FCY-UPP fusion in the root cap . 5FC treatment abolishes expression of genes in the root cap ( as visualized by a strong reduction in GFP signal ) while keeping cells alive ( as visualized by the absence of PI staining in the nuclei ) . The heat shock treatment control promotes cell death ( visualized as PI nuclear staining ) . Images are focused on the surface of the lateral root cap . Magnifications of the images are shown in the inserts . Red: PI , green: GFP . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 00510 . 7554/eLife . 14577 . 006Figure 1—figure supplement 3 . Effect of 5FC ( 3 . 8 mM ) on primary root growth in the Q0171>>FCY-UPP line . Plantlet phenotypes are shown after the 4-day transfer to 5FC or DMSO ( control ) . Arrows indicate the tips of primary roots at the beginning ( blue ) and end ( black ) of the respective treatment . Primary root growth is shown at the beginning ( Day 0 ) and after 2 and 4 days of treatment . DMSO , a solvent of 5FC , was used as the control treatment . Values are means ± SD of 15 plantlets . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 006 Pi translocation absorbed by the whole root system contributes to the presence of Pi in the root tip , as revealed by the accumulation of radiotracer in the root tip within 30 min of 33Pi application to the middle of the root system ( Figure 1—figure supplement 1 ) . Consequently , we chose genetic approaches to decipher the role of the root tip in Pi uptake , since the small size of this tissue prevented application of 33Pi at the root tip in a precise or exclusive manner . To bypass the functional redundancy of Pi transporters ( comprising 9 distinct loci in Arabidopsis ) and their overlapping expression at the tissue and cell levels ( Nussaume et al . , 2011 ) , we produced lines in which elimination of the root cap was activated . Previously , this was achieved by introducing a diphtheria toxin A-chain gene driven by a root cap promoter ( Tsugeki and Fedoroff , 1999 ) . However , due to the essential functions of the root cap in several physiological processes , these plants exhibited roots with abnormal differentiation and very poor growth rate . Therefore , we employed a conditional approach using the negative marker FCY ( cytosine deaminase; Stougaard , 1993 ) coupled with uracil phosphoribosyltransferase ( UPP; Tiraby et al . , 1998 ) . These enzymes convert the innocuous 5-fluorocytosine ( 5FC ) to cytotoxic 5-fluorouracil and 5-fluoroUMP ( Figure 1—figure supplement 2A ) , which stop cellular activities . Genes encoding a FCY-UPP fusion were specifically expressed in root cap cells using the Arabidopsis GAL4/UAS binary expression system ( Laplaze et al . , 2005 ) . The FCY-UPP construct under control of a UAS promoter was introduced in a GAL4 line ( Q0171 ) selected for its specific root cap expression . This line also expresses GFP under the control of a UAS promoter ( Figure 1—figure supplement 2 ) , for visualizing tissues in which the transactivation took place ( Figure 1—figure supplement 2 ) . As expected , a five-day treatment with 5FC drastically reduced the expression of GFP in the root cap ( Figure 1—figure supplement 2 ) ; the low GFP signal may result from high GFP protein stability or low residual expression . Root tip cell viability was not affected , as revealed by propidium iodide staining ( PI; Figure 1—figure supplement 2 ) . In living cells , PI staining is restricted to the cell wall ( as in the non-control plants ) , whereas it is localized to nuclei in dead tissues , as in the heat-shocked WT control ( see inserts in Figure 1—figure supplement 2 ) . Pi absorption was visualized using a real-time radioisotope imaging system developed for plant nutrient uptake studies ( Kanno et al . , 2012 ) . Roots were treated with short pulses of 33Pi radiotracer and examined by live radioisotope microimaging . The ability of the root cap to promote Pi uptake is illustrated in Figure 1C . Expression of FCY-UPP in the root cap combined with 5FC treatment clearly abolished the 33Pi accumulation observed in either the WT or the transgenic line not treated with 5FC . Treating the WT line with 5FC had no effect on either plant growth ( results not shown ) or Pi accumulation , demonstrating its innocuous nature when FCY-UPP is not expressed ( Figure 1C ) . This result establishes the existence of an unexpected active Pi uptake process at the root cap level . Nevertheless , this FCY-UPP-based system cannot be used to quantify the physiological effect of root cap ablation on Pi status , since arrested root growth was observed within 2 days of 5FC treatment . This alters the root architecture and prevents any precise quantitative comparison with the WT control ( Figure 1—figure supplement 3 ) . As stated above , the high redundancy of Pi transporters poses technical difficulties to investigating their roles in a selected tissue . We used a PHF1 mutant ( phf1-1 ) to circumvent this obstacle , as this mutation strongly reduces PHT1 accumulation in the PM , resulting in a 70–80% reduction in Pi uptake ( Bayle et al . , 2011; Gonzalez et al . , 2005 ) . Consequently , the phf1-1 mutant exhibits phosphate starvation traits in Pi-rich medium , although its growth is only slightly reduced . This offers an appropriate genetic tool for targeted PHF1 complementation to restore Pi uptake in specific tissues . Using the same GAL4/UAS system described above , we back-crossed the GAL4 enhancer trap driving expression in the root cap ( line Q0171 ) in a phf1-1 background . The specific complementation of the root cap was obtained by introducing the UAS-PHF1 construct , producing the phf1 Q0171>>PHF1 line . The proper targeting of Pi transporters to the PM in the root tip of resulting plants was confirmed by introducing the fluorescent marker mCherry fused to the PHT1;4 gene driven by the constitutive 35S promoter . This produced a strong fluorescent signal in the columella and lateral root cap , where the GFP marker driven by GAL4/UAS was also observed ( Figure 2A ) . The proper targeting of the PHT1;4-mCherry fusion protein was validated by its colocalization with the PM-specific FM4-64 dye ( Figure 2—figure supplement 1A , B ) . This confirms that an effective restoration of PHT1 targeting in the root cap PM has taken place in the phf1 Q0171>>PHF1 line ( Figure 2—figure supplement 1A , C ) . A very low , diffuse fluorescence signal could also be detected in other tissues ( Figure 2—figure supplement 1C ) , corresponding to a previously reported low level of ER-retained protein in the phf1-1 mutant ( Gonzalez et al . , 2005; Bayle et al . , 2011 ) . The root cap specificity of the phf1 Q0171>>PHF1 complementation therefore provides a unique opportunity to investigate the effect of localized Pi uptake in the root cap with physiological studies . 10 . 7554/eLife . 14577 . 007Figure 2 . Root cap complementation of the phf1-1 mutant in the phf1 QO171>>PHF1 line . ( A ) Accumulation of PHT1;4 in the plasma membrane ( red; 35S:PHT1;4-mCherry ) correlates with PHF1 complementation in cells expressing GFP . Scale bars: 50 μm . Lower panels display magnified views of the mCherry image in the columella ( CC ) and lateral root cap ( LC ) cells . ( B ) Visualization of 33P absorption ( blue ) by real-time imaging . The image displays a time course during a 10 min period . ( C ) Quantification of radioactivity ( 200 kBq application ) in the root apex ( 0 . 2 mm from the tip ) . Values are means ± SD . 3 root tips were analyzed . ( D ) Quantification of 33P along the root after 1 min . Measurements were taken from the tip to a distance of 1 . 8 mm at 0 . 2 mm intervals . A representative graph is shown . The experiment was performed eight times giving the same trend . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 00710 . 7554/eLife . 14577 . 008Figure 2—figure supplement 1 . Colocalization of the plasma membrane marker FM4-64 and PHT1;4-mCherry in the phf1 Q0171>>PHF1 line . ( A ) Images of lateral root cap cells . For arrow , see ( B ) . ( B ) Intensity profile of FM4-64 and mCherry fluorescence signals along the white arrow indicated in ( A ) . ( C ) Images of a root epidermis cell . Yellow: FM4-64 , red: mCherry . Scale bar: 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 008 Visualization ( Figure 2B ) and quantification ( Figure 2C ) of Pi absorption at the root apex of the phf1Q0171>>PHF1 , phf1-1 and WT lines were performed by real-time imaging . Our time course experiment revealed rapid Pi absorption after only 1 min following 33P addition ( Figure 2B ) . We observed full restoration of Pi absorption at the phf1Q0171>>PHF1 root tip similar to the WT level , whereas no labeling was detected in the phf1-1 mutant during the same time period ( Figure 2B , C ) , demonstrating the successful functional complementation of Pi uptake activity . Real-time quantification of 33P uptake revealed a significant role for the root tip in Pi absorption as compared to the mature zone of the WT root ( Figure 2D ) . WT plants exhibited a very low level of radiotracer ( Figure 2D ) in the region between these two parts of the root ( corresponding to the elongation and differentiation zones; approximately 500–900 µm from the apex ) , indicating poor Pi absorption activity . Root labeling occurs in the differentiated zone ( ≥1 . 2 mm from the root tip ) as a result of uptake through the epidermis . The signal present in the WT was impaired in both phf1-1 and phf1Q0171>>PHF1 ( Figure 2D ) . This confirms the specificity of the root cap complementation in phf1Q0171>>PHF1 , since the 33P quantification only describes a specific enrichment in the phf1Q0171>>PHF1 root tip ( Figure 2D ) . Similarly , in the WT , the accumulation of P33 was high in the root tip . Previous experiments have revealed that Pi starvation drives major transcriptional regulation in the plant genome ( Thibaud et al . , 2010 ) . PHT1;4 is a well-established marker , and is highly induced under Pi deficiency ( Muchhal et al . , 1996; Misson et al . , 2004 ) . To determine its contribution to Pi sensing , we used a translational fusion construct comprising a GUS reporter gene associated with endogenous PHT1;4 ( Misson et al . , 2004 ) . This marker was present in a previously isolated phf1-2 mutant background , and was introduced into diverse enhancer lines ( Figure 3—figure supplement 2 ) . As a consequence , the complemented lines ( phf1J1092>>PHF1 , phf1Q0171>>PHF1 and phf1J0481>>PHF1 ) lost the strong induction of PHT1;4-GUS expression observed in the phf1-2 mutant when Pi is present in the growth medium . This complementation effect was confirmed at the molecular level by analyzing the expression of several Pi starvation markers previously identified in a whole-genome transcriptomic analysis ( Thibaud et al . , 2010 ) . This approach differentiated systemic and locally responsive genes according to their respective response to internal Pi status or to available Pi present in the growth medium . Most of the markers regulated by external phosphate were not affected , and were observed to react identically in WT , phf1-1 and phf1Q0171>>PHF1 backgrounds ( Figure 3—figure supplement 3A , Figure 3—figure supplement 3A—source data 1 ) . Analysis of the systemically regulated Pi starvation-induced genes revealed a different situation , in which a general repression of these markers occurred after transfer of the WT and phf1Q0171>>PHF1 lines into Pi-rich medium after 2 days ( results not shown ) or 3 days ( Figure 3D and Figure 3—figure supplement 3B , Figure 3—figure supplement 3B—source data 1 ) . As previously reported ( Gonzalez et al . , 2005 ) , these systemic markers are strongly induced when the phf1-1 mutant line is grown in high Pi ( Figure 3D and Figure 3—figure supplement 3B ) , whereas these genes are repressed in the WT . This confirms the result obtained with the GUS reporter gene fused with the systemically regulated PHT1;4 ( Figure 3—figure supplement 2 ) . The repression of the systemically regulated genes was partially rescued in the root cap PHF1-complemented line . This repression was equivalent to half the repression observed in the WT . This is not proportional to the additional Pi uptake related to root cap complementation ( at least 20% ) , confirming the existence of a non-proportional relationship between Pi uptake , growth and systemic regulation . In order to investigate whether a significant role for the root cap in Pi import also exists outside of Arabidopsis , we finally examined two very different plants: a monocot ( rice ) and the wild legume Lotus japonicus ( Fabaceae ) . Clear labeling of the root cap was observed in these specimens by applying the aforementioned protocol consisting of a short pulse of 33P radiotracer ( Figure 4 and Figure 4—figure supplement 1 ) . The migration of Pi is clearly visible in both plants beginning at the extremity of the root tip ( 1 min labeling ) and moving gradually toward the differentiated tissues of the root ( after 3 and 5 min in rice and lotus; Figure 4A and Figure 4—figure supplement 1 respectively ) . The quantification of 33P after 1 min ( Figure 4B and Figure 4—figure supplement 1 ) indicates that Pi uptake takes place in the root cap , as in Arabidopsis . These results provide evidence that Pi uptake at the root tip is a feature shared by several disparate plant species . 10 . 7554/eLife . 14577 . 018Figure 4 . Imaging of 33P uptake at the root tip in Oryza sativa . ( A ) Time course of 33P uptake ( 200 kBq application ) . ( B ) Quantification of 33P ( after 1 min absorption ) along the root in 0 . 2 mm intervals , extending 2 mm from the tip . Values are means ± SD of 4 plantlets . Scale bar: 0 . 5 mm . 33P level is represented as color intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 01810 . 7554/eLife . 14577 . 019Figure 4—figure supplement 1 . Imaging of 33P uptake at the Lotus japonicus root tip . ( A ) Time course of 33P uptake ( 200 kBq application ) . ( B ) Quantification of 33P ( after 1 min absorption ) along the root in 0 . 2 mm intervals , extending 2 mm from the tip . Values are means ± SD of 4 plantlets . Scale bar: 0 . 5 mm . 33P level is shown as color intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 14577 . 019
We have established the existence of a highly active Pi uptake process in the root cap through a combined approach utilizing genetics ( i . e . conditional ablation and the specific complementation of Pi transport ) with real-time imaging of radioisotopes . Our results indicate that the root cap contributes approximately 20% of the total amount of Pi absorbed from the medium by the roots . This Pi contributes to the plant’s total Pi pool , significantly sustaining the overall development of the plant . This is demonstrated by the 80 to 180% increase in biomass production observed in root cap-complemented lines as compared to the phf1-1 mutant , during the vegetative growth of plants in soil . The Arabidopsis root cap therefore appears to be a “hot spot” for Pi absorption , despite its small surface area . Interestingly , this confirms the hypothesis of the 17th century plant anatomist Nehemiah Grew , who drew a connection between plant nutrition and the distal end of the root tip ( Grew , 1682 ) . At the root tip , only the cap seems to play an important role in Pi acquisition . Indeed , the elongation zone ( distal to the cap ) displays poor Pi absorption despite its essential metabolic activity . Various factors can be proposed to explain this observation . For instance , the fast rate of root growth reduces the period spent by cells in this area by 6 to 7 hr ( Beemster and Baskin , 1998 ) , limiting the time allocated to protein production . Additionally , the cell elongation process in this region could contribute to diluting the activity of the produced PHT1 transporters . These features do not affect the root cap , which is located below the elongation and meristematic zones of the root and is composed of cells that are 10 to 20 times smaller than mature epidermis cells ( Beemster and Baskin , 1998 ) and that have low turn-over ( Dolan et al . , 1993 ) . This likely contributes to the high cellular concentration of PHT1 proteins in the membrane of this tissue . The root cap structure of Arabidopsis , composed of 180–260 cells ( Dolan et al . , 1993 ) , can appear simple in comparison to other plant species . For example , there are as many as 4 , 000–21 , 000 cells in the maize root cap ( Clowes , 1976 ) . This disparity can be explained by the presence of additional cell layers in the columella , with 20 in maize ( Iijima et al . , 2008 ) and 16 in rice plantlets ( Wang et al . , 2014a ) , in comparison to only 5 in Arabidopsis ( Dolan et al . , 1993 ) . Although this promotes the formation of a slightly larger cap ( see Lotus japonicus in Figure 4—figure supplement 1 ) with more cell layers , it does not significantly increase the external surface in contact with the medium . Despite these species-specific features , the root cap has also been observed to absorb Pi in rice and Lotus . The high concentration of Pi transporters in the root cap could be due to the physicochemical properties of phosphate . Pi interacts strongly with many cations and exhibits a very weak diffusion coefficient in soil ( from 10−12 to 10−15 m2 s−1; Schachtman et al . , 1998 ) . Pi mobility is consequently very low , which promotes a Pi-depleted zone around the root as a result of its absorption by plants ( Kanno et al . , 2012 ) . This suggests that the root cap may provide a temporal advantage to the plant ( « first come , first served » ) , as the Pi stock available to the growing root becomes limited . This feature could therefore be beneficial in natural soil with low Pi availability and high Pi heterogeneity , which may explain the higher biomass gain observed in our soil experiments as compared to hydroponic conditions . Other micro- or macronutrients may also be absorbed by the root cap , although their identification will require further analyses . Nevertheless , an absence of tools to monitor the presence and movements of specific ions could pose technical challenges . Real-time isotope imaging is one practical approach to circumvent this issue . However , the limited detectable range of elements such as conventional ß-ray and some gamma-ray or X-ray emitters ( Kanno et al . , 2012 ) prevents studying the dynamics of many important plant nutrients at the microscopic level . Plant nutrient carriers are known to be highly redundant , which poses a bottleneck for genetic studies . For many nutrients , redundancy is not only linked to gene duplications , since different transporter families have been described for some nutrients ( e . g . nitrate ) , reinforcing the difficulty in finding common regulators . Genetic tools that can impair the activity of a complete family of transporters remain very rare , and are currently restricted to Pi nutrition , such as the phf1-1 mutant used here . Previous spatial gene expression studies have investigated the presence of ion transporters in the root cap for iron ( IRT1; Vert et al . , 2002 ) , potassium ( ATkT3/KUP4/TRH1; Vicente-Agullo et al . , 2004 ) , nitrate ( NRT1-1; Remans et al . , 2006 ) and boron ( BOR1; Takano et al . , 2010 ) . For boron , accumulation of this element around the quiescent center was predicted using a model based on the distribution of several boron transporters ( Takano et al . , 2010 ) and confirmed by chemical analysis ( Shimotohno et al . , 2015 ) , thereby indicating a role for the root tip in boron uptake . Defects in the nitrate NRT1-1 ( Krouk et al . , 2010 ) or potassium TRH1 ( Vicente-Agullo et al . , 2004 ) carriers affect root morphology ( gravitropism , modification of lateral root development , etc . ) . This is related to the ability of these proteins to competitively transport auxin . To date , their presence in the root cap appears to be linked to a developmental role in the adaptation of root morphology to the presence of nitrate or potassium in the environment . Nevertheless , the role of these proteins in nutrition remains to be elucidated . In the case of Pi , it is well documented that the primary root growth is disconnected from Pi import and Pi internal status ( for review see: Svistoonoff et al . , 2007; Thibaud et al . , 2010; Peret et al . , 2011 ) . The present study confirms this view , as Pi absorption by the root cap has not been found to modify root architecture , although it significantly affects plant Pi nutrition . As we have shown , plant growth varies according to the medium used ( in vitro , hydropony , soil ) but in all cases restoring Pi absorption in the root cap promoted a significant gain in biomass . Indeed for plantlets grown in vitro , a 20% increase in Pi uptake generated a 40% gain in biomass for the root tip-complemented line within 4 days . For mature plants grown for a longer period , this gain increased and could even reach 180% as observed with soil experiments . The level of Pi derived from root cap absorption ( with regard to the size of this tissue ) contributed also significantly to the modification of Pi homeostasis and to regulation of gene expression , as revealed by changes in the expression of several transcriptomic markers regulated by internal Pi status ( Thibaud et al . , 2010 ) . This observation is in agreement with previous reports that internal Pi may be involved in Pi homeostasis regulation ( Liu et al . , 1998; Lv et al . , 2014; Wang et al . , 2014b; Puga et al . , 2014 ) . The transcriptional response was not linearly connected to Pi uptake , since absorption in the root cap of the complemented line restored half of the regulation of Pi starvation systemic markers . As previously shown , this confirms that a small amount of Pi absorbed by the root ( at the root tip level , as shown in this study ) can regulate gene expression in the whole plant ( Thibaud et al . , 2010 ) . Charles Darwin characterized the root tip as a plant brain , based on its role in environmental perception and gravitropism ( Darwin and Darwin , 1880 ) . Numerous studies since then have revealed how the root cap responds to various abiotic stresses including: light , modulation of root architecture in response to phosphate ( Svistoonoff et al . , 2007 ) , and potassium or nitrate supply ( Arnaud et al . , 2010 ) . The present study establishes essential roles for this tissue in Pi nutrition , as well as Pi homeostasis adaptation . Finally , this study demonstrates that the root cap contributes to the plant’s adaptation to soil Pi presence in a disproportionate way , considering its diminutive size .
Surface-sterilized seeds were sown in vitro on square Petri plates containing modified Murashige and Skoog medium diluted 10x ( MS/10 ) ( from Arnaud et al . , 2014 ) , supplemented with 2 μM iron and phosphate ( 0 or 500 μM NaH2PO4 for low-Pi ( -P ) orhigh-Pi ( +P ) respectively ) . Low-phosphate medium containing 13 μM Pi ( Pi present in the agar ) was supplemented with NaCl to maintain Na concentration at 500 μM . Plantlets were grown vertically in a growth chamber under a 16 hr photoperiod at 23°C:21°C ( light:dark ) . Growth conditions are described in further detail for each experiment . Rice ( Oryza sativa var . Nipponbare ) and Lotus japonicus ( var . MG-20 ) seeds were surface-sterilized then germinated in distilled water for 3 d . Seedlings were then grown hydroponically on -P medium for 10 days . Seedlings were grown in a growth chamber under a 16 hr photoperiod at 28°C for rice and 23°C for Lotus japonicus . Nine-day-old in vitro Arabidopsis plantlets ( in +P ) were transferred to washed sand distributed in pots ( 12 to 24 plants per genotype ) . The nutrient solution ( MS/10 supplemented with 10 μM Pi ) was replaced every 3 to 4 days . The cambisol surface layer ( 0–30 cm ) was air-dried , gently ground , and sieved ( <2 mm ) for experiments with soil . The composition of this loamy soil ( 36 . 2% silt , 16 . 5% clay , and 47 . 2% sand ) was determined according to normalized methods ( Soil Analyze laboratory ARRAS , NF EN ISO/CEI 17025: 2005 , INRA , France ) to contain ( in mg . g-1 DW ) : organic C: 25 . 4 , N: 2 . 14 , P: 0 . 89 , Ca: 7 . 68 , K: 9 . 75 , Na: 4 . 96 , Fe: 24 , and Si: 348 , with a pH ( water ) of 6 . 9 . The soil was mixed with fine sand ( 1 part sand:2 parts soil , w/w ) and distributed in pots . Nutrient solution ( MS/10 without Pi ) was provided every day by immersion ( 20 min ) . Seeds were sown on the soil surface , and 24 to 30 plantlets per genotype were kept after 10 days . Plants were grown on Pi-depleted medium for 7 days and then transferred to +P medium for 4 days . All experiments were performed in triplicate . Ten rosettes were individually weighed for biomass determination . Biomass production was also measured on mature plants while growing hydroponically in sand or in soil for 3 to 6 weeks . Rosettes were harvested when flower buds appeared and were individually weighed ( 10 to 23 rosettes ) . Pi uptake in the whole plant was measured after transfer for 3 days to +P medium supplemented with 33P ( 5 . 5 kBq/mL ) . Ten plantlets were individually analyzed for radioactivity . A preliminary experiment demonstrated that the phf1 UAS:PHF1 line and the phf1-1 mutant displayed similar influxes ( 16 . 5 ± 4 . 7 and 14 . 2 ± 4 . 3% , respectively , as compared to the WT ) . Free Pi in roots and leaves was measured after transferring plants from -P ( 7 days ) to +P for 4 days . Pools of 10 to 20 plants were analyzed in triplicate . Frozen material was homogenized in a grinder ( Mixer Mill MM400 , Retsch; Germany ) , resuspended and homogenized in MES buffer ( 0 . 17 M , pH 5 . 8 , 10 μL per mg of fresh weight ) . After centrifugation , the supernatants were analyzed in a 96-well plate and triplicates of 5 – 20 μL subsamples were diluted into a final volume of 145 μL . Phosphate content was then measured using a Malachite green protocol ( Misson et al . , 2004 ) that was modified such that 30 μL of each reagent were added , and measurements were performed at 595 nm with a microplate reader ( Biorad , Model 3550; USA ) . Phosphate concentrations were calculated using a calibration curve ( performed with a KH2PO4 solution ) and were expressed per root or rosette fresh weight . An amplified product ( see Supplementary file 1A for oligonucleotides ) containing the PHT1;4 promoter ( 2 . 6 kb upstream of the start codon ) along with the UTR and genomic sequence was cloned into the pENTR/D-TOPO vector using a pENTR directional TOPO cloning kit ( Invitrogen; USA ) . The cloned fragment was then transferred into the Gateway vector pGWB4 ( Nakagawa et al . , 2007 ) to create a translational fusion with sGFP . Homozygous lines ( in a WS ecotype background ) were selected and designated as PHT1;4-GFP lines . The UAS:PHF1 construct was created by amplifying PHF1 cDNA ( see Supplementary file 1A for primer sequences ) . The resulting PCR product was cloned into the pGEMT vector ( Invitrogen ) . PHF1 cDNA was cloned into the pBI-UAS-KNAT4 plasmid between the UAS and Nos terminator sequences ( Truernit et al . , 2006 ) using the BamHI and SacI restriction sites to replace the KNAT4 gene . This construct was introduced into the phf1-1 background ( ecotype Col-0 , Gonzalez et al . , 2005 ) and homozygous transformants ( referred to as phf1 UAS:PHF1 ) were selected . Enhancer trap lines ( ecotype C24 ) were introgressed into the phf1-1 background by at least 3 successive backcrosses . The selected GAL4 lines include Q0171 ( specific to the lateral root cap and columella ) , J1092 ( specific to the root cap and epidermal initial cells ) , and J0481 ( specific to the root cap and epidermis ) . Each line was used to express mGFP5 under UAS promoter control . For each GAL4 driver , a homozygous line was crossed with the phf1 UAS:PHF1 line . Successive self-crossings were performed to obtain homozygous phf1 QO171>>PHF1 , phf1 J1092>>PHF1 , and phf1 J0481>>PHF1 lines . These lines were then crossed with the phf1-2 pht1;4–1 double mutant . This double mutant was previously obtained by EMS mutagenesis of the pht1;4–1 mutant ( Misson et al . , 2004 ) . PHT1;4 cDNA was cloned without its stop codon ( see Table S1 for primer sequences ) into pENTR/D-TOPO , resulting in pEN L1-PHT1;4-L2 . The mCherry sequence was amplified by PCR ( see Table S1 for primer sequences ) from pEN L3-mCherry-HA-L2 ( Addgene; USA ) and cloned into pDONR P2r-P3 using the BP reaction to produce pEN R2-mCherry-L3 . A 35S:PHT1;4-mCherry fusion was produced by multi-site Gateway reaction ( Karimi et al . , 2007 ) using the LR reaction between the following vectors: pEN-L4-2-R1 ( containing 35S; Karimi et al . , 2007 ) , pENTR/D-TOPO , pEN R2-mCherry-L3 and pB7m34GW . The product was then introduced into Col-0 , phf1-1 and the phf1 QO171>>PHF1 line . Successive self-crossings were performed to obtain homozygous lines . FCY1 ( from S . cerevisiae ) and UPP ( from E . coli ) were amplified by PCR ( see Table S1 for primer sequences ) . After purification , an additional PCR was performed to obtain the FCY-UPP fragment , which was cloned into the pDONR207 vector using Gateway® technology , resulting in pEN L1-FCY-UPP-L2 . A UAS:FCY-UPP fusion was created by multi-site Gateway reaction ( Karimi et al . , 2007 ) between the pEN L1-FCY-UPP-L2 , pEN-L4-UAS-R1 and pB7m24GW vectors . The final construct was used to transform Col-0 plants . A homozygous line was subsequently selected and crossed to the Q0171 line ( backcrossed 3 times in Col-0 ) . Experiments were performed with plants from F1 seeds , referred to as Q0171>>FCY-UPP . All plant transformants were generated by the floral dip method as previously described ( Clough and Bent , 1998 ) , following introduction of the construct into Agrobacterium tumefaciens . To visualize 33P absorption , 7-day-old Arabidopsis seedlings grown in high-phosphate medium were incubated for 1 day in liquid medium supplemented with 33P ( 400 Bq/mL ) and exposed against an imaging plate ( Fujifilm; Japan ) for 4 days at -80°C . Radioluminographic images of the seedlings were then scanned using the FLA-5000 imaging analyzer ( Fujifilm; Japan ) and analyzed using Image Gauge v4 . 0 ( Fujifilm; Japan ) . Plants were grown in +P medium for 7 days . The middle of the root was isolated from the medium with Vaseline and a 10-µl drop of 33Pi solution ( 1 µM Pi , including 200 kBq 33P ) was applied for 30 min . 33P signal in the root was detected using the Micro Real-time Radio Imaging system ( Kanno et al . , 2012 ) with an EMCCD camera iXon3 888 ( Andor; USA ) . Seven-day-old plantlets grown in -P medium were transferred to glass slides covered with 0 . 1% agar , containing 1 µM Pi solution supplemented with 33P ( 200 kBq/10 µL ) . Real-time imaging of 33P uptake was performed using the Imaging System as described above . Images were acquired after 1 to 10 min , and radioactive signal was quantified using the AQUACOSMOS software ( Hamamatsu Photonics; Japan ) on a selected root tip zone ( 0 . 2 mm ) . A separate labeling procedure was used to quantify a broader part of the root ( 2 mm ) . Roots of 7-day-old plantlets were immersed in 1 µM Pi solution containing 2000 kBq 33P for 1 min . Samples were then rinsed in a solution containing 1 mM 2 , 4-dinitrophenol . This inhibitor of ATP production was used to block cellular metabolic activity and to prevent ionic movement ( including 33Pi ) . Successive images along the root were then recorded as described above . Radioactive signal quantification was performed with AQUACOSMOS on successive 0 . 2 mm portions starting from the root tip . 13-day-old rice and lotus plantlets were fed 0 . 5% agar , containing 1 µM Pi solution supplemented with 33P ( 200 kBq ) . Successive images were obtained during 5 min and radioactive signal was quantified after 1 min , as described above . To analyze Pi uptake in plantlets treated with 5FC , seedlings were grown for 3 days on +P and transferred for one day to +P medium containing 3 . 87 mM 5FC or DMSO ( as a control ) . Subsequently , samples were transferred to -P medium containing 5FC or DMSO for 3 days before imaging as described above . Q0171>>FCY-UPP plantlets were grown in +P medium and then treated with 3 . 87 mM 5FC or DMSO for 4 days . Images of the plates were taken when transferred , and after 2 and 4 days . Primary root length was quantified using the ImageJ software with the NeuronJ plugin ( version 1 . 46r , http://imagej . nih . gov/ij ) . RNA extraction , purification , reverse transcription and qPCR analyses were performed as previously described ( Thibaud et al . , 2010 ) . Primer efficiency factor was measured for each gene , and GapC1 , ROC3 and AT1G32050 were used as reference genes . Primer sequences are provided in Supplementary file 1B . GUS staining was performed as previously described ( Misson et al . , 2004 ) on plants grown for 10 days in -P medium supplied with 2 μM FeCl2 . For visualization , seedlings were either placed in water and observed under a stereomicroscope ( MZ16 , Leica Microsystems; Germany ) or between coverslips and photographed under the microscope ( LMD6000 , Leica Microsystems; Germany ) . For luminescence imaging , seeds from the proPHT1;1:LUC line ( Castrillo et al . , 2013 ) were sown on the +P medium with 2 μM Fe . After 4 days , plants were transferred to fresh medium containing 5 μM Pi supplemented with 50 μM Luciferin ( D-Luciferin Firefly Potassium Salt , Biosynth ) . Root tips were excised 7 days after transfer and placed between a cover slip and a thin film of medium containing 5 μM Pi with 50 μM Luciferin . Root tips were imaged using a dedicated luminescence microscope ( Luminoview , Olympus ) connected to a cooled back-illuminated CCD camera ( IkonM , Andor ) . For sGFP imaging in the PHT1;4:GFP and PHF1:GFP lines , plants were grown in -P medium for 11 or 4 days , respectively . Plantlets were incubated for 3 min in 20 μg/mL propidium iodide ( PI ) for cell wall staining ( at room temperature ) . For mCherry and mGFP5 imaging , plants were grown in -P medium ( +10 μM Fe ) for 12 days . Tips of secondary roots were observed by confocal microscopy . GFP , PI and mCherry imaging were performed on either a TCS SP2 ( Leica; Germany ) or LSM780 ( Zeiss; Germany ) confocal microscope . For confocal imaging , GFP was excited at 488 nm ( argon laser line ) , and PI ( or mCherry ) was excited at 561 nm ( diode-pump solid-state laser ) . Fluorescence was detected with the LSM780 confocal microscope using the following settings: sGFP ( GaAsP , 491 – 545 nm ) , mGFP5 ( PMT , 492 – 522 nm ) , mCherry ( GaAsP , 607 – 696 nm ) , and PI ( PMT , 586 – 685 nm ) . Detection of mCherry with the TCS SP2 confocal microscope was between 599 and 651 nm . Characterization of mGFP5 localization in the enhancer trap lines ( phf1 QO171>>PHF1 , phf1 J1092>>PHF1 , and phf1 J0481>>PHF1 ) was performed on 5-day-old plantlets grown in -P stained with PI . Q0171>>FCY-UPP plantlets were grown for 7 days in +P and then treated with 3 . 87 mM 5-fluorocytosine ( 5FC ) or DMSO ( 5FC solvent ) for 5 days , or by heat shock at 95°C for 5 min in water . Plants were then treated with PI . Images were acquired with a macroscope equipped with structured illumination ( Axiozoom V16 , + ApoTome2 , Zeiss; Germany ) . Fluorophores were excited using a mercury lamp ( GFP: Ex filter 470 nm BP40 , Em 525 nm BP50; PI: Ex filter 572 nm BP25 , Em 629 nm BP62 ) . Images are composed of maximum intensity z-series projections of 26 to 33 images at 2-μm intervals ( software: Zen 2012 , version 1 . 1 . 2 . 0 , Zeiss ) . Localization of the PHT1;4-mCherry fusion protein to the PM was analyzed in the phf1 QO171>>PHF1 background with 1 μg/mL FM4-64 for 5 min at room temperature ( Interchim; France ) . Co-imaging of FM4-64 and mCherry was performed on a LSM780 confocal microscope in spectral mode ( Ex: 561 nm , Em range: 563 – 696 nm ) . Linear unmixing was applied to separate mCherry and FM4-64 components , using reference spectra acquired separately and the Zen software ( Zen 2012 SP1 , version 8 . 1 , Zeiss ) . Fluorescence intensity profiles were subsequently calculated ( software: Zen 2012 , version 1 . 1 . 2 . 0 , Zeiss ) . All statistical analyses were performed with GraphPad prism 6 software ( version 6 . 0f ) . Values were tested for normality ( D’Agostino-Pearson omnibus normality test ) and variance homogeneity ( Brown-Forsythe and Bartlett’s tests ) . Unpaired t-test ( Student’s test ) was run to compare treatments . | All plants need phosphate to grow because it is a major component of DNA and many other biological molecules . Most of the Earth’s soil is poor in phosphate , and so farmland is routinely supplemented with fertilizers to provide crops with this essential nutrient . However , phosphate fertilizers are becoming scarce and their quality is expected to decline in the near future . Plant biologists must therefore determine how to adapt plants to a restricted supply of this resource , in order to sustain high crop yields for the growing world population . Plants are known to absorb phosphate through specific protein-based transporters located in the cells that make up the outer layer of roots . These proteins are highly concentrated at the root tip , and while this specialized tissue is well-known for perceiving gravity and light , it had not been shown to play a role in phosphate absorption . Kanno , Arrighi et al . have now used genetically modified Arabidopsis plants to demonstrate that phosphate can be taken up via the small cells that surround the root tip . The experiments showed that the absorbed phosphate rapidly reaches the leaves within minutes , helps the plant grow and modifies its metabolism . As the root tip can accumulate high amounts of phosphate in order to sustain its own activity , it was important to distinguish uptake of phosphate from the environment from redistribution of phosphate already within the plant . Kanno , Arrighi et al . tackled this issue through the development of a new radioactive micro-imaging technique . Phosphate transporters are also present within the cell layers within the root , but their purpose and activity are not well described . Further studies are needed to analyze the role of other root cell layers in phosphate uptake and transport , and the newly developed techniques will help decipher the mechanisms involved . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2016 | A novel role for the root cap in phosphate uptake and homeostasis |
Fumarase is distributed between two compartments of the eukaryotic cell . The enzyme catalyses the reversible conversion of fumaric to L-malic acid in mitochondria as part of the tricarboxylic acid ( TCA ) cycle , and in the cytosol/nucleus as part of the DNA damage response ( DDR ) . Here , we show that fumarase of the model prokaryote Bacillus subtilis ( Fum-bc ) is induced upon DNA damage , co-localized with the bacterial DNA and is required for the DDR . Fum-bc can substitute for both eukaryotic functions in yeast . Furthermore , we found that the fumarase-dependent intracellular signaling of the B . subtilis DDR is achieved via production of L-malic acid , which affects the translation of RecN , the first protein recruited to DNA damage sites . This study provides a different evolutionary scenario in which the dual function of the ancient prokaryotic fumarase , led to its subsequent distribution into different cellular compartments in eukaryotes .
The enzyme fumarase ( classII , fumarate hydratase in higher eukaryotes ) is a conserved protein in all organisms from bacteria to human with respect to its sequence , structure , and enzymatic activity ( Akiba et al . , 1984 ) . Fumarase is a dual targeted protein in eukaryotes; its echoforms are distributed between mitochondria and the cytosol/nucleus ( Yogev et al . , 2011; Yogev et al . , 2010 ) . Moonlighting of proteins is also a well-known phenomenon , which is defined by single proteins that can perform different functions in the cell ( Gancedo et al . , 2016; Espinosa-Cantú et al . , 2015 ) . Fumarase is also a moonlighting protein since it performs functions in the tricarboxylic acid ( TCA ) cycle in mitochondria and it participates in the DNA damage response ( DDR ) in the nucleus ( Boukouris et al . , 2016 ) . In the TCA cycle , fumarase converts fumaric acid to L-malic acid while in the nucleus it catalyses the opposite reaction , thereby supplying fumaric acid as a signaling molecule for the DDR ( Yogev et al . , 2010 ) . In human cells , fumaric acid has been shown to inhibit certain histone dimethylases and prolyl hydroxylases ( Jiang et al . , 2015; Gottlieb and Tomlinson , 2005; Isaacs et al . , 2005 ) . Worth mentioning is that fumarase associates with different protein partners in the two compartments ( in the mitochondria where it is part of the TCA cycle , it interacts with other TCA cycle enzymes such as malate dehydrogenase , while in the cytosol/nucleus , it interacts with components of the DDR such as kinases and histones ) . Two questions with regard to the evolution of fumarase come to mind; ( 1 ) how and when did dual targeting of the protein evolve ? And ( 2 ) how and when did dual function of the protein evolve ? With regard to the first question , fumarase has been shown to distribute between the mitochondria and the cytosol/nucleus by different mechanisms in different eukaryotes ( Yogev et al . , 2011 ) . In Arabidopsis , there are two fumarase-encoding genes , which are highly homologous , besides the fact that one encodes a mitochondrial targeting signal ( MTS ) , while the other lacks it ( Pracharoenwattana et al . , 2010 ) . In human , there is a single gene , however , it makes multiple mRNAs , which either encode or lack a MTS . In these cases , described above , two different types of mRNA are made which determine dual localization of fumarase echoforms ( Dik et al . , 2016 ) . In yeast ( S . cerevisiae ) , mitochondrial and cytosolic echoforms of fumarase , are encoded by a single nuclear gene ( FUM1 ) and follow an intriguing mechanism of protein subcellular localization and distribution ( Yogev et al . , 2011 ) . Translation of all FUM1 messages initiates only from the 5'-proximal AUG codon and results in a single translation product that contains the MTS ( Yogev et al . , 2007 ) . The precursor of yeast fumarase is first partially translocated into mitochondria , so that the N-terminal signal is cleaved . While a subset of these molecules continues to be fully translocated into the organelle , the rest are folded into an import-incompetent conformation and are released by the retrograde movement back into the cytosol . Thus , protein folding is the driving force for fumarase dual targeting in yeast , a mechanism termed reverse translocation ( Karniely and Pines , 2005; Kalderon and Pines , 2014 ) . Dual localization is a very abundant phenomenon in eukaryotes , and in fact , we estimate that a third of the yeast mitochondrial proteome is dual targeted ( Ben-Menachem et al . , 2011 ) . Therefore , dual targeting or dual localization of proteins is a major outcome of gene expression and as such , the evolutionary pressures governing this phenomenon are of fundamental importance . In a major discovery , we have shown that dual-targeted proteins are significantly more evolutionary conserved than exclusive mitochondrial proteins . We reached this conclusion by employing codon usage bias , propensity for gene loss , phylogenetic relationships , conservation analysis at the DNA level , and gene expression ( Dinur-Mills et al . , 2008; Kisslov et al . , 2014 ) . This has changed the way we think about dual targeting since we now assume that the majority of dual targeted proteins have discrete functions in the different subcellular compartments , regardless of their dual-targeting mechanism . Thus we hypothesize that dual targeting is maintained due to separate selective pressures administered by the different compartments to maintain the functions associated with the protein sequences ( Kisslov et al . , 2014 ) . Here , we use the enzyme fumarase as a paradigm of this evolution and show for the first time that the single fumarase gene of B . subtilis is involved in the bacterial DDR , in addition to its role in the TCA cycle . This finding suggests that dual function of fumarase , in the bacterial progenitor , preceded the dual targeting that we find in eukaryotes . Intriguingly , fumarase in bacteria , similarly to eukaryotes , appears to be linked to the DDR by metabolite signaling , however the active molecule in this case is L-malic acid and not fumaric acid .
The conservation of fumarase dual function in eukaryotes without conservation of the mechanism of dual targeting has brought us to examine the following hypothesis: Dual function of fumarase in bacteria preceded dual targeting in eukaryotes; in other words , it occurred in the prokaryotic ancestor prior to evolving of eukaryotes . Thus , dual function was the driving force for fumarase dual localization in the eukaryotic cell . To examine this question we employed the Gram positive bacterium Bacillus subtilis , which contains a single class II fumarase gene ( fum-bc ) . As shown in Figure 1A ( compare the right and left panels ) , B . subtilis deleted for the citG gene ( fum-bc ) which encodes fumarase , exhibits very poor growth on defined medium ( S7 ) containing a low level of glucose ( 0 . 1% ) that requires respiration and the TCA cycle for efficient growth . Thus , as expected , Fum-bc has a crucial function within the TCA cycle . Intriguingly , fum-bc deleted strains are sensitive to DNA damaging agents such as ionized radiation ( IR ) ( Figure 1A , middle panel ) or methyl methanesulfonate ( MMS ) ( Figure 1B ) . We expressed Fum-bc from the ectopic amyE locus ( Δfum+Fum-bc , Figure 1—figure supplement 1 ) and this strain exhibits similar resistance to MMS treatment as the wild type ( Figure 1B , compare the first and third rows ) . The slight difference in resistance may be since fumarase is expressed from the amyE locus and not from its native promoter . Quantitative experiments following colony forming units ( CFU ) are shown in Figure 1Ci and 1Cii . The difference between wild type and Δfum strains is insignificant ( compare the two left bars of Figure 1Ci and Figure 1Cii ) while the difference between these strains following DNA damage is highly significant ( bars 4 and 5 , p<0 . 01 ) . These results suggest , as we hypothesized , that dual function of fumarase can already be found in prokaryotes . To examine the ability of Fum-bc to perform its two functions ( TCA and DDR ) in a eukaryotic model , we took advantage of our experience with the yeast Saccharomyces cerevisiae . fum-bc was cloned into a yeast expression vector under the control of the GAL promoter . The levels of fumarase in cultures grown in galactose medium of wild type yeast , yeast deleted for the chromosomal FUM1 and such a deletion strain expressing bacterial fum-bc , are shown in Figure 1—figure supplement 2 . The fum-bc gene was expressed in a yeast strain deleted for the endogenous FUM1 gene ( Δfum1 ) . This strain was grown on glucose ( dextrose ) as a control medium that does not require respiration for growth , and on ethanol which does require respiration and the function of the TCA cycle . As shown in Figure 1D , yeast strains deleted for the endogenous FUM1 and expressing fum-bc can partially complement the lack of yeast fumarase for growth on ethanol ( compare row 4 [fum-bc expression] to row 2 [no fumarase expression] ) . Fum-bc does not contain a mitochondrial targeting sequence and it is not targeted to , or imported into mitochondria . We have performed subcellular fractionation of yeast cells expressing fum-bc . The bacterial protein is located only in the ‘cytosol’ and not in mitochondria ( Figure 1—figure supplement 3 ) . From previous studies , it turns out that fumarase , located outside mitochondria , can nevertheless function in the TCA cycle ( Sass et al . , 2003; Stein et al . , 1994 ) . The explanation for this is that the metabolites fumarate and malate can enter and exit the organelle ( via specific inner membrane transporters ) thereby completing the TCA cycle . To examine the activity of Fum-bc in the DNA damage response , the yeast FUMm strain was employed . The FUMm strain harbors a chromosomal FUM1 deletion ( Δfum1 ) and a FUM1 ORF ( open reading frame ) insertion in the mitochondrial DNA , thereby allowing exclusive mitochondrial fumarase ( Fum1 ) expression , but lacking extra-mitochondrial ( cytosolic/nuclear ) fumarase ( Yogev et al . , 2010 ) . Accordingly , the FUMm strain exhibits a functional TCA cycle and the ability to respire; however , it displays sensitivity to DNA damaging agents such as ionized radiation . When fum-bc is expressed in the yeast FUMm strain , it can complement the lack of extra-mitochondrial fumarase with respect to sensitivity to ionized radiation ( Figure 1E , compare row 3 [cytosolic fum-bc expression] to row 2 [no cytosolic fumarase expression] ) . These results support our hypothesis that bacterial fumarase has the capacity to function both in the TCA cycle and the DNA damage response . To examine the role of fumarase in the DDR , Bacillus subtilis cells were grown to log phase and then incubated in the presence or absence of MMS . Cell extracts were analyzed with time by western blot using a mixture of anti yeast fumarase ( anti yFum ) and anti human fumarase-FH ( anti hFum ) antibodies . As shown in Figure 2A the level of Fum-bc gradually increases ( compare the top panel to the SigmaA loading control in the lower panel . After 60 min of MMS treatment , the amount of fumarase rises two fold ( Figure 2B ) . This finding is consistent with the proposed function of Fum-bc in the DDR . As shown In Figure 2—figure supplement 1 , the levels of other TCA cycle enzymes , citrate synthase 2 ( citZ ) and isocitrate dehydrogenase ( ICDH ) , do not change significantly upon treatment with MMS . We next asked whether the appearance and localization of Fum-bc changes upon induction of the DNA damage response . For this , we monitored fluorescence in B . subtilis strains harboring genomic Fum-bc-GFP fusions , in combination with DAPI for DNA staining and FM4-64 for membrane staining . Fum-bc-GFP retains full fumarase activity in cell extracts ( Figure 2—figure supplement 2 ) . As shown in Figure 2C , in untreated cells ( top panels ) Fum-bc does not generally colocalize with the bacterial DNA ( top right panel ) ; only 13% of Fum-bc-GFP foci showed colocalization with the DNA ( Figure 2D ) . Following treatment with MMS , we detected full colocalization of Fum-bc-GFP fluorescence with the DNA DAPI stain ( bottom right panel ) ; more than 95% of the foci showed colocalization with the DNA ( Figure 2D ) . Upon DNA damaging treatment , fumarase clearly coincides with the condensed DNA and in fact as shown in the two lower right panels of Figure 2C , the Fum-bc-GFP fluorescence and the DNA DAPI stain perfectly coincide during this process . Without DNA damaging treatment the fum-bc-GFP foci and DAPI stained DNA do not regularly superimpose . Nevertheless one does see some co-staining which can be explained by simple coincidence or naturally occurring low random DNA damage for example at DNA replication sites . Furthermore , we found that without induction of double strand breaks ( DSBs ) ( -MMS ) most of the cells show between three to four Fum-bc-GFP foci , while after induction of DSB ( +MMS ) most of the cells show one to two extensive foci that overlap with the DAPI stained DNA ( Figure 2E ) . Thus , Fum-bc appears to be recruited to the DNA during the DNA damage response supporting a role for the bacterial fumarase in this response . To examine whether the role of Fum-bc in the DDR requires its enzymatic activity , we first identified mutations within the fumarase active site that may abolish its enzymatic activity . According to the literature ( Weaver et al . , 1997; Alam et al . , 2005; Kokko et al . , 2006 ) fumarase has a known active site in eukaryotes ( S . cerevisiae ) and in prokaryotes ( Escherichia coli ) . Based on sequence similarity between eukaryotic and prokaryotic fumarases , we created two separate substitution mutations within the suspected Fum-bc active site; H186N and H127R . H186N has been suggested to be a residue of the active site of E . coli fumarase while H127R has been suggested to be a residue of the active site of the eukaryotic S . cerevisiae fumarase . We created fum-bc point mutations corresponding to the above single amino acid changes in the respective active sites , and examined expression of the proteins by western blot ( Figure 3—figure supplement 1 ) . While extracts of cells expressing only the H186N mutated fumarase , were essentially devoid of enzymatic activity , those expressing H127R displayed 40% of the wild type activity ( Figure 3—figure supplement 2 ) . Accordingly , cells expressing H186N exhibit sensitivity to MMS and defective growth on S7 medium ( low glucose , Figure 3A , right panel ) , while H127R grows normally on S7 plates and is not sensitive to MMS ( Figure 3—figure supplement 3 ) . This indicates that enzymatic activity is required for both DDR and respiration-related functions of fumarase . Fumarase catalyses the reversible conversion of fumaric acid to L-malic acid as part of the TCA cycle in mitochondria . In yeast and human cells , defective for extra mitochondrial fumarase , the sensitivity to DNA damage can be complemented by fumaric acid , added to the growth medium in the form of an ester ( monoethyl fumarate , which is cleaved in the cells to form the free acid ) ( Yogev et al . , 2010; Jiang et al . , 2015 ) . To examine if products or substrates of the fumarase reaction in B . subtilis may complement the lack of fumarase in the DDR , bacteria were grown in the presence of organic acids added to the medium . As shown in Figure 3B , B . subtilis cells deleted for the fum-bc gene are protected from the DNA damaging treatment with MMS , by L-malic acid ( compare the third and fourth rows of the right and left panels ) . In contrast , succinic and citric acids have no protective effect ( two middle panels respectively ) . Since , fumaric acid is not soluble; we also examined the capacity of esters of the other organic acids - to protect the bacterial cells against MMS . While diethylmalate protects the cells from MMS the other organic acids ( monoethyl fumarate and monoethyl succinate ) , had a much weaker effect ( Figure 3—figure supplement 4 ) . Organic acids , and in particular L-malic acid , appear to play a role in the DNA damage response in B . subtilis . To correlate changes in the intracellular levels of these organic acids we employed GC-MS of cell lysates . As shown in Figure 3C , following induction of DNA damage with MMS , the relative levels of L-malic acid increase while those of succinic and fumaric acids decrease ( compare the two left sets of bars ) . This clearly fits the role of L-malic as a DDR signaling molecule and that succinic and fumaric acids do not have such a role . In B . subtilis , strains deleted for fum-bc , the single fumarase gene , accumulate higher levels of fumaric and succinic acid ( Figure 3C ) . This is expected from a block in the TCA cycle at the conversion step of succinic to fumaric acid and subsequently to L-malic acid by fumarase ( succinic - > fumaric - > L malic ) . Interestingly , upon treatment with MMS the levels of fumaric and succinic acids are even higher ( compare the fourth and third sets of bars ) , indicating an induced flow of metabolites through TCA cycle upon DNA damage . To further implicate Fum-bc expression and its function in the DNA damage response , we decided to examine certain B . subtilis DDR components in conjunction with fumarase . RecN appears to be one of the first proteins recruited to DNA damage sites in live cells ( Cardenas et al . , 2014; Alonso et al . , 2013 ) . B . subtilis cells deleted for the fumarase gene exhibit an alteration in the localization and appearance of RecN ( Figure 4A , see description below ) . RecN appears to be the first protein detected as discrete foci ( repair centers ) in live cells in response to DNA double strand breaks ( Cardenas et al . , 2014 ) . It is cytoplasmically located in untreated cells and upon treatment with DNA damaging agents is recruited to damage sites followed by RecO and then RecF ( Alonso et al . , 2013 ) . As shown in Figure 4B , a strain deleted for RecN shows weak DNA damage sensitivity to MMS ( Sanchez et al . , 2007 ) , while a double knock out of RecN and fumarase exhibits an additive effect with the cells exhibiting much higher sensitivity ( Figure 4B ) . This effect can be reversed by addition of L-malic acid to the medium ( Figure 4B , right panel , compare rows 4 and 2 to row 1 ) . Thus , according to the results with RecN , we conclude that fumarase is involved in the resistance to DNA damage and its absence can be complemented by L-malic acid . To examine whether the appearance and localization of RecN-GFP changes upon knockout of the fumarase gene and/or induction of the DNA damage response , we created B . subtilis strains harboring genomic RecN-GFP fusions . Following treatment with MMS of wild type cells , there appears to be only a small increase in the proportion of cells containing RecN foci ( less than 5% ) , nevertheless , this increase is statistically significant ( Figure 4—figure supplement 1 ) , and the appearance of these foci remained unchanged under MMS treatment ( Figure 4A , compare the RecN-GFP untreated control to RecN-GFP MMS ) . In strains deleted for the fumarase gene , we observed a similar number of RecN associated foci ( Figure 4A , third row , RecN-GFP Δfum ) when compared to the RecN-GFP control ( fourth row ) . In the presence of MMS ( DNA damage ) strains deleted for the fumarase gene displayed a drastic change in the number of cells which contain foci ( a two fold increase ) and in their appearance ( Figure 4A and Figure 4—figure supplement 1 , compare RecN-GFP Δfum + MMS to untreated RecN-GFP Δfum ) . Upon DNA damage the RecN-GFP fluorescence , in the Δfum strain , does not appear as discrete foci but rather this fluorescence coincides with the DNA DAPI stain of the condensed B . subtilis chromosomes . Thus , in the absence of fumarase , RecN appears to be recruited differently to the DNA during the DNA damage response . In addition to the DNA damage sensitivity and subcellular appearance , we wished to examine the effect of fumarase on RecN levels in the cell . Cells deleted for the fumarase gene were grown to early logarithmic phase ( OD = 0 . 4 ) , MMS was added for 30 min and then cell lysates were subjected to western blot analysis ( we could not detect RecN without induction of DSB- data not shown ) . As shown in Figure 5A , RecN is expressed about three fold higher in the strain deleted for the fumarase gene when compared to the wild type ( Figure 5A , compare lane 2 to lane 1 of the top panel , quantification , Figure 5B ) . In contrast , when L-malic acid is added to the medium the levels of RecN are essentially the same in wild type and strain deleted for the fumarase gene ( Figure 5A , upper panel , compare the two right lanes , quantitation Figure 5B ) . Worth mentioning is the fact that we have tried to coimmunoprecipitate fumarase and RecN with no positive indications ( see the Supplementary methods ) . Together these data support the notion that fumarase affects RecN function by producing the metabolite L-malic acid and not by direct interaction of the proteins . The primary goal of this research was to implicate prokaryotic fumarase in the DNA damage response . The finding that fumarase and L-malic acid affect the expression and subcellular appearance of RecN fully supports this hypothesis . Nevertheless , we decided to also examine at which level of gene expression , do fumarase and L-malic acid affect RecN cellular levels . Cells deleted for the fumarase gene and the corresponding wild type were grown to early logarithmic phase ( OD = 0 . 4 ) , MMS was added for 30 min and then the levels of RecN mRNA were determined . We employed quantitative RT-PCR ( see materials and methods ) and observed no difference between the mRNA levels ( data not shown ) . To further make the point that transcription does not play a role in RecN higher levels in Δfum strains , we grew B . subtilis strains in the presence of rifampicin . Rifampicin inhibits bacterial RNA polymerase and blocks transcription initiation , thus , in its presence; changes in protein synthesis do not result from changes in transcription . As shown in Figure 5C and D , cells treated with MMS in the presence of rifampicin for 10 or 20 min , revealed significantly higher levels of RecN in the Δfum strain than in the wild type strain . These results indicate that the increased synthesis of RecN , in the absence of fumarase , is due to translation and not transcription . To rule out the possibility that the different RecN levels in strains result from differences in protein stability , we assessed RecN-protein turnover in the presence of chloramphenicol following induction of DNA damage . Chloramphenicol prevents protein chain elongation by inhibiting the peptidyl transferase activity of the bacterial ribosome . As presented in Figure 5E the decrease in RecN levels in this experiment shows that both Δfum strain and wild type exhibit similar RecN turnover kinetics ( Figure 5F ) .
How did moonlighting and dual targeting of proteins evolve in eukaryotes ? The notion is that a protein with a single function , activity or location , acquired new traits through evolution . With regard to fumarase the notion was that the enzyme acquired its second function in the DNA damage response after endosymbiosis and the creation of mitochondria . An example of the acquisition of novel functions after endosymbiosis are roles assumed by yeast Hsp60 and aconitase in mitochondrial genome stability through binding of mitochondrial DNA ( Chen et al . , 2005; Kaufman et al . , 2000 ) . Fumarase has been shown to distribute between the mitochondria and the cytosol/nucleus by different mechanisms in different eukaryotes ( Yogev et al . , 2011 ) . It seems unlikely that the dual targeting/dual function arose independently in different eukaryotic ancestors . A different possibility is that dual function arose prior to dual localization and actually it was the function that was the driving force for the evolution of fumarase dual targeting . Our results support a second function for fumarase ( in addition to its function in the TCA cycle ) in the DDR of bacteria: We have shown that dual-targeted proteins are significantly more evolutionary conserved than exclusive mitochondrial proteins , strongly suggesting that dual function drives the evolution of dual targeting or at least its maintenance ( Ben-Menachem et al . , 2011; Kisslov et al . , 2014 ) . Our model depicts fumarase as a protein with dual function in the bacterial ancestor of mitochondria . Upon transfer of the endosymbiont gene into the eukaryotic nucleus , the selective pressure due to the two functions that are ‘needed to be carried out in different subcellular compartments’ , resulted in dual targeting of this enzyme . A similar model can be assigned to the dual-targeted yeast aconitase , which is a component of the TCA cycle in mitochondria and the glyoxylate shunt in the cytosol ( Chen et al . , 2005; Rouault et al . , 1991 ) . This is a simpler example , compared to fumarase , in which aconitase functions in two parallel metabolic pathways that coexist in the prokaryotic cytosol . The main goal of this study was to determine that B . subtilis Fum-bc has both TCA cycle and DDR-associated functions . Nevertheless , we have also gained insight into some mechanistic features of fumarase function within the DDR . The most exciting finding is that intracellular signaling of the DDR is achieved via L-malic acid , the product of the reaction catalyzed by fumarase . This conclusion was reached not only due to the fact that L-malic acid , added to the medium , can complement Δfum strains upon DNA damage , but it is also consistent with other data in this study . B . subtilis fumarase enzymatic activity is required for its DNA damage protective function . In addition , upon DNA damage B . subtilis accumulates L-malic acid versus lower levels of fumaric and succinic acids . Furthermore , upon DNA damage Δfum cells accumulate extremely high levels of fumaric ( and succinic ) acids as though the cells are ‘trying to make more L-malic acid’ in order to signal DNA damage . This finding is intriguing since in yeast and human cells , the signaling molecule associated with fumarase , with respect to the DDR , is fumaric not L-malic acid as we find for B . subtilis ( Yogev et al . , 2011; Jiang et al . , 2015 ) . Thus , although the fumarase protein sequence and the dual function/targeting of the enzyme are conserved , the signaling metabolite is different . In other words , when we talk about conservation of function of fumarase in the DDR we do not mean that all aspects of dual function are conserved but rather that the metabolic pathway with specific organic acid intermediates are recruited . This raises questions on how intermediates of primary metabolism were chosen during evolution as signaling molecules in different organisms . Our results indicate that fumarase does not interact directly with RecN , but rather , the effect is via L-malic acid . How does L-malic acid precisely exert its effect on the DDR ? A number of possibilities come to mind; the organic acid binds components of the DDR directly , thereby , modulating their activity . A good example of such a scenario are succinate and fumarate that can inhibit prolyl hydroxylases ( PHDs ) , resulting in the stabilization of HIF1-α and activation of downstream hypoxic pathways in human cells ( Gottlieb and Tomlinson , 2005; Isaacs et al . , 2005; Pollard et al . , 2005; Sudarshan et al . , 2007 ) . Another example are local concentrations of fumarate produced by phosphorylated fumarase ( bound to histone H2A . Z ) which inhibits KDM2B , histone dimethylase , which in turn results in enhancement of histone H3 dimethylation and downstream activation of the DDR ( Jiang et al . , 2015 ) . Another possibility is that L-malic acid affects components of the DDR by affecting their expression . RecN fits the profile of an L-malic acid target since its expression is affected by the acid; the level and appearance of RecN is altered in Δfum cells following DNA damage which is correlated with lower levels of L-malic acid in the cells . RecN does not appear to be regulated at the mRNA level , transcription or protein stability . We claim that L-malic acid affects the translation of RecN mRNA , since upon DNA damage , B . subtilis cells lacking fumarase , have three fold higher amounts RecN protein in cell extracts . Importantly , this higher level of RecN can be reversed by growth of the cells in the presence of L-malic acid . To summarize these notions , the activity of fumarase and L-malic acid are required for an efficient DNA damage response and their effect on RecN has two consequences: ( 1 ) A change in the localization of RecN , ( upon DNA damage induction ) , from foci throughout the cell to co-localization with the condensed DNA and ( 2 ) A two fold over-expression of RecN at the protein level . While we do not know how lack of fumarase and accumulation of L-malic acid affect the localization of RecN we do know that the change in expression of RecN occurs at the level of translation . How could L-malic acid affect RecN translation ? There are three different ways to temporally regulate gene expression at the translational level: through trans-acting proteins , through cis-acting mRNA elements , acting as riboswitches ( Kirchner and Schneider , 2017; Perez-Gonzalez et al . , 2016 ) and through transacting RNAs ( small RNA ) ( Kim et al . , 2009 ) . There are no known riboswitches in the RecN gene yet riboswitches have been detected in the upstream gene , ahrC , of the RecN operon , which appears to be highly regulated in Bacillus ( Dar et al . , 2016 ) . The ahrC gene participates in the metabolism of arginine and the riboswitches were identified by term-seq which is quantitative mapping of all exposed RNA 3′ ends in bacteria . This allowed unbiased , genome-wide identification of genes that are regulated by premature transcription termination . Small untranslated RNA SR1 , from the Bacillus subtilis genome , is a regulatory RNA involved in fine-tuning of arginine catabolism ( Gimpel et al . , 2012 ) . SR1 is an sRNA that acts as a base-pairing regulatory RNA on the ahrC mRNA . The interaction of SR1 and ahrC mRNA does not lead to degradation of ahrC mRNA , but inhibited translation at a post-initiation stage ( Heidrich et al . , 2006 ) . Overexpression of RecN has been shown to be lethal for B . subtilis cells . Thus , one of the roles of L-malic acid may be to maintain appropriate RecN levels .
B . subtilis strains are listed in Supplementary file 1 , S . cerevisiae strains are listed in Supplementary file 2 of the supplemental material . The plasmids and primers referred to in this study are described in Plasmid construction . All general methods were carried out as described previously ( Harwood and Cutting , 1990 ) . Molecular biological methods for Bacillus . Wiley , Chichester , United Kingdom ) . Cultures were inoculated at an optical density at 600 nm ( OD600 ) of 0 . 05 from an overnight culture , and growth was carried out at 37°C in LB medium . During logarithmic phase ( OD600 of 0 . 4 to 0 . 6 ) , 0 . 5% xylose or 0 . 1% IPTG was added to induce citG ( fumarase ) expression , as indicated . Pfum-GFP ( fum-GFP-kan ) , containing the 3’ region of fum-bc fused to gfp , was constructed by amplifying the 3’ region by PCR using primers: F GAATTC TTC CAT GAT AAA TGT GCT GT R CTCGAG CGC CTT TGG TTT TAC CAT G , which replaced the stop codon with a XhoI site . The PCR-amplified DNA was digested with EcoRI and XhoI and was cloned into the EcoRI and XhoI sites of pKL168 ( kan ) ( Lemon and Grossman , 1998 ) , which contains the in frame gfp coding sequence . Pfum-bc ( amyE::fum-spc ) , containing flanking amyE sequences and the spc gene , was constructed by amplifying the citG gene by PCR using primers: F GTCGAC ATG GAA TAC AGA ATT GAA CGA R GCTAGC G CAG CCG TTC TTC CTA TTA . The PCR-amplified DNA was digested with SalI and NheI and cloned into the SalI and NheI sites of pDR111 ( amyE::spc ) . PH127R , pDR150 [amyE::fum-bc ( spec ) ] is an ectopic integration containing the xylose-inducible promoter . pDR150 was generated by site-directed mutagenesis using the KAPAHiFiTM kit , using primers: F CGT CCA AAT GAT GAC GTG AAC P R A ATC GTT TGA TCA GAG TTC TTC P PH186N , pDR150 [amyE::fum-bc ( spec ) ] is an ectopic integration containing the xylose-inducible promoter . pDR150 was generated by site-directed mutagenesis using the KAPAHiFiTM kit , using primers: F GAT CTT CAG GAT GCT ACG R CGT GCG TCC GAT TTT GAC Pyfum yeast expression vector yep51 , was constructed by amplifying the citG gene by PCR using primers: F GTCGAC ATG GAA TAC AGA ATT GAA CGA R GGATCC CGC CTT TGG TTT TAC CAT G . The PCR-amplified DNA was digested with SalI and BamHI and cloned into the SalI and BamHI sites of YEp51 . Samples ( 0 . 5 mL ) of a given culture were removed , centrifuged briefly , and resuspended in 10 µL of PBS × 1 ( Phosphate-Buffered Saline ) supplemented with 1 µg/mL FM4–64 ( Molecular Probes , Invitrogen ) . Cells were visualized and photographed using an Axioplan2 microscope ( Zeiss ) equipped with CoolSnap HQ camera ( Photometrics , Roper Scientific ) or an Axioobserver Z1 microscope ( Zeiss ) equipped with a CoolSnap HQII camera ( Photometrics , Roper Scientific ) . System control and image processing were performed using MetaMorph 7 . 2r4 software ( Molecular Devices ) . GC-MS analysis of three organic acids was performed using gas chromatograph ( Agilent 7890A ) coupled to the mass selective ( Agilent 5975C MSD ) . The gas chromatograph was equipped with the CTC COMBI PAL autosampler . Mass spectrometer was operated in SIM mode ( single ion monitoring ) . Plasma samples were dissolved in water following the addition of isotopically labeled succinic acid – D6 . The samples were cleaned by SPE ( Phenomenex Strata X-AW ) and dried over a stream of nitrogen . Acids were chemically derivatized by trimethyl silylation before GC-MS analysis . For RecN half-life determination , strains RecN-GFP and RecN-GFP , Δfum were grown to an OD600 = 0 . 4 at 37°C in LB . MMS ( 0 . 07 v/v ) was added to an aliquot , and the cells were incubated for 30 min . Then , rifampicin ( 100 μg/ml ) or chloramphenicol ( 20 μg/ml ) were added . Aliquots were then collected at variable times , and cell growth was halted by addition of NaN3 ( 10 μM ) . Cells were harvested in lysis buffer containing: 10 mM Tris pH 8 , 10 mM MgCl2 , 0 . 2 mg/ml AEBSF ( MegaPharm-101500 ) , 0 . 5 mg/ml Lysozyme ( USBiological-L9200 ) , 5 µg/ml DnaseI ( Sigma DN25 ) . Protein concentrations were determined using Bradford analyses . Samples were separated on 10% SDS-PAGE gels , and then transferred onto PVDF membranes ( Millipore ) . The following primary antibodies were used: Polyclonal anti-yeast fumarase and anti-human FH were generated in rabbits injected with the purified proteins . Monoclonal anti GFP was a product of Roche ) . Monoclonal anti SigmaA was kindly provided by M . Fujitas lab , . Polyclonal anti ICDH and anti citZ were product of kerafast . Blots were incubated with the appropriate IgG-HRP-conjugated secondary antibody . Protein bands were visualized using the ECL immunoblotting detection system ( GE Healthcare ) and developed on an ImageQuant LAS4000 mini Fuji luminescence imagining system . For the analysis of protein expression , bands from at least three independent experiments were quantified by densitometry using Image J analysis software . Total RNA was extracted by using a FastRNA blue kit ( MP ) according to manufacturer's instructions . For integrity assessment and purification level of extracted RNA , 2% agarose gel electrophoresis as well as spectrophotometric assays were performed . The extracted RNA was reverse-transcribed into cDNA by using a Maxima First Strand cDNA Synthesis Kit with dsDNase ( Thermo scientific ) . To reveal the modification in the expression of RecN , real-time reverse transcriptase PCR ( real-time RT PCR ) was exploited using SYRB Green dye ( Thermo scientific ) and the Mic system ( Bio Molecular Systems ) . The raw data were further normalization to the yoxA gene . Three independent experiments were conducted . Sequences of the primers used in the current study were as follows: yoxA: F ATACAATGCGGACGGAAAAC R GGCTCCAGCACTTGTAAACC RecN F CAGGCTCCTTGAACTGCTG R CGT CAG TTC CTC AAT AAT GGC RecN ( set2 ) F TGC ATT ACA CAC CTG CCT CA R CGC TAC CTT TTC CTG CTT AG When more than two groups were compared , statistical analysis was performed by one-way repeated measure analysis of variance with Duncan’s test . When only two groups were compared , significance was analyzed by the paired t test . | Living cells make an enzyme called fumarase . It converts a chemical called fumaric acid into L-malic acid . This is a crucial step in primary metabolism and aerobic respiration , the process of using oxygen to release energy for life . Yet it is not the only role that fumarase plays . In the cells of eukaryotes such as plants , animals and even baker’s yeast , aerobic respiration happens inside compartments called mitochondria . Yet fumarase is also found in the nucleus , which contains the cell’s genetic material . Inside the nucleus , this enzyme takes part in the DNA damage response that senses and repairs damage to the genetic code . Simpler organisms , like bacteria , do not have mitochondria or a nucleus . Instead , all their reactions take place inside the main space within the cell . The current model for the evolution of fumarase is that the enzyme evolved in an ancient bacterium for the production of energy . Then , in more complex organisms , becoming split between the mitochondria and the nucleus allowed it to take on a second role in the DNA damage response . Singer et al . now challenge that model , and show that fumarase takes part in DNA damage repair in bacteria too . Bacillus subtilis has one fumarase gene , known as fum-bc . Singer et al . showed that , without this gene , the bacteria do not grow well under conditions where they need to use aerobic respiration . But , the bacteria also became sensitive to DNA-damaging agents such as ionizing radiation or a chemical called methyl methanesulfonate . Singer et al . then expressed the bacterial fum-bc gene in baker’s yeast , Saccharomyces cerevisiae . This organism has mitochondria and a cell nucleus . With the yeast's own fumarase gene switched off , the bacterial fumarase was able to take on both roles – aerobic respiration and the DNA damage response . In bacteria grown with the DNA-damaging chemical , the level of fumarase started to rise . A fluorescent tag revealed that it also changed location , moving close to the bacteria’s DNA . As such , even in bacteria , fumarase has two roles . Further experiments showed that the L-malic acid made by fumarase affects the production of a protein called RecN , and it is this protein that triggers DNA repair . These findings shed new light on the evolution of fumarase , and suggest that its dual role evolved before its dual location in eukaryotes . The next step is to find out exactly how L-malic acid affects the production of RecN . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2017 | Bacterial fumarase and L-malic acid are evolutionary ancient components of the DNA damage response |
We previously discovered a new osteogenic growth factor that is required to maintain adult skeletal bone mass , Osteolectin/Clec11a . Osteolectin acts on Leptin Receptor+ ( LepR+ ) skeletal stem cells and other osteogenic progenitors in bone marrow to promote their differentiation into osteoblasts . Here we identify a receptor for Osteolectin , integrin α11 , which is expressed by LepR+ cells and osteoblasts . α11β1 integrin binds Osteolectin with nanomolar affinity and is required for the osteogenic response to Osteolectin . Deletion of Itga11 ( which encodes α11 ) from mouse and human bone marrow stromal cells impaired osteogenic differentiation and blocked their response to Osteolectin . Like Osteolectin deficient mice , Lepr-cre; Itga11fl/fl mice appeared grossly normal but exhibited reduced osteogenesis and accelerated bone loss during adulthood . Osteolectin binding to α11β1 promoted Wnt pathway activation , which was necessary for the osteogenic response to Osteolectin . This reveals a new mechanism for maintenance of adult bone mass: Wnt pathway activation by Osteolectin/α11β1 signaling .
The maintenance of the adult skeleton requires the formation of new bone throughout life as a result of the differentiation of skeletal stem/progenitor cells into osteoblasts . Leptin Receptor+ ( LepR+ ) bone marrow stromal cells are the major source of osteoblasts and adipocytes in adult mouse bone marrow ( Zhou et al . , 2014 ) . These cells arise postnatally in the bone marrow , where they are initially rare and make little contribution to the skeleton during development , but expand to account for 0 . 3% of cells in adult bone marrow ( Mizoguchi et al . , 2014; Zhou et al . , 2014 ) . Nearly all fibroblast colony-forming cells ( CFU-F ) in adult mouse bone marrow arise from these LepR+ cells , and a subset of LepR+ cells form multilineage colonies containing osteoblasts , adipocytes , and chondrocytes , suggesting they are highly enriched for skeletal stem cells ( Zhou et al . , 2014 ) . LepR+ cells are also a critical source of growth factors that maintain hematopoietic stem cells and other primitive hematopoietic progenitors in bone marrow ( Ding and Morrison , 2013; Ding et al . , 2012; Himburg et al . , 2018; Oguro et al . , 2013 ) . To identify new growth factors in the bone marrow , we performed RNA-seq analysis on LepR+ cells and looked for transcripts predicted to encode secreted proteins with sizes and structures similar to growth factors and whose function had not been studied in vivo . We discovered that Clec11a , a secreted glycoprotein of the C-type lectin domain superfamily ( Bannwarth et al . , 1999; Bannwarth et al . , 1998 ) , was preferentially expressed by LepR+ cells ( Yue et al . , 2016 ) . Prior studies had observed Clec11a expression in bone marrow but inferred based on colony-forming assays in culture that it was a hematopoietic growth factor ( Hiraoka et al . , 1997; Hiraoka et al . , 2001 ) . We made germline knockout mice and found it is not required for normal hematopoiesis but that it is required for the maintenance of the adult skeleton ( Yue et al . , 2016 ) . The mutant mice formed their skeleton normally during development and were otherwise grossly normal as adults but exhibited significantly reduced osteogenesis and bone volume beginning by 2 months of age ( Yue et al . , 2016 ) . Recombinant protein promoted osteogenic differentiation by bone marrow stromal cells in vitro and in vivo ( Yue et al . , 2016 ) . Based on these observations we proposed to call this new osteogenic growth factor , Osteolectin , so as to have a name related to its biological function . Osteolectin/Clec11a is expressed by a subset of LepR+ stromal cells in the bone marrow as well as by osteoblasts , osteocytes , and hypertrophic chondrocytes . The discovery of Osteolectin offers the opportunity to better understand the mechanisms that maintain the adult skeleton; however , the Osteolectin receptor and the signaling mechanisms by which it promotes osteogenesis are unknown . Several families of growth factors , and the signaling pathways they activate , promote osteogenesis , including Bone Morphogenetic Proteins ( BMPs ) , Fibroblast Growth Factors ( FGFs ) , Hedgehog proteins , Insulin-Like Growth Factors ( IGFs ) , Transforming Growth Factor-betas ( TGF-βs ) , and Wnts ( reviewed by Karsenty , 2003; Kronenberg , 2003; Wu et al . , 2016 ) . Bone marrow stromal cells regulate osteogenesis by skeletal stem/progenitor cells by secreting multiple members of these growth factor families ( Chan et al . , 2015 ) . The Wnt signaling pathway is a particularly important regulator of osteogenesis , as GSK3 inhibition and β-catenin accumulation promote the differentiation of skeletal stem/progenitor cells into osteoblasts ( Bennett et al . , 2005; Dy et al . , 2012; Hernandez et al . , 2010; Krishnan et al . , 2006; Kulkarni et al . , 2006; Rodda and McMahon , 2006 ) . Consistent with this , mutations that promote Wnt pathway activation increase bone mass in humans and in mice ( Ai et al . , 2005; Balemans et al . , 2001; Boyden et al . , 2002 ) while mutations that reduce Wnt pathway activation reduce bone mass in humans and in mice ( Gong et al . , 2001; Holmen et al . , 2004; Kato et al . , 2002 ) . The Wnt pathway can be activated by integrin signaling . There are 18 integrin α subunits and 8 β subunits , forming 24 different functional integrin heterodimer complexes ( Humphries et al . , 2006; Hynes , 1992 ) . Integrin signaling promotes Wnt pathway activation through Integrin-Linked Kinase ( ILK ) -mediated phosphorylation of GSK3 and nuclear translocation of β-catenin ( Burkhalter et al . , 2011; Delcommenne et al . , 1998; Novak et al . , 1998; Rallis et al . , 2010 ) . Conditional deletion of Ilk or Ptk2 ( which encodes Focal Adhesion Kinase , FAK ) from osteoblast progenitors reduces osteogenesis and depletes trabecular bone in adult mice ( Dejaeger et al . , 2017; Sun et al . , 2016 ) , suggesting a role for integrins in adult osteogenesis . Conditional deletion of β1 integrin from chondrocytes or skeletal stem/progenitor cells impairs chondrocyte function and skeletal ossification during development ( Aszodi et al . , 2003; Raducanu et al . , 2009; Shekaran et al . , 2014 ) . Activation of αvβ1 signaling by Osteopontin ( Chen et al . , 2014 ) or α5β1 signaling by Fibronectin ( Hamidouche et al . , 2009; Moursi et al . , 1997 ) promotes the osteogenic differentiation of mesenchymal progenitors . Germline deletion of integrin α10 leads to defects in chondrocyte proliferation and growth plate function ( Bengtsson et al . , 2005 ) and germline deletion of integrin α11 leads to defects in tooth development ( Popova et al . , 2007 ) . However , little is known about which integrins are required for adult osteogenesis in vivo .
The Osteolectin/Clec11a gene first appeared in bony fish and is conserved among bony vertebrates ( Yue et al . , 2016 ) . Osteolectin contains a glutamic acid-rich sequence , an alpha-helical leucine zipper , and a C-type lectin domain ( Figure 1A and B ) ( Bannwarth et al . , 1999; Bannwarth et al . , 1998 ) . To generate hypotheses regarding potential Osteolectin receptors , we examined the amino acid sequence and found two integrin-binding motifs , RGD ( Gardner and Hynes , 1985; Pierschbacher and Ruoslahti , 1984; Plow et al . , 1985 ) and LDT ( Fong et al . , 1997; Viney et al . , 1996 ) in human ( Figure 1A ) and mouse Osteolectin ( Figure 1B ) . One or both of these motifs were conserved across Osteolectin sequences in all bony vertebrates ( Figure 1C ) . This raised the possibility that the Osteolectin receptor might be an integrin . Given that bone marrow stromal cells undergo osteogenesis in response to Osteolectin ( Yue et al . , 2016 ) , we examined the expression of all α and β integrins in mouse bone marrow stromal cells by RNA-seq analysis . Among the genes that encode α integrins , Itga1 ( encoding α1 ) , Itga6 ( encoding α6 ) , Itga11 ( encoding α11 ) and Itgav ( encoding αv ) , were strongly expressed by bone marrow stromal cells ( Figure 1D ) . Among the genes that encode β integrins , only Itgb1 ( encoding β1 ) was strongly expressed ( Figure 1E ) . Itga1 , Itga6 , and Itgav were strongly expressed by both LepR+ cells and endothelial cells , and and are widely expressed in non-osteogenic cells , where they have known ligands ( Belkin et al . , 1990; Defilippi et al . , 1991; Lee et al . , 2006; Mahabeleshwar et al . , 2006; Yang et al . , 2008 ) , suggesting they are less likely to encode the Osteolectin receptor ( Figure 1F ) . In contrast , Itga11 was expressed exclusively by LepR+ cells , not by endothelial cells or other bone marrow cells ( Figure 1F ) . Quantitative reverse transcription PCR ( qRT-PCR ) analysis of sorted bone marrow cells ( Supplementary file 1 shows the markers used to isolate these cells ) showed that Itga11 was highly expressed by LepR+CD45-Ter119-CD31- stromal cells and Col2 . 3-GFP+CD45-Ter119-CD31- osteoblasts but not by any hematopoietic stem or progenitor population ( Figure 1G ) . The expression patterns of Itga11 and Itgb1 were thus consistent with a potential role in osteogenesis . Consistent with our results , integrin α11 is expressed by human bone marrow stromal cells in a way that correlates with osteogenic potential in culture ( Kaltz et al . , 2010 ) ; however , α11 is not known to regulate osteogenesis . Integrin α11 heterodimerizes with integrin β1 ( Velling et al . , 1999 ) and the only known ligand for α11β1 is collagen ( Popova et al . , 2004; Velling et al . , 1999 ) . Few cells express integrin α11 , and it has been studied less than other integrins . Itga11 deficient mice are growth retarded and have smaller bones , but this was thought to reflect a defect in incisor development that leads to malnutrition ( Popova et al . , 2007 ) . To test whether α11β1 binds Osteolectin , we overexpressed Flag-tagged human Osteolectin in MC3T3-E1 pre-osteoblast cells and immunoprecipitated with anti-Flag beads . The anti-Flag beads pulled down the tagged Osteolectin along with endogenous integrin α11 and integrin β1 ( Figure 1H ) . We then tested the affinity of recombinant human Osteolectin for multiple recombinant human integrin complexes by a microtiter well binding assay . Osteolectin selectively bound to integrin α11β1 and α10β1 , but not to other integrins , including αVβ1 , αVβ3 , α4β1 , α9β1 , αIIbβ3 or αMβ2 ( Figure 1I ) . Integrin α10 is the gene most closely related to α11 . Integrin α10 is expressed by osteoblasts and chondrocytes ( Bengtsson et al . , 2005; Engel et al . , 2013 ) but only at a low level by bone marrow stromal cells ( Figure 1D ) . The dissociation constant ( kd ) of Osteolectin for α10β1 and α11β1 was 0 . 3 nM whereas the kd of Osteolectin for other integrins was >100 nM . The kd of human Pro-Collagen 1α for α11β1 was also high ( 0 . 1 nM; Figure 1J ) ; however , in contrast to Osteolectin , addition of Pro-Collagen 1α to culture , either by adding it to the culture medium or using it to coat the plates , did not promote osteogenesis by MC3T3-E1 cells or two primary human bone marrow stromal cell lines ( hBMSC#1 or hBMSC#2; Figure 1K ) . A peptide that inhibits the binding of integrins to RGD-containing ligands , RGDS ( Arg-Gly-Asp-Ser ) ( Gardner and Hynes , 1985; Plow et al . , 1985; Ruoslahti , 1996 ) , inhibited the binding of integrin α11β1 to Osteolectin ( Figure 1L ) and the osteogenic response of MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells to Osteolectin in culture ( Figure 1M ) . MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells secrete Osteolectin into the culture medium ( Figure 2A ) , consistent with our observation that Osteolectin is synthesized by a subset of LepR+ bone marrow stromal cells ( Yue et al . , 2016 ) . Deletion of Osteolectin from these cell lines reduced their osteogenic differentiation in osteogenic differentiation medium ( Figure 2B and C ) , demonstrating that autocrine Osteolectin production is part of what drives osteogenesis by these cells in culture . To assess the signaling mechanisms by which Osteolectin promotes osteogenesis , we treated parental MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells with recombinant Osteolectin and assessed the levels of phosphorylated PI3-kinase , Akt , and GSK3 . The most prominent change we observed was a dramatic increase of GSK3 phosphorylation at Ser21/9 within 30 to 60 min of Osteolectin treatment in all three cell lines ( Figure 2D ) . Phosphorylation at Ser21/9 inhibits the GSK3-mediated degradation of β-catenin , increasing β-catenin levels and promoting the transcription of Wnt pathway target genes ( Cross et al . , 1995; Peifer et al . , 1994; Yost et al . , 1996 ) . We did not observe an increase in β-catenin levels within 1 hr of Osteolectin treatment , but did detect increased β-catenin levels in all three cell lines within 24 hr of Osteolectin treatment ( Figure 2E ) . The transcription of several Wnt target genes , including Axin2 ( Jho et al . , 2002; Lustig et al . , 2002; Yan et al . , 2001; Yan et al . , 2009 ) , Lef1 ( Filali et al . , 2002; Gaur et al . , 2005; Hovanes et al . , 2001 ) , Runx2 ( Dong et al . , 2006; Gaur et al . , 2005 ) , and Alkaline phosphatase ( Rawadi et al . , 2003 ) , were activated within 24 hr of Osteolectin treatment ( Figure 2F ) . Osteolectin deficient MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells had lower levels of phospho-GSK3 and β-catenin as compared to parental cells ( Figure 2G ) as well as significantly lower levels of Wnt target genes ( Figure 2H ) . These data demonstrate that Osteolectin promotes Wnt pathway activation in osteogenic cells . To test whether Wnt pathway activation phenocopies the effects of Osteolectin , we evaluated the effects of AZD2858 , which inhibits GSK3 function and promotes β-catenin accumulation ( Berg et al . , 2012 ) . As expected ( Gilmour et al . , 2013; Marsell et al . , 2012; Sisask et al . , 2013 ) , AZD2858 increased GSK3 phosphorylation and β-catenin levels ( Figure 3A ) as well as osteogenic differentiation ( Figure 3B ) by MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells . Osteolectin also increased GSK3 phosphorylation , β-catenin levels , and osteogenic differentiation by MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells ( Figure 3A and B ) . However , when the two agents were added together , there was no further promotion of osteogenic differentiation beyond the effects of the individual agents ( Figure 3B ) . These data suggest that Osteolectin and GSK3/β-catenin act in the same pathway to promote osteogenic differentiation by mesenchymal progenitors . To test whether Osteolectin requires β-catenin to promote osteogenesis , we evaluated an inhibitor of Wnt pathway signaling , IWR-1-endo , which depletes β-catenin by stabilizing Axin2 in the β-catenin destruction complex ( Chen et al . , 2009 ) . As expected , Osteolectin increased β-catenin levels and osteogenesis by MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells while IWR-1-endo reduced β-catenin levels and osteogenesis ( Figure 3C and D ) . When added together , IWR-1-endo blocked the effect of Osteolectin on β-catenin levels and osteogenesis ( Figure 3C and D ) . This suggests that Osteolectin requires β-catenin to induce osteogenesis by mesenchymal progenitors . To test if Integrin α11 is required for the osteogenic response to Osteolectin , we used CRISPR/Cas9 to delete Itga11 from MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells . Deletion of Itga11 significantly reduced osteogenesis by each cell line in culture ( Figure 4A and B ) . Addition of recombinant Osteolectin to culture significantly promoted osteogenesis by parental , but not Itga11 deficient , MC3T3-E1 , hBMSC#1 , and hBMSC#2 cells ( Figure 4A and B ) . Integrin α11 is therefore required by mouse pre-osteoblast cells and human bone marrow stromal cells to undergo osteogenesis in response to Osteolectin . The Itga11 deficient MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells also exhibited lower levels of GSK3 phosphorylation and β-catenin as compared to parental control cells ( Figure 4C ) . Addition of recombinant Osteolectin to culture increased the levels of phosphorylated GSK3 and β-catenin in parental control cells but not in Itga11 deficient cells ( Figure 4C ) . We also observed significantly lower levels of Wnt target gene transcripts in the Itga11 deficient cells ( Figure 4D ) . Addition of recombinant Osteolectin to culture significantly increased the levels of Wnt target gene transcripts in parental control cells but not in Itga11 deficient cells ( Figure 4D ) . Integrin α11 is therefore required by mouse pre-osteoblast cells and human bone marrow stromal cells to activate Wnt pathway signaling in response to Osteolectin . Given that Collagen 1α bound to α11β1 ( Figure 1J ) but did not promote osteogenesis by MC3T3-E1 cells , hBMSC#1 cells , or hBMSC#2 cells ( Figure 1K ) , we tested whether Collagen 1α promoted Wnt pathway activation in these cells . Addition of recombinant Osteolectin to culture increased levels of phosphorylated GSK3 , β-catenin , and Wnt target gene transcripts in MC3T3-E1 cells , hBMSC#1 cells , and hBMSC#2 cells ( Figure 4E and F ) ; however , addition of Pro-Collagen 1α to these cells did not seem to have any effect on the levels of phosphorylated GSK3 , β-catenin , or Wnt target gene transcripts in these cells ( Figure 4E and F ) . We also added Osteolectin or Pro-Collagen 1α to freshly isolated mouse bone marrow stromal cells in culture . Both Pro-Collagen 1α and Osteolectin promoted Focal Adhesion Kinase ( FAK ) phosphorylation at Tyrosine 397 within one hour of addition to culture without affecting total FAK levels ( Figure 4G ) , consistent with the activation of integrin signaling ( Cooper et al . , 2003 ) . Although Pro-Collagen 1α had not detectably affected GSK3 phosphorylation in MC3T3-E1 cells , hBMSC#1 cells , or hBMSC#2 cells ( Figure 4E ) , it did increase the levels of phosphorylated GSK3 in primary mouse bone marrow stromal cells , as did Osteolectin ( Figure 4G and H ) . Osteolectin , but not Pro-Collagen 1α , promoted β-catenin accumulation within 24 hr of treatment ( Figure 4H ) . To more precisely assess the time required for Osteolectin to promote β-catenin accumulation , we treated primary mouse bone marrow stromal cells with Osteolectin or phosphate-buffered saline ( PBS ) then assessed the levels of nuclear versus cytosolic/membrane associated β-catenin 2 , 4 , 6 , 12 , and 24 hr later . Osteolectin treatment did not significantly affect the levels of cytosolic/membrane associated β-catenin at any time point , but did increase nuclear β-catenin within 4 hr of treatment ( Figure 4I ) . Osteolectin treatment did not significantly affect Ctnnb1 ( which encodes β-catenin ) transcript levels in these cells ( Figure 4J ) , suggesting that the increase in nuclear β-catenin reflected an inhibition of proteasomal degradation . Together , these data suggest that Osteolectin promoted integrin α11 signaling , leading to Wnt pathway activation , accumulation of nuclear β-catenin , and increased transcription of Wnt target genes involved in osteogenesis . Exogenous Pro-Collagen 1α also appeared to activate integrin signaling , at least in primary mouse bone marrow stromal cells ( Figure 4G ) , but not accumulation of nuclear β-catenin ( Figure 4H ) or increased transcription of Wnt target genes ( Figure 4F ) , offering a potential explanation for its failure to promote osteogenesis . Itga11-deficiency blocked the ability of Osteolectin to promote FAK phosphorylation in primary bone marrow stromal cells , but did not affect the ability of Pro-Collagen 1α to promote FAK phosphorylation ( Figure 4K ) . This suggests that Osteolectin promotes integrin signaling in an α11-dependent manner but that Pro-Collagen 1α promotes integrin signaling in an α11-independent manner . This was expected as bone marrow stromal cells express multiple integrins that are capable of functioning as collagen receptors , including α1β1 , α10β1 and αVβ3 ( Figure 1D and E ) ( Davis , 1992; Gullberg and Lundgren-Akerlund , 2002 ) . To test whether Integrin α11 is necessary for osteogenesis in vivo we generated mice bearing a floxed allele of Itga11 ( Figure 5—figure supplement 1A–C ) , then conditionally deleted it from skeletal stem and progenitor cells in the bone marrow using Lepr-Cre . Only 5% of osteoblasts derive from LepR+ cells at two months of age but this number increases to approximately 50% by 10 months of age ( Zhou et al . , 2014 ) . Lepr-Cre; Itga11fl/fl mice did not exhibit the defects in incisor development ( data not shown ) or the growth retardation observed in germline Itga11-/- mice ( Popova et al . , 2007 ) . Lepr-Cre; Itga11fl/fl mice appeared grossly normal ( Figure 5A ) , with body lengths ( Figure 5B ) , body masses ( Figure 5C ) , and femur lengths ( Figure 5D ) that did not significantly differ from sex-matched littermate controls . However , qRT-PCR analysis showed that LepR+ bone marrow cells from Lepr-Cre; Itga11fl/fl mice had an approximately 85% reduction in Itga11 transcript levels as compared to LepR+ cells from control mice ( Figure 5—figure supplement 1D ) . Serum Osteolectin levels did not significantly differ between Lepr-Cre; Itga11fl/fl mice and littermate controls at 2 or 12 months of age but were modestly higher in Lepr-Cre; Itga11fl/fl mice than in controls at 6 months of age ( Figure 5E ) . This demonstrates that Integrin α11 is not required for the synthesis or secretion of Osteolectin . To test whether deletion of Itga11 from LepR+ cells affected osteogenesis in vivo , we performed micro-CT analysis of the distal femur from 2 , 6 , and 12 month old Lepr-Cre; Itga11fl/fl mice and sex-matched littermates . Consistent with the observation that LepR+ cells contribute little to skeletal development prior to 2 months of age ( Zhou et al . , 2014 ) , we observed no significant difference in trabecular bone parameters between Lepr-Cre; Itga11fl/fl mice and sex-matched littermates at 2 months of age ( Figure 5F–L ) . However , LepR+ cells and Osteolectin are necessary for adult osteogenesis ( Yue et al . , 2016; Zhou et al . , 2014 ) . Consistent with this , 6 and 12-month-old male and female Lepr-Cre; Itga11fl/fl mice had significantly reduced trabecular bone volume as compared to sex-matched littermate controls ( Figure 5F and G ) . At 6 and 12 months of age , male and female Lepr-Cre; Itga11fl/fl mice also tended to have lower trabecular number ( Figure 5H ) and trabecular thickness ( Figure 5I ) than sex matched littermate controls . Calcein double labelling showed that the mineral apposition rate was significantly reduced in trabecular bone from Lepr-Cre; Itga11fl/fl mice as compared to sex-matched littermates at 6 and 12 months of age ( Figure 5M ) . Levels of Procollagen type 1 N-terminal Propeptide ( P1NP ) , a marker of bone formation , were also significantly lower in the serum of Lepr-Cre; Itga11fl/fl mice as compared to littermate controls at 6 and 12 months of age ( Figure 5O ) . Integrin α11 is , therefore , required by LepR+ cells and their progeny for normal rates of trabecular bone formation and maintenance of trabecular bone volume during adulthood , phenocopying the accelerated trabecular bone loss in adult Osteolectin deficient mice ( Yue et al . , 2016 ) . While Lepr-Cre; Itga11fl/fl mice had significantly reduced rates of bone formation as compared to sex-matched littermates ( Figure 5M ) , they did not significantly differ in the urinary bone resorption marker deoxypyridinoline at 6 or 12 months of age ( Figure 5N; this was not tested in 2-month-old mice because no difference in bone parameters was observed at that age ) . This suggests that , like Osteolectin , Integrin α11 promotes bone formation but does not regulate bone resorption ( Yue et al . , 2016 ) . Osteolectin deficiency has a milder effect on cortical bone as compared to trabecular bone , with no significant reduction in cortical bone until after 10 months of age ( Yue et al . , 2016 ) . Consistent with this , femur cortical bone parameters did not significantly differ between Lepr-Cre; Itga11fl/fl mice and sex-matched littermates at 2 or 6 month of age ( Figure 6A–F ) . However , cortical bone mineral density was significantly lower in male and female Lepr-Cre; Itga11fl/fl mice as compared to sex-matched littermates at 12 months of age ( Figure 6F ) . Calcein double labelling revealed that the mineral apposition rate was significantly reduced in cortical bone from male and female Lepr-Cre; Itga11fl/fl mice as compared to sex-matched littermate controls at 6 and 12 months of age ( Figure 6G ) . Deletion of Integrin α11 from LepR+ cells thus reduces the rate of cortical bone formation during adulthood , slowly leading to a thinning of cortical bone that became apparent in the femurs at 12 months of age , phenocopying the slow loss of cortical bone in adult Osteolectin deficient mice ( Yue et al . , 2016 ) . To test whether Integrin α11 is necessary for the maintenance or the proliferation of skeletal stem/progenitor cells in the bone marrow , we cultured at clonal density enzymatically dissociated femur bone marrow cells from Lepr-Cre; Itga11fl/fl and sex-matched littermate control mice at 2 and 6 months of age . We observed a slight , but statistically significant , reduction in the frequency of cells that formed CFU-F colonies in Lepr-Cre; Itga11fl/fl mice at 2 months of age , though no significant difference was apparent at 6 months of age ( Figure 7A ) . We observed no significant difference in the number of cells per colony at either age ( Figure 7B ) . Integrin α11 is therefore not required for the maintenance of CFU-F in vivo or for their proliferation in vitro . To test whether integrin α11 regulates the differentiation of bone marrow stromal cells , we cultured CFU-F from Lepr-Cre; Itga11fl/fl and littermate control mice at clonal density , then replated equal numbers of cells from Itga11 deficient and control colonies into osteogenic or adipogenic culture conditions ( Figure 7C and D ) . We also centrifuged 2 × 105 CFU-F cells from Lepr-Cre; Itga11fl/fl and control colonies to form pellets and then cultured them in chondrogenic medium ( Figure 7E ) . Consistent with the decreased osteogenesis from Itga11 deficient mesenchymal cell lines in culture ( Figure 4A ) and the reduced osteogenesis in Lepr-Cre; Itga11fl/fl mice in vivo ( Figure 5 and Figure 6 ) , bone marrow stromal cells from Lepr-Cre; Itga11fl/fl mice formed significantly less bone in culture as compared to control colonies ( Figure 7C ) . This demonstrates that , like Osteolectin , Integrin α11 promotes osteogenesis by bone marrow stromal cells . We did not detect any difference between Lepr-Cre; Itga11fl/fl and control colonies in adipogenic or chondrogenic differentiation ( Figure 7D and E ) . This is also consistent with the Osteolectin deficiency phenotype , which reduced osteogenesis in vitro and in vivo without having any detectable effect on adipogenesis or chondrogenesis ( Yue et al . , 2016 ) . To test if Integrin α11 is necessary for the osteogenic response of bone marrow stromal cells to Osteolectin , we cultured CFU-F from the bone marrow of 2-month-old Lepr-Cre; Itga11fl/fl mice and littermate controls then added osteogenic differentiation medium with or without recombinant mouse Osteolectin . Osteolectin significantly increased osteogenic differentiation by control colonies , but Lepr-Cre; Itga11fl/fl colonies underwent significantly less osteogenesis and did not respond to Osteolectin ( Figure 7F and G ) . Osteolectin treatment also significantly increased the levels of the Wnt target gene transcripts Alp , Axin2 , Lef1 , and Runx2 in cells from control mice , but not Lepr-Cre; Itga11fl/fl mice ( Figure 7—figure supplement 1A ) . Bone marrow stromal cells thus require Integrin α11 to undergo osteogenesis in response to Osteolectin . To test if bone marrow stromal cells require Integrin α11 to undergo osteogenesis in response to Osteolectin in vivo , we administered daily subcutaneous injections of recombinant mouse Osteolectin to 2-month-old Lepr-Cre; Itga11fl/fl and littermate control mice for 28 days . Osteolectin is functionally important for bone maintenance by 2 months of age given that Osteolectin deficient mice exhibit a significant reduction in trabecular bone volume at 2 months of age ( Yue et al . , 2016 ) . Consistent with our prior study ( Yue et al . , 2016 ) , in the distal femur metaphysis of control mice , Osteolectin treatment significantly increased trabecular bone volume ( Figure 7H and I ) , trabecular bone number ( Figure 7J ) , and trabecular connectivity density ( Figure 7M ) , while significantly reducing trabecular spacing ( Figure 7L ) . However , Osteolectin treatment had no significant effect on these parameters in Lepr-Cre; Itga11fl/fl mice ( Figure 7H–N ) . Osteolectin treatment also significantly increased the levels of the Wnt target gene transcripts Alp , Lef1 , and Runx2 in LepR+ cells isolated from the bone marrow of control mice , but not Lepr-Cre; Itga11fl/fl mice ( Figure 7—figure supplement 1B ) . LepR+ bone marrow stromal cells and their progeny thus require Integrin α11 to undergo osteogenesis in response to Osteolectin in vivo . Neither Osteolectin administration nor Itga11 deficiency had any significant effect on cortical bone parameters in this relatively short-term experiment performed in young mice ( data not shown ) . To test if bone marrow stromal cells from Lepr-Cre; Itga11fl/fl mice retained the capacity to undergo osteogenesis upon Wnt pathway activation , we cultured these cells from 2-month-old Lepr-Cre; Itga11fl/fl and littermate control mice and treated half of the cultures with the Wnt pathway agonist , AZD2858 . In control cultures , bone marrow stromal cells from Lepr-Cre; Itga11fl/fl mice underwent significantly less osteogenesis as compared to stromal cells from control mice ( Figure 7O ) . Addition of AZD2858 significantly increased osteogenic differentiation from both Lepr-Cre; Itga11fl/fl and control stromal cells . In cultures treated with DMSO control , Lepr-Cre; Itga11fl/fl stromal cells had lower levels of phosphorylated GSK3 and β-catenin as compared to control stromal cells ( Figure 7P ) . AZD2858 increased the levels of phosphorylated GSK3 and β-catenin in both Lepr-Cre; Itga11fl/fl and control stromal cells ( Figure 7P ) . Lepr-Cre; Itga11fl/fl stromal cells thus retain the ability to undergo osteogenesis in response to Wnt pathway activation , even though they do not respond to Osteolectin .
Our data demonstrate that integrin α11 is a physiologically important receptor for Osteolectin , mediating its effect on osteogenesis . Integrin α11 is expressed by LepR+ skeletal stem cells and osteoblasts but shows little expression in non-osteogenic cells ( Figure 1G ) . Osteolectin bound selectively to α11β1 integrin , with nanomolar affinity ( Figure 1I and J ) , and promoted Wnt pathway activation in bone marrow stromal cells ( Figure 2 ) . Integrin α11 was required in bone marrow stromal cells for Wnt pathway activation and osteogenesis in response to Osteolectin ( Figure 4A–D ) . Blocking Wnt pathway activation in bone marrow stromal cells blocked the osteogenic effect of Osteolectin ( Figure 3C–D ) . Conditional deletion of Itga11 from LepR+ cells phenocopied the effect of Osteolectin deficiency ( Yue et al . , 2016 ) : in both cases the mice were grossly normal but exhibited accelerated bone loss during adulthood , particularly in trabecular bone ( Figure 5 ) . Like Osteolectin deficiency ( Yue et al . , 2016 ) , Itga11 deficiency significantly reduced the rate of bone formation in adult mice ( Figure 5M and O ) without affecting the rate of bone resorption ( Figure 5N ) . Bone marrow stromal cells from Lepr-Cre; Itga11fl/fl mice differentiated normally to adipocytes and chondrocytes ( Figure 7A–E ) but exhibited reduced osteogenic differentiation and did not respond to Osteolectin in vitro ( Figure 7C , F and G and Figure 7—figure supplement 1A ) or in vivo ( Figure 7H–N and Figure 7—figure supplement 1B ) . Nonetheless , bone marrow stromal cells from Lepr-Cre; Itga11fl/fl mice retained the ability to form bone in response to a chemical inhibitor of GSK3 , which activates the Wnt pathway ( Figure 7O and P ) . We conclude that integrin α11 is required by skeletal stem/progenitor cells to undergo osteogenesis in response to Osteolectin . Multiple factors promote osteogenesis by activating the Wnt pathway , including Wnts ( Boyden et al . , 2002; Cui et al . , 2011; Gong et al . , 2001 ) , BMPs ( Chen et al . , 2007; Rawadi et al . , 2003 ) , hedgehog proteins ( Mak et al . , 2006 ) , and parathyroid hormone ( Bonnet et al . , 2012; Kulkarni et al . , 2005; Wan et al . , 2008 ) . Our data suggest that Osteolectin contributes to Wnt pathway activation in osteogenic stem/progenitor cells along with other factors . While deficiency for integrin α11 phenocopied the effects of Osteolectin deficiency , we do not rule out a potential role for α10 integrin in mediating certain effects of Osteolectin . α10β1 also bound Osteolectin with nanomolar affinity ( Figure 1I ) . Integrin α11 is more highly expressed than α10 by LepR+ cells ( Figure 1D ) ; however , integrin α10 is expressed by chondrocytes ( Bengtsson et al . , 2005; Reinisch et al . , 2015 ) . While we did not observe any cartilage defects in Osteolectin deficient mice ( Yue et al . , 2016 ) , Osteolectin may promote the differentiation of hypertrophic chondrocytes into bone in adult mice , such as during fracture healing . Therefore , α10 integrin may mediate the effects of Osteolectin on hypertrophic chondrocytes while integrin α11 may mediate the effects of Osteolectin on skeletal stem/progenitor cells . It also remains possible that osteolectin has other receptors . Osteolectin may not be the only osteogenic ligand for integrin α11 . Collagen is a known ligand for α11β1 integrin ( Popova et al . , 2007 ) . We found that collagen binds to α11β1 with nanomolar affinity ( Figure 1J ) and actives integrin signaling ( Figure 4G and 4H , but we did not detect any effect of exogenous collagen on β-catenin accumulation ( Figure 4F ) or osteogenic differentiation ( Figure 1K ) . This suggests that collagen may bind α11β1 in a way that regulates cell adhesion and migration but not osteogenic differentiation , at least in skeletal stem/progenitor cells . Alternatively , endogenous collagen may bind α11β1 differently than exogenous collagen , potentially promoting osteogenesis . Since Lepr-Cre; Itga11fl/fl mice delete Itga11 in postnatal bone marrow cells that exhibit little contribution to the skeleton prior to two months of age ( Zhou et al . , 2014 ) , it remains untested whether integrin α11 regulates osteogenesis during fetal or early postnatal development . If so , this would raise the possibility of a distinct osteogenic ligand for α11 during development as Osteolectin deficient mice do not appear to exhibit defects in skeletal development ( Yue et al . , 2016 ) . While integrin α11 is not widely expressed by non-osteogenic cells , integrin α11 may have non-osteogenic functions in certain other cell types , or during development , in cells that are not competent to undergo osteogenesis . Integrin α11 is expressed by periodontal ligament fibroblasts and is required for the migration of these cells during ligament development , leading to a failure of tooth eruption in germline Itga11 deficient mice ( Popova et al . , 2007 ) . This raises the possibility that collagen binding to α11β1 may have biologically distinct consequences in cells that are not competent to form bone . Given that integrins can function as mechanosensors ( Schwartz , 2010 ) , our data raise the possibility that integrin α11 mediates the osteogenic response to mechanical loading in bones . Interestingly , it was recently discovered that skeletal stem cells in the developing jaw undergo osteogenesis in response to mechanical forces by activating FAK , suggesting the involvement of integrins in this process ( Ransom et al . , 2018 ) . In conclusion , we identify integrin α11 as an Osteolectin receptor and a new regulator of osteogenesis and adult skeleton maintenance . The identification of a new ligand/receptor pair that regulates the maintenance of the adult skeleton offers the opportunity to better understand the physiological and pathological mechanisms that influence skeletal homeostasis .
Lepr-cre mice were described previously ( DeFalco et al . , 2001 ) and obtained from the Jackson Laboratory ( Stock No: 008320 ) . Lepr-cre mice were backcrossed at least eight times onto a C57BL/Ka background . To generate Itga11fl/fl mice , CleanCap Cas9 mRNA ( TriLink ) and sgRNAs ( transcribed using MEGAshortscript Kit ( Ambion ) , purified using the MEGAclear Kit ( Ambion ) ) , and recombineering plasmids were microinjected into C57BL/Ka zygotes . Chimeric mice were genotyped by restriction fragment length poly-morphism ( RFLP ) analysis and confirmed by Southern blotting and sequencing of the targeted allele . Founders were mated with C57BL/Ka mice to obtain germline transmission then backcrossed with wild-type C57BL/Ka mice for at least three generations prior to analysis . This study was performed in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All procedures were approved by the UTSW Institutional Animal Care and Use Committee ( protocol number 2016–101334 G ) . Cell lines used in this study included mouse preosteoblast MC3T3-E1 cells ( Subclone 4 , ATCC CRL-2593 ) , human bone marrow stromal cells from ATCC ( PCS-500–012; referred to as hBMSC#1 ) , and human bone marrow stromal cells from Lonza ( PT-2501 , hBMSC#2 ) . The identity of MC3T3-E1 cells has been authenticated by ATCC , based on their expression of osteoblast marker genes including Bsp , Ocn , Pth and Pthrp . The identity of hBMSC#1 cells has been authenticated by ATCC using cell surface markes for these cells , including CD105 , CD73 , and CD44 . The identity of hBMSC#2 cells has been authenticated by Lonza using cell surface markes for these cells , including CD105 , CD166 , CD73 , and CD44 . We found no contamination of these cells from yeast , fungi , gram-positive or gram-negative bateria using the Cell Culture Contamination Detection Kit ( Molecular Probes ) . MC3T3-E1 cells were cultured in Alpha Minimum Essential Medium with ribonucleosides , deoxyribonucleosides , 2 mM L-glutamine and 1 mM sodium pyruvate , but without ascorbic acid ( GIBCO , A1049001 ) , and supplemented with 10% fetal bovine serum ( Sigma , F2442 ) and penicillin-streptomycin ( HyClone ) . MC3T3-E1 cells were used for experiments before passage 20 . hBMSC cells were cultured in low glucose DMEM ( Gibco ) supplemented with 20% fetal bovine serum ( Sigma , F2442 ) and penicillin-streptomycin ( HyClone ) , and were used for experiments before passage 16 . As described previously ( Yue et al . , 2016 ) , mouse and human Osteolectin cDNA were cloned into pcDNA3 vector ( Invitrogen ) containing a C-terminal 1XFlag-tag and transfected into HEK293 cells with Akashi et al . ( 2000 ) ( Invitrogen ) . Stably expressing cell lines were selected using 1 mg/ml G418 ( Sigma ) then cultured in DMEM plus 10% FBS ( Sigma ) , and 1% penicillin/streptomycin ( Invitrogen ) . Culture medium was collected every two days , centrifuged to eliminate cellular debris , and stored with 1 mM phenylmethylsulfonyl fluoride ( Sigma ) at 4°C to inhibit protease activity . One liter of culture medium was filtered through a 0 . 2 µm membrane ( Nalgene ) to eliminate cellular debris before being loaded onto a chromatography column containing 2 ml Anti-FLAG M2 Affinity Gel ( Sigma ) , with a flow rate of 1 ml/min . The column was sequentially washed using 20 ml of high salt buffer ( 20 mM Tris-HCl , 300 mM KCl , 10% Glycerol , 0 . 2 mM EDTA ) followed by 20 ml of low salt buffer ( 20 mM Tris-HCl , 150 mM KCl , 10% Glycerol , 0 . 2 mM EDTA ) and finally 20 ml of PBS . The FLAG-tagged Osteolectin was then eluted from the column using 10 ml 3X FLAG peptide ( 100 mg/ml ) in PBS . Eluted protein was concentrated using Amicon Ultra-15 Centrifugal Filter Units ( Ultracel-10K , Millipore ) , then quantitated by SDS-PAGE and colloidal blue staining ( Invitrogen ) and stored at −80°C . Recobinant human Pro-Collagen 1α was purchased from R and D Systems , and we removed the His tag using TEV protease ( Sigma ) . After cleavage , we purified the untagged Pro-Collagen 1α using Ni-NTA agarose columns ( Thermo Fisher Scientific ) to separate it from the cleaved His tag , the His-tagged Pro-Collagen 1α , and the His-tagged TEV protease . To add recombinant proteins in culture , recombinant human or mouse Osteolectin or Pro-Collagen 1α was added to osteogenic differentiation medium ( described below ) . Unless otherwise specified , we used 30 ng/ml recombinant Osteolectin for in vitro assays . For in vivo use , recombinant mouse Osteolectin ( 50 μg/kg of body mass ) was subcutaneously injected daily into 2-month-old female Lepr-Cre; Itga11fl/fl or littermate Itga11fl/fl control mice for 28 days . Mice receiving control injections received an equal volume of PBS . Cells were cultured until confluent , then transferred into osteogenic differentiation medium with or without Osteolectin or small molecule inhibitors of Wnt pathway components . Prior to extracting proteins , cells were washed with PBS and then lysis buffer was added containing 50 mM Tris-HCl , 150 mM NaCl , 1% NP-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 1 mM sodium vanadate , 0 . 5 mM sodium fluoride , and cOmplete Mini EDTA-free Protease Inhibitor Cocktail ( Sigma ) . The cells were scraped off the plate in the lysis buffer , transferred to an Eppendorf tube on ice , incubated for 20 min with occasional vortexing , then centrifuged at 17 , 000xg for 10 min at 4°C to clear cellular debris . The cell lysates were Western blotted with the indicated antibodies and immunoreactive bands were detected using ECL reagent ( Pierce ) . For some experiments , a Nuclear/Cytosol Fractionation Kit ( Biovision ) was used to separate the nuclear and cytosolic/membrane fractions of cell lysates . Antibodies used in this study include anti-Phospho-PI3 Kinase p85 ( Tyr458 ) /p55 ( Tyr199 ) , anti-Phospho-Akt ( Ser473 ) , anti-Phospho-GSK-3α/β ( Ser21/9 ) , anti-Phospho-FAK ( Y397 ) , anti-β-Catenin , anti-β-Actin , anti-GSK-3β ( 27C10 ) , anti-Histone H3 , and anti-rabbit IgG , HRP-linked antibody from Cell Signaling , anti-mouse Osteolectin ( AF3729 ) and anti-human Osteolectin ( AF1904 ) antibodies from R and D Systems , anti-Adiponectin ( ab181699 ) and anti-Integrin α11 antibody ( ab198826 ) from Abcam . For co-immunoprecipitation experiments , human Osteolectin cDNA was cloned into pcDNA3 vector ( Invitrogen ) containing a C-terminal 1XFlag-tag , then transfected into MC3T3-E1 cells with Morrison et al . , 2000 ( Invitrogen ) . After 48 hr , cells were solubilized in lysis buffer and cellular debris was cleared by centrifugation as described above , then lysates were immunoprecipitated with anti-FLAG M2 Affinity Gel ( Sigma ) . After incubation of lysates with M2 Affinity Gel for 2 hr at 4°C , the gel was centrifuged and washed six times with lysis buffer . Immunoprecipitates were analyzed by western blotting . For quantitative reverse transcription PCR ( qRT-PCR ) , cells were lysed using TRIzol LS ( Invitrogen ) . RNA was extracted and reverse transcribed into cDNA using SuperScript III ( Invitrogen ) . qRT-PCR was performed using a Roche LightCycler 480 . The primers used for qRT-PCR analysis of mouse RNA include: Osteolectin: 5’-AGG TCC TGG GAG GGA GTG-3’ and 5’-GGG CCT CCT GGA GAT TCT T-3’; Actb: 5’-GCT CTT TTC CAG CCT TCC TT-3’ and 5’-CTT CTG CAT CCT GTC AGC AA-3’; Lef1: 5’-TGT TTA TCC CAT CAC GGG TGG-3’ and 5’-CAT GGA AGT GTC GCC TGA CAG-3’; Runx2: 5’-TTA CCT ACA CCC CGC CAG TC-3’ and 5’-TGC TGG TCT GGA AGG GTC C-3’; Axin2: 5’-GAG TAG CGC CGT GTT AGT GAC T-3’ and 5’-CCA GGA AAG TCC GGA AGA GGT ATG-3’; Alp: 5’-CCA ACT CTT TTG TGC CAG AGA-3’ and 5’-GGC TAC ATT GGT GTT GAG CTT TT-3’ , Rankl: 5’-CAG CAT CGC TCT GTT CCT GTA-3’ and 5’-CTG CGT TTT CAT GGA GTC TCA-3’ , Itga11: 5’-TGC CCC AAT GGA AAC CAA TG-3’ and 5’-CAC TCG TGC GAC CAG AGA G-3’ , Dmp1: 5’-TGG GAG CCA GAG AGG GTA G-3’ and 5’-TTG TGG TAT CTG GCA ACT GG-3’ , Ctnnb1: 5’-CAT CTA CAC AGT TTG ATG CTG CT-3’ and 5’-GCA GTT TTG TCA GTT CAG GGA-3’ . The primers used for qRT-PCR analysis of human RNA include: Osteolectin: 5’-ACA TCG TCA CTT ACA TCC TGG GC-3’ and 5’-CAC GCG GGT GTC CAA CG-3’; Actb: 5’-ATT GGC AAT GAG CGG TTC-3’ and 5’-CGT GGA TGC CAC AGG ACT-3’; Lef1: 5’-TGC CAA ATA TGA ATA ACG ACC CA-3’ and 5’-GAG AAA AGT GCT CGT CAC TGT-3’; Runx2: 5’-GAA CCC AGA AGG CAC AGA CA-3’ and 5’-GGC TCA GGT AGG AGG GGT AA-3’; Axin2: 5’- CAA CAC CAG GCG GAA CGA A-3’ and 5’- GCC CAA TAA GGA GTG TAA GGA CT-3’; Alp: 5’-GTG AAC CGC AAC TGG TAC TC-3’ and 5’-GAG CTG CGT AGC GAT GTC C-3’ , Dmp1: 5’-CTC CGA GTT GGA CGA TGA GG-3’ and 5’-TCA TGC CTG CAC TGT TCA TTC-3’ . To genotype Itga11 floxed mice the following primers were used: 5’- AATTCAGTGCCGATCCTCCAGTGTC-3’ , 5’-CCCTTGCTTCCTTCTGCTGTCACTT-3’ ( Itga11fl allele: 370 bp; Itga11+ allele: 280 bp ) . Integrin binding assays were performed as described ( Nishiuchi et al . , 2006 ) . Microtiter plates were coated with 10 nM recombinant human Osteolectin , recombinant human Pro-Collagen 1α , or Bovine Serum Albumin ( BSA , Sigma A3156 ) overnight at 4°C , and then blocked with 10 mg/ml BSA . 6xHis tagged recombinant human integrin heterodimers were purchased from R and D Systems . The plates were incubated with integrins in TBS buffer ( 50 mM Tris-Cl , pH 7 . 5 150 mM NaCl ) with 1 mM MnCl2 , then washed with TBS containing 1 mM MnCl2 , 0 . 1% BSA , and 0 . 02% Tween 20 , followed by quantification of bound integrins by an enzyme-linked immunosorbent assay using an anti-His tag monoclonal antibody ( Thermo Fisher Scientific , clone 4E3D10H2/E3 ) followed by a horseradish peroxidase-conjugated anti-mouse secondary antibody . After washing , bound HRP was detected using SureBlue TMB Microwell Peroxidase Substrate ( KPL ) and the reaction was stopped with TMB stop solution ( KPL ) . The optical density was measured at 450 nm . MicroCT analysis was performed using the same settings as previously described ( Yue et al . , 2016 ) . Based on previously described methods ( Bouxsein et al . , 2010 ) , mouse femurs were dissected , fixed overnight in 4% paraformaldehyde ( Thermo Fisher Scientific ) and stored in 70% ethanol at 4°C . Femurs and lumbar vertebrae were scanned at an isotropic voxel size of 3 . 5 μm and 7 μm , respectively , with peak tube voltage of 55 kV and current of 0 . 145 mA ( μCT 35; Scanco ) . A three-dimensional Gaussian filter ( s = 0 . 8 ) with a limited , finite filter support of one was used to suppress noise in the images , and a threshold of 263–1000 was used to segment mineralized bone from air and soft tissues . Trabecular bone parameters were measured in the distal metaphysis of the femurs . The region of interest was selected from below the distal growth plate where the epiphyseal cap structure completely disappeared and continued for 100 slices toward the proximal end of the femur . Contours were drawn manually a few voxels away from the endocortical surface to define trabecular bones in the metaphysis . Cortical bone parameters were measured by analyzing 100 slices in mid-diaphysis femurs . Numbers of experiments noted in figure legends reflect independent experiments performed on different days . Mice were allocated to experiments randomly and samples processed in an arbitrary order , but formal randomization techniques were not used . Prior to analyzing the statistical significance of differences among treatments we tested whether data were normally distributed and whether variance was similar among treatments . To test for normality , we performed the Shapiro–Wilk tests . To test whether variability significantly differed among treatments we performed F-tests ( for experiments with two treatments ) or Levene’s median tests ( for experiments with more than two treatments ) . When the data significantly deviated from normality ( p < 0 . 01 ) or variability significantly differed among treatments ( p < 0 . 05 ) , we log2-transformed the data and tested again for normality and variability . If the transformed data no longer significantly deviated from normality and equal variability , we then performed parametric tests on the transformed data . If the transformed data still significantly deviated from normality or equal variability , we performed non-parametric tests on the non-transformed data . Data from the same cell culture experiments were always paired for statistical analysis . Mouse littermates were paired for statistical analysis . To assess the statistical significance of a difference between two treatments , we used paired two-tailed Student’s t-tests ( when a parametric test was appropriate ) or Wilcoxon’s tests ( when a non-parametric test was appropriate ) . To assess the statistical significance of differences between more than two treatments , we used one-way or two-way repeated measures ANOVAs ( when a parametric test was appropriate ) followed by post-hoc tests including Dunnett’s , Sidak’s , and Tukey’s tests depending on the experimental settings and planned comparisons , or multiple Wilcoxon’s tests followed by Holm-Sidak’s method for multiple comparisons adjustment ( when a non-parametric test was appropriate ) . Relative mRNA levels were always log2-transformed before any statistical tests were performed . All statistical analyses were performed with Graphpad Prism 7 . 02 . All data represent mean ±standard deviation ( *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 ) . As previously described ( Yue et al . , 2016 ) , mouse femurs and tibias were cut at both ends to flush out intact marrow plugs . Both the flushed plugs and crushed bone metaphyses were subjected to two rounds of enzymatic digestion in prewarmed digestion buffer containing 3 mg/ml type I collagenase ( Worthington ) , 4 mg/ml dispase ( Roche Diagnostic ) and 1 U/ml DNase I ( Sigma ) in HBSS with calcium and magnesium , at 37°C for 15 min each round . During each round of digestion , the suspension was vortexed six times for 10 s each time at speed level three using a Vortex-Genie two to promote more complete dissociation . Dissociated cells were transferred into a tube with staining medium ( HBSS without calcium and magnesium +2% fetal bovine serum ) and 2 mM EDTA to stop the digestion . Cells were then centrifuged , resuspended in staining medium , and passed through a 90 μm nylon mesh to filter undigested plugs or bone . To form CFU-F colonies , freshly dissociated bone marrow cell suspensions were plated at clonal density in 6-well plates ( 5 × 105 cells/well ) or 10 cm plates ( 5 × 106 cells/dish ) with DMEM ( Gibco ) plus 20% fetal bovine serum ( Sigma F2442 ) , 10 mM ROCK inhibitor ( Y-27632 , Selleck ) , and 1% penicillin/streptomycin ( Invitrogen ) at 37°C in gas-tight chambers ( Billups-Rothenberg ) with 1% O2 and 6% CO2 ( with balance Nitrogen ) to maintain a low oxygen environment that promoted survival and proliferation ( Morrison et al . , 2000 ) . The CFU-F culture dish was rinsed with HBBS without calcium and magnesium and replenished with freshly made medium on the second day after plating to wash out contaminating macrophages . Cultures were then maintained in a gas-tight chamber that was flushed daily for 1 min with a custom low oxygen gas mixture ( 1% O2 , 6% CO2 , balance Nitrogen ) . The culture medium was changed every 4 days . To count CFU-F colonies , the cultures were stained with 0 . 1% Toluidine blue in 4% formalin solution eight days after plating . The osteogenic potential of primary CFU-F cells , human bone marrow stromal cells , and MC3T3-E1 cells was assessed by plating the cells into 48- well plates ( 25 , 000 cells/cm2 ) . On the second day after plating , the culture medium was replaced with osteogenic differentiation medium ( StemPro Osteogenesis Differentiation kit , Gibco ) . Cells were maintained in the differentiation medium , with medium change every other day for 14 days for primary CFU-F cells and MC3T3-E1 cells before differentiation was assessed . For human bone marrow stromal cells , the culture medium was changed every 3 days for 21 days . Osteoblastic differentiation was detected by staining with Alizarin red S ( Sigma ) . To quantitate Alizarin red staining , the stained cells were rinsed with PBS , and extracted with 10% ( w/v ) cetylpyridinium chloride in 10 mM sodium phosphate , pH 7 . 0 for 10 min at room temperature . Alizarin red in the extract was quantitated by optical density measurement at 562 nm . The adipogenic potential of CFU-F cells was assessed by plating them into 48-well plates ( 25 , 000 cells/cm2 ) . On the second day after plating , the culture medium was replaced with adipogenic differentiation medium ( StemPro Adipogenesis Differentiation Kit , Gibco ) and the cultures were allowed to differentiate for 14 days , with culture medium changed every 3 days . Adipocyte differentiation was detected by staining with Oil red O ( Sigma ) . To quantitate the amount of Oil red O staining , cells were rinsed with PBS , and extracted with 100% isopropanol for 10 min at room temperature . Oil red O in the extract was quantitated by optical density measurement at 500 nm . The chondrogenic potential of CFU-F cells was assessed by centrifuging 2 × 105 cells to form cell pellets , which were then cultured in chondrogenic medium ( StemPro chondrogenesis differentiation kit; Gibco ) for 21 days . The culture medium was changed every 3 days . Chondrocyte formation within the cell pellets was assessed by cryosectioning and Toluidine blue staining as described ( Robey et al . , 2014 ) . As previously described ( Egan et al . , 2012 ) , mice were injected intraperitoneally with 10 mg/kg body weight of calcein , dissolved in 0 . 15 M NaCl plus 2% NaHCO3 in water , at day 0 and day 7 . Mice were sacrificed on day 9 . Mouse tibias were fixed overnight in 4% paraformaldehyde at 4°C , dehydrated in 30% sucrose at 4°C for two days and sectioned without decalcification ( 7 µm sections ) . Mineral apposition rates were determined as previously described ( Egan et al . , 2012 ) . The surface used to quantify the trabecular bone mineral apposition rate was 100 μm distal to the growth plate and 50 μm in from the endosteal cortical bone of the femur . The surface used to quantify cortical bone mineral apposition rate was the medial endosteal cortical bone surface of the femur . The bone resorption rate was determined by measuring urinary levels of deoxypyridinoline ( DPD ) using a MicroVue DPD ELISA Kit ( Quidel ) . The DPD values were normalized to urinary creatinine levels using the MicroVue Creatinine Assay Kit ( Quidel ) . The bone formation rate was determined by measuring serum levels of Procollagen type 1 N-terminal Propeptide ( P1NP ) using the Rat/Mouse P1NP ELISA kit ( Immunodiagnostic Systems ) . The ELISA assay for Osteolectin was described previously ( Yue et al . , 2016 ) . | Throughout our lives , our bones undergo constant remodeling . Cells called osteoclasts break down old bone and cells called osteoblasts lay down new . Normally , the two cell types work in balance but if the rate of breakdown outpaces new bone formation the skeleton can become weak . This weakness leads to a condition called osteoporosis , in which people suffer from fragile bones . Osteoporosis is hard to reverse , in part because our ability to encourage new bone to form is limited . In 2016 , researchers discovered a protein called osteolectin , which promotes new bone formation during adulthood by helping skeletal stem cells transform into bone cells . But so far , it has been unclear how osteolectin achieves this . To investigate this further , Shen et al . – including some researchers involved in the 2016 study – marked osteolectin with a molecular tag and tested what it bound on the surface of mouse and human bone marrow cells . The experiments revealed that osteolectin binds to a specific receptor protein called α11 integrin , which can only be found on skeletal stem cells and the osteoblasts they give rise to . Once osteolectin binds to the receptor , it activates a signaling pathway that induces the stem cells to develop into osteoblasts . Mice that lacked either osteolectin or α11 integrin produced less bone and lost bone tissue faster as adults . Osteolectin could potentially be useful in the treatment of osteoporosis or broken bones . Since only skeletal stem cells and osteoblasts cells produce α11 integrin , osteolectin would specifically target these cells without affecting cells that do not form bones . A next step will be to assess how well osteolectin compares to existing treatments for fragile bones . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2019 | Integrin alpha11 is an Osteolectin receptor and is required for the maintenance of adult skeletal bone mass |
The coordination of cell proliferation and migration in growing tissues is crucial in development and regeneration but remains poorly understood . Here , we find that , while expanding with an edge speed independent of initial conditions , millimeter-scale epithelial monolayers exhibit internal patterns of proliferation and migration that depend not on the current but on the initial tissue size , indicating memory effects . Specifically , the core of large tissues becomes very dense , almost quiescent , and ceases cell-cycle progression . In contrast , initially-smaller tissues develop a local minimum of cell density and a tissue-spanning vortex . To explain vortex formation , we propose an active polar fluid model with a feedback between cell polarization and tissue flow . Taken together , our findings suggest that expanding epithelia decouple their internal and edge regions , which enables robust expansion dynamics despite the presence of size- and history-dependent patterns in the tissue interior .
Writing in 1859 , physiologist Rudolf Virchow presented the concept of the ‘Zellenstaat’ or ‘Cell State , ’ describing tissues as ‘a society of cells , a tiny well-ordered state’ ( Virchow , 1855 ) . This social framework motivated Abercrombie and Heaysman , 1954 work on cellular behavior that elucidated how encounters between cells can regulate locomotion and proliferation via contact inhibition . Since then , concerted interdisciplinary effort has been brought to bear on understanding how cell-cell interactions give rise to the complex collective behaviors driving so many crucial biological processes . One of the most foundational collective behaviors is collective cell migration—the directed , coordinated motion of cellular ensembles that enables phenomena such as gastrulation , wound healing , and tumor invasion ( Friedl and Gilmour , 2009 ) . Given this importance , considerable effort spanning biology , engineering , and physics has been directed towards understanding how local cellular interactions can give rise to globally coordinated motions ( Alert and Trepat , 2020; Hakim and Silberzan , 2017 ) . Studies of collective cell migration are most often performed using epithelial tissues due to their fundamental role in multicellular organisms and strong cell-cell adhesion , which in turn gives rise to elegant , cohesive motion . Moreover , given that epithelia naturally form surfaces in vivo , studying epithelial layers in vitro has a physiological basis that can inform our understanding of processes such as healing ( Poujade et al . , 2007 ) , envelopment ( Steinberg , 2007 ) , and boundary formation ( Dahmann et al . , 2011 ) . These features have made epithelia both the gold standard in collective cell migration studies , and one of the most well-studied models for biological collective behaviors . Due to the complexity of collective behaviors , much effort has gone towards reductionist assays that restrict degrees of freedom and ensemble size to simplify analysis and interpretation . One such approach is to confine a tissue within predefined boundaries using micropatterning to create adhesive and non-adhesive regions ( Doxzen et al . , 2013; Deforet et al . , 2014; Notbohm et al . , 2016; Pérez-González et al . , 2019; Peyret et al . , 2019; Petrolli et al . , 2019 ) . Such confinement mimics certain in vivo contexts such as constrained tumors as well as aspects of compartmentalization during morphogenesis ( Lecuit and Lenne , 2007 ) . Alternately , many studies have explored the expansion of tissues that initially grow into confluence within confinement but are later allowed to migrate into free space upon removal of a barrier . A popular assay of this type relies on rectangular strips of tissue that are allowed to expand in one or both directions ( Poujade et al . , 2007; Trepat et al . , 2009; Petitjean et al . , 2010; Reffay et al . , 2011; Nnetu et al . , 2012; Serra-Picamal et al . , 2012; Zhang et al . , 2017; Uroz et al . , 2018; Tlili et al . , 2018 ) , where averaging along the length of the strip can reveal coordinated population-level behaviors such as complex migration patterns , non-uniform traction force fields , and traveling mechanical waves . Other studies have focused on the isotropic expansion of micro-scale ( < 500 μm diameter ) circular tissues using the barrier stencil technique ( Jang et al . , 2017 ) as well as photoswitchable substrates ( Rolli et al . , 2012 ) . Still more work has explored approaches to induce directional migration , from geometric cues to applied electric fields ( Vedula et al . , 2012; Cohen et al . , 2014 ) . In contrast to micro-scale confinement assays , other work has focused on large , freely-expanding tissues of uncontrolled initial size and shape , which grow from either single cells ( Puliafito et al . , 2012; Huergo et al . , 2011 ) or cell-containing droplets ( Lee et al . , 2013; Beaune et al . , 2014 ) . Related experiments track long-term growth of cell colonies via images taken once per day over several days , but this low temporal resolution cannot access timescales over which migration is important ( Huergo et al . , 2011; Simpson et al . , 2013 ) . Thus , there is still a lack of assays to study long-term expansion and growth of large-scale tissues with precisely-controlled initial conditions , especially initial tissue size , shape , and density . To address this gap , we leveraged bench-top tissue patterning ( Poujade et al . , 2007; Cohen et al . , 2016 ) to precisely pattern macro-scale circular epithelia of two sizes ( >1 mm in diameter ) and performed long-term , high frequency , time-lapse imaging after release of a barrier . To elucidate the consequences of size effects on the tissue , we tracked every cell , relating the overall expansion kinetics to cell migration speed , cell density , and cell-cycle dynamics . We find that , whereas the tissue edge dynamics is independent of the initial conditions , the tissue bulk exhibits size-dependent patterns of cell proliferation and migration , including large-scale vortices accompanied by dynamic density profiles . Together , these data comprise the first comprehensive study of macro-scale , long-term epithelial expansion , and our findings demonstrate the importance of exploring collective cell migration across a wider range of contexts , scales , and constraints .
We began by characterizing the overall expansion and growth of tissues with the same cell density but different initial diameters of 1 . 7 mm and 3 . 4 mm ( a 4X difference in area , with tissues hereafter referred to as either ‘small’ or ‘large’ ) , using an MDCK cell line stably expressing the 2-color FUCCI cell-cycle marker ( Sakaue-Sawano et al . , 2008; Streichan et al . , 2014; Uroz et al . , 2018; Beaune et al . , 2014; Benham-Pyle et al . , 2016 ) . We patterned the tissues by culturing cells in small and large circular silicone stencils for ∼18 hr ( Cohen et al . , 2016; Poujade et al . , 2007 ) , whereupon stencils were removed and tissues were allowed to freely expand for 46 hr ( Figure 1A , Figure 1—video 1 ) , while images were collected at 20 min intervals using automated microscopy ( see Materials and Methods ) . Our cell seeding conditions and incubation period were deliberately tuned to ensure that the stencils did not induce contact inhibition of proliferation prior to stencil removal ( checking FUCCI to ensure the tissue was not arrested in G1 ) . Upon stencil removal , tissues expanded while maintaining their overall circular shape throughout the 2 day experiment . Unless otherwise noted , cell density at stencil removal was ∼2700 cells/mm2 , a value consistent with active and growing confluent MDCK epithelia ( Streichan et al . , 2014; Uroz et al . , 2018 ) . First , we measured relative areal increase ( Figure 1B ) and relative cell number increase ( Figure 1—figure supplement 1 ) of small and large tissues . By 46 hr , small and large tissues had increased in area by 6 . 4X and 3 . 3X , respectively , while cell number increased by 9 . 2X and 5 . 5X , respectively . Since proliferation outpaces area expansion in long-term growth , average tissue density increased by the end of the experiment . The evolution of average tissue density was more complex , however , as small tissues experienced a density decrease from 4 to 12 hr while large tissues exhibited a monotonic increase in cell density ( Figure 1C ) . Accordingly , at any given time after stencil removal , large tissues had a higher density than small tissues . Non-monotonic density evolution has been observed in thin epithelial strips ( Poujade et al . , 2007 ) and likely arises from competition between migration and proliferation dynamics , which we discuss later . We then related area expansion to the kinematics of the tissue edge . To quantify edge motion , we calculated the average radial velocity of the tissue boundary , vr ( t ) , at 1 hr intervals over 46 hr ( Materials and methods ) . We found that vr is independent of both tissue size and a wide range of initial cell densities , in all cases reaching ∼30 μm/h after ∼16 hr ( Figure 1D ) . Before reaching this constant edge velocity , vr ramps up during the first 8 hr after stencil removal , and , notably , overshoots its long-time value by almost 30% . We hypothesize that the overshoot is due to the formation of fast multicellular finger-like protrusions that emerge at the tissue edge in the early stages of expansion and then diminish ( Figure 1—video 2 ) . This hypothesis is supported by a recent model showing that edge acceleration ( as observed during the first 8 hr in Figure 1D ) leads to finger formation ( Alert et al . , 2019 ) . It is remarkable that the edge radial velocity vr ( t ) is independent of the initial tissue size and density , especially considering that cell density evolution shows opposite trends at early stages of expansion for small and large tissues ( Figure 1C ) . This observation suggests that the early stages of epithelial expansion are primarily driven by cell migration rather than proliferation or density-dependent decompression and cell spreading . The observation that vr is independent of tissue size ought to explain why small tissues have faster relative area expansions than large tissues . We hypothesized that the relation between tissue size and areal increase could be attributed primarily to the perimeter-to-area ratio . Assuming a constant edge velocity vn normal to the tissue boundary , the tissue area increases as dA=Pvndt , where P is the perimeter of tissue and dt is a small time interval . Thus , the relative area increase dA/A= ( P/A ) vndt scales as the perimeter-to-area ratio , which is inversely proportional to the radius for circular tissues , so the relative area increases faster for smaller tissues ( Figure 1B ) . To verify that the perimeter-to-area ratio is proportional to the relative area increase , we analyzed elliptical tissues with the same area and cell density but different perimeters ( Figure 1—video 3 ) . Increasing the perimeter-to-area ratio of a tissue by increasing its aspect ratio indeed leads to faster relative area expansion ( Figure 1E ) . A simple , edge-driven expansion model with linear increase of the tissue major and minor axes predicts A ( t ) /A ( 0 ) = ( a+vnt ) ( b+vnt ) / ( ab ) , where a and b are the initial major and minor axes of the tissue . This model fits our data well assuming the same edge speed vn≃29 . 5 μm/h for all tissues ( Figure 1E ) . This observation suggests that edge speed is mostly independent of edge curvature . However , we measure a smaller edge speed at the major axes of ellipses , which are high-curvature points with radius of curvature rc≲0 . 75 mm ( Figure 1—figure supplement 2 ) . Such high curvatures are concentrated around the major axes of our elliptical tissues . However , most of the tissue edge has a smaller curvature , and therefore advances at a curvature-independent speed . Further , even high curvature regions blunt due to expansion over time ( see Figure 1—video 3 ) . As a result , our model with a single edge speed vn≃29 . 5 μm/h is sufficient to capture the area expansion of both circular and elliptical tissues ( Figure 1E ) . Together , our findings demonstrate that epithelial shape and size determine area expansion dynamics via the perimeter-to-area ratio . This relationship results from the fact that tissues exhibit a constant , size-independent , migration-driven edge speed normal to tissue boundary . Since initial tissue size does not affect boundary dynamics , but does impact the relative growth and expansion of the tissue , we hypothesize that cells in the tissue bulk exhibit tissue size-dependent behaviors . Having demonstrated the role of the boundary in the expansion of large-scale epithelia , we sought to relate tissue areal expansion rate to internal collective cell migration dynamics . We used Particle-Image-Velocimetry ( PIV , Materials and methods ) to obtain flow fields describing cell migration within freely expanding epithelia ( Poujade et al . , 2007; Petitjean et al . , 2010; Angelini et al . , 2010; Cohen et al . , 2014; Aoki et al . , 2017 ) . We constructed kymographs ( Materials and Methods ) to display the full spatiotemporal flow patterns of the tissue ( Figure 2A , B; Serra-Picamal et al . , 2012; Zhang et al . , 2017 ) , averaging over the angular direction and over 16 tissues ( for representative kymographs , see Figure 2—figure supplement 1 ) . We also separately show time evolution ( Figure 2C ) and spatial profiles ( Figure 2D ) of speed and radial velocity to compare small and large tissues . Kymographs of speed and radial velocity reveal the existence of an edge region of fast , outward , radial cell motion ( Figure 2A , B ) , with speeds similar to the radial edge velocity reported in Figure 1D . Up to ∼500 μm from the tissue edge , the speed and radial velocity profiles are practically identical for small and large tissues ( Figure 2D ) , showing that cell motion near the tissue edge is independent of tissue size . The tissue centers , in contrast , exhibit size-dependent behaviors . For both small and large tissues , a wave front of cell speed and radial velocity propagates toward the tissue centers at ∼90 μm/h ( Figure 2A and B , dashed lines ) . This is approximately 3X faster than the tissue edge speed , consistent with previously described waves of strain rate in cell monolayers ( Serra-Picamal et al . , 2012 ) . Soon after the wave of radial velocity reaches the center , it retreats , leaving a region of low radial velocity that increases in extent in the center of both small and large tissues ( Figure 2B ) . This decrease of radial velocity is accompanied by a reduction in cell speed in the center of large tissues but not in small tissues , in which cell speed remains high until 36 hr ( Figure 2A and C Bottom ) . We examine the behavior of this high-speed but low-radial-velocity central region of small tissues in the next section . The propagation of low radial velocity out from the center of small tissues coincides with the formation and expansion of a millimeter-scale , persistent vortex ( see Figure 3A , Figure 3—video 1 for representative vortex ) . These large vortices are observed in both small and large tissues ( Figure 3—video 2 ) , but they only reach tissue-spanning sizes in small tissues . To visualize the form and scale of these vortices , we tracked individual cell motion and colored cell trajectories according to their orientation ( Püspöki et al . , 2016 ) for a representative small and large tissue vortex ( see Figure 3A and Materials and Methods ) . We plotted trajectories for the time periods that the vortex was most apparent , which was 20–40 hr in the small tissue ( Figure 3A , left ) and 10–30 hr in the large tissue ( Figure 3A , right ) . During the vortex period in small tissues , cell trajectories are primarily radial in the boundary zone , but mainly tangential in the entire central zone ( Figure 3A left , see Figure 3—figure supplement 1 for vortex trajectory quantification ) . To understand the emergence of the vortices , we build on a continuum physical model of tissue spreading that describes the cell monolayer as a two-dimensional compressible active polar fluid ( Blanch-Mercader et al . , 2017; Pérez-González et al . , 2019; Alert et al . , 2019 ) . Consistent with our velocity measurements ( Figure 2C ) , we assume that cells at the edge zone are radially polarized and motile , whereas cells in the bulk of the tissue are unpolarized and non-motile . We describe cell polarization at a coarse-grained level via a polarity field 𝐩 that obeys the following dynamics ( Alert and Trepat , 2020 ) : ( 1 ) ∂t𝐩=𝐡γ+νs𝐯 . Here , γ is the rotational viscosity that damps polarity changes . Respectively , 𝐡=-a𝐩+K∇2𝐩 is the so-called molecular field that governs polarity relaxation: the first term drives the polarity to zero , and the second term opposes spatial variation of the polarity field . As a result of these terms , the radial polarity at the tissue edge decays over a length scale Lc=K/a into the tissue bulk . With respect to previous models of tissue spreading , we add the last term in Equation 1 , which couples the polarity to the tissue velocity field 𝐯 . This coupling is a generic property of active polar fluids interacting with a substrate ( Brotto et al . , 2013; Kumar et al . , 2014; Oriola et al . , 2017; Maitra et al . , 2020 ) . Previous works in agent-based models showed that similar polarity-velocity alignment interactions ( Alert and Trepat , 2020 ) can lead to waves ( Petrolli et al . , 2019 ) , flocking transitions ( Szabó et al . , 2006; Henkes et al . , 2011; Basan et al . , 2013; Malinverno et al . , 2017; Giavazzi et al . , 2018 ) , and vortical flows ( Rappel et al . , 1999; Camley et al . , 2014; Li and Sun , 2014; Segerer et al . , 2015; Barton et al . , 2017; Lin et al . , 2018 ) in small , confined , and polarized tissues . Here , using a continuum model , we propose that cell polarity not only aligns with but is also generated by tissue flow , and we ask whether this polarity-velocity coupling can lead to large-scale spontaneous flows in the unpolarized bulk of unconfined tissues . To determine the flow field 𝐯 , we impose a balance between internal viscous stresses in the tissue , with viscosity η , and external cell-substrate forces , including viscous friction with coefficient ξ , active traction forces with coefficient Ta , and the cell-substrate forces associated with the polarity-velocity coupling νs: ( 2 ) η∇2𝐯=ξ𝐯-Ta𝐩-νs𝐡 . This force balance predicts that even if cell polarity , and hence active traction forces , are localized to a narrow boundary layer of width Lc∼50 μm ( Blanch-Mercader et al . , 2017; Pérez-González et al . , 2019 ) , cell flow can penetrate a length ∼λ=η/ξ into the tissue . Based on our measurements ( Figure 2D ) , we estimate λ∼0 . 5-1 mm , which is larger than the velocity correlation length of ∼200 μm in the tissue bulk ( Petitjean et al . , 2010 ) . A linear stability analysis of Equations 1 and 2 shows that perturbations of wave number q around the quiescent ( 𝐯=0 ) and unpolarized ( 𝐩=0 ) state grow with a rate ( 3 ) Ω ( q ) =-aγ ( 1+Lc2q2 ) +Taνs-aνs2 ( 1+Lc2q2 ) ξ ( 1+λ2q2 ) . This result shows that , if Taνs>a ( ξ/γ+νs2 ) , the unpolarized state of an active polar fluid described by Equations 1 and 2 is unstable ( Ω>0 ) to perturbations of wavelength longer than a critical value 2π/qc given by Ω ( qc ) =0 ( Figure 3B ) . This analysis suggests that , for tissues larger than this critical value ∼2π/qc , the quiescent tissue bulk becomes unstable and starts to flow spontaneously at large scales , consistent with the emergence of large-scale vortices . The mechanism of this instability is the positive feedback between flow-induced cell polarization and the flows due to migration of polarized cells . The fact that a critical size of the order of millimeters is required for this long-wavelength instability might explain why large-scale vortices have not been observed in previous studies , which considered smaller tissues . To quantify the kinematics of the large-scale vortical flows , we obtained the vorticity field ω ( 𝐫 , 𝐭 ) =∇×𝐯 ( 𝐫 , 𝐭 ) . Before averaging over tissues , we took the dominant direction of rotation of each tissue to correspond to positive vorticity . This direction was counterclockwise in 51 . 5% of tissues and clockwise in 49 . 5% of tissues , with a sample size of 68 . With this convention , the vortex core always has positive vorticity . Accordingly , the outer region of the vortex exhibits negative vorticity ( Figure 3C , see Figure 3—figure supplement 2 for kymographs and heatmaps of vorticity representative tissues ) , which corresponds to the counter-rotation that occurs when the central vortical flow transitions to the outer radial flow ( Figure 3A , left ) . We define a characteristic vortex radius as the radial position of the center of the negative-vorticity region , which is ∼1 mm at 36 hr in small tissues ( Figure 3C , black bars ) . To analyze vortex dynamics across different tissues with varying vortex positioning , and to quantitatively capture the onset and strength of vortices , we calculated the enstrophy spectrum ℰ ( q , t ) =|ω~ ( 𝐪 , t ) |2 , where ω~ ( 𝐪 , t ) =∫ ( d𝐫/A ) ω ( 𝐫 , t ) ei𝐪⋅𝐫 are the spatial Fourier components of the vorticity field ω ( 𝐫 , t ) ( Alert et al . , 2020 ) . The enstrophy spectrum is the power spectral density of the vorticity field as a function of the wave-vector modulus q , and therefore provides a measure of the vortex intensity at a length scale 2π/q . The kymographs of the enstrophy spectrum show that most of the vortex’s intensity is found at a characteristic length scale of ∼1 mm ( Figure 3—figure supplement 3 ) . For each tissue we characterized the maximal vortex strength by the maximum value of ℰ ( q , t ) as well as its associated wavelength 2π/q and time of occurrence . We represented these three quantities on a scatter plot , which shows that vortices in small tissues have generally higher intensity than those in large tissues ( Figure 3D ) . Vortices in small tissues are also larger relative to tissue size , since the absolute size of vortices in small and large tissues is similar ( Figure 3D ) . Furthermore , vortex strength peaks several hours later in small tissues than in large tissues ( Figure 3D ) . We hypothesized that this difference is due to large tissues featuring a faster density increase than small tissues ( Figure 1C ) . To test this hypothesis , we varied the initial cell density of small tissues and observed that the time of maximum vortex intensity decreases with increasing density ( Figure 3E , Figure 3—figure supplement 3 ) . These results prompted us to examine spatiotemporal cell density evolution . Given that cell density appears to affect vortex formation and is known to control contact inhibition of locomotion and proliferation ( Schnyder et al . , 2020 ) , we explored the spatiotemporal evolution of cell density . Constructing average kymographs in the same way as for speed , radial velocity , and vorticity , we observe that the vortex region in the center of small tissues is accompanied by an unexpected local density minimum ( Figure 4A ) . Strikingly , snapshots of small and large tissues reveal that large-scale vortices occur in low-density regions , regardless of location within the tissue ( Figure 4—figure supplement 1 ) . However , given that vortices in large tissues are often off-centered , the low-density region does not appear in their average kymograph of cell density ( Figure 4A ) . To investigate the effects of initial conditions , we tracked the density evolution of the center and boundary zones across tissues with different starting densities and sizes , grouping initial densities into three ranges as before ( Figure 4B and C ) . As with the average density in Figure 1C , the density monotonically increases in large tissues centers but is non-monotonic in small tissues . Notably , the cell density at the center of small tissues of different initial cell densities reach a common minimum during the 16–32 hr time period ( Figure 4B ) , which includes the vortex onset time . At the boundary zone , the long-time evolution of the cell density is independent of initial tissue size and density ( Figure 4C ) . This common long-time evolution is reached at about 12 hr ( Figure 4C ) , which coincides with the time at which the edge radial velocity stabilizes upon the overshoot ( Figure 1D ) . To understand the unexpected transient density decrease at the center of small tissues , we sought to explain it as the result of combined advective transport based on the measured radial flow fields 𝐯r ( 𝐫 , t ) and homogeneous cell proliferation at a rate k ( 𝐫 , t ) =k0 throughout the tissue . To test this hypothesis , we solved the continuity equation for the cell density field ρ ( 𝐫 , t ) , ( 4 ) ∂ρ∂t=-∇⋅ ( ρ𝐯 ) +k0ρ , using the average radial velocity profiles vr ( r , t ) measured by PIV ( Figure 2D ) , and a proliferation rate k0=1 . 04 h−1 , which corresponds to a cell doubling time of 16 hr ( Materials and methods ) . This minimal model recapitulates the major features of the evolving density profiles for both small and large tissues ( compare Figure 4D with Figure 4A ) . Therefore , the unexpected formation of a central low-density region results from the combination of outward tissue flow and proliferation within the colony . However , further research is required to determine the biophysical origin of the non-monotonic density evolution . Moreover , having assumed a density-independent proliferation rate , our model predicts a cell density in the center of large tissues higher than the one measured at the end of the experiment , and it does not quantitatively reproduce the cell density profiles at the edge regions . These discrepancies suggest that more complex cell proliferation behavior is required to fully recapitulate the density dynamics in expanding cell monolayers . To better understand how tissue expansion affects cell proliferation , we analyzed the spatiotemporal dynamics of cell-cycle state . Our cells stably express the FUCCI markers , meaning that cells in the G0-G1-S phase of the cell cycle ( referred to here as G1 ) fluoresce in red ( shown as magenta ) , and cells in the S-G2-M phase of the cell cycle ( referred to here as G2 ) fluoresce in green ( Sakaue-Sawano et al . , 2008 ) . Additionally , immediately-post-mitotic cells do not fluoresce and appear dark . Small and large tissues are initially well mixed with green and magenta cells , confirming that cells are actively cycling throughout the tissue at the time of stencil removal ( Figure 5—figure supplement 1 ) . During tissue expansion , spatiotemporal patterns of cell-cycling behavior emerge ( Figure 5A , Figure 5—video 1 ) . To quantitatively investigate these cell-cycle patterns , we obtained the local fractions of G1 , G2 , and post-mitotic cells by evaluating cell cycle state for each cell nucleus ( see Materials and Methods ) . We then overlaid kymographs of the G1 and G2 cell-cycle-state fractions ( Figure 5B ) and plotted the time evolution of G1 , G2 , and post-mitotic fractions together ( Figure 5C , D ) . Immediately after stencil removal , we observe a cell division pulse in all tissues , which manifests in a decrease in G2 and increase in post-mitotic fraction ( Figure 5C , D ) . After about 12 hr of tissue expansion , the boundary region becomes primarily populated by rapidly-cycling cells ( Figure 5B , C ) , which results in a predominance of cells in this region that either have recently divided ( post-mitotic , black ) or are likely to divide soon ( G2 , green ) . The high numbers of post-mitotic cells indicate that cells in G1 rapidly proceed to mitosis . Given that the edge radial speed overshoots during the first 12 hr of tissue expansion ( Figure 1D ) , future work is necessary to characterize the effect of cell cycling on edge motion at early stages of expansion . In the central region of small tissues ( Figure 5B left , D left ) , we observe cell-cycling dynamics similar to the boundary region . Thus , in the tissue-spanning vortex of small tissues , cells are also rapidly cycling . The fraction of cells in G1 only starts to increase at ∼40 hr ( Figure 5D left ) , coinciding with the weakening of the vortex ( Figure 3C left ) . In contrast , the center zone of large tissues undergoes strong cell-cycle arrest at the G1-G2 transition at about 30 hr , also coinciding with the weakening of the vortex in large tissues ( Figure 5B right , D right ) . Cells already past G1 at this time continue to division and re-enter G1 , evidenced by the steady increase in local fraction of G1 accompanied by a steady decrease in G2 after 30 hr . Similar cell-cycle arrests were previously reported both in growing epithelia ( Streichan et al . , 2014 ) and in spreading 3D cell aggregates ( Beaune et al . , 2014 ) . Before the onset of cell-cycle arrest , the center of large tissues exhibits large-scale coordinated cell-cycling dynamics in the form of anti-phase oscillations , with peaks in G2 fraction accompanied by troughs in G1 fraction ( Figure 5B right , D right ) . Finally , we sought to link cell-cycle dynamics to the kinematics of tissue expansion by studying correlations between local measurements of cell cycle , cell speed , and cell density ( Figure 5E ) . Here , each point represents one PIV window , with color indicating its average cell-cycle state . As expected , cell speed is negatively correlated with cell density . Further , in large tissues , the cell-cycle state transitions from G1-dominated to G2-dominated when cell density increases above ∼5000 cells/mm2 and cell speed falls below ∼12 μm/h ( Figure 5E right ) . In this regime , the decrease of cell speed with increasing cell density bears similarities to previously-reported glass transitions and contact inhibition of locomotion ( Angelini et al . , 2011; Zimmermann et al . , 2016; Garcia et al . , 2015 ) . Small tissues , by contrast , lack the G1-dominated , slow , high-density cell population ( Figure 5E , left ) found in the center of large tissues . Taken together , our findings emphasize that cell cycling , cell flow , and cell density patterns are inextricably linked and depend on the initial size of an expanding tissue .
We began this study by asking how changes in initial size affect the long-term expansion and growth of millimeter-scale epithelia . By means of high spatiotemporal resolution imaging and precisely controlled initial conditions , our assays systematically dissected tissue expansion and growth from the overall boundary kinematics ( Figure 1 ) to the internal flow patterns ( Figures 2 , 3 and 4 ) and cell-cycle dynamics ( Figure 5 ) . While we demonstrated that ‘small’ tissues increase in area relatively much faster than do ‘large’ tissues , our data suggest a surprising and stark decoupling of the outer and inner regions of an expanding epithelium . Notably , the behaviors of the edge zones are largely independent of tissue size , cell density , and history , while interior dynamics depend strongly on these factors . Unexpectedly , the overall tissue growth and expansion dynamics ( Figure 1 ) could be attributed to one dominant feature: these epithelia expanded at the same edge speed regardless of initial tissue size , shape , and cell density . The only exception is the major axes of ellipses , where the normal edge speed is smaller when the radius of curvature is rc<0 . 75 mm . This observation , combined with the fact that the velocity penetration length is 500 mm ( Figure 2D ) , suggests that a tissue must be 1 mm in diameter for the tissue edge to move independently of bulk flows . As a result of this robust edge motion , the areal expansion rate of the tissue is dictated by its perimeter-to-area ratio . To further emphasize the decoupling of the boundary and internal dynamics of epithelia , consider that the key findings in Figure 1 neither predict nor depend upon the radically different internal dynamics we observed within ‘small’ and ‘large’ tissues . For instance , despite the roiling vortices occupying large portions of ‘small’ tissues and the pronounced , large-scale contact inhibition of ‘large’ tissues–two antithetical phenomena–no hints of these behaviors can be detected in the motion of the boundary . Critically , the type and timing of internal dynamics are dictated not by the current size but by the expansion history of a given tissue . While a small tissue eventually expands to reach the initial size of a large tissue , it exhibits different internal dynamics from the large tissue at this size ( Figure 6 ) . This difference in internal dynamics is perhaps easiest to observe in spatiotemporal evolution of cell cycle ( Figure 6D , Figure 5B dashed boxes ) ; the small-tissue footprint from 30 to 46 hr closely matches the large-tissue footprint from 0 to 16 hr , but the cell cycle distribution during these time periods bears almost no similarities . This applies as well to other important bulk properties of the tissue ( Figure 6A–C ) , as cell cycle is tightly linked to cell speeds and density ( Figure 5E ) . For example , at equal current sizes , the center of initially-small tissues features high vorticity with decreasing cell speed whereas initially-large tissues exhibit low vorticity and increasing cell speed ( Figure 6A , B ) . Respectively , at equal current sizes , while absolute cell densities in the tissue centers share some overlap , it is notable that the rate of density change at the tissue center is increasing faster in initially-small tissues than in initially-large tissues ( Figure 6C ) . However , the most striking differences in cell density evolution occur not at equal current sizes but during the early stages of tissue expansion: whereas the cell density at the center of large tissues increases at all times , the center of small tissues features a marked density decrease between ∼8 and ∼24 hr ( Figure 4A , B ) . Overall , while edge dynamics are stereotyped and conserved across different sizes , our findings suggest that initial tissue size impacts the bulk dynamics by altering the constraints under which the tissue grows . We expect that tissues with sizes between our two choices would exhibit similar edge dynamics and internal patterns that cross over between our small and large tissues . The vortices are a particularly striking example of such size- and history-dependent internal patterns ( Figure 3 , Figure 6B ) . Our active fluid model suggests that the vortices emerge from a dynamical instability of the tissue bulk , which occurs when the tissue reaches a critical size . Thus , whereas the instability itself is a bulk phenomenon independent of the tissue edge , edge-driven expansion allows small tissues to reach the critical size that triggers the instability . In addition , our data suggest a strong correlation between vortex formation and the development of non-monotonic density profiles . Not only did small tissues exhibit co-occurrence of vortices with density decreases in the tissue center , but also off-center vortices in large tissues always co-localized with a local density decrease ( Figure 4—figure supplement 1 ) . Our model does not currently describe cell density , and hence cannot explain the relationship between vortex formation and local density decreases . Thus , our experimental findings call for the development of more detailed models that couple cell density to both the velocity and the polarity fields , accounting for how density gradients influence cell polarization ( Alert and Trepat , 2020 ) . The pronounced decoupling between boundary and internal dynamics in epithelia confers stability to the overall expansion of the tissue , making it robust to a wide range of internal perturbations . From the perspective of collective behavior , we speculate that such robust boundary dynamics may be beneficial in a tissue such as an epithelium whose teleology is to continuously expand from its free edges to sheath organ surfaces . Further , the ability to accurately predict epithelial expansion with a single parameter , the edge speed , will have practical uses in experimental design and tissue-engineering applications . Finally , given that many of the phenomena presented here only occurred due to the millimetric scale of our unconfined tissues and the long duration of the experiments , our results showcase the value of pushing the boundaries of large-scale , long-term studies on freely-expanding tissues .
All experiments were performed with MDCK-II cells expressing the FUCCI cell-cycle marker system as received from: Streichan et al . , 2014 . After treatment with Mycoplasma Removal Agent ( MPI Biological ) , cells tested negative for mycoplasma ( MycoProbe , R and D Systems ) . We cultured cells in MDCK media consisting of low-glucose ( 1 g/L ) DMEM with phenol red ( Gibco , USA ) , 1 g/L sodium bicarbonate , 1% streptomycin/penicillin , and 10% FBS ( Atlanta Biological , USA ) . Cells were maintained at 37°C and 5% CO2 in humidified air . We coated tissue-culture plastic dishes ( BD Falcon , USA ) with type-IV collagen ( MilliporeSigma , USA ) by incubating 150 μL of 50 μg/mL collagen on the dish under a glass coverslip for 30 min at 37°C , washing three times with deionized distilled water ( DI ) , and allowing the dish to air-dry . We then fabricated silicone stencils with cutouts of desired shape and size and transferred the stencils to the collagen coated surface of the dishes . Stencils were cut from 250 μm thick silicone ( Bisco HT-6240 , Stockwell Elastomers ) using a Silhouette Cameo vinyl cutter ( Silhouette , USA ) . We then seeded the individual stencils with cells suspended in media at 1000 cells/mL . Suspended cells were concentrated at ∼2 . 25×106 cells/mL and pipetted cells into the stencils at the appropriate volume . Care was taken not to disturb the collagen coating with the pipette tip . To allow attachment of cells to the collagen matrix , we incubated the cells in the stencils for 30 min in a humidified chamber before flooding the dish with media . We then incubated the cells for an additional 18 hr to allow the cells to form monolayers in the stencils , after which the stencils were removed with tweezers . Imaging began 30 min after stencil removal . Media without phenol red was used throughout seeding and imaging to reduce background signal during fluorescence imaging . All imaging was performed with a 4X phase contrast objective on an automated , inverted Nikon Ti2 with environmental control ( 37°C and humidified 5% CO2 ) using NIS Elements software and a Nikon Qi2 CMOS camera . Phase contrast images were captured every 20 min , while RFP/GFP channels were captured every 60 min at 25% lamp power ( Sola SE , Lumencor , USA ) and 500 ms exposure time . No phototoxicity was observed under these conditions for up to 48 hr . Final images were composited from 4 × 4 montages of each dish using NIS Elements . Tissues were segmented to make binary masks using a custom MATLAB ( Mathworks ) script . Tissue edge radial velocity was measured from the binary masks within more than 200 discrete sectors of the tissue; the edge radial velocity of all sectors were averaged to arrive at the tissue average edge radial velocity . Radial velocity at each sector was calculated for each timepoint as the rate of change of the average extent of the boundary pixels of the sector , utilized a rolling average of 3 timepoints ( 1 hr ) to account for capture phase offsets resulting from capturing phase and fluorescence images at different frequencies . Sectors originated from the center of each tissue at the initial timepoint and were ∼20 µm wide at the edge of the tissue at the starting point . Curvature at the major and minor axes of growing tissues was approximated at each time-point by fitting an ellipse to the tissue footprint and taking the radius of curvature at the minor and major axes as b2/a and a2/b , respectively , where a is the major semi-axis length and b is the minor semi-axis length . Normalized χ2 values in Figure 1E were calculated as 1N∑i=1N ( ui-μi ) 2σi2 , where N is the number of time-points in the curve , ui are the model predictions , and μi and σi are the mean and standard deviation of the measured values , respectively . With these definitions , a fit with χ2<1 is good . The P-value in Figure 3D was calculated using a Mann-Whitney U test , and the two-tailed p-value of p<10-4 indicates that the large and small vortex power data indeed come from different populations . The FUCCI system contains a period after M-phase where cells go dark , making FUCCI unreliable for cell counting . Instead , we developed and trained a convolutional neural network to reproduce nuclei from 4X phase contrast images using our in-house Fluorescence Reconstruction Microscopy tool ( LaChance and Cohen , 2020 ) . The output of this neural network was then segmented in ImageJ to determine nuclei footprints and centroids . Tissue velocity vector fields were calculated from 2 × 2 resized phase contrast image sequences using the free MATLAB package PIVLab ( Thielicke and Stamhuis , 2014 ) with the FFT window deformation algorithm . We used a 1 st pass window size of 64 × 64 pixels and second pass of 32 × 32 pixels , with 50% pixel overlaps . This resulted in a 115 × 115 μm window . The window size was chosen to be smaller than the velocity-velocity correlation length but large enough to enable fast computation of PIV fields for many tissues . As seen in Figure 2—figure supplement 1 , using a window size of 57 × 57 μm , which contains only a few cells , yields higher resolution velocity fields but does not qualitatively affect the measured speed and radial velocity . We focus on large-scale features of the velocity field , which are not affected by choosing a smaller PIV window size . Local density was also calculated for each PIV window by counting the number of approximate nucleus centroids in that window . Data from PIV were smoothed in time with a moving average of 3 time points centered at each timepoint as before . First , we constructed kymographs for individual tissues using distance from the tissue center as the spatial index for each measurement window corresponding to a kymograph pixel . We did not plot kymograph pixels for which more than 95% of the measurements at that distance were beyond the tissue footprint . We then averaged the individual tissue kymographs , aligning by the centers . We first generated a plot of all relevant trajectories ( Tinevez et al . , 2017 ) colorized randomly in grayscale using a custom MATLAB ( Mathworks ) script . We then used the Fiji plugin OrientationJ on this plot to colorize the resulting image according to orientation ( Püspöki et al . , 2016 ) . To test whether the observed spatiotemporal evolution of density ρ ( r , t ) could be explained by flow of material ( rather than divisions , extrusions , and cell death ) , we solved the continuity equation for a homogenous tissue in a circular geometry with spatiotemporal evolution of average radial velocity vr ( r , t ) as measured from PIV in experiments ( Figure 2B ) . The continuity equation is ( 5 ) ∂ρ∂t=-∇⋅𝐣+k0ρ , where a homogeneous cell proliferation rate k0=1 . 04h-1 is assumed throughout the tissue , which corresponds to the cell doubling time of 16 hr . The current density is 𝐣=ρ𝐯𝐫-D∇ρ , where we included a diffusion term with a small diffusion constant D=0 . 22mm2/h for numerical stability . The continuity Equation ( 5 ) was discretized using the finite volume method ( Eymard et al . , 2000 ) , which is briefly summarized below . The tissue domain was divided into an inner circle Ω0 of radius r1/2=12Δr and circular annuli Ωi with inner radii ri-1/2= ( i-12 ) Δr and outer radii ri+1/2= ( i+12 ) Δr , respectively , where i=1 , 2 , 3 , … and Δr=115μm corresponds to the width of 1 window in the PIV analysis . The continuity Equation ( 5 ) was then integrated over the inner circle Ω0 and circular annuli Ωi as ( 6a ) 1A0∫0r1/2 ( 2πrdr ) ∂ρ∂t=1A0∫0r1/2 ( 2πrdr ) [−∇⋅j+k0ρ] , ( 6b ) 1Ai∫ri−1/2ri+1/2 ( 2πrdr ) ∂ρ∂t=1Ai∫ri−1/2ri+1/2 ( 2πrdr ) [−∇⋅j+k0ρ] , where A0=πr1/22 is the area of the inner circle Ω0 and Ai=πri+1/22-πri-1/22 is the area of the circular annulus Ωi . The integrals in Equation ( 6a , b ) can be approximated as ( 7a ) ∂ρ ( 0 , t ) ∂t=−2πA0r1/2j ( r1/2 , t ) +k0ρ ( 0 , t ) , ( 7b ) ∂ρ ( ri , t ) ∂t=−2πAi[ri+1/2j ( ri+1/2 , t ) −ri−1/2j ( ri−1/2 , t ) ]+k0ρ ( ri , t ) . Here , density profiles ρ ( ri , t ) are evaluated at ri=iΔr for all i=0 , 1 , 2 , … . Current densities are evaluated as j ( ri+1/2 , t ) =ρ ( ri+1/2 , t ) vr ( ri+1/2 , t ) -D[ρ ( ri+1 , t ) -ρ ( ri , t ) ]/Δr for all i=0 , 1 , 2 , … , where ρ ( ri+1/2 , t ) =[ρ ( ri , t ) +ρ ( ri+1 , t ) ]/2 and vr ( ri+1/2 , t ) =[vr ( ri , t ) +vr ( ri+1 , t ) ]/2 . Density profiles ρ ( ri , t ) were then obtained by integrating Equation ( 7 ) with the forward Euler method using a time step Δt=20 min to align with experimental data collection of radial velocity profiles vr ( ri , t ) from Figure 2B . The initial conditions were ρ ( ri , 0 ) =2700 cells/mm2 for ri<rtissue and ρ ( ri , 0 ) =0cells/mm2 for ri>rtissue , where rtissue is the radius of tissue at the beginning of experiment . For comparison with experimental data ( see Figure 4 ) , we thresholded the kymographs of simulated density at 100cells/mm2 , which corresponds to much lower density than a confluent tissue . For panels ( E ) and ( F ) in Figure 4—figure supplement 1 , we applied a Fourier low-pass filter on vorticity fields , retaining only large-scale vorticity fluctuations ( with wavelengths longer than 1 mm ) . We excluded the tissue edge region ( 500 μm from the boundary ) that is outward polarized and does not exhibit vortical flows . Each point in panels ( E ) and ( F ) corresponds to a point in the filtered vorticity field , plotted against the cell density in that point . The Fucci system consists of an RFP and GFP fused to proteins Cdt1 and Geminin , respectively ( Sakaue-Sawano et al . , 2008 ) . Cdt1 levels are high during G1 and low during the rest of the cell cycle , while Geminin levels are high during the S , G2 , and M phases ( Sakaue-Sawano et al . , 2008; Streichan et al . , 2014 ) . After capturing the appropriate fluorescence images , preprocessing was implemented identically for GFP and RFP channels to normalize channel histograms . To determine local cell cycle fraction , we determined the median value of RFP and GFP signal for each cell nucleus and manually selected thresholds for RFP and GFP signals separately to classify cell cycle for each cell as G0-G1-S ( RFP above threshold ) , S-G2-M ( RFP below threshold and GFP above threshold ) , or postmitotic ( RFP and GFP below threshold ) . Local cell cycle fraction of each state could then be easily computed for each PIV pixel . Note that S phase ( both RFP and GFP signals above threshold ) did not prove to be a reliable feature for segmentation . Data for representative small , large , and ellipse tissues ( Heinrich et al . , 2020 ) and analysis Matlab scripts ( Heinrich , 2020 ) have been made available ( copy archived at https://github . com/elifesciences-publications/FreelyExpandingTissues ) . | Cells do not exist in isolation . Instead , they form tissues , where individual cells make contact with their neighbors and form microscopic ‘architectures’ . Epithelia are a type of tissue where cells are arranged in flat sheets , and are found in organs such as the lining of the kidney or the skin . Tissues need to grow , especially early in life . If tissues are damaged – for example , if the skin is cut or grazed – cells also need to divide ( to create new healthy cells ) and move as a group ( to close the wound ) . Such coordinated motions result in cells exhibiting distinct group behaviors , similar to those observed within crowds of people or schools of fish . If coordination breaks down , problems can happen such as uncoordinated tissue growth seen in cancer . However , how cell movements are coordinated is still not fully understand . For example , researchers know that cells’ positions within a group can determine how they behave , meaning that even the same type of cell could behave differently at the edge or center of a tissue . This suggests that the initial size and shape of a tissue should influence its subsequent growth and behavior; however , the nature of this influence is still largely unknown . Heinrich et al . therefore wanted to determine the differences in the way larger and smaller tissues grow . Microscope imaging was used to track the growth of circular , artificial tissues made from single-layered sheets of dog kidney cells grown in the laboratory . Comparing how quickly the tissues expanded revealed that the area of tissue circles that started out smaller increased at a much faster rate than that of tissue circles that were larger to begin with . This turned out to be because the edges of the tissues grew at a constant speed , independent of their initial size or shape , but circles with a smaller area have a larger proportion of cells on their edges . The motions of the cells at the center of the tissues had no effect on how the edges of the tissue grew . A final observation was that the way tissues of a given size behaved depended on whether they had grown to be that size , or they started off that big . These results shed light on how groups of cells interact in growing tissues . In the future , this information could be used to predict how different tissues grow over time , potentially helping scientists engineer better artificial tissues or organs for transplantation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"physics",
"of",
"living",
"systems"
] | 2020 | Size-dependent patterns of cell proliferation and migration in freely-expanding epithelia |
Disruption of protein folding in the endoplasmic reticulum ( ER ) activates the unfolded protein response ( UPR ) —a signaling network that ultimately determines cell fate . Initially , UPR signaling aims at cytoprotection and restoration of ER homeostasis; that failing , it drives apoptotic cell death . ER stress initiates apoptosis through intracellular activation of death receptor 5 ( DR5 ) independent of its canonical extracellular ligand Apo2L/TRAIL; however , the mechanism underlying DR5 activation is unknown . In cultured human cells , we find that misfolded proteins can directly engage with DR5 in the ER-Golgi intermediate compartment , where DR5 assembles pro-apoptotic caspase 8-activating complexes . Moreover , peptides used as a proxy for exposed misfolded protein chains selectively bind to the purified DR5 ectodomain and induce its oligomerization . These findings indicate that misfolded proteins can act as ligands to activate DR5 intracellularly and promote apoptosis . We propose that cells can use DR5 as a late protein-folding checkpoint before committing to a terminal apoptotic fate .
Proper folding of transmembrane and secreted proteins is critical to cell function and intercellular communication . Quality control of protein folding begins in the endoplasmic reticulum ( ER ) and responds to increased protein-folding demand during physiological or pathophysiological stresses . Accumulation of unfolded or misfolded proteins in the ER , known as ER stress , activates the unfolded protein response ( UPR ) – a network of intracellular signaling pathways that initially mount cytoprotective response to restore ER homeostasis but can ultimately switch to a pro-apoptotic program under irresolvable stress ( Walter and Ron , 2011; Tabas and Ron , 2011 ) . Two key UPR sensors , IRE1 and PERK coordinate the decision between cell survival and death through the delayed upregulation of the apoptosis-initiating protein death receptor 5 ( DR5 ) ( Lu et al . , 2014; Chang et al . , 2018 ) . During ER stress , IRE1 and PERK oligomerize upon directly binding to misfolded proteins , leading to their activation ( Karagöz et al . , 2017; Wang et al . , 2018 ) . PERK activation causes the selective translation of ATF4 and CHOP , which , in addition to upregulating genes that enhance the folding capacity of the ER , promotes the transcription of pro-apoptotic DR5 ( Harding et al . , 2003; Yamaguchi and Wang , 2004 ) . The pro-apoptotic signal is initially counteracted by regulated IRE1-dependent mRNA decay ( RIDD ) that degrades DR5 mRNA ( Lu et al . , 2014 ) . Upon prolonged ER stress , PERK exerts negative feedback on IRE1 activity attenuating RIDD , thus de-repressing DR5 synthesis to drive cell commitment to apoptosis ( Chang et al . , 2018 ) . DR5 is a pro-apoptotic member of the tumor necrosis factor receptor ( TNFR ) superfamily that signals from the plasma membrane into the cell in response to extracellular cues ( Sheridan et al . , 1997; Walczak et al . , 1997; Ashkenazi , 1998 ) . It is constitutively expressed in various tissue types and forms auto-inhibited dimers in its resting state , analogous to other members of the TNFR family ( Spierings et al . , 2004; Pan et al . , 2019; Vanamee and Faustman , 2018 ) . In its canonical mode of activation , binding of the homotrimeric extracellular ligand TRAIL ( also known as Apo2L ) ( Wiley et al . , 1995; Pitti et al . , 1996 ) assembles DR5 into higher-order oligomers ( Hymowitz et al . , 1999; Mongkolsapaya et al . , 1999; Valley et al . , 2012 ) . Consequently , DR5 forms intracellular scaffolds in which its cytosolic death domains recruit the adaptor protein FADD and pro-caspase 8 into the ‘death-inducing signaling complex’ ( DISC ) ( Kischkel et al . , 2000; Sprick et al . , 2000; Jin et al . , 2009; Dickens et al . , 2012 ) . Upon DISC-mediated dimerization , pro-caspase 8 molecules undergo regulated auto-proteolysis to form active initiator caspase 8 ( Muzio et al . , 1998 ) . Activated caspase 8 frequently induces the intrinsic mitochondrial apoptotic pathway by truncating Bid , a pro-apoptotic Bcl2 protein , to cause Bax-mediated permeabilization of the mitochondrial outer membrane ( Wei et al . , 2001; LeBlanc et al . , 2002 ) . While DR5 and caspase 8 are both required for apoptosis during ER stress , we ( Lu et al . , 2014; Lam et al . , 2018 ) , along with other independent groups ( Cazanave et al . , 2011; Iurlaro et al . , 2017; Dufour et al . , 2017 ) , found unexpectedly that TRAIL is dispensable for this DR5 activation . Indeed , upon ER stress , most newly synthesized DR5 molecules never make it to the plasma membrane but remain intracellular and thus inaccessible to extracellular ligands ( Lu et al . , 2014; Iurlaro et al . , 2017 ) . Given that at physiological levels DR5 is auto-inhibited until activated by a ligand , it remained a mystery how DR5 is activated in response to ER stress , prompting us to interrogate its intracellular mechanism of activation .
To examine the mechanism of cell death driven by an unmitigated protein folding burden , we induced the exogenous expression of a GFP-tagged form of the glycoprotein myelin protein zero ( MPZ ) in epithelial cells ( Figure 1A ) . MPZ initially folds in the ER and then travels to the plasma membrane to mediate membrane adhesion in myelin-forming Schwann cells , where it is normally expressed . Mutations of MPZ that impair folding and cause its intracellular retention activate the UPR , leading to apoptosis in a manner dependent on CHOP ( Pennuto et al . , 2008 ) . We found that in epithelial cells , titration of even non-mutant , GFP-tagged MPZ to high expression levels resulted in its intracellular accumulation , indicating a compromised MPZ folding state ( Figure 1A ) . Folding-compromised MPZ induced a dose-dependent upregulation of the UPR transcriptional target genes CHOP , BiP , and DR5 ( Figure 1—figure supplement 1A ) . Upregulated DR5 was retained intracellularly ( Figure 1A , Figure 1—figure supplement 1B ) and occurred concomitantly with cleavage of caspase 8and its downstream target caspase 3 ( Figure 1B ) . By contrast , low levels of MPZ-GFP expression that exhibited proper plasma membrane localization did not induce caspase 8 or 3 activity ( Figure 1A , B ) . To determine when caspase 8 became active relative to cytoprotective UPR signaling , we assessed IRE1 activity during high MPZ-GFP expression through analysis of XBP1 mRNA splicing . As expected , IRE1-mediated XBP1 mRNA splicing initiated a few hours post-transfection with MPZ-GFP and later attenuated ( Figure 1—figure supplement 1C ) . The upregulation of DR5 , caspase activity , and PARP cleavage ( another indicator of apoptotic progression ) occurred after the attenuation of IRE1 activity , consistent with the hallmarks of terminal pro-apoptotic UPR signaling ( Figure 1—figure supplement 1D–1E ) . To determine if DR5 was required for apoptosis during this sustained protein misfolding stress , we acutely depleted DR5 mRNA by siRNA prior to overexpressing MPZ-GFP . Knockdown of DR5 significantly reduced PARP cleavage and annexin V staining following overexpression of MPZ-GFP ( Figure 1C , D ) , which was not observed in control experiments expressing cytosolic GFP . To determine if upregulation of DR5 was sufficient to induce apoptosis , we increased DR5 levels in the absence of ER stress through ectopic expression of CHOP . Comparable levels of CHOP-induced DR5 protein in the absence of ER stress drove drastically lower levels of PARP cleavage and trypan blue staining ( demarking apoptotic cells ) compared to the presence of misfolded-protein stress ( Figure 1—figure supplement 2A and C–D ) . These results show that DR5 activation does not occur spontaneously after its upregulation but requires additional input signals conveyed by ER stress . To assess the molecular composition of activated DR5 assemblies formed in response to ER stress , we measured caspase 8 activity in cell extracts fractionated through size exclusion chromatography . We detected increased caspase 8 activity in high-molecular w8 ( MW ) fractions of cells transfected with MPZ-GFP relative to GFP ( Figure 1E ) . The fractions contained DR5 complexes and co-eluted with full-length MPZ-GFP but not GFP-degradation products ( Figure 1E , lanes 2 and 4 ) . Pull-down of DR5 from cell lysates enriched for FADD and MPZ-GFP ( Figure 1—figure supplement 3A ) , suggesting that the co-elution of DR5 and MPZ-GFP in the high MW fractions resulted from their physical association . To test if MPZ physically interacted with activated DR5 complexes , we immunoprecipitated MPZ-GFP and detected DR5 , FADD , and caspase 8 ( both full-length p55 and its cleaved form p43 ) ( Figure 1F , Figure 1—figure supplement 3B ) . Furthermore , MPZ-GFP immunoprecipitates contained 2–3-fold more caspase 8 activity compared to empty beads ( Figure 1G , Figure 1—figure supplement 3C ) , indicating that they contained assembled DISC in a similar degree as seen after affinity purification of TRAIL-ligated DR5 ( Hughes et al . , 2013 ) . In contrast , pull-down of cytosolic GFP did not enrich for DR5 , FADD , or caspase activity ( Figure 1F , G ) , confirming the selectivity for ER-folded MPZ-GFP . To determine if misfolded proteins generally induced caspase activity through association with DR5 , we overexpressed GFP-tagged forms of two other ER-trafficked proteins , rhodopsin ( RHO ) and proinsulin ( INS ) , which are also associated with CHOP-dependent cell death pathologies ( Chiang et al . , 2016; Oyadomari et al . , 2002 ) . Sustained overexpression of both RHO-GFP and INS-GFP upregulated BiP and CHOP mRNAs ( Figure 1—figure supplement 4A ) and induced XBP1 mRNA splicing ( Figure 1—figure supplement 4B ) . Both proteins formed SDS-insoluble aggregates and induced PARP cleavage and annexin V staining in a DR5-dependent manner ( Figure 1—figure supplement 4C–4E ) . By contrast , immunoprecipitation of RHO-GFP enriched for DR5 protein and caspase 8 activity more robustly than INS-GFP ( Figure 1—figure supplement 5 ) , despite inducing DR5-dependent apoptosis to a similar extent . This indicates that misfolded proteins differ in their propensity to directly engage the DR5-assembled DISC , and that other misfolded substrates—caused by the ectopically overexpressed ER-trafficked protein—may mediate direct DR5 activation . Thus , as exemplified by MPZ and RHO , a selective subset of misfolded proteins in the secretory pathway can engage DR5 to form oligomeric complexes that induce caspase 8 activation . To explore where within in the cell DR5 associated with misfolded protein , we used confocal imaging of fixed cells for immunofluorescence . These analyses revealed that intracellular MPZ-GFP and DR5 appeared in discrete puncta that often overlapped ( Figure 2A ) . DR5 siRNA knockdown eliminated the DR5 signal , confirming the specificity of the DR5 antibody ( Figure 2—figure supplement 1A , right panel ) . Similarly , overexpression of RHO also resulted in intracellular puncta that frequently co-localized with DR5 clusters ( Figure 2—figure supplement 1B ) . Quantification of the mean Pearson’s correlation per cell demonstrated statistically significant overlap with DR5 signal for both GFP-tagged MPZ and RHO ( Figure 2—figure supplement 1C ) , indicating that these misfolded proteins accumulate in the same compartment as DR5 . Previous findings suggested that DR5 is retained near the Golgi apparatus during ER stress ( Lu et al . , 2014 ) . We confirmed co-localization with the purported Golgi marker RCAS1 , as previously reported ( Figure 2—figure supplement 1D ) . However , we observed little overlap in DR5 staining with another cis-Golgi marker , giantin ( Figure 2E ) . To resolve this discrepancy , we employed subcellular fractionation as an orthogonal biochemical approach . Separating organelle membranes revealed that RCAS1 , DR5 , and MPZ-GFP co-sedimented in fractions containing ERGIC53 , a marker of the ER-Golgi intermediate compartment ( ERGIC ) , but not with those containing giantin ( Figure 2B ) . Notably , a portion of FADD , a cytosolic protein expected to exclusively remain in the topmost , cytosolic fraction , migrated into the second fraction of the gradient , indicating its association with the ERGIC membranes . Consistent with the presence of FADD , the first and second ERGIC-associated fractions harbored the majority of the caspase 8 activity in the cell lysate ( Figure 2C ) , indicating the presence of active DR5 DISCs . Moreover , immunofluorescence with quantification of the mean correlation per cell demonstrated the co-localization of DR5 with the ERGIC rather than with the Golgi ( Figure 2D and F ) . To determine when DR5 accumulates at the ERGIC relative to misfolded proteins , we compared the immunofluorescence of cells fixed at 20 hr ( before the onset of caspase activity ) and at 24 hr post-transfection ( after the onset of caspase activity , Figure 1—figure supplement 1E ) . Intracellular puncta of MPZ appeared at 20 hr , preceding the appearance of DR5 clusters at 24 hr ( Figure 2—figure supplement 2A ) . Between 20 and 24 hr , the correlation of DR5 and ERGIC53 increased , whereas the correlation of MPZ with ERGIC53 remained steady , indicating that DR5 accumulated after saturation of MPZ levels at the ERGIC ( Figure 2—figure supplement 2B–2C ) . By contrast , the mean Pearson’s correlation with giantin approached zero for both MPZ and DR5 at 24 hr post-transfection ( Figure 2—figure supplement 2B , Figure 2F ) . These results confirm the localization of DR5 and misfolded protein at the ERGIC under conditions of unmitigated ER stress . With evidence of a physical association between misfolded protein and active DR5 oligomers at the ERGIC , we asked how misfolded proteins and DR5 interact . Considering the precedence that ( i ) DR5 binds unstructured peptides mimicking TRAIL ( Kajiwara , 2004; Pavet et al . , 2010 ) and ( ii ) that UPR sensors can directly bind misfolded protein to sense ER stress ( Karagöz et al . , 2017; Wang et al . , 2018; Gardner and Walter , 2011 ) , we hypothesized that DR5 may directly recognize unstructured regions of misfolded proteins through its ectodomain ( ECD ) that would project into the ERGIC lumen . Probing a peptide array with purified recombinant Fc-tagged DR5 ECD revealed promiscuous recognition of amino acid sequences throughout the ectodomain of MPZ and within extracellular loops of RHO ( Figure 3A , Figure 3—figure supplement 1A–1B ) . Quantification of the relative signal intensity revealed that DR5-binding sequences were enriched for aliphatic and aromatic residues whereas polar and acidic residues were excluded ( Figure 3—figure supplement 1C ) , reminiscent of qualities that become surface-exposed in misfolded or unfolded proteins . To validate the specificity of DR5 interactions on the array , we performed pull-down assays on the MPZ-derived peptide exhibiting the strongest signal ( spots C18-C19 in Figure 3A , hereon referred to as MPZ-ecto ) with recombinant Fc-tagged DR5 ECD versus TNFR1 ECD as a selectivity control . The MPZ-ecto peptide bound specifically to the DR5 ECD but not the TNFR1 ECD ( Figure 3B ) . Under equilibrium conditions , interaction with MPZ-ecto peptide quenched fluorescently labeled DR5 ECD but not fluorescently labeled TNFR1 ECD , yielding an apparent binding affinity of K1/2 = 109 μM±11 μM with a Hill coefficient of 2 . 6 ( Figure 3C , Figure 3—figure supplement 2A ) . Adding excess unlabeled DR5 ECD restored fluorescence ( Figure 3—figure supplement 2B ) , indicating that the quenching reflected a specific and reversible interaction between the DR5 ECD and the MPZ-ecto peptide . Moreover , mutation of two aromatic amino acids ( both Tyr ) to disfavored acidic amino acids ( Glu ) abrogated binding ( Figure 3C ) , demonstrating that the interaction is sequence-specific . The Hill coefficient of 2 . 6 suggested cooperative binding . Therefore , we tested if the DR5 ECD forms oligomers in the presence of peptide . In the absence of peptide , the addition of a chemical cross-linker captured dimers of FLAG-tagged DR5 ECD ( Figure 3D , Figure 3—figure supplement 2C ) , consistent with pre-ligand assembled dimers previously observed for members of the TNFR family ( Clancy et al . , 2005; Siegel et al . , 2000; Chan et al . , 2000 ) . With increasing concentration of peptide ( up to 200 μM ) , crosslinking revealed multimers of the DR5 ECD ( Figure 3D ) , indicating that the peptide acts as a ligand to template assembly of DR5 oligomers . Interestingly , excess peptide ( 400 μM ) dissociated higher-order oligomers of DR5 , suggesting a lower valency of interaction when the DR5 concentration becomes limiting . To examine the DR5 oligomerization at saturating peptide concentrations by an orthogonal method , we fractionated DR5 ECD-peptide complexes using size exclusion chromatography . At 100 μM MPZ-ecto ( ~K1/2 ) , DR5 ECD co-eluted with the peptide as higher-order oligomers near the void volume ( 7–8 ml ) and as apo-dimers centered at 14 ml , as shown in the Coomassie blue-stained gel for DR5 and fluorescence scan for fluorescein-labeled MPZ-ecto peptide ( Figure 3E , F , green outline ) . This elution pattern was similar to that of the DR5 ECD-TRAIL complex , for which both proteins co-eluted near the void volume ( Figure 3—figure supplement 2E–2F ) . However , with excess MPZ-ecto peptide at 400 μM ( 4-times K1/2 ) , the proportion of higher-order oligomers of DR5 ECD and the peptide diminished and re-distributed to later eluting fractions at 12–15 ml ( Figure 3—figure supplement 2G–2H , teal outline ) , indicating disassembly into smaller oligomers of DR5 ECD and pointing at the reversibility of the higher-order DR5-peptide assemblies . Importantly , the non-binding peptide bearing the Tyr-to-Glu substitutions did not co-migrate with or induce the oligomerization of DR5 ECD ( Figure 3E , F , magenta outline ) . Since mutating the Tyr residues to Glu on the MPZ-ecto peptide proved sufficient to disrupt the DR5 ECD interaction in solution , we tested the ability of this minimal MPZ-derived sequence to bind to and activate DR5 in cells . To this end , we generated constructs that replaced the ectodomain of MPZ with either the MPZ-ecto peptide , the peptide sequence with Tyr-to-Glu substitutions , or the peptide with all its aromatic residues changed to Glu to further deplete DR5-favored amino acid side chains revealed in the peptide array ( Figure 4A ) . In a titration of MPZ-ecto peptide expression , the WT peptide sequence induced more PARP cleavage and caspase activity than similar or higher levels of the peptides containing Glu substitutions ( Figure 4B , compare lanes 5 , 7 , and 10 , Figure 4C ) . The Glu-containing peptides also induced reduced PARP cleavage in another epithelial cell type , HepG2 ( Figure 4—figure supplement 1A ) . Acute knockdown of DR5 reduced PARP cleavage during MPZ-ecto peptide expression , while exogenous FLAG-tagged DR5 expression restored PARP cleavage ( Figure 4—figure supplement 1B ) . Of note , depletion of DR5 resulted in detection of higher levels of the MPZ-ecto peptide , likely because cells with this protein-folding burden were not eliminated . Expressing comparable levels of the MPZ-ecto peptide and its variants ( using conditions of lanes 5 , 7 , and 10 in Figure 4B ) induced XBP1 mRNA splicing and transcription of CHOP and BiP mRNAs , indicating that the presence of these peptides perturb ER protein folding homeostasis to a similar degree ( Figure 4D , Figure 4E ) . Immunofluorescence showed that the MPZ-ecto peptide localized to the plasma membrane and within intracellular puncta that partially overlapped with ERGIC signal , although to a lesser extent than overexpressed full-length MPZ ( Figure 4F , Figure 4—figure supplement 2C ) . The Glu-containing mutant peptides were similarly distributed within cells with no significant difference in their average correlation with ERGIC signal ( Figure 4—figure supplement 2A–2B ) . DR5 , in all three conditions , also showed a positive correlation with the ERGIC marker ( Figure 4—figure supplement 2E ) . To determine if DR5 interacted with the MPZ-ecto peptide or its mutants , we immunoprecipitated the GFP-tagged peptides . Pulldown of the MPZ-ecto peptide enriched for DR5 relative to the Glu-containing mutant peptides ( Figure 4G , Figure 4—figure supplement 3A ) . Consistent with this specific enrichment of DR5 for the WT sequence , PARP cleavage and caspase activity measured in cell lysates were increased with the WT MPZ-ecto relative to the mutants ( Figure 4B , C ) . To confirm that the expression of MPZ-ecto peptide induces apoptotic cell death , we measured annexin V staining in the absence and presence of the pan-caspase inhibitor z-VAD ( Figure 4H , Figure 4—figure supplement 2C–2D ) . As expected , expressing the MPZ-ecto peptide increased annexin V staining relative to the empty vector but treatment with zVAD diminished the extent of annexin V staining ( Figure 4H ) . Importantly , cells expressing the Glu-containing mutant peptides exhibit decreased annexin V staining , demonstrating that DR5 binding of exposed polypeptides on misfolded protein is important for driving apoptosis .
Our data identify misfolded protein as the ER stress factor that switches upregulated DR5 from its inactive auto-inhibited dimer state to active multimeric clusters to initiate DISC assembly and apoptosis at the ER-Golgi intermediate complex . We have examined the mechanism of apoptosis induction by the sustained expression of three different candidate ER-trafficked proteins associated with CHOP-dependent disease pathologies: MPZ , RHO , and INS ( Pennuto et al . , 2008; Chiang et al . , 2016; Oyadomari et al . , 2002 ) . In epithelial cells , overexpression of each protein induces apoptosis in a DR5-dependent manner . Consistent with previous reports of ectopic CHOP expression in the absence of ER stress ( McCullough et al . , 2001; Han et al . , 2013; Southwood et al . , 2016 ) , CHOP-driven upregulation of DR5 alone did not account for the apoptosis observed during the overexpression of an ER-trafficked protein . For MPZ and RHO , the intracellular , misfolded pools of each protein physically associated with the DR5-caspase 8 complex . For proinsulin , which weakly associated with DR5 but triggered apoptosis to a similar extent , we believe it is likely that overexpression of this singular protein perturbed the folding of endogenous trafficking substrates and thereby provided other , perhaps more favored , misfolding substrates to directly engage DR5 . This latter scenario is likely to occur under pharmacologically induced ER stress as well . The interaction between misfolded protein and DR5 bridges the long-standing mechanistic gap of why CHOP expression ( and subsequent upregulation of its downstream factors ) is necessary but not sufficient to drive cell death . Through characterizing the interaction between the DR5 ECD and peptide sequences of ER-trafficked proteins , we demonstrate that DR5 promiscuously binds to exposed hydrophobic stretches of misfolded proteins with an affinity in the range of 100 μM and in a highly cooperative manner . To grasp how such a high concentration of misfolded protein could occur in the ERGIC , it is important to consider that the compartment is composed of vesicles and tubules measuring 60–100 nm in diameter and <500 nm in length . In a back-of-the-envelope calculation , we estimate that reaching 100 μM in a vesicle with a diameter of 100 nm would require only 32 molecules ( Sesso et al . , 1994; Fan et al . , 2003 ) . The measured affinities are therefore well within physiological range . Quantitative fluorescence microscopy of living COS7 cells has indicated up to 100 molecules of a GFP-tagged viral glycoprotein in a 100 nm vesicle ( Hirschberg et al . , 1998 ) , providing experimental evidence that surpassing concentrations of 100 μM is indeed physiologically relevant . In fact , the ‘low’ affinity between DR5 and misfolded proteins is likely a necessary feature that prevents aberrant DR5 oligomerization and activation in the crowded lumenal environment of membrane-bound compartments , as we previously established for other unfolded protein sensors , such as IRE1 ( Gardner et al . , 2013; Gardner and Walter , 2011; Karagöz et al . , 2017 ) . Given that misfolded receptors can be exported from the ER when quality control mechanisms are overwhelmed ( Satpute-Krishnan et al . , 2014; Sirkis et al . , 2017 ) , detection of misfolded proteins by DR5 downstream of the ER likely serves to prevent the cell from displaying or secreting dysfunctional proteins that would be detrimental in a multicellular context . While IRE1 and PERK act as initial UPR sensors in the ER , DR5 acts as a late sensor of misfolded protein at the ERGIC during unmitigated ER stress . Thus , intracellular DR5 triggers apoptosis to enforce a terminal quality control checkpoint for secretory and transmembrane proteins . We postulate that other members of the TNFR family , for exmple DR4 , which has been reported to play a role in cell death during Golgi stress ( van Raam et al . , 2017 ) , may respond similarly to intracellular stimuli . Although extensive research has focused on the therapeutic activation of death receptors including DR5 ( Ashkenazi , 2015 ) , limited strategies exist to inhibit such receptors despite their demonstrated role in apoptosis-mediated disease progression ( Vunnam et al . , 2017 ) . Namely , DR5-mediated apoptosis in hepatocytes has been linked to non-alcoholic fatty liver disease , while CHOP-dependent apoptosis in Schwann cells—wherein a role for DR5 has yet to be investigated—may contribute to diabetic peripheral neuropathies ( Cazanave et al . , 2011; Sato et al . , 2015 ) . Our finding that the assembly and disassembly of DR5 ECD oligomers can be controlled by a peptide raises the possibility that intracellular DR5 activation could be inhibited through small molecule ligand-induced dissociation of DR5 clusters to prevent apoptosis and thus preserve cell viability in the face of unresolved ER stress . From the work herein , this notion now emerges as a promising strategy to interfere therapeutically with deleterious death receptor function .
HCT116 cells ( ATCC CCL-247 ) and HepG2 cells ( ATCC CRL-10741 ) were cultured in DMEM with high glucose ( Sigma D5796 ) supplemented with 10% FBS ( Life technologies # 10082147 ) , 2 mM L-glutamine ( Sigma G2150 ) , 100 U penicillin , and 100 μg/mL streptomycin ( Sigma P0781 ) . Cells were incubated at 37°C , 5% CO2 for growth and transfections . All cell lines were authenticated by DNA fingerprint STR analysis by the American Type Culture Collection ( ATCC ) . All cell lines were visually inspected using DAPI DNA staining and tested negative for mycoplasma . Thapsigargin was purchased from Sigma and used at 100 nM in 0 . 1% DMSO unless otherwise indicated . Unpaired two-tailed t-tests for data sets were performed using GraphPad Prism 6 . 0 , where the variance between two data sets was non-significant , unless otherwise indicated . For each 6-well sample seeded with 400 , 000 cells the evening prior to transfection , the final transfection mixture put onto the cells was composed of 2 ml OptiMEM I ( Thermo Fisher Scientific #31985070 ) , 5 μl of Lipofectamine-LTX ( Life Technologies #15338100 ) , 1000 ng of total DNA ( supplemented with the empty vector in cases of variable MPZ-GFP ) . Plasmid preparations were preformed fresh for each transfection to maximize reproducibility . Unless otherwise noted , 1 . 0 μg of plasmid containing GFP-tagged ER-trafficked protein was used , while 0 . 25 μg of GFP supplemented with 0 . 75 μg of empty vector was sufficient to yield GFP protein levels in excess of ER-trafficked GFP fusions . To prepare the transfection mixture for one well of a 6-well plate , 5 μl of Lipofectamine-LTX and 1000 ng of plasmid were each diluted separately into 200 μl of OptiMEM I and then combined and incubated at RT for 15 min , as adapted from the manufacturer’s protocol . Growth media for each 6-well sample was replaced with 1 . 5 ml of OptiMEM I and the 400 μl transfection mixture was added dropwise to each well and incubated for 24 hr ( see Cell line culture conditions ) . For transfections in 15 cm dishes or 8-well ibidi slides , the transfection mixture was scaled to the number of cells plated ( i . e . 10 μg of plasmid for 4 million cells , or 62 . 5 ng of plasmid for 25 , 000 cells ) . For experiments shown in Figure 1 and Figure 2—figure supplement 1A , the siRNA oligonucleotides against DR5 and a non-targeting control were purchased from Dharmacon ( ON-TARGETplus Human TNFRSF10B 8795 siRNA # L-004448-00-0005 and ON-TARGETplus Non-targeting siRNA #2 # D-001810-02-05 ) . The siRNA transfection was performed as previously described in ( 2 ) using Lipofectamine RNAiMAX . For the knockdown in Figure 4—figure supplement 1B , we synthesized custom siRNAs from Dharmacon siRNA ( DR5siRNA-2: 5' AAG ACC CUU GUG CUC GUU GUC UU 3' , Nt siRNA: 5’ AAA CCU UGC CGA CGG UCU ACC UU 3’ ) . The siRNA transfection was performed 24 hr previous to the DR5L-FLAG and MPZ ecto-peptide-GFP plasmid co-transfection for a knockdown of 48 hr total . Cells were grown in 6-well plates and harvested with 0 . 5 ml of TRIzol reagent ( Life Technologies ) per manufacturer’s protocol to extract RNA . For cDNA synthesis , 500 ng of total RNA was reverse transcribed with 2 μl of the SuperScript VILO master mix ( Life Technologies # 11755050 ) in a total reaction volume of 10 μl following the manufacturer’s protocol for reaction temperature and duration . The reverse transcription product was diluted to 200 μl with 10 mM Tris-HCl pH 8 . 2 and used at 1:100 dilution for subsequent RT-PCR reactions . 2 μl of cDNA was added to 0 . 2 μM of forward and reverse primers ( Hs_XBP1_Fwd: 5’ -GGAGTTAAGACAGCGCTTGG- 3’; Hs_XBP1_Rev: 5’ -ACTGGGTCCAAGTTGTCCAG-3’ ) , 0 . 2 mM of each dNTP , 0 . 5 units of Taq DNA polymerase ( Thermo Scientific ) . The reaction was set at an annealing temperature of 60 . 5°C with an extension time of 30 s for 26 cycles . The products were then visualized on a 3% agarose gel ( comprised of a 1:1 mixture of low-melting point agarose and standard agarose ) stained by 1:10000 SybrSAFE . PCR samples were prepared as described by manufacturer’s protocols from iQ SYBR Green Supermix ( BioRad #17088800 ) . For each experiment , qRT-PCR reactions were set up in triplicate and run using a CFX96 Real Time System ( Bio-Rad ) . Quantitation cycles were determined with CFX Manager 3 . 0 software ( Bio-Rad ) and then normalized to GAPDH as an internal control . Media and cells for each sample were collected by cell scraping into cold PBS followed by a subsequent wash with 1 ml of cold PBS . The cell pellet was then resuspended in cold lysis buffer ( 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X100 , 1x Roche protease inhibitor ) and lysed via needle shearing on ice . Samples were centrifuged at 1000xg for 5 min , and the supernatants were mixed with sample buffer to a final concentration of 1% SDS , 62 . 5 mM Tris-HCl pH 6 . 8 , 10% glycerol , 0 . 1% bromophenol blue , 50 mM DTT . Samples were boiled at 95°C and loaded on SDS-PAGE gels ( GenScript ) . Samples were subsequently transferred onto nitrocellulose membranes , blocked with Odyssey buffer ( Licor ) for 1 hr at RT , and probed with primary antibodies diluted 1:1000 ( unless otherwise specified ) in Licor Odyssey buffer supplemented with 0 . 1% Tween 20% at 4°C overnight . Cells harvested from a 6-well plate were resuspended in 100 μl of lysis buffer ( 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X-100 ) and lysed via needle shearing ( 25G , 10 passes ) followed by a 30 min incubation on ice . Supernatant was collected after centrifuging at 1000xg for 5 min . For the caspase activity assay , 10 μl of supernatant diluted with lysis buffer ( 1:25 , 1:50 , 1:100 to stay within linear range of luminescence measurements ) was incubated with 10 μl of the luminogenic caspase glo 8 substrate ( Promega #PRG8200 ) in a 384-white walled plate ( Corning 3574 ) for 45 min at RT before measuring luminescence on a Spectramax-M5 plate reader . Cells seeded in a 6-well plate were transfected with the specified plasmid for 24 hr and harvested via trypsinization in complete media to allow a 30 min recovery at 37°C . The samples were then centrifuged at 500xg for 5 min and resuspended In 150 μl of a 1:20 dilution of Annexin V-AlexaFluor 647 conjugate ( Thermo Scientific #A23204 ) . Samples were transferred to a 96-well U-bottom plate and incubated in the dark for 15 min at RT . The distribution of apoptotic cells was determined by flow cytometry on a BD LSR II . Cells seeded in a 6-well plate were transfected with the specified plasmid for 24 hr and harvested via trypsinization in complete media and ultimately resuspended in 400 μl of media to obtain a concentration of 1 . 0–3 . 0 × 106 cells/ml . The cell suspension was then mixed 1:1 with 0 . 4% Trypan blue ( Sigma-Aldrich ) and incubated at 37°C for 30 min . The percentage of cells staining positive for was then quantified using the Countess II FL Automated Cell Counter ( ThermoFisher , catalog # AMQAF1000 ) default brightfield settings with disposable slides . Size-exclusion fractionation of cell lysate was performed as previously described in Lu et al . ( 2014 ) with minor modifications . In a 15 cm dish , 4 million HCT116 cells were seeded 18 hr prior to transfection ( see Transient transfections for protein expression ) . 24 hr post-transfection , cells were harvested by collecting all media ( to collect detached , dying cells ) and scraping the dish in 3 ml of cold PBS . The cell suspension was centrifuged at 500xg to pellet the sample . Cells were washed with an additional 1 ml of cold PBS , pelleted , and flash frozen in liquid N2 prior to lysis . Cells were resuspended in 600 μl of lysis buffer ( 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X-100 ) with 1x protease inhibitor ( Roche ) and lysed through a 25G needle with 10 passes . Lysates were clarified by centrifugation at maximum speed for 15 min at 4°C , and the supernatant ( 400 μl ) was loaded onto a SuperDex200 10/300 GL column equilibrated with lysis buffer at a flow rate of 0 . 35 ml/min . Fractions were collected in 0 . 5 ml aliquots . For the caspase 8 activity assay , 10 μl of each fraction was incubated with 10 μl of the caspase glo 8 substrate ( Promega #PRG8200 ) in a 384-well white walled plate ( Corning 3574 ) for 45 min at RT before measuring luminescence on a Spectramax-M5 plate reader . For Western blots , every three fractions were pooled for a total of 1 . 5 ml and subjected to TCA precipitation to concentrate the protein content . Samples were then analyzed by SDS-PAGE and immunoblotted for GFP and DR5 . IP for DR5 was performed as previously described in Lu et al . ( 2014 ) , using anti-DR5 mAb 5C7-conjugated agarose beads gifted by David Lawrence of the Ashkenazi lab at Genentech Inc . 15 cm dishes were seeded with 4 million HCT116 cells and allowed to recover for 18 hr prior to transfection . Samples were then transfected for 24 hr with the GFP-tagged ER trafficked protein , cytosolic GFP , or empty vector . To harvest apoptotic and living cells for each sample , all media and washes were collected . Cells were scraped in 2 ml of cold PBS and combined with the media and washes . Samples were centrifuged to pellet cells , washed with 1 ml cold PBS , and flash frozen in liquid N2 prior to lysis . The cell pellets were resuspended in 750 μl of 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X-100 with 1x Roche protease inhibitor cocktail ( if used subsequently for caspase glo 8 assay ) or with 5x Roche protease inhibitor cocktail ( if used for immunoblotting ) . The cells were lysed by mechanical shearing through a 25G needle for 13–15 plunges followed by incubation on ice for 30 min . The lysate was centrifuged at 2000xg for 5 min to remove debris and then incubated with 30 μl of GFP-Trap magnetic agarose beads ( Chromotek gtma-20 ) for 4 hr at 4°C . The beads were then washed with 750 μl of lysis buffer for 10 min at 4°C for three rounds . For measuring caspase activity , a fifth of the beads was resuspended in 40 μl of 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% Triton X-100 . 10 μl of the resuspended mixture was incubated with 10 ul of the caspase glo 8 luminogenic substrate ( Promega #PRG8200 ) for 45 min at RT before measuring luminescence . For Western blots , samples were eluted by adding 35 μl of non-reducing SDS gel loading buffer ( 50 mM Tris-HCl ( pH 6 . 8 ) , 2% SDS , 0 . 1% bromophenol blue , 10% glycerol ) and incubated for 10 min at 70°C . The beads were then magnetically removed from the eluted sample before adding DTT to a final concentration of 25 mM . The entire sample was loaded onto the gel and immunoblotted for DR5 , caspase 8 , and GFP , in that order . HCT116 were seeded at 25000 cells per well in an 8-well glass-bottom μSlide ( Ibidi 80827 ) pre-coated with Collagen IV ( Sigma-Aldrich C6745 , 0 . 03 mg/ml in PBS incubated for 30 min at RT , and then rinsed off with PBS x 3 ) in growth media 18 hr prior to transfection ( see above for protocol ) . 24 hr post-transfection , the cells were fixed using 4% paraformaldehyde ( PFA , Electron Microscopy Sciences ) or methanol , depending on the antibody combination used for immunostaining ( see table below ) . Samples were protected from light after fixation to preserve GFP fluorescence . Images were analyzed using CellProfiler ( Carpenter et al . , 2006 ) . From the 405 nm channel , cell nuclei were identified as primary objects using maximum correlation thresholding ( MCT ) and shape to distinguish clumped objects . Secondary objects were outlined using propagation from the nuclei ( DAPI ) after applying MCT from the 488 nm , 561 nm , and 633 nm channels . Tertiary objects for each cell was identified as the mask of each nucleus subtracted from the mask of its corresponding secondary object in each channel , yielding masks for intracellular GFP-tagged protein from the 488 nm channel , for intracellular DR5 from the 561 nm channel , and for the organelle marker in the 633 nm channel . For overlap between DR5 and MPZ/RHO-GFP , Pearson’s correlation coefficients were calculated within the MPZ/RHO mask for each cell to filter out untransfected cells . For overlap between DR5/MPZ and organelle markers , Pearson’s correlation coefficients were calculated within tertiary objects for the organelle marker in GFP+ cells . Statistical analyses for all data sets were performed using GraphPad Prism 6 . 0 . This subcellular fractionation protocol was adapted from Xu et al . ( 2015 ) . Cells transfected with MPZ-GFP were harvested from a 15 cm dish as described in the Size Exclusion Chromatography section above . The cell pellet was resuspended in 400 μl of homogenization buffer ( 10 mM triethanolamine-acetic acid , pH 7 . 4 , 0 . 25 M sucrose , 1 mM sodium EDTA , protease inhibitor cocktail ( Roche ) ) and lysed via mechanical shearing through a 25-gauge needle on a 1 ml syringe with 13 passes . ( Note on protease inhibitor: 4x protease inhibitor was used to lyse samples used for Western blotting to minimize degradation of proteins , while 1x protease inhibitor was used in sample lysis for caspase glo 8 assay so that excess inhibitor would not interfere with cleavage of luminescent substrate . ) Homogenized sample was then centrifuged at 2000xg for 15 min at 4°C to remove unlysed cells and the nuclear fraction . 250 μl of the supernatant was loaded via capillary action onto a 9–25% iodixanol gradient in a 13 × 51 mm polyclear centrifuge tube ( Seton Scientific # 7022–29426 ) prepared using the BioComp Gradient Master ( Model 107 ) . The 9% and 25% layers of the gradient were made by diluting 60% iodixanol ( OptiPrep Density Medium , Sigma Aldrich # D1556 ) into cell suspension medium ( 0 . 85% ( w/v ) NaCl , 10 mM Tricine-NaOH , pH 7 . 4 ) . The loaded gradients were then centrifuged in SW55Ti rotor at 44800 rpm for 2 hr at 4°C . After deceleration without braking , 300 ul fractions were collected from the top of the tube . Fractions from samples lysed with 4x protease inhibitor were subjected to TCA precipitation to concentrate protein content for immunoblotting . Fractions from samples lysed with 1x protease inhibitor were used for the caspase glo 8 assay ( 10 μl of each fraction + 10 μl of the caspase glo 8 reagent ( Promega #PRG8200 ) incubated for 45 min at RT prior to measuring luminescence ) . Peptide arrays were purchased from the MIT Biopolymers Laboratory as described previously in Karagöz et al . ( 2017 ) and Gardner and Walter ( 2011 ) . The sequences of each protein were tiled from N- to C-terminus in 18-amino-acid-long peptides shifting by three amino acids from the previous spot . The arrays were incubated in methanol for 10 min , and then washed in binding buffer ( 50 mM HEPES pH 7 . 2 , 250 mM NaCl , 10% glycerol ) for 10 min at room temperature x 3 . 500 nM of Fc-tagged DR5 extracellular domain ( ECD , a kind gift from Scot Marsters ) was incubated with the array in binding buffer at room temperature for 1 hr . Then , the array was washed three times for 10 min with binding buffer to remove unbound protein . The bound DR5 ECD was then transferred onto a nitrocellulose membrane via a semi-dry transfer apparatus ( Owl HEP-1 ) at 80 mA for 45 min at 4°C . The membrane was then blocked with 1xPBST with 5% milk for 1 hr at room temperature and probed with 1:1000 anti-Fc ( One World Lab #603–510 ) overnight at 4°C followed by 1:10000 anti-mouse IgG-HRP ( Promega #W4021 ) for 1 hr at RT . The membrane was imaged using chemi-luminescence with the Bio-Rad Universal Hood II Gel Doc below saturating pixel intensities . The signal of each spot was quantified using Max Quant and normalized to the maximum intensity of all the spots on the same array . Peptides were ordered from GenScript at >95% purity and stored with desiccant at −20°C . Peptides were dissolved as a highly concentrated stock solution in anhydrous DMSO and diluted 1:50 in aqueous buffer to measure the stock concentration using absorbance at 280 nm . All solution mixtures containing peptide and protein had a final concentration of 0 . 5% DMSO . Recombinant Fc-tagged DR5 and TNFR1 ECD proteins were generated and purified at Genentech Inc by S . Marsters . Fc-tagged proteins were incubated with Dynabeads Protein G ( 20 μg per 12 . 5 μl of beads for each sample ) in 1xPBS for 1 . 5 hr at RT . Beads were then washed with 1xPBS via magnetic pulldown , followed by two 10 min washes with 20 mM HEPES pH 7 . 2 , 100 mM KOAc , 0 . 2% Tween-20 at RT to remove unbound Fc-tagged protein . Protein bound-beads were incubated with 50 μl of 100 μM peptide in 20 mM HEPES pH 7 . 2 , 100 mM KOAc , 0 . 2% Tween-20 for 1 hr at RT . Beads were then washed with 50 μl of buffer three times , and samples were eluted with 25 μl of non-reducing sample buffer ( 62 . 5 mM Tris-HCl pH 6 . 8 , 2 . 5% SDS , 10% glycerol ) by incubating at 65°C for 15 min to elute the complex . Human DR5 ECD ( long isoform , residues 72–213: KRSSPSEGLCPPGHHISEDGRDCISCKYGQDYSTHWNDLLFCLRCTRCDSGEVELSPCTTTRNTVCQCEEGTFREEDSPEMCRKCRTGCPRGMVKVGDCTPWSDIECVHKESGTKHSGEVPAVEETVTSSPGTPASPCSLSG ) and TNFR1 ECD ( residues 22–211: IYPSGVIGLVPHLGDREKRDSVCPQGKYIHPQNNSICCTKCHKGTYLYNDCPGPGQDTDCRECESGSFTASENHLRHCLSCSKCRKEMGQVEISSCTVDRDTVCGCRKNQYRHYWSENLFQCFNCSLCLNGTVHLSCQEKQNTVCTCHAGFFLRENECVSCSNCKKSLECTKLCLPQIENVKGTEDSGTT ) were cloned into a pFastBac HTB vector containing a Gp67 ( glycoprotein ) signal peptide and 6xHis tag to force secretion of the expressed protein . The pFastBac HTB constructs were recombined into bacmid DNA using the Bac-to-Bac baculovirus expression system according to manufacturer’s protocols ( Life Technologies ) . Lab archivePlasmid descriptionVectorResistanceConstruct used in figure ( s ) :pPW3408Gp64-His6x-DR5 long ECD-FLAGpFastBacHTAmpR3D-3FpPW3410Gp64-His6x-DR5 long ECDpFastBacHTAmpR3CpPW3411Gp64-His6x-TNFR1 ECDpFastBacHTAmpR3C Table S6: Constructs used to generate bacmid for recombinant protein purification . SF21 were grown in SF-900 II media supplemented with 10% FBS at 28°C in disposable Erlenmeyer flasks rotating at 150 rpm and transfected using Cellfectin II ( Thermo Fisher Scientific ) according to manufacturer’s protocols . The baculovirus was amplified two more times at a low M . O . I . prior to infection of SF21 with a 1:50 dilution of the virus for protein expression . After 72–96 hr of infection , the SF21 suspension was centrifuged at 2000xg for 15 min ( two rounds ) to collect the media containing the secreted ECD protein . The media was then further clarified through a 0 . 2 um filter before loading onto a HisTrapFF column equilibrated with 25 mM imidazole pH 7 . 4 , 150 mM NaCl ( Buffer A ) at a flow rate of 3 . 0 ml/min . The column was washed with 20 CV of Buffer A at a flow rate of 4 . 0 ml/min . To elute , the concentration of imidazole was increased through a linear gradient from 0–100% of Buffer B ( 500 mM imidazole pH 7 . 4 , 150 mM NaCl ) in 7 CV . Fractions containing the His-tagged ECD were then concentrated and further purified on a SuperDex200 10/300 GL column ( GE Healthcare ) equilibrated with 30 mM HEPES pH 7 . 2 , 150 mM NaCl . For long-term storage , the protein was diluted into 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 10% glycerol and flash frozen in liquid N2 before storing at −80°C . Recombinant DR5 and TNFR1 ECD were labeled with AlexaFluor647 NHS Ester ( Succinimidyl Ester ) ( Life Technologies # A37573 ) in a 3:1 dye:protein molar ratio in 30 mM HEPES pH 7 . 2 , 150 mM NaCl , 1% DMF overnight at 4°C protected from light . The labeled proteins were re-loaded onto a SuperDex200 10/300 GL column to remove the excess dye , yielding labeling efficiencies of 59% and 49% for DR5 ECD and TNFR1 ECD , respectively . Labeled proteins were diluted into buffer with 10% glycerol and flash frozen in liquid N2 for long-term storage at −80°C . To set up reactions , the unlabeled peptide titration ( 10 μl each ) was made from a two-fold dilution series of the highest peptide concentration assayed . To each peptide sample , 10 μl of AlexaFluor647-labeled ECD protein was added to give a final concentration of 200 nM ECD protein in 20 mM HEPES pH 7 . 2 , 100 mM KOAc , 0 . 1% Tween-20 . Samples were incubated for 30 min at RT protected from light before measuring fluorescence . Samples were then loaded by capillary action onto premium capillaries ( Nanotemper Technologies , cat . #MO-K025 ) . Fluorescence was measured on a Monolith NT . 115 Instrument ( NanoTemper Technologies , Germany ) using the capillary scan function at 25°C . The half-maximum binding constant ( K1/2 ) and Hill coefficient were determined by fitting the data points on Prism 6 . 0 to the model equation: Y = Bmax*X^h/ ( K1/2^h + X^h ) , where Y is the % quenching , X is the concentration of peptide , Bmax is the maximum % quenching , and h is the Hill slope . To set up the peptide titration series , two-fold dilutions were made from the highest peptide concentration used into 20 mM HEPES pH 7 . 2 , 150 mM NaCl . An equal volume of C-terminal FLAG-tagged DR5 ECD was added to each peptide sample or buffer alone for a final concentration of 10 μM DR5 ECD . Reactions were equilibrated at RT for 30 min prior to the addition of 100 μM BS3 for 20 min at RT . Excess crosslinker was quenched by adding Tris-HCl pH 7 . 4 to a final concentration of 100 mM and incubated for 15 min at RT before analysis by SDS-PAGE . The resulting gel was then transferred onto a nitrocellulose membrane ( 120 V , 2 hr for high MW species ) and blotted with anti-FLAG to visualize discrete crosslinked species . The specified concentration of protein and peptide were incubated overnight at 4°C . The following day , 200 ul of each sample was loaded onto a SuperDex200 10/300 GL column equilibrated with 30 mM HEPES pH 7 . 2 , 150 mM NaCl at a flow rate of 0 . 5 ml/min . Fractions were collected in 1 ml aliquots . For subsequent analysis of each fraction by SDS-PAGE , 1 ml fractions were concentrated to 50 μl using a 3 kDa MW cutoff and loaded onto a gel . The fluorescein-labeled peptide was visualized using the fluorescence detection mode on a Typhoon 9400 Variable Mode Imager ( GE Healthcare ) . Protein was visualized by staining with Coomassie . | Proteins are chains of building blocks called amino acids , folded into a flexible 3D shape that is critical for its biological activity . This shape depends on many factors , but one is the chemistry of the amino acids . Because the internal and external environments of cells are mostly water-filled , correctly folded proteins often display so-called hydrophilic ( or ‘water-loving’ ) amino acids on their surface , while tucking hydrophobic ( or ‘water-hating’ ) amino acids on the inside . A compartment within the cell called the endoplasmic reticulum folds the proteins that are destined for the outside of the cell . It can handle a steady stream of protein chains , but a sudden increase in demand for production , or issues with the underlying machinery , can stress the endoplasmic reticulum and hinder protein folding . This is problematic because incorrectly folded proteins cannot work as they should and can be toxic to the cell that made them or even to other cells . Many cells handle this kind of stress by activating a failsafe alarm system called the unfolded protein response . It detects the presence of incorrectly shaped proteins and sends signals that try to protect the cell and restore protein folding to normal . If that fails within a certain period of time , it switches to signals that tell the cell to safely self-destruct . That switch , from protection to self-destruction , involves a protein called death receptor 5 , or DR5 for short . DR5 typically triggers the cell’s self-destruct program by forming molecular clusters at the cell’s surface , in response to a signal it receives from the exterior . During a failed unfolded protein response , DR5 seems instead to act in response to signals from inside the cell , but it was not clear how this works . To find out , Lam et al . stressed the endoplasmic reticulum in human cells by forcing it to fold a lot of proteins . This revealed that DR5 sticks to misfolded proteins when they leave the endoplasmic reticulum . In response , DR5 molecules form clusters that trigger the cell's self-destruct program . DR5 directly recognized hydrophobic amino acids on the misfolded protein’s surface that would normally be hidden inside . When Lam et al . edited these hydrophobic regions to become hydrophilic , the DR5 molecules could no longer detect them as well . This stopped the cells from dying so easily when they were under stress . It seems that DR5 decides the fate of the cell by detecting proteins that were incorrectly folded in the endoplasmic reticulum . Problems with protein folding occur in many human diseases , including metabolic conditions , cancer and degenerative brain disorders . Future work could reveal whether controlling the activation of DR5 could help to influence if and when cells die . The next step is to understand how DR5 interacts with incorrectly folded proteins at the atomic level . This could aid the design of drugs that specifically target such receptors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2020 | Misfolded proteins bind and activate death receptor 5 to trigger apoptosis during unresolved endoplasmic reticulum stress |
Spatial organization of the transcriptome has emerged as a powerful means for regulating the post-transcriptional fate of RNA in eukaryotes; however , whether prokaryotes use RNA spatial organization as a mechanism for post-transcriptional regulation remains unclear . Here we used super-resolution microscopy to image the E . coli transcriptome and observed a genome-wide spatial organization of RNA: mRNAs encoding inner-membrane proteins are enriched at the membrane , whereas mRNAs encoding outer-membrane , cytoplasmic and periplasmic proteins are distributed throughout the cytoplasm . Membrane enrichment is caused by co-translational insertion of signal peptides recognized by the signal-recognition particle . Time-resolved RNA-sequencing revealed that degradation rates of inner-membrane-protein mRNAs are on average greater that those of the other mRNAs and that this selective destabilization of inner-membrane-protein mRNAs is abolished by dissociating the RNA degradosome from the membrane . Together , these results demonstrate that the bacterial transcriptome is spatially organized and suggest that this organization shapes the post-transcriptional dynamics of mRNAs .
In eukaryotic systems , the spatial organization of the transcriptome plays a fundamental role in regulating the post-transcriptional fate of RNA . Such organization leads to spatially localized translation and degradation of mRNAs , which are essential for a diverse set of biological behaviors including cell motility , cellular polarization , and stress response ( Balagopal and Parker , 2009; Buxbaum et al . , 2015; Holt and Schuman , 2013; Martin and Ephrussi , 2009 ) . By contrast , spatial localization has not been considered to play a significant role in the post-transcriptional dynamics of bacterial mRNAs . Early measurements of the dynamics of a handful of synthetic mRNAs within bacterial cells suggest that mRNAs are more or less uniformly distributed inside the cells . Single fluorescently labeled synthetic mRNAs have been observed to diffuse freely throughout the cytoplasm in E . coli ( Golding and Cox , 2004; 2006 ) , and the local fluorescent signals from labeled mRNAs appear to fluctuate in time in a manner consistent with free diffusion ( Le et al . , 2005; Valencia-Burton et al . , 2009 ) . However , recent evidence has begun to reveal that some native mRNAs do not diffuse freely throughout the cell but are rather localized to specific cellular compartments ( Campos and Jacobs-Wagner , 2013; Nevo-Dinur et al . , 2012 ) . Different spatial patterns of mRNAs have been identified . In one of the patterns , mRNAs have been observed to reside in the vicinity of the DNA loci from which they were transcribed . This pattern was observed for groESL , creS , divJ , ompA , and fljK transcripts in C . crescentus and the lacZ transcript in E . coli ( Montero Llopis et al . , 2010 ) . Similarly , the average distributions of the lacI mRNA in E . coli cells appear to show enrichment in a cellular region at which the lacI chromosome loci is also enriched ( Kuhlman and Cox , 2012 ) . A second , distinct pattern has been observed in which mRNAs do not reside near the DNA loci from which they are transcribed , but instead reside in the cellular compartment where their encoded proteins are localized . For example , the E . coli bglGFB , lacY , and ptsG mRNAs , which encode the inner-membrane proteins BglF , LacY , and PtsG , have been found enriched near the cell membrane ( Fei et al . , 2015; Nevo-Dinur et al . , 2011 ) ; the bglG fragment of the E . coli bglGFB transcript and the B . subtilis comE transcript , which encode the polar localized BglG and ComEC proteins , respectively , have been observed enriched at the cell pole or at the septa of sporulating cells ( Nevo-Dinur et al . , 2011; dos Santos et al . , 2012 ) . In addition to these patterns , higher order mRNA structures have also be suggested , such as a helical RNA distribution near the cell membrane ( Valencia-Burton et al . , 2009 ) . Because of the disparate spatial patterns that have been observed previously and the relatively small number of RNAs that have been investigated , it remains unclear how mRNAs are spatially organized inside bacterial cells and whether any of the observed spatial organizations is a genome-wide property or a special property of a small number of genes . The molecular mechanisms responsible for mRNA localization in bacterial also remain incompletely understood . Several lines of biochemical evidence have revealed that mRNAs encoding inner-membrane proteins are , in part , translated at the membrane by the co-translational insertion of inner-membrane proteins into the membrane ( Driessen and Nouwen , 2008 ) . Thus , these mRNAs should spend at least a portion of their lifetimes near the cell membrane . However , co-translational insertion has yet to be linked to any of the reported RNA localization patterns . Instead , a recent study has suggested a translation-independent membrane localization mechanism for RNA based on the observation that inhibition of translation of the bglF transcript does not disrupt its membrane localization ( Nevo-Dinur et al . , 2011 ) . This observation has led to the suggestion that an RNA zip-code in combination with unknown zip-code-binding proteins directs this bacterial RNA localization ( Nevo-Dinur et al . , 2011 ) , similar to the established RNA localization mechanisms in eukaryotes ( Buxbaum et al . , 2015 ) . However , the proteins responsible for identifying this putative zip-code have not been described in bacteria nor has this localization mechanism been extended to any other bacterial mRNAs . Similarly , no mechanism is known for the retention of mRNA near the chromosomal locus from which they were transcribed . Finally , it remains unknown what physiological consequences spatial organization might have on the post-transcriptional dynamics of mRNAs in bacterial cells . Interestingly , translation and RNA processing enzymes are not uniformly distributed in bacteria . For example , ribosomal proteins tend to be excluded from the nucleoid and enriched in the cell periphery and cell poles in both E . coli and B . subtilis ( Bakshi et al . , 2015; Robinow and Kellenberger , 1994 ) . Core components of the RNA degradation machinery have been found enriched at the cell membrane in E . coli ( Mackie , 2012 ) and B . subtilis ( Lehnik-Habrink et al . , 2011 ) , and in the nucleoid in C . crescentus ( Montero Llopis et al . , 2010 ) . Even components of the trans-translation pathway , a pathway responsible for the resolution of defective transcripts , appear to cycle between the cytoplasm and membrane as a function of cell cycle in C . crescentus ( Russell and Keiler , 2009 ) . Such non-uniform distributions of RNA-interacting proteins give rise to the possibility that spatial organization may play an important role in shaping the post-transcriptional dynamics of mRNAs , if the mRNAs are themselves not uniformly distributed . However , no evidence has been described yet for the role of spatial organization in the post-transcriptional fate of bacterial mRNAs . Here we probe the presence , mechanism , and physiological consequences of the spatial organization of mRNAs in E . coli at the transcriptome scale . We developed a method to directly image the spatial organization of large but defined fractions of the transcriptome , and our measurements revealed transcriptome-scale spatial organizations of mRNAs that depended on the cellular locations of their targeted proteins: mRNAs encoding inner-membrane proteins were found enriched at the membrane whereas mRNAs encoding cytoplasmic , periplasmic and outer-membrane proteins were found relatively uniformly distributed throughout the cytoplasm . Genomic organization , on the other hand , did not appear to play a major role in the organization of the transcriptome in E . coli . We further demonstrated that co-translational insertion of signal peptides recognized by the signal-recognition-particle ( SRP ) was responsible for this membrane localization of inner-membrane-protein mRNAs . To explore the physiological consequences of this transcriptome-scale organization , we used time-resolved next-generation sequencing to measure mRNA lifetimes and found that the mRNAs encoding inner-membrane proteins were selectively destabilized compared to mRNAs encoding outer-membrane , cytoplasmic and periplasmic proteins . Finally , to elucidate potential mechanisms for this selective destabilization , we imaged the distribution of all of the enzymes associated with RNA processing in E . coli and observed that members of the RNA degradosome are enriched on the membrane . A genetic perturbation that removed these enzymes from the membrane preferentially stabilized mRNAs encoding inner-membrane proteins , suggesting that their physical proximity to the membrane-bound RNA degradosomes may be responsible for the native destabilization of these mRNAs .
To enable the direct measurement of the spatial distribution of mRNAs at the transcriptome scale , we developed a method that allows us to directly measure the spatial distribution of large but defined portions of the transcriptome . In particular , by selectively staining large mRNA groups which share common properties , we not only avoided inference of transcriptome-scale organization from measurements of only a few mRNAs but also directly tested the role of general mRNA properties on spatial organization ( Figure 1A ) . 10 . 7554/eLife . 13065 . 003Figure 1 . The E . coli transcriptome is spatially organized with inner-membrane-protein mRNAs enriched at the membrane . ( A ) A scheme illustrating fluorescent labeling and imaging of large but defined populations of mRNAs simultaneously instead of imaging one mRNA species at a time . ( B ) The required complex FISH probe sets are generated via enzymatic amplification of array-derived custom oligonucleotide pools containing tens of thousands of unique sequences . Subsets of these oligopools are selected via PCR , amplified and converted into RNA via in vitro transcription , converted back into DNA via reverse transcription with a fluorescently labeled primer . The RNA templates are removed by alkaline hydrolysis . I1 and I2 represent PCR primers unique for each probe set . RTP represents a reverse transcription primer common to all probe sets . TR ( targeting region ) represents the portion of the oligo complementary to one of the RNAs of interest . ( C ) Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of example fixed E . coli cells stained with FISH probes against all mRNAs encoding inner-membrane proteins that are in the abundance range of 3–30 copies per cell . The STORM images of the middle section ( 300-nm thick ) of cells are shown here . 3D-STORM images of the entire cells as well as images of mRNAs in other abundance ranges are shown in Figure 1—figure supplement 1 . ( D ) Average short-axis ( left ) and long axis ( right ) cross-section images of inner-membrane-protein mRNAs derived from 611 cells computationally normalized to a common width and a common length and then aligned . ( E ) Density profile of inner-membrane-protein mRNAs constructed from the middle slice ( 150 nm ) of the average long-axis cross-section image shown in ( D , right ) . The x-axis is normalized to the radius of the cell . ( F , G ) Same as ( C , D ) but for mRNAs encoding cytoplasmic proteins in the abundance range of 3–30 per cell and the average cross-section images were derived from 319 individual cells . ( H ) Same as ( E ) but for mRNAs encoding cytoplasmic proteins ( red ) , periplasmic proteins ( purple ) , and outer-membrane proteins ( cyan ) . The cytoplasmic , periplasmic and outer-membrane-protein distributions were derived from 319 , 338 and 194 cells , respectively . Scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 00310 . 7554/eLife . 13065 . 004Figure 1—figure supplement 1 . Spatial organization of mRNAs that encode proteins residing in different cellular locations and are in different RNA abundance ranges . ( A ) mRNAs encoding inner membrane proteins . Top left panels: Stacked phase contrast ( gray ) and 3D-STORM images ( color ) of example E . coli cells stained with FISH probes against all mRNAs predicted to encode inner membrane proteins , grouped into two expression ranges: 1/3–3 copies per cell and 3–30 copies per cell . The z-positions of the molecules are color-coded according to the colored z scale . Bottom left panels: Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of the middle 300-nm section of the cells depicted in the top panels . Less than 10 inner-membrane-protein mRNA species are expressed in the 30–300 copies per cell range , and hence are not shown here . Top right panel: Average long-axis cross-section image of mRNA encoding inner-membrane proteins in the concentration range of 3–30 copies per cell . Bottom right panel: Density profile for all mRNAs that encode cytoplasmic proteins in the 1/3–3 copies per cell ( blue ) and 3–30 copies per cell ( green ) abundance ranges . Density profile is as defined in Figure 1E . The average density and axial distributions were derived from measurements of tens to hundreds of cells . 247 and 27 mRNA species were stained as part of the 1/3–3 and 3–30 copies per cell groups , respectively . ( B ) Same as ( A ) but for mRNAs encoding cytoplasmic proteins . Here the mRNAs are grouped into three expression ranges: 1/3–3 copies per cell , 3–30 copies per cell , and 30–300 copies per cell . The average long-axis cross-section image of mRNA is for the 30–300 copies per cell group , and the density profiles are for all mRNAs that encode cytoplasmic proteins in the 1/3–3 copies per cell ( blue ) , 3–30 copies per cell ( green ) , and 30–300 copies per cell ( cyan ) abundance ranges . 620 , 105 , and 62 mRNA species were stained as part of the 1/3–3 , 3–30 , and 30–300 copies per cell groups , respectively . ( C ) Same as ( A ) and ( B ) but for mRNAs encoding periplasm proteins . Only mRNAs in the expression range of 1/3–3 copies per cell are shown . Less than 10 periplasmic mRNA species are expressed in either the 3–30 or 30–300 copies per cell range . 44 mRNA species were stained within this group . ( D ) Same as ( C ) but for mRNAs encoding outer-membrane proteins . 23 mRNA species were stained within this group . ( E ) Stacked phase contrast and 3D-STORM images for cells stained with anti-sense control probes derived from cytoplasmic-protein mRNAs ( left ) or inner-membrane-protein mRNAs ( right ) , both in the 3–30 copies per cell abundance range . These antisense probes have the same number of unique probes as the sense probes used in ( A ) and ( B ) but have targeting regions that are the reverse complement of the original targeting regions and , thus , have limited targets within the cell . ( F ) The average number of single-molecule localizations per cell observed for all depicted groups shown in ( A–E ) . Error bars represent SEM . The numbers in the label depict the copy number range . Scale bars: 2 µmDOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 00410 . 7554/eLife . 13065 . 005Figure 1—figure supplement 2 . Being polycistronic with an inner-membrane-protein message can confer partial membrane enrichment to an mRNA . ( A ) Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of example E . coli cells stained with FISH probes to all mRNAs encoding cytoplasmic proteins in the 3–30 copies per cell abundance range . The mRNAs are subdivided into two groups: those not polycistronic ( NP ) to mRNAs encoding inner-membrane proteins ( left ) and those polycistronic ( P ) to mRNAs encoding inner-membrane proteins ( right ) . ( B ) Average long-axis cross-section images for the mRNAs . Top panel: for mRNAs not polycistronic to mRNAs encoding inner-membrane proteins . Bottom panel: for mRNAs polycistronic to mRNAs encoding inner-membrane proteins . ( C ) Density profiles for mRNAs not polycistronic to mRNAs encoding inner-membrane proteins ( red ) and for mRNAs polycistronic to mRNAs encoding inner-membrane proteins ( blue ) . Density profile is as defined in Figure 1E . ( B ) and ( C ) are average cell images and density profiles derived from hundreds of cells in each case . ( D–F ) same as ( A–C ) but for mRNAs encoding periplasmic proteins . Scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 005 Our method is based on single-molecule fluorescence in-situ hybridization ( FISH ) ( Femino et al . , 1998; Raj et al . , 2008 ) . A challenge to simultaneously imaging large populations of mRNAs is the generation of complex FISH probe sets for labeling such RNA populations , each comprising hundreds to thousands of unique oligonucleotide probes . To overcome this challenge , we used an Oligopaint-based approach and took advantage of the ability of array-based synthesis to generate complex oligonucleotide libraries ( Beliveau et al . , 2012 ) . Specifically , we designed libraries comprising sequences that target the desired RNAs flanked by primers that allow the selection , enzymatic amplification , and fluorescent tagging of defined subsets of these libraries , each of which target a specific group of mRNAs ( Figure 1B ) . We then amplified these oligonucleotide templates to generate complex but defined sets of FISH probes by a high-throughput enzymatic amplification method ( Chen et al . , 2015b; Murgha et al . , 2014 ) . We first used this approach to test whether the transcriptome of E . coli is organized based on the intracellular locations of the encoded proteins . To this end , we designed several FISH probe sets , each targeting a specific population of mRNAs whose encoded proteins reside within one of the four cellular compartments in E . coli: cytoplasm , inner membrane , periplasm , and outer membrane . To control for the large differences in mRNA abundance within a group , we further sub-divided each group by mRNA abundance . Within the final groups , no single mRNA species was predicted to produce more than ~10% of the signal from the imaged group . We fixed and labeled E . coli cells with these probes , and imaged the mRNA distributions using three-dimensional stochastic optical reconstruction microscopy ( 3D-STORM ) ( Huang et al . , 2008; Rust et al . , 2006 ) . The mRNA distributions showed a clear distinction between different mRNA groups: mRNAs that encode inner-membrane proteins were strongly enriched at the membrane ( Figure 1C–E and Figure 1—figure supplement 1A ) whereas mRNAs encoding cytoplasmic proteins were more or less uniformly distributed throughout the cytoplasm ( Figure 1F–H and Figure 1—figure supplement 1B ) , except for some cases where we observed a moderate depletion of mRNAs from the nucleoid ( Figure 1—figure supplement 1B ) . This difference is evident not only in the distributions of mRNAs in individual cells ( Figure 1C , F ) but also in the average mRNA distributions over several hundred imaged cells after normalization of the cell dimensions ( Figure 1D , E , G , H ) . These different spatial distributions did not depend on the abundance range of the stained mRNAs ( Figure 1—figure supplement 1A , B ) . Notably , mRNAs that encode periplasmic proteins and outer-membrane proteins , which reside within nanometers of inner-membrane proteins , did not show a strong enrichment at the membrane . Instead , these mRNAs were found distributed more or less throughout the cytosol ( Figure 1H and Figure 1—figure supplement 1C , D ) , like those mRNAs that encode cytoplasmic proteins . Interestingly , among these latter groups , the subset of mRNAs that were polycistronic with inner-membrane-protein mRNAs also exhibited membrane enrichment ( Figure 1—figure supplement 2 ) , which explains the slight membrane enrichment observed in the mRNA populations encoding periplasm and outer-membrane proteins ( Figure 1—figure supplement 1C , D ) . For all groups , the number of RNA localizations that we detected using mRNA-targeting probes was much larger ( ~10–100 fold ) than that detected when using anti-sense probes with reverse compliment sequences ( Figure 1—figure supplement 1E , F ) , indicating highly specific labeling . The bacterial genome is spatially organized with defined genomic loci occupying defined locations within the cell as a function of the division cycle ( Wang et al . , 2013 ) . To test whether this genomic organization plays a role in the spatial organization of the transcriptome , we constructed multiple FISH probe sets , each labeling the specific population of mRNAs that are transcribed from one of twenty different 100-kb chromosomal regions ( Figure 2 ) . Because such 100-kb regions occupy small volumes within the nucleoid ( Wang et al . , 2013 ) , we would expect each mRNA probe set to produce one or a few bright fluorescent foci ( one for each copy of the chromosome ) within cells if these mRNAs resided near their corresponding DNA loci . Instead , the majority of these mRNA populations were uniformly distributed throughout the cytoplasm ( Figure 2 ) . Interestingly , a subset of these mRNA groups showed some membrane enrichment , and these groups were enriched for mRNAs that encode inner-membrane proteins . 10 . 7554/eLife . 13065 . 006Figure 2 . Genomic organization does not play a major role in the organization of the E . coli transcriptome . Stacked phase contrast ( gray ) and STORM cross-section images ( color ) for example fixed E . coli stained for all mRNAs transcribed from discrete 100-kb genomic loci in the abundance range of 1/3–3 and/or 3–30 copies per cell . The label marks the genomic region and abundance range studied in each case . Several cases show significant membrane enrichment , and these cases correspond to loci enriched in mRNAs that encode inner-membrane proteins . Scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 006 The substantial fraction of the transcriptome probed in these measurements suggests the generality of our observed spatial patterns . In our study of the relationship between the spatial organization of the mRNAs and their encoded proteins , we stained 27% of all E . coli mRNAs and 76% of those actually expressed under the probed growth condition ( >0 . 3 copies per cell ) . Similarly , in the study of the role of locations of the genomic loci on the spatial organization of their encoded mRNAs , we probed mRNAs transcribed from nearly half of the E . coli chromosome . Thus , we conclude that the patterns we observe here reflect the general behavior of E . coli mRNAs: mRNAs that encode inner-membrane proteins are enriched near the membrane whereas mRNAs that encode proteins that reside in the cytoplasm , periplasm and outer-membrane tend to be distributed throughout the cytoplasm , although we cannot rule out the possibility that there are exceptions to these general behaviors for some mRNAs . By contrast , the spatial organization of the genome does not appear to play a major role in shaping the spatial organization of the E . coli transcriptome . Next we investigated the mechanism responsible for establishing this general spatial organization of the E . coli transcriptome . Given the correlation observed between the spatial organization of the mRNAs and their encoded proteins , we reasoned that the pathways involved in directing proteins to different cellular locations might be involved in establishing the spatial distribution of mRNAs . In bacteria , there are two major pathways responsible for protein targeting ( Driessen and Nouwen , 2008 ) : the signal recognition particle ( SRP ) -dependent pathway and the SecB-dependent pathway . Most inner-membrane proteins use the SRP pathway , whereas most outer-membrane and periplasmic proteins use the SecB pathway . The SRP pathway is believed to co-translationally insert proteins into the membrane; therefore , mRNAs subject to this pathway would be translated , in part , at the membrane . By contrast , proteins destined for the SecB pathway are translated in the cytosol . Thus , the co-translational membrane insertion of the inner-membrane proteins via the SRP pathway would provide a simple explanation for the membrane enrichment observed for mRNAs encoding these proteins . However , this mechanism has not been linked to the previously observed mRNA distribution patterns and , for the one E . coli mRNA ( bglF ) whose membrane localization mechanism has been probed , it has instead been suggested that its membrane localization is translation independent and , thus , cannot be established by SRP-dependent co-translational insertion ( Nevo-Dinur et al . , 2011 ) . Therefore , a critical test of the mRNA localization mechanism is needed . The choice of SRP or SecB pathway is dictated by signal peptides near the N-terminus of the protein ( Driessen and Nouwen , 2008 ) . Thus , to test the role of co-translational insertion in the spatial localization of mRNAs , we created a series of fusion constructs between a test mRNA that encodes the fluorescent protein mMaple3 ( Wang et al . , 2014 ) and native signal-peptide sequences derived from different E . coli proteins ( Figure 3A ) . We inserted these fusion genes into the chromosome and measured the spatial distribution of their mRNAs using FISH labeling and 3D-STORM imaging . Fusion to different signal-peptide sequences clearly directed the fusion mRNAs to different locations in the cell: mRNAs that were fused to the SRP signal sequences derived from the inner-membrane proteins FhuB , CcmH , and AcrB ( Huber et al . , 2005 ) were almost exclusively localized at the membrane ( Figure 3B–D ) , whereas those mRNAs that were fused to the SecB signal sequences derived from the periplasmic proteins GlpQ , LivJ , PhoA , and MalE ( Huber et al . , 2005 ) were distributed throughout the cytoplasm ( Figure 3E , F ) . As further evidence that the membrane localization of mRNAs is driven by the SRP targeting pathway , the signal-peptide sequence derived from TolB , one of the rare periplasmic proteins that uses the SRP pathway ( Huber et al . , 2005 ) , also directed the mMaple3 mRNA to the membrane ( Figure 3C ) . To determine if translation of the signal peptide is required for the membrane enrichment induced by the SRP signal sequences , we removed the start codon from the fusion mRNAs and found that this perturbation removed these mRNAs from the membrane ( Figure 3C , D ) . Thus , translation of the N-terminal signal peptide that target proteins to the SRP pathway is required for directing these mRNAs to the membrane . 10 . 7554/eLife . 13065 . 007Figure 3 . SRP-dependent co-translational insertion of signal peptides plays a major role in the membrane localization of inner-membrane-protein mRNAs . ( A ) Fusion constructs between different signal peptides and mMaple3 . ( B ) Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of example E . coli cells expressing mMaple3 fused to the signal peptide from an SRP-dependent protein FhuB . The cells were stained with FISH probes against mMaple3 . ( C ) Left: Average long-axis cross-section images of cells expressing mMaple3 fused to SRP-dependent signal peptides derived from FhuB , CcmH , AcrB , and TolB . Right: Average long-axis cross-section images of cells expressing mMaple3 fusions to AcrB and TolB signal peptides without the start codon ( -AUG ) . ( D ) Density profiles derived from the average long-axis cross-section images of mMaple3 fusions to the AcrB signal peptide with ( red ) and without ( blue ) the start codon . Density profile is as defined in Figure 1E . ( E ) Stacked phase contrast and STORM cross-section images of example E . coli cells expressing mMaple3 fused to a signal peptide derived from a SecB-dependent protein GlpQ . The cells were stained with FISH probes against mMaple3 . ( F ) Average long-axis cross-section images of cells expressing mMaple3 fused to SecB-dependent signal peptides derived from GlpQ , LivJ , PhoA and MalE . ( G ) Stacked phase contrast and STORM cross-section images for example E . coli cells treated with the translation-initiation-inhibitor kasugamycin . The cells were stained with the FISH probe set against inner-membrane-protein mRNAs in the abundance range of 3–30 copies per cell . ( H ) Average long-axis cross-section images of cells in the presence ( +Kas ) and absence ( -Kas ) of kasugamycin . The cells were stained with the FISH probes against inner-membrane-protein mRNAs in the abundance range of 3–30 copies per cell . Average long-axis cross-section images in C , F and H were derived from all measured cells , tens to hundreds of cells in each case . Scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 00710 . 7554/eLife . 13065 . 008Figure 3—figure supplement 1 . mRNA for the inner-membrane protein BglF is enriched at the membrane in a translation-dependent fashion . ( A ) Cartoon depiction of an mRNA expressing a fusion between full length , wild-type bglF and mMaple3 . The green hashed region represents a putative SRP signal , i . e . the first transmembrane domains in bglF . ( B ) Left panel: Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of example E . coli cells stained with FISH probes to mMaple3 . Right panel: Average long-axis cross section images derived from hundreds of cells . ( C ) Density profile for the cross section in ( B ) . Density profile is as defined in Figure 1E . ( D–F ) As in ( A–C ) but for bglF construct in which the start codon has been replaced by a stop codon ( TAA ) . ( G–I ) As in ( A–C ) but for a bglF construct in which the codon at position 202 has been replaced with a stop codon ( TAA ) . The resulting construct expresses only the portion of bglF upstream of the putative SRP signal sequence . This blgF derivative is identical to a previously published derivative that was reported to be membrane enriched ( Nevo-Dinur et al . , 2011 ) . All constructs were expressed from the AttB site on the chromosome . Scale bars: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 008 To test the translation-dependence of the membrane localization of mRNAs at the transcriptome scale , we used kasugamycin , a translation-initiation inhibitor , to release the cellular pool of mRNAs from ribosomes ( Schluenzen et al . , 2006 ) and re-measured the spatial distribution of the endogenous mRNAs encoding inner-membrane proteins . The membrane enrichment observed for these mRNAs was abolished by the kasugamycin treatment ( Figure 3G , H ) , indicating that this transcriptome-scale pattern is translation dependent . Taken together , our results suggest that co-translational insertion of the inner-membrane proteins mediated by the SRP plays the major role in the membrane localization of mRNAs encoding these proteins . Because our results contrast the translation-independent membrane-targeting mechanism previously proposed for bglF ( Nevo-Dinur et al . , 2011 ) , we re-examined the localization mechanism of bglF . bglF is an inner-membrane protein and , as such , contains a SRP signal peptide in the N-terminal region of the protein ( Figure 3—figure supplement 1A ) , and the nucleotide sequence encoding this signal peptide overlaps with the RNA region previously proposed to direct this transcript to the membrane via a translation-independent mechanism ( Nevo-Dinur et al . , 2011 ) . To test the role of translation of this signal region on the localization of bglF mRNA , we created several fusion constructs between bglF derivatives and mMaple3 that are either translationally fully competent or translationally inhibited by insertion of stop codons before the SRP signal . Translationally competent bglF mRNAs were found enriched at the membrane ( Figure 3—figure supplement 1A–C ) , as expected; however , constructs with stop-codon insertions that disrupted the translation of the SRP signal region , but which still contain the RNA sequence that encodes it , were no longer enriched at the membrane ( Figure 3—figure supplement 1D–I ) . Thus , our results suggest that translation of the SRP signal sequence is required for bglF membrane localization , consistent with the transcriptome-wide mechanisms described above . We next asked if this spatial organization has any physiological consequences on the post-transcriptional dynamics of E . coli mRNAs . To address this question , we used time-resolved RNA sequencing ( Chen et al . , 2015a; Geisberg et al . , 2014; Kristoffersen et al . , 2012; Munchel et al . , 2011; Rabani et al . , 2011 ) to simultaneously measure the degradation kinetics of all mRNA species in E . coli — a technique that we refer to as τ-seq hereafter . Briefly , we inhibited transcription initiation with the antibiotic rifampicin and then measured RNA abundance with RNA-seq at various time points after this treatment . After an initial period of delay determined by the time required to complete the transcription started prior to rifampicin addition ( Chen et al . , 2015a ) , the abundance of each individual mRNA species decays exponentially to a stable baseline ( Figure 4—figure supplement 1A ) . From these decay curves , we extracted the half-life for each mRNA using a simple model for RNA decay ( Materials and methods ) . Both the decay curves and the extracted half-lives were highly reproducible between biological replicates ( Figure 4—figure supplement 1A–D ) . To identify potential effects of localization on the lifetime of mRNAs , we sorted the measured half-lives of mRNAs into four groups based on the predicted locations of the encoded protein ( cytoplasmic , periplasmic , inner-membrane and outer-membrane ) . Within each group , we observed significant variation between half-lives for individual mRNAs , as previously observed ( Bernstein et al . , 2002; Chen et al . , 2015a; Selinger et al . , 2003 ) . Despite this spread , mRNAs that encode cytoplasmic , periplasmic , or outer-membrane proteins had half-lives that were similarly distributed and the lifetime distributions of these three groups were statistically indistinguishable according to a two-sided Kolmogrov-Smirnov test ( Figure 4A , B ) . By contrast , mRNAs that encode inner-membrane proteins were degraded substantially more rapidly , on average , than the other three groups of mRNAs , exhibiting a statistically significantly different lifetime distribution ( Figure 4A , B ) . 10 . 7554/eLife . 13065 . 009Figure 4 . Inner-membrane-protein mRNAs are preferentially destabilized relative to mRNAs encoding cytoplasmic , periplasmic , and outer-membrane proteins . ( A ) Scatter plot ( grey symbols ) of the half-lives of individual E . coli mRNA species grouped based on the predicted locations of the proteins that they encode . Each data point represents one mRNA species . Blue colored shapes represent the probability distributions for these data points . ( B ) Average half-lives of the mRNA groups depicted in ( A ) . The p-value was determined with a two-sided Kolmogrov-Smirnov test . ( C , D ) Same as ( A , B ) but for cells after treatment with kasugamycin . See Figure 4—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 00910 . 7554/eLife . 13065 . 010Figure 4—source data 1 . RNA abundance measurements versus time and half-lives derived from these data for wild-type E . coli in the presence and absence of kasugamycin . ‘Sample’ specifies the strain , the presence or absence of kasugamycin , and the biological replicate ( one of two ) . ‘Predicted location’ is the predicted location of the protein as determined by the PsortB 3 . 0 server . ‘Gene name’ specifies the E . coli gene or small RNA name . ‘Abundance N min’ provides the copy number of the specific RNA per cell at the specified time point . ‘Decay rate’ provides the measured decay rate , e . g . log ( 2 ) / half-life , in units of 1 per minute . ‘95% Confidence interval’ is the 95% confidence interval of the decay rate . The decay rate and its confidence interval are only provided if the accuracy in the determination of specific decay rates met our criteria: the error as estimated from the 95% confidence interval range is less than one half of the fit value . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01010 . 7554/eLife . 13065 . 011Figure 4—figure supplement 1 . Reproducibility of τ-seq measurements between biological replicates . ( A ) Example decay time courses of mRNAs as derived from τ-seq measurements shown together with fits to these decays . Data correspond to the two biological replicates of the measurements of the WT strain mg1655 ( replicate 1 and 2 are shown in blue crosses and cyan circles , respectively ) . Solid lines represent the fits to Equation ( 1 ) in the Supplemental Experimental Procedures . ( B ) Scatter plot of the initial mRNA abundance measurement ( 0 min time point ) between two replicates for all genes . The Pearson correlation coefficients for the log10 abundances is 0 . 927 derived from 4243 RNAs . ( C ) Scatter plot of the final mRNA abundance ( 20 min time point ) between two replicates for all genes for the WT strain . The Pearson correlation coefficients for the log10 abundances is 0 . 913 derived from 3984 RNAs . ( D ) Scatter plot of half-lives between two replicates for the WT strain . The Pearson correlation coefficients for the log10 values is 0 . 85 derived from 2181 RNAs . Only half-lives for which the error of the decay rate ( determined as 1/4 of the 95% confidence interval returned by the fit ) was less than half of the fit value itself are reported . If the half-life for an mRNA passed this criterion in both replicates , the reported value is the average of these two half-lives . 2181 RNAs had half-lives that passed this criterion in both replicates . See Figure 4—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01110 . 7554/eLife . 13065 . 012Figure 4—figure supplement 2 . The ratio of mRNA half-lives in the presence and absence of kasugamycin . Scatter plot ( grey symbols ) of the log2 ratios of the half-lives measured for all mRNAs measured in the presence of kasugamycin relative to that measured in its absence . The ratios are grouped based on the predicted locations of the encoded proteins . The associated probability distributions ( blue ) are also shown . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 012 To determine if the native destabilization of the inner-membrane-protein mRNAs was related to the membrane localization of these mRNAs , we treated cells with kasugamycin to remove mRNAs from the membrane and repeated the τ-seq measurements . This treatment preferentially stabilized the mRNAs that encode inner-membrane proteins ( Figure 4—figure supplement 2 ) , and after the treatment the average lifetime of this group of mRNAs became comparable to those of the other three groups ( Figure 4C , D ) . These results suggest that membrane localization is important to the native destabilization of these mRNAs . To further correlate mRNA lifetime with cellular localization without the global perturbation to cellular metabolism introduced with kasugamycin , we measured how the lifetimes of mRNAs were affected by fusion with signal-peptide sequences that target mRNAs to different cellular locations . However , it is known that the degradation rates of mRNAs depends on their sequences ( Mackie , 2012 ) ; thus , sequence changes to the mRNAs caused by such fusions could lead to additional , sequence-dependent changes in lifetime , complicating the interpretation of such measurements . To overcome this challenge , we developed an approach to measure the lifetimes of a large number of fusion RNAs so that the average effect that arises from the cellular localization could be determined . Specifically , we exploited massively multiplexed cloning ( Kosuri and Church , 2014 ) to create a large library of fusion constructs comprising ~4800 distinct signal-peptide sequences fused to five different test mRNAs ( mMaple3 , neo , bla , lacZ , and phoA; Figure 5A ) . The ~4800 signal-peptide sequences include the following groups: i ) sequences of all 775 predicted E . coli SRP signal peptides; ii ) sequences of all 431 predicted E . coli SecB signal peptides; iii ) sequences from the N-terminal region of 400 cytoplasmic proteins; iv ) synthetic , non-native encodings of i ) –iii ) , which still encode the designated peptide sequences but with synonymous codons; and v ) replicates of i ) –iii ) but with the first two codons , including the start codon , replaced with a pair of stop codons . Based on the results in Figure 3 , we expect the group i ) signal sequences to send the test mRNAs to the membrane , whereas group ii ) signal sequences would not send the mRNAs to the membrane . Group iii ) serves as an additional control that also should not send mRNAs to the membrane . We included group ( iv ) to test whether the effect on mRNA degradation rates is determined by the amino acid or mRNA sequence of the signal peptides and group ( v ) to test whether this effect requires translation . 10 . 7554/eLife . 13065 . 013Figure 5 . Targeting mRNAs to the membrane reduces their lifetimes . ( A ) Schematic diagram describing the construction of ~24 , 000 unique fusions of signal peptide sequences and test genes , and the measurement of the lifetime of the mRNA for each of these fusion constructs . Only the variable signal peptide region is amplified and sequenced; thus , it also serves as a unique barcode for each construct . ( B ) The mRNA half-lives of all fusion constructs between various signal peptides ( SRP , red; SecB , blue; and cytoplasmic-control , cyan ) and different test genes ( neo , bla , mMaple3 , phoA and lacZ ) . The mRNA lifetimes of fusion constructs with SRP signal peptides are statistically significantly different from those of the fusion constructs with SecB signal peptides or cytoplasmic controls , as determined by a two-sided Kolmogrov-Smirnov test . These p-values are 4×10–20 , 2×10–22 , 2×10–15 , 1×10–21 , , and 3×10–10 for difference between the SRP and SecB fusions for neo , bla , mMaple3 , phoA , and lacZ , respectively . ( C ) As in ( B ) but for the fusion mRNAs in which the start codon is replaced by a stop codon ( -AUG ) . Colored symbols in ( B ) and ( C ) represent lifetimes of individual mRNA species , and black bars represent the mean for each group . All error bars represent standard error of the mean . See Figure 5—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01310 . 7554/eLife . 13065 . 014Figure 5—source data 1 . RNA abundance measurements versus time and half-lives derived from these data for all signal-peptide fusions . ‘Test gene’ is the name of the fusion gene . ‘Signal-peptide gene’ is the gene from which the signal peptide was derived . ‘Signal-peptide type’ is the type of signal peptide . If 'Spike-In RNA' is listed , then the data are for one of the four spike in RNAs . ‘Encoding’ specifies the encoding of the signal peptide . Native indicates that the nucleotide sequence is the native E . coli sequence; Synthetic indicates that the native E . coli codons have been exchanged at random with synonymous codons; and ‘No translation’ indicates that the first two codons , including the start codon , of the native E . coli sequence have been replaced with stop codons . ‘Abundance N min’ provides the average copy number of the specific RNA per cell at the specified time point . ‘Decay rate’ provides the measured decay rate , e . g . log ( 2 ) / half-life , in units of 1 per minute . ‘95% Confidence interval’ is the 95% confidence interval of the decay rate . The decay rate and its confidence interval are only provided if the accuracy in the determination of specific decay rates met our criteria: the error as estimated from the 95% confidence interval range is less than one half of the fit value . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01410 . 7554/eLife . 13065 . 015Figure 5—figure supplement 1 . Half-lives of fusion constructions between five test mRNAs and the native or synthetic encodings of various signal peptides . The measured half-lives for all fusions to the native ( N ) and synthetic ( S ) encodings of SRP ( red ) , SecB ( blue ) , and cytoplasmic control ( cyan ) signal peptides are shown . Symbols: measured half-lives of individual fusion mRNAs . Bars: mean ± SEM for all mRNAs within the indicated group . The synthetic encodings conserve the amino acid sequence but scramble the nucleic acid sequence by randomly selecting codons from all synonymous codons with a weight set by the frequency with which each codon appears in the E . coli genome . Changing the encoding scheme to synthetic encodings did not produce statistically significant changes in the measured half-lives for any of the SRP fusions , but lowered the average half-lives of the SecB and the cytoplasmic controls in some cases . The average half-lives of the fusions to the synthetic encoding of the SRP signal peptides are still smaller than those of the fusions to the synthetic encoding of either the SecB or control peptides in all cases . See Figure 5—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01510 . 7554/eLife . 13065 . 016Figure 5—figure supplement 2 . Effect of sequence bias on the half-lives of the SRP-fusion mRNAs . ( A ) The percent of G , C , A , T nucleotides in each group of signal sequences . Symbols: percentage in individual fusion mRNAs . The box and whiskers represents the 50% and 95% quartiles , respectively . The individual panels correspond to native encodings of the SRP signal peptides ( red ) , the SecB signal peptides ( blue ) , and the cytoplasmic control peptides ( cyan ) . The underrepresentation of nucleotide A in SRP signal peptide sequences is consistent with the previous finding of the underrepresentation of A in sequences encoding hydrophobic residues ( Prilusky and Bibi , 2009 ) , which are enriched in transmembrane domains . Identical skews were observed for the synthetic encodings of these signal peptides . ( B ) Average half-lives for beta lactamase ( bla ) fusions to all native ( top ) and synthetic ( bottom ) encodings of the SRP signal peptides ( red ) , SecB signal peptides ( blue ) , and cytoplasmic controls ( cyan ) , measured as a function of the number of each type of nucleotide in the signal sequences ( G , C , A , and T from left to right ) . Only results for bla fusions are shown but similar behaviors are observed for all five of the test genes ( bla , neo , mMaple3 , phoA , and lacZ ) . ( C ) The average ratio of SecB-fusion half-lives to SRP-fusion half-lives ( blue ) and the average ratio of the cytoplasmic control fusion half-lives to the SRP fusion half-lives ( cyan ) as a function of the number of each type of nucleotide in the signal sequences . The dashed lines represent a ratio of 1 . The average is performed across all five test genes . Data in ( B ) and ( C ) have been binned in 3 nucleotide increments , and error bars represent the standard error of the mean . The mRNA half-life appears to depend on the nucleotide compositions , in particular the A and G content , in the signal sequences . However , for a fixed number of each nucleotide , SRP fusions have smaller average half-lives than the SecB and cytosolic control groups in most cases . Thus , SRP fusions destabilize the mRNAs relative to either the SecB or cytoplasmic control fusions even after controlling for nucleotide usage . See Figure 5—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 016 We next used τ-seq to measure the lifetimes of these ~24 , 000 fusion mRNAs , utilizing the signal peptide as a variable barcode to identify each of the different constructs ( Figure 5A ) . For all five test genes , fusions to SRP signal-peptide sequences ( Figure 5B , red ) had substantially lower average mRNA lifetimes than fusions to either SecB signals or the cytoplasmic controls ( Figure 5B , blue and cyan ) , and in all cases the SRP fusions were the only statistically distinguishable set of half-lives as determined by a two-sided Kolmogrov-Smirnov test . This difference was not abolished by replacing the native codons of these signal-peptide sequences with randomly selected synonymous codons ( Figure 5—figure supplement 1 ) . Notably , inhibition of translation by deletion of the start codon removed the stability differences between fusions to SRP signal sequences and fusions to SecB or cytoplasmic controls ( Figure 5C ) . Taken together , these results support a model in which translation of SRP signal-peptide sequences that direct mRNAs to the membrane in a translation-dependent manner preferentially destabilize these mRNAs . However , the observed destabilization in the SRP-fusions need not be entirely due to mRNA localization . For example , it has been previously recognized that the RNA sequences encoding signal peptides have a sequence bias ( Prilusky and Bibi , 2009 ) . Indeed , we find a correlation between the nucleotide content of the different fusion sequences and the lifetime of the fusion ( Figure 5—figure supplement 2 ) : nucleotide sequences encoding SRP-dependent signal peptides tend to be depleted of adenosine ( A ) , and mRNA sequences with a lower A content tend to have a lower lifetime . This observation reveals that sequence bias of the RNA encoding the signal peptides also plays a role in the relative destabilization observed between the SRP fusions and all other fusions ( Figure 5B ) and raises the possibility that such sequence bias might explain a part of the destabilization observed for native mRNAs that encode inner-membrane proteins ( Figure 4A , B ) . However , after controlling for this sequence bias in the signal peptide fusions , we still observed a relative destabilization of the SRP fusions as compared to other fusion groups for sequences containing the same number of A , T , G , or C nucleotides ( Figure 5—figure supplement 2 ) , supporting the model in which a portion of this destabilization is due to the membrane enrichment of these mRNAs as established by translation of the SRP-dependent signal peptide . Interestingly , the two different modes of translation inhibition that we employed produced different global effects on mRNA lifetimes , with the average lifetime increased upon inhibition of ribosome assembly on mRNAs via kasugamycin treatment ( Figure 4D ) and decreased upon removal of the start codon ( Figure 5C ) . These contrasting changes in mRNA stability suggest that the coupling between translation and degradation may be more complicated than previously anticipated ( Mackie , 2012 ) . Nonetheless , despite their differential effects on mRNA lifetimes , both treatments abrogated the stability differences between mRNAs encoding inner-membrane proteins and mRNAs encoding proteins in other cellular compartments , supporting the model in which the native destabilization of the inner-membrane-protein mRNAs arises from the translation-dependent membrane localization of these mRNAs . It is also worth noting that the signal-peptide fusion experiments also revealed a surprising degree of variability in the rate at which mRNAs are degraded in E . coli . The nucleic acid sequences encoding these signal peptides account for no more than ~5% of the total mRNA sequences , yet , for each test gene and each group of signal peptides , variation in this portion of the mRNA can cause up to ten-fold differences in the lifetime ( Figure 5B ) . Such a high sensitivity of lifetime to sequence suggests that the cell could fine tune lifetimes through modest changes to sequence . Finally , it has been proposed that the nucleotide sequences encoding signal peptides as well as those flanking such regions have been evolutionarily optimized to introduce translational pauses that facilitate membrane targeting and co-translational insertion ( Fluman et al . , 2014 ) . Specifically , it has been shown that sequences that cause translational pauses are enriched in regions flanking the sequences encoding the SRP-signal-peptide , and this observation has led to the proposal that such sequence-induced pauses may help improve membrane targeting and prevent the translation of cytotoxic membrane proteins in the cytoplasm ( Fluman et al . , 2014 ) . However , we observe membrane targeting ( Figure 3 ) and the corresponding decrease in half-life ( Figure 5 and Figure 5—figure supplement 1 ) in constructs that do not contain such pause sequences: the constructs to which we fuse the signal peptides are not membrane proteins and , thus , will not have these sequences , and one set of our SRP fusions uses synthetic encodings of the SRP sequences which would likely eliminate or weaken any nucleotide-sequence-based signals . Thus , our results indicate that such cis-acting nucleotide sequence features are not required for membrane localization; though , we cannot rule out the possibility that they improve the performance of SRP targeting . Next , we investigated the mechanism responsible for the preferential destabilization of mRNAs localized at the membrane . We reasoned that spatial organization of the mRNA processing and degradation enzymes might play a role in this effect since several RNA degradation enzymes have been found on the membrane in E . coli ( Mackie , 2012 ) . However , out of the roughly 20 enzymes involved in RNA processing and degradation , the localization of only a handful of these enzymes have been studied previously ( Mackie , 2012 ) . Thus , to understand the full extent of spatial organization in this pathway , we created C-terminal fusions between each of these proteins and the monomeric , photoactivatable fluorescent protein mMaple3 ( Wang et al . , 2014 ) at the native chromosomal locus and measured the distribution of these enzymes using 3D STORM in live cells . Of the 24 enzymes measured , we observed that only four — the endonuclease RNase E , the 3’-5’ exonuclease PNPase , the RNA helicase RhlB , and the poly-adenylation enzyme PAPI — were enriched at the membrane , whereas the remaining proteins were largely uniformly distributed throughout the cell ( Figure 6A , B and Figure 6—figure supplement 1 ) . These four enzymes are part of a multi-enzyme complex called the RNA degradosome ( Mackie , 2012 ) , which is known to bind to the membrane via a short amphipathic helix , segment A , that is internal to RNase E ( Khemici et al . , 2008 ) . To confirm that all four enzymes are indeed enriched at the membrane due to the membrane anchoring of RNase E , we constructed a strain in which segment A of RNase E was removed ( ΔA ) and found that these enzymes were no longer localized at the membrane ( Figure 6C , D ) . 10 . 7554/eLife . 13065 . 017Figure 6 . Membrane localization of RNA degradation enzymes is required for the preferential destabilization of inner-membrane-protein mRNAs . ( A ) Stacked phase contrast ( gray ) and STORM cross-section images ( color ) of example E . coli cells expressing mMaple3 fused to RNase E , RhlB , PNPase , and PAPI in the wild-type background . ( B ) Density profiles of RNase E , RhlB , PNPase , and PAPI in the wild-type background . Density profile is as defined in Figure 1E . ( C , D ) Same as ( A , B ) but for the ΔA strains where the membrane anchor of RNase E , segment A , is deleted . ( E ) Scatter plot ( grey symbols ) of half-lives of E . coli mRNAs in the ΔA strain grouped based on the predicted locations of the encoded proteins , shown together with the associated probability distributions ( red ) . ( F ) Average half-lives for the mRNA groups depicted in ( E ) . Error bars represent standard error of the mean . Scale bars: 2 µm . Average density profiles in B and D were derived from all measured cells , tens to hundreds of cells for each strain . See Figure 6—source data 1 for all abundance data versus time and the fit decay rates used to derive half-lives . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01710 . 7554/eLife . 13065 . 018Figure 6—source data 1 . RNA abundance measurements versus time and half-lives derived from these data for the mutant E . coli strain . ‘Sample’ specifies the strain and the biological replicate ( one of two ) . ‘Predicted location’ is the predicted location of the protein as determined by the PsortB 3 . 0 server . ‘Gene name’ specifies the E . coli gene or small RNA name . ‘Abundance N min’ provides the average copy number of the specific RNA per cell at the specified time point . ‘Decay rate’ provides the measured decay rate , e . g . log ( 2 ) / half-life , in units of 1 per minute . ‘95% Confidence interval’ is the 95% confidence interval of the decay rate . The decay rate and its confidence interval are only provided if the accuracy in the determination of specific decay rates met our criteria: the error as estimated from the 95% confidence interval range is less than one half of the fit value . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01810 . 7554/eLife . 13065 . 019Figure 6—figure supplement 1 . The spatial distribution of RNA processing enzymes in E . coli . Stacked phase contrast images of the cells ( gray ) and STORM cross-section images of the enzymes ( color ) are plotted with average long-axis cross-section images of the enzymes in all imaged cells on the right . The average long-axis cross-section images are derived from hundreds of imaged cells for each strain . Plotted are the Endonucleases RNase E , RNase G , and RNase III; the exonucleases PNPase , RNase II , oligoribonuclease ( Orn ) , and RNase R; the RNA helicases RhlB , RhlE , SrmB , and DeaD; the decapping enzyme , RppH; the polyadenylation enzyme PAPI; the metabolic enzymes Enolase and polyphosphate kinase ( Ppk ) ; the RNA chaperone Hfq; the accessory protein for the tmRNA pathway , SmpB; the RNase regulators RraA , RraB , and RssB; and the stable RNA processing enzymes RNase P , RNase BN , tRNase Z , and RNase D . Scale bars: 2 µm . The only enzymes found enriched on the membrane are RNase E , PNPase , RhlB , and PAPI . Enolase and Hfq have binding sites on the C-terminal domain of RNase E ( Mackie , 2012 ) . However , because of the significantly higher expression level of Enolase as compared to RNase E , we would not expect to see a clear membrane enrichment of this enzyme even if it completely saturated the binding site on RNase E . By contrast , the more modest expression levels of Hfq suggest that this enzyme does not bind significantly to RNase E under these growth conditions , consistent with a recent report ( Persson et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 01910 . 7554/eLife . 13065 . 020Figure 6—figure supplement 2 . The ratio of half-lives between a degradosome mutant and the wild-type strain . Scatter plot ( grey symbols ) of the log2 ratios of the half-lives measured for all mRNAs in the ΔA strain over the half-lives measured for the WT strain . The ratios are grouped based on the predicted locations of the encoded proteins . The associated probability distributions ( red ) are also shown . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 020 To determine if the spatial organization of the RNA degradosome plays a role in the native destabilization of inner-membrane-protein mRNAs , we exploited the fact that deletion of segment A completely abrogates membrane localization of RNA degradation enzymes and repeated our τ-seq measurements in the ΔA strain . As expected from the sensitivity of the enzymatic activity of RNase E to lipid binding ( Murashko et al . , 2012 ) , deletion of segment A led to a global stabilization of mRNAs . However , not all four groups of mRNAs were equally affected . Remarkably , this perturbation preferentially stabilized mRNAs ( Figure 6—figure supplement 2 ) encoding inner-membrane mRNAs to the degree that the lifetimes of this group were no longer statistically distinct from the lifetimes of the other three groups ( Figure 6E , F ) . In total , these observations favor a model in which the spatial proximity between the membrane bound RNA degradosome and membrane-localized mRNAs leads to a specific increase in the turnover rates of these mRNAs ( Figure 7 ) . 10 . 7554/eLife . 13065 . 021Figure 7 . A model for the molecular mechanisms underlying the membrane localization of mRNAs encoding inner-membrane proteins and the role of this membrane localization in mRNA degradation . Translation of SRP-dependent signal peptides recruits SRP and directs mRNAs to the membrane , where the nascent polypeptide is co-translationally inserted in the membrane pore , SecYEG . Proximity of the membrane-bound RNA degradosome to these membrane-localized mRNAs leads to a preferential destabilization of these mRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 13065 . 021
In this work , we examined the spatial organization of the transcriptome of the model bacterium E . coli using a novel FISH-based RNA imaging approach that allowed us to measure the spatial organization of large and defined fractions of the transcriptome directly . These measurements revealed a transcriptome-scale spatial organization of the E . coli mRNAs: mRNAs that encode inner-membrane proteins are strongly enriched at the membrane while mRNAs that encode cytoplasmic , periplasmic and outer-membrane proteins are relatively diffusely distributed throughout the cytosol ( Figure 1 ) . In total , we imaged 75% of the expressed E . coli transcriptome; thus , we conclude that the distribution patterns observed here are the basal patterns of the spatial organization of mRNAs , although we cannot rule out the possibility that the distributions of some individual mRNAs may deviate from these transcriptome-wide patterns . We also elucidated the molecular mechanism that gives rise to the observed spatial organization . Our experiments show that the membrane localization of mRNAs encoding inner-membrane proteins depends on translation and is most likely caused by co-translational insertion of the membrane proteins mediated by the SRP pathway ( Figure 3 ) . These results are consistent with previous biochemical studies showing that co-translational insertion is the dominant pathway for targeting these proteins to the bacterial membrane ( Driessen and Nouwen , 2008 ) and are in keeping with the notion derived from previous nucleoid imaging studies that these membrane proteins can be translated and inserted into the membrane while their mRNAs are still being transcribed ( Bakshi et al . , 2015 ) . We further discovered that the spatial organization of the E . coli transcriptome has a physiological consequence on the post-transcriptional dynamics of mRNAs encoding inner-membrane proteins: these mRNAs are degraded more rapidly , on average , than mRNAs that encode cytoplasmic , periplasmic and outer-membrane proteins ( Figure 4 ) . Moreover , this native destabilization of the inner-membrane protein mRNAs depends on their membrane localization ( Figures 4 and 5 ) and on the membrane localization of the RNA degradosome ( Figure 6 ) . Removal of this degradation machinery from the membrane preferentially stabilized inner-membrane protein mRNAs and equalized the stability of all four groups of mRNAs ( Figure 6 ) . Thus , our measurements suggest a model in which proximity between the membrane-bound degradation machinery and the mRNAs encoding inner-membrane-proteins , which are localized at the membrane by co-translational membrane insertion of the proteins , is at least in part responsible for the preferential destabilization of this group of mRNAs ( Figure 7 ) . It is possible that the transcriptome organization that we observe might also play a role in shaping the dynamics of other aspects of the life of a bacterial mRNA , e . g . transcription and translation . However , because our measurements do not discriminate between nascent versus fully formed RNA or RNAs that are or are not being translated , the data we present here provide little insight into these processes . Finally , there are reasons to believe that the proximity-enhanced degradation mechanism that we discovered for E . coli may be found in a wide range of prokaryotes . Several studies have revealed that a surprisingly disparate set of prokaryotes anchor core components of their RNA degradation machinery to the membrane . Both the gram positive bacterium B . subtilis and the archaeon S . solfataricus have membrane-bound degradosomes ( or exosomes in the case of archaea ) . Remarkably , each organism uses a unique localization mechanism that differs from that utilized by E . coli ( Lehnik-Habrink et al . , 2011; Mackie , 2012; Roppelt et al . , 2010 ) , suggesting that this organization is the result of convergent evolution . Moreover , the SRP-pathway is also broadly conserved in all forms of life . Based on these observations , we anticipate that the co-translational insertion of integral membrane proteins may lead to similar membrane localization of their mRNAs and therefore an enhanced turnover of these mRNAs in a wide range of bacterial organisms . Our results thus reveal that prokaryotes , like eukaryotes , can also use spatial organization to modulate the post-transcriptional fate of RNAs .
We generated our FISH probes using enzymatic amplification of array-based oligonucleotides libraries ( Beliveau et al . , 2012; Chen et al . , 2015b; Murgha et al . , 2014 ) ( Figure 1B ) . Individual template molecules were designed by concatenating the following sequences: i ) an index primer ( I1 ) unique to the specific probe group , ii ) the common reverse transcription primer ( RTP ) P9 ( CAG GCA TCC GAG AGG TCT GG ) , iii ) the site for the nicking enzyme Nb . BsmI , iv ) the reverse complement of the targeting region ( TR ) designed to hybridize to a specific cellular RNA , v ) the reverse complement of the nicking enzyme Nb . BsrDI , and vi ) a second index priming site ( I2 ) unique to the specific probe group . Targeting regions were designed using OligoArray 2 . 0 ( Rouillard et al . , 2003 ) and all annotated E . coli transcripts ( K-12 mg1655; NC90013 . 2 ) with the following constraints: 30-nt length , a 80–85°C melting-temperature range for the duplex formed between the targeting region and the cellular RNA , a 50–60% GC content range , and a 75°C maximum melting temperature for secondary structure and cross-hybridization between different targeting regions . Index primers were designed by truncating the members of an existing library ( Xu et al . , 2009 ) of 240 , 000 oligonucleotides , each 25-nt long , to 20-nt length and selecting oligos for melting temperatures of 65–70°C , GC content of 50–60% , the absence of contiguous runs of 4 or more identical bases , the presence of a 3’ GC clamp , i . e . 2–3 G/C within the final 5 nucleotides , and low homology ( <12-nt homology ) to other primers as well as the E . coli transcriptome , the T7 promoter ( TAA TAC GAC TCA CTA TAG GG ) , and the common reverse transcription primer ( P9 ) using BLAST+ ( Camacho et al . , 2009 ) . To test the role of protein localization , mRNAs were grouped into three abundance ranges ( 1/3–3 mRNAs/cell; 3–30 mRNAs/cell; and 30–300 mRNAs/cell ) and six cellular locations as predicted by the pSortB 3 . 0 server ( Yu et al . , 2010 ) ( http://www . psort . org/psortb/ ) : cytoplasm , inner membrane , periplasm , outer membrane , extracellular , or unknown . The extracellular and unknown groups were not studied in this work . To test the role of polycistrons in RNA localization , mRNAs groups were further subdivided by whether or not a given message is polycistronic with a message encoding inner-membrane proteins . All designed targeting regions for mRNAs within each group were utilized to make probes for that group with the exception of mRNAs that encode cytoplasmic proteins . 1/3 of the possible targeting regions for each gene that encodes a cytoplasmic protein were selected at random to limit the number of probes required for these stains . To test the role of genome organization , targeting regions for mRNAs transcribed from every other 100-kb region of the genome , e . g . 100–200 kb , 300–400 kb , etc . , and within the abundance ranges of 1/3–3 mRNAs/cell or 3–30 mRNAs/cell were used . Multiple probe template sets were combined into large oligopools , and these pools were synthesized via CustomArray ( http://customarrayinc . com/ ) . Template sequences are provided in Supplementary file 1 . Template subsets were amplified and labeled using the following procedure ( Figure 1B ) . First each subset was selected and amplified with limited-cycle PCR . These templates were then amplified using in vitro transcription . The RNA products were then converted back into DNA with reverse transcription using a fluorescently labeled primer ( Alexa647-P9 ) . The template RNA was removed with alkaline hydrolysis , and the probes were column purified ( Zymo Oligo Clean and Concentrator; D4060 ) using a published protocol ( Chen et al . , 2015b ) . Antisense control probe sets ( Figure 1—figure supplement 1E ) were created by PCR amplifying the complex oligopool with index primers in which the T7 promoter was switched to index primer 1 . Probes were then produced using an Alex647-labeled index primer 2 as the reverse transcription primer . Overnight cultures of E . coli were diluted 1:200 in Lennox Luria Broth ( LB ) and grown at 32°C with shaking ( 250 rpm ) to an optical density at 600 nm ( OD600 ) of 0 . 3 . Cells were fixed , permeabilized , and stained as described previously ( Skinner et al . , 2013 ) utilizing either complex FISH probes targeting individual groups of RNA described above or single-molecule FISH ( smFISH ) probes against mMaple3 . smFISH probes to mMaple3 were designed as described previously ( Skinner et al . , 2013 ) . Kasugamycin-treated cells were harvested 15 min after the addition of kasugamycin ( Sigma; K4013 ) to a final concentration of 1 mg/mL . Cells were affixed to the surface of custom imaging chambers coated with 0 . 1% v/v plolyethyleneimine ( Sigma; P3143 ) for 15 min at room temperature . Samples were imaged on a home-built STORM microscope described elsewhere ( Huang et al . , 2008 ) . Alexa647 was excited with a 657-nm laser and reactivated with a 405-nm laser . Laser powers at 657 nm and 405 nm were 100 mW and 1 mW at the microscope backport , respectively . Oblique-incidence illumination was used for all measurements . The sample was imaged with a 100× , 1 . 40 NA , UPlanSApo Ph3 oil immersion objective ( Olympus ) and an EM-CCD camera ( Andor; iXon-897 ) . Z calibration was performed by imaging Alexa-647-labeled antibodies affixed to a coverslip scanned along the optical axis with an objective positioner ( Mad City Labs; Nano-F ) . These data were analyzed using the previously reported 3D STORM method ( Huang et al . , 2008 ) and the open source software zee-calibrator ( http://zhuang . harvard . edu/software . html ) . Images were analyzed with the algorithm 3D-daoSTORM ( Babcock et al . , 2012 ) and rendered with custom software written in Matlab ( https://github . com/ZhuangLab/matlab-storm ) . Phase contrast images were collected before and after each STORM image . Individual cells were identified and internal coordinate systems constructed using the phase contrast images and a custom implementation of previous algorithms ( Guberman et al . , 2008; Sliusarenko et al . , 2011 ) . Cell boundaries were identified with sub-pixel resolution from the contour of constant intensity corresponding to the region of steepest descent in the phase image . The two regions of largest curvature in this boundary were identified as the cell poles , and a center line was created between these poles . The boundary of the cell was then divided into 100 regions of equal arc length , and corresponding regions on each side of the cell were connected to form cellular ‘ribs’ . Spurious or filamentous cells were eliminated from subsequent analysis by discarding cells whose cell boundary lengths along each side of the center line were not within 20% of each other and by discarding cells whose total areas were larger than 3 µm2 . ( See Source code 1 ) . Single-molecule localizations were mapped to the coordinate system of each cell based on their relative position to the centerline and the closest cellular ribs . Cell-to-cell variations in width , length , and curvature were removed by normalizing this coordinate system by the length of the center line and the length of the individual ribs . This transformation effectively maps each cell to a cylinder of a common length and a common radius , where X , Y , and Z correspond to the position along the center line; the distance from the center line in the imaging plane; and the distance from the center line along the optical axis . Average short-axis cross-section images , such as that in Figure 1D ( left ) , were rendered from all molecules from all cells with Y positions within the central 150-nm thick slice of the cell . Average long-axis cross-section images such as that in Figure 1D ( right ) were rendered from all molecules from all cells with X positions in the middle 80% range to remove molecules at poles . Cross-sectional density profiles in Figure 1E , H , Figure 3D , and Figure 6B , D were created from a histogram of all localizations along the normalized Y direction falling within the middle 50% of the normalized Z range and the middle 80% of normalized X range . Cells were harvested as a function of time after rifampicin addition from E . coli cultures grown as described above to an OD600 of 0 . 4 using a previously published protocol ( Bernstein et al . , 2002 ) . In vitrotranscribed RNAs ( spike-ins ) were added for normalization between time points , and total RNA was harvested using the RNAsnap protocol ( Stead et al . , 2012 ) . The sequences of the spike-in RNAs are available upon request . DNase I was used to remove genomic contamination , and rRNA was removed using the Gram-Negative RiboZero kit ( Epicentre; MRZGN126 ) . Sequencing libraries were constructed using the RNA Ultra Directional Kit ( New England Biolabs; E7420 ) . 50-bp or 75-bp single-ended sequencing of τ-seq samples was performed on either the Illumina HiSeq2000 or the NextSeq500 . All sequencing data are available via GEO accession GSE75818 . Sequencing data were aligned to the mg1655 genome ( NC_000913 . 2 ) using bowtie 0 . 12 . 9 ( Langmead et al . , 2009 ) . Reported counts per mRNA were determined by summing counts corresponds to the region between the start and stop codons of each gene . The abundance of the in vitro spike-ins in combination with the published conversion between OD600 and cell number ( Volkmer and Heinemann , 2011 ) were used to initially calibrate absolute copy numbers per cell . Using this calibration , it was determined that the stable RNA species , tmRNA , had an average copy number of 597 ± 27 ( STD across the 8 time points from the first replicate of the wild-type strain in the absence of kasugamycin ) . The final calibration was performed by fixing the tmRNA concentration at all time points to this value , thereby eliminating small variations in RNA extraction efficiency between samples . All decay profiles were fit with an expression that incorporates three features: i ) a delayed onset of the exponential decay; ii ) a period of exponential decay; and iii ) a stable baseline . The delayed onset of decay arises because rifampicin is an initiation inhibitor not an elongation inhibitor; thus , there is a period of time during which RNAs continue to be transcribed ( Chen et al . , 2015a ) . During this period of time , transcription continues to replenish degraded RNAs and the system is effectively at steady-state . Thus , we fit the number of RNA molecules as a function of time , N ( t ) , with the following piecewise function: ( 1 ) N ( t ) =Nf+N01t≤αe-k ( t-α ) t>α where N0 + Nf is the initial number of mRNA molecules , Nf , is the number of mRNA molecules in the stable baseline , k is the rate of exponential decay , and α is the duration of the initial delay before net decay begins . Conceptually , α is related to the time required for the last round of polymerases bound prior to the rifampicin treatment to complete synthesis of the given gene . The duration of this delay depends linearly on the distance between the specific portion of an mRNA being measured and the promoter; thus , messages that are at the end of polycistronic mRNAs will have a larger α value than messages that are at the beginning of polycistronic messages or are not members of polycistronic messages ( Chen et al . , 2015a ) . Our observations indeed confirmed this prediction . Equation ( 1 ) can be viewed as an approximation for more complicated models that restrict when RNA degradation can begin , i . e . co- or post-transcriptionally , or incorporate the finite time required for RNAP polymerase to transcribe a message of a given length ( Chen et al . , 2015a ) . Such additional complications soften the boundary between the constant and exponential decay phases by introducing additional piece-wise components that contain linear or quadratic decays ( Chen et al . , 2015a ) . A non-linear least squares algorithm was used to fit the natural logarithm of Equation ( 1 ) to the natural logarithm of the τ-seq decay profiles . This logarithmic transformation equalized the weighting of all abundance measurements in the fitting routine . Reported half-lives are determined from the fit decay rates via τ=log ( 2 ) /k . Half-lives are reported only if the error ( estimated as 1/4 of the 95% confidence interval of this value returned by the fit ) of the corresponding decay rate is less than half of the measured decay rate . Where applicable the reported half-lives are the average across two biological replicates . Measured half-lives larger than our final time point ( 20 min ) were also excluded because Equation ( 1 ) was unreliable in fitting such decay curves given the time resolution of our measurement . The signal-peptides in these libraries were designed by submitting all gene sequences from the annotated mg1655 genome ( NC_000913 . 2 ) to the following servers: pSortB 3 . 0 ( Yu et al . , 2010 ) ( http://www . psort . org/psortb/ ) , signalP 4 . 0 ( Petersen et al . , 2011 ) ( http://www . cbs . dtu . dk/services/SignalP/ ) and TMHMM 2 . 0 ( Krogh et al . , 2001 ) ( http://www . cbs . dtu . dk/services/TMHMM/ ) . SRP-dependent proteins were defined as proteins with more than one TMHMM-predicted transmembrane domain and which were predicted to reside within the inner-membrane by pSortB . The signal peptide was derived from a 30-amino-acid region centered on the first TM domain . If this region exceeded the N-terminus of the protein , the first 30 amino acids at the N-terminus of the protein were used as the signal peptide . SecB-dependent proteins were defined as proteins predicted to contain a SecB-dependent signal peptide via the signalP server . Because of the similarity between N-terminal transmembrane domains and SecB-dependent signal sequences ( both are highly hydrophobic ) , some N-terminal transmembrane domains that are identified as SecB signal peptides via signalP are also identified as transmembrane domains via TMHMM , and are thus likely SRP-dependent proteins . To eliminate these spurious SecB signals , we removed predicted SecB signals derived from proteins that TMHMM predicted to have two or more transmembrane domains . The non-native SecB-dependent protein , beta lactamase , was also included in this set . The cytoplasmic control proteins were selected at random from proteins predicted to reside in the cytosol by pSortB and which were not included in either of the SRP or SecB groups . The first 30 amino acids of the SecB-dependent and cytosolic control proteins were used as the signal peptide sequence . Three encodings were used for each signal sequence: i ) the native E . coli encoding , ii ) a synthetic encoding , and iii ) an untranslated encoding . The synthetic encoding was generated by replacing each of the 30 codons in the native encoding by randomly selected synonymous codons using the codon usage across the E . coli genome as relative selection weights . The untranslated encoding was generated by replacing the first two codons of the native E . coli encoding with a pair of TAA stop codons . A complex oligopool containing the desired signal peptide sequences was synthesized by CustomArray , amplified via PCR , and inserted via Gibson assembly ( Gibson et al . , 2009 ) into pZ-series plasmids ( Lutz and Bujard , 1997 ) containing the desired genes . All signal peptides were linked to the test proteins via a common flexible linker , GGSGGS . The sequences for the signal peptides are provided in Supplementary file 2 . RNA samples were prepared as described in the “τ-Seq measurements of endogenous mRNAs” section , albeit with a different set of in vitro spike-in molecules . Sequences are available upon request . cDNA was constructed for the signal peptide region only and amplified using a mixture of primers targeting the common flexible linker and differing only in the length of a stretch of random nucleotides . This stretch of random nucleotides introduced a random length offset that was required to overcome sequencing challenges with the NextSeq500 due to regions of low complexity in these libraries . The cDNA was amplified and sequenced , and relative abundances for each library member were determined using the abundance of the spike-in molecules . The resulting decay curves were fit with an exponential decay to a stable baseline . All sequencing data are available via GEO accession GSE75818 . All plasmids were created with Gibson assembly ( Gibson et al . , 2009 ) and are based on the pZ-series plasmids ( Lutz and Bujard , 1997 ) . Chromosomal integrations were created using the lambda red recombination system ( Datta et al . , 2006 ) . All plasmids and strains reported here are summarized in Supplementary file 3 and are available upon request . Overnight cultures were diluted 1:10000 into MOPS minimal defined media supplemented with 0 . 2% w/v glucose and 34 µg/mL chloramphenicol and grown at 32°C to an OD600 of 0 . 2 . This medium was used to reduce autofluorescence in imaging measurements . Cells were concentrated thirty-fold and spotted onto sub-micron patterned agarose pads containing grooves to control cell density and orient cells ( Moffitt et al . , 2012 ) . Cells were imaged at room temperature on a home-built microscope and published protocols ( Huang et al . , 2008; Wang et al . , 2011 ) . Briefly , mMaple3 was excited with a 561-nm laser and activated with a 405-nm laser , utilizing 100 mW and 1 mW at the microscope backport , respectively . Z-calibration was conducted with antibodies conjugated to Cy3 . STORM Z-calibration , image reconstruction , and image rendering were conducted as described in the “FISH-staining and STORM imaging of RNA” section above . | Within a cell , molecules of messenger RNA ( mRNA ) encode the proteins that the cell needs to survive and thrive . The amount of mRNA within a cell therefore plays an important role in determining both the amount and types of proteins that a cell contains and , thus , the behavior of the cell . In eukaryotic organisms , like humans , it has been established that it is not just the amount of mRNA that influences cell behavior , but also where the mRNA molecules are found within the cell . However , in bacteria , which are much smaller than human cells , it has long been believed that the location of an mRNA within the cell does not affect its behavior . Despite this , recent studies that have looked at small numbers of bacterial mRNAs have shown that some of these molecules are found in larger numbers than usual at certain sites inside cells . This suggests that location may actually affect the activity of some bacterial mRNAs . But do similar localization patterns occur for all of the thousands of different mRNAs that bacteria can make ? To address this question , Moffitt et al . developed an approach that allows large , defined sets of mRNAs to be imaged in bacteria . Using this approach to study E . coli revealed that a considerable fraction of all the mRNAs that these bacteria can make locate themselves at specific sites within a cell . For example , mRNAs that encode proteins that reside inside the cell’s inner membrane are found enriched at this membrane . This localization also plays an important role in the life of these mRNAs , as they are degraded more quickly than those found elsewhere in the cell . This enhanced degradation rate arises partly because the enzymes that break down mRNA molecules are also found at the membrane . Thus , bacteria can shape the process by which an mRNA is made into protein by controlling where in a cell the mRNA is located . The next steps are to understand why bacteria use cell location to influence the rate of mRNA degradation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2016 | Spatial organization shapes the turnover of a bacterial transcriptome |
Trafficking of myelin-reactive CD4+ T-cells across the brain endothelium , an essential step in the pathogenesis of multiple sclerosis ( MS ) , is suggested to be an antigen-specific process , yet which cells provide this signal is unknown . Here we provide direct evidence that under inflammatory conditions , brain endothelial cells ( BECs ) stimulate the migration of myelin-reactive CD4+ T-cells by acting as non-professional antigen presenting cells through the processing and presentation of myelin-derived antigens in MHC-II . Inflamed BECs internalized myelin , which was routed to endo-lysosomal compartment for processing in a time-dependent manner . Moreover , myelin/MHC-II complexes on inflamed BECs stimulated the trans-endothelial migration of myelin-reactive Th1 and Th17 2D2 cells , while control antigen loaded BECs did not stimulate T-cell migration . Furthermore , blocking the interaction between myelin/MHC-II complexes and myelin-reactive T-cells prevented T-cell transmigration . These results demonstrate that endothelial cells derived from the brain are capable of enhancing antigen-specific T cell recruitment .
In the neuro-inflammatory disorder multiple sclerosis ( MS ) , the trafficking of immune cells into the brain is a crucial step not only in the onset but also the progression of the disease ( Lublin et al . , 2014; Sospedra and Martin , 2005 ) . Once within the central nervous system ( CNS ) , myelin-reactive T-cells induce severe neuronal and tissue damage and degeneration ( Trapp et al . , 1998 ) . During the initiation of the disease , myelin-specific CD4+ T cells are differentiated into effector T helper ( Th ) 1 or Th17 cells ( Bailey et al . , 2007; Bielekova et al . , 2000 ) . There is compelling evidence that Th1 and Th17 cells , separately or in cooperation , mediate deleterious responses in MS ( Carbajal et al . , 2015 ) . To enter the CNS , immune cells have to cross the blood-brain barrier ( BBB ) , which is composed of highly specialized brain endothelial cells ( BECs ) that are sealed by closely regulated tight junctions ( Tietz and Engelhardt , 2015 ) . Lymphocyte migration into the CNS parenchyma is a multi-step process that requires close contact between lymphocytes and BECs ( Ransohoff et al . , 2003 ) . These cell-cell contacts are mediated by cell surface molecules on both the lymphocytes and BECs . During inflammation BECs upregulate the expression of the adhesion molecules Inter-Cellular Adhesion Molecule ( ICAM ) −1 , Vascular cell adhesion molecule ( VCAM ) −1 and Activated leukocyte cell adhesion molecule ( ALCAM ) , which are necessary for the firm adhesion of lymphocytes to the endothelium as well as for the trans-migration process ( Larochelle et al . , 2011 ) . Although these adhesion molecules expressed by BECs have been shown important for the transendothelial migration of leukocytes , the complexity of this interaction and the molecules involved remain poorly understood . Several studies provided evidence for an antigen-specific component in the transmigration process of encephalitogenic T-cells ( Archambault et al . , 2005; Galea et al . , 2007; Ludowyk et al . , 1992 ) . Using an adoptive transfer model of murine MS ( experimental autoimmune encephalomyelitis , EAE ) it was demonstrated that only activated myelin-specific CD4+ T-cells accumulated in the CNS parenchyma , while non-CNS-specific T-cells failed to infiltrate ( Archambault et al . , 2005 ) . Furthermore , expression of MHC-II on the recipient cells appeared to be required for CNS infiltration , as the myelin-specific T-cells did not transmigrate over CNS vascular endothelium when adoptively transferred in MHC-II deficient mice . However , whether the antigen-specific signal was provided by APCs or BECs was not elucidated . Also infiltration of CD8+ T-cells into the brain was shown to be an antigen-specific process: haemagglutinin-specific CD8+ T-cells were only detected in the CNS upon intra-cerebral injection of cognate , but not control , peptides in an haemagglutinin T-cell receptor transgenic mouse ( Galea et al . , 2007 ) . A role for BECs in providing the antigen signal to T-cells was claimed by showing luminal expression of MHC-I on BECs and the fact that intra-venous injection of a blocking MHC-I antibody significantly reduced the CD8+ T-cell infiltration . However , since a soluble , nominal MHC-I epitope was used as antigen , exogenous binding of these peptides to MHC-I molecules on other cells in the CNS , or even in CNS-draining lymph nodes cannot be excluded ( Weller et al . , 1996 ) . Thus , so far no direct evidence is provided for a role of BECs in processing and presenting CNS-derived antigens during inflammatory conditions . BECs have been shown to express MHC-I while MHC-II is virtually absent . Inflammation induced activation of BECs causes increased expression of MHC-II as well as of the co-stimulatory molecule CD40 and enhanced their ability to stimulate the proliferation of allogeneic T-cells in-vitro ( Wheway et al . , 2013 ) . Although BECs have been shown to take up soluble antigens by macro-pinocytosis and clathrin-coated pits ( Wheway et al . , 2013 ) , not much is known about their capacity to process and present internalized antigens . We therefore explored the potential of BECs as antigen-presenting cells and determined whether antigen-presentation by BECs contributes to transmigration of myelin-reactive T-cells . We here demonstrate that inflamed BECs take up and process myelin via the endo-lysosomal degradation pathway in a time-dependent manner . Importantly , these myelin-derived antigens are presented in de-novo expressed MHC-II molecules and facilitate the migration of antigen-specific Th1 and Th17 pathogenic T-cells through the brain endothelium . Better insight into the events that trigger T-cell migration into the brain is crucial for our understanding of MS pathogenesis and will aid the development of new treatments to prevent T-cell infiltrating the CNS .
To determine if BECs play a role in antigen-specific migration of CD4+ T cells by acting as APCs , we first assessed the expression of molecules necessary for antigen presentation and co-stimulation . Resting , non-inflamed , human BECs express MHC-I and PD-L1 while MHC-II , CD40 and VCAM−1 are expressed at low levels ( Figure 1A ) . Upon inflammatory activation , BECs express high levels of VCAM−1 , and significantly increased the expression levels of MHC-II ( Figure 1A , B ) . Similarly , CD40 expression was increased upon activation . Both MHC-I and PD-L1 were highly expressed on resting as well as on activated BECs . Expression of the classical co-stimulatory molecules CD80 and CD86 were undetectable on resting and activated BECs ( data not shown ) . Comparable changes in phenotype were observed when BECs were activated using IFN-γ instead of TNFα ( Figure 1—figure supplement 1 ) Together , these results confirm and extend previous findings ( Wheway et al . , 2013 ) and indicate that BECs are equipped to present antigens under inflammatory conditions . Up-regulation of MHC class II molecules via inflammation induced CIITA activity has been associated with increased susceptibility of EAE , yet how increased MHC-II expression contributes to actual disease has so far not been described ( Reith et al . , 2005 ) . 10 . 7554/eLife . 13149 . 003Figure 1 . Human brain endothelial cells internalize myelin particles . Confluent monolayers of brain endothelial cells ( BECs ) were stimulated with 5 ng/ml TNFα for 24 hr . ( A ) Expression of MHC-I , MHC-II , CD40 , PD-L1 and VCAM−1 was determined by flow cytometry . Histograms depict expression of indicated markers in resting ( grey solid line ) and activated ( black solid line ) BECs . Dashed lines indicate isotype controls . ( B ) The MFI of expression of the indicated markers is shown . Data are presented as the mean ± SD of duplicate values ( n = 5 independent experiments ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ( Student t-test ) . ( C–E ) Fluorescent labeled human myelin was added to resting or activated BECs for 4 hr or 24 hr and uptake was analyzed by ( C ) flow cytometry or ( D–E ) imaging flow cytometry . ( C ) Representative facs plots of myelin uptake by BECs , numbers in plots indicate the MFI of myelin-positive cells . The percentage of myelin-positive resting and activated BECs at 4 and 24 hr after loading with antigen is shown in a graph . ( D–E ) Myelin-positive BECs internalized between 1–3 particles/cell . On average , BECs acquired 2–3 myelin particles/cell . Activation of BECs did not affect the number of internalized particles . The average number of internalized myelin particles per cell is shown in a bar graph . Data presented are the means of triplicate values ± SEM of at least three independent experiments . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ( Student t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13149 . 00310 . 7554/eLife . 13149 . 004Figure 1—figure supplement 1 . Human brain endothelial cells increase MHC and costimulatory molecule expression upon activation by IFN-γ . Confluent monolayers of brain endothelial cells ( BECs ) were stimulated with 10 ng/ml IFN-γ for 24 hr . Expression of MHC-I , MHC-II , CD40 and VCAM−1 was determined by flow cytometry . Histograms depict expression of indicated markers in resting ( orange line ) and activated ( green line ) BECs . Red and blue lines indicate isotype controls . DOI: http://dx . doi . org/10 . 7554/eLife . 13149 . 00410 . 7554/eLife . 13149 . 005Figure 1—figure supplement 2 . Brain endothelial cells internalize myelin particles . BECs were loaded with fluorescently labeled human myelin and internalization of myelin particles was assessed by imaging flow cytometry . To determine internalization scores , a mask was designed based on the surface of BECs in the brightfield image . This mask was then eroded to exclude the cell membrane . The resulting mask was applied to the fluorescence channel . The internalization score , interpreted as a ratio of the intensity of the intracellular space versus the intensity of the whole cell , was calculated on this mask using the internalization feature of the Ideas v6 . 0 software ( AMNIS Merck Millipore ) . Cells that have internalized antigens have positive scores , as depicted here for BECs . DOI: http://dx . doi . org/10 . 7554/eLife . 13149 . 005 Since myelin-derived antigens are the major target of auto-reactive T-cells in MS , we investigated if BECs can take up and process myelin . We therefore incubated BECs with fluorescent labeled myelin for different time-points under resting and inflammatory conditions and determined myelin uptake by flow cytometry . As depicted in Figure 1C a time-dependent increase in the proportion of myelin+ BECs was observed . Moreover , this process is not significantly affected by treatment with inflammatory stimuli as activated BECs showed a similar amount of internalized myelin as resting BECs . Using imaging flow cytometry , we assessed that BECs that were able to capture myelin increased the number of myelin particles over time to a maximum of three myelin particles/cell after a 24 hr incubation ( Figure 1D ) . Moreover , the average amount of myelin particles per cell was the same in both resting and inflammatory conditions , again , demonstrating that this process is not significantly affected by treatment with inflammatory stimuli ( Figure 1D , E ) . Of note , in order to measure whether the localization of the myelin signal was intracellular or membrane-bound , we designed a mask that excludes the cell membrane and calculated a ratio of the amount of fluorescence located in the mask vs the total amount of fluorescence , as previously reported ( Garcia-Vallejo et al . , 2015 ) . The results indicate that the myelin fluorescence signal was intracellular , demonstrating that BECs are able to efficiently internalize myelin ( Figure 1—figure supplement 2 ) . The endo-lysosomes are the typical antigen-processing compartments of APCs ( Blum et al . , 2013; Roche and Furuta , 2015 ) . This intracellular route allows optimal processing of exogenous protein antigens and transfer of antigen-derived peptides to the MHC-II compartment for loading and subsequent presentation to CD4+ T-cells . To determine whether internalized myelin is shuttled to these compartments in BECs , myelin-treated BECs were stained with antibodies against EEA1 ( a marker of early endosomes ) and LAMP1 ( a marker of late endosomes and lysosomes ) to measure co-staining with myelin using imaging flow cytometry . We observed that myelin co-localized with both EEA1 and LAMP1 as shown by a high co-localization score ( Figure 2A , B ) . The co-localization with both markers was higher at 24 hr of exposure to myelin compared to 4 hr . Since the increase of the co-localization score for myelin-EEA1 was not as strong as shown for myelin-LAMP1 at 24 hr ( Figure 2A , B ) , this suggests that at that time point the majority of myelin was present in lysosomes . However , non-internalized myelin fragments that are attached to the cell membrane , could potentially be 'internalized' as a consequence of trypsinization of adherent BECs . To demonstrate that myelin is actively taken up by BECs , we analyzed myelin uptake and intracellular routing in adherent BECs using confocal laser scanning microscopy . Similar to our experiments using imaging flow cytometry , we observe that 24 hr after loading of adherent BECs a proportion of cells show internalized myelin . Furthermore , it is clear that internalized myelin is present within LAMP1 positive vesicles and not with EEA1 positive organelles ( Figure 2C–I ) . 10 . 7554/eLife . 13149 . 006Figure 2 . Myelin particles are preferably routed to the endo-lysosomes . Resting or activated BECs were loaded with Atto-633 labeled myelin for 4 hr or 24 hr . Uptake of myelin particles and their co-localization with early endosomal ( EEA1 ) or endosomal/lysosomal ( LAMP1 ) compartments was analyzed by imaging flow cytometry and quantified using the brightfield similarity R3 feature ( see methods for details ) . Myelin particles co-localized with ( A ) EEA1 and ( B ) LAMP1 in both resting ( grey bars ) and activated ( black bars ) BECs . Graphs represent the mean of triplicate values ± SEM of n = 3 independent experiments . ( C–G ) Adherent BECs were loaded with Atto−633-labeled myelin and 24 hr later , co-localization of myelin ( in red ) with EEA1 ( in green , upper panels ) or LAMP1 ( in green , lower panels ) was analyzed using CSLM . Nuclei were visualized with Hoechst ( in blue ) and the cytoskeletal F-actin bundles are shown in yellow . Representative images of adherent brain endothelial cells with subcellular localization of myelin with EEA1 ( C , E ) or LAMP1 ( D , F ) . A magnification of indicated areas is shown in E–F . A cross-sectional study focusing in an myelin-rich area demonstrates the presence of the antigen surrounded by LAMP1 staining , indicating its presence within lysosomes . ( G–H ) Histograms were created for a selected area ( indicated by a line ) using ImageJ software ( NIH , USA ) . Histograms were created from each fluorochrome and overlays were made by the program . ( I ) Quantification of myelin positive early-endosomal and lysosomal compartments . Percentage of myelin fragments associated with each marker was determined using ImageJ software ( N = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13149 . 006 Together , these data suggest that myelin enters the endosomal/lysosomal pathway when internalized by BECs . Furthermore , it is clear that the efficiency of myelin uptake and internal routing to compartments associated with antigen processing by BECs is not affected by inflammation per se . This is in contrast to their professional counterparts: only in an immature state DCs possess high antigen internalization and processing capacities . Activation induced maturation of DCs strongly reduces these functions , and increases the presentation of antigens in MHC molecules ( Inaba et al . , 2000; Jin et al . , 2004 ) . Taking the lack of human myelin-specific T-cell clones as well as HLA-matching issues with BECs into account , we used murine BECs ( mBECs ) and MOG35-55-specific CD4+ T-cells from 2D2 transgenic mice ( Bettelli et al . , 2003 ) as a model system to elucidate whether myelin antigens are processed and presented by BECs to facilitate T-cell transmigration . Notably , an increased expression of MHC-II was observed on cerebral blood vessels of mice in the active phase of EAE when compared to control adjuvant injected mice ( CFA; not shown ) , demonstrating that mBECs , similar to the human counterparts , are properly equipped to present antigens to pathogenic CD4+ T cells . Since in the brain of MS patients and of EAE mice mainly Th1 and Th17 effector cells have been found ( Carbajal et al . , 2015 ) , we generated MOG-specific Th1 and Th17 in-vitro ( Figure 3A ) and used them in a trans-well setting with myelin-loaded activated mBECs . To allow sufficient antigen processing , mBECs were loaded with myelin in the presence of TNFα 24 hr prior the co-culture with T-cells . To control for antigen-specificity , mBECs were loaded with the non-CNS antigen ovalbumin ( OVA ) . Loading of mBECs with OVA did not significantly induce the migration of any of the MOG-specific T-cell subsets , similar to medium-control mBECs ( Figure 3B , C ) . However , when mBECs were loaded with myelin , a significant increase in migrated Th1 and Th17 cells was observed , demonstrating that processing and presentation of myelin-derived peptides by mBECs specifically leads to migration of antigen-specific T-cells . Addition of an MHC-II-blocking antibody during the migration period significantly reduced the trans-migration of both Th1 and Th17 cells ( Figure 3D , E ) , further providing evidence that presentation of myelin-derived antigens in MHC-II by mBECs facilitates T-cell migration . Using the nominal epitope for 2D2 T-cells ( i . e . MOG35–55 ) to load mBECs with , similar results were obtained as with myelin-loaded mBECs ( Figure 3F ) . Moreover , our observation that OVA-specific Th1 and Th17 only trans-migrated when encountering OVA-loaded BECs and not when co-cultured with MOG35–55-loaded or medium control BECs substantiates the finding that T-cell migration over the BEC monolayer occurs in an antigen-specific manner ( Figure 3G , H ) . These data demonstrate that brain endothelial cells can internalize antigen and promote antigen-specific T cell transmigration in vitro . 10 . 7554/eLife . 13149 . 007Figure 3 . Migration of myelin-specific T-cells depends on presentation of myelin-antigens in MHC-II by BECs . ( A ) Th1 and Th17 subsets were generated in-vitro from naive CD4+CD62Lhigh 2D2 T cells . Expression of IFN-γ , Il−17 , IL−10 , T-bet and RORγT was determined using qRT-PCR . Data are the means of triplicate values ± SEM of three independent experiments . ( B ) mBECs were seeded onto trans-wells , activated with TNFα and loaded with myelin for 24 hr . As a control , BECs loaded with the CNS-unrelated antigen OVA or unloaded BECs were used . ( B ) Th1 or ( C ) Th17 2D2 T-cells were added to the upper compartment and T-cell migration was quantified by flow cytometry 3 hr later using fluorescent labelled beads as reference . To block antigen recognition by T-cells , an MHC-II blocking antibody was added to mBECs one hour prior addition of the ( D ) Th1 or ( E ) Th17 cells . The MHC-II blocking antibody was present during the 3 hr incubation with the T-cells . ( F ) Transmigration of Th1 and Th17 cells over a monolayer of MOG35-55 pulse-loaded activated mBECs was analyzed using Transwells . Migration was assessed in the presence of an MHC-II blocking antibody or control antibody . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ( ANOVA with Bonferroni correction ) . The average frequency of T cells that transmigrated in the control setting are 10 . 8% ± 1 . 2 for 2D2 Th1 and 11 . 6% ± 0 . 4 for 2D2 Th17 . ( G , H ) Th1 and Th17 subsets were generated in-vitro from naive CD4+CD62Lhigh OT-II T cells . mBECs were seeded onto trans-wells , activated with TNF and loaded with OVA for 24 hr . As a control , BECs loaded with MOG35-55 or unloaded BECs were used . Th1 ( G ) or Th17 ( H ) OT-II T-cells were added to the upper compartment and T-cell migration was quantified by flow cytometry 3 hr later using fluorescent labelled beads as reference . Average frequency of OT-II Th1 and OT-II Th17 that transmigrated in the control settings are 7 . 9% ± 1 . 9 and 12 . 5% ± 1 . 4 , respectively . **p<0 . 01 , ***p<0 . 001 ( ANOVA with Bonferroni correction ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13149 . 007 Mice lacking a functional class-II restricted antigen processing machinery are resistant to both active and adoptive transfer EAE ( Tompkins et al . , 2002 ) , suggesting that proper processing of antigens is essential for disease initiation . Although these results could be due to the lack of activation of auto-reactive T-cells by peripheral APCs , the failure to induce disease by adoptive transfer of activated T-cells in this study could also be explained by the lack of a functional antigen processing machinery in BECs since trafficking of injected , ex-vivo activated , T-cells into the brain is impaired . Thus , together with our novel data , it seems likely that antigen-presentation by the brain endothelium facilitates the entry of antigen-specific CD4+ effector T cells into the brain . This phenomenon has also been proposed , but was never demonstrated , in other diseases suffering from the infiltration and destruction of tissue by auto-reactive T-cells . In Type 1 Diabetes , pancreatic islet antigen expression was shown to be a key factor in governing the ability of the autoantigen-specific T-cells to accumulate in the pancreatic islets ( Hamilton-Williams et al . , 2003; Van Halteren et al . , 2005 ) . Importantly , using human T-cell clones and humanized mice , it was demonstrated that only beta cell-specific T-cells reached the pancreatic islets where they destroyed the insulin-producing beta cells . By contrast , diabetes unrelated T-cells retained at the peri-vascular sites ( Unger et al . , 2012 ) , demonstrating that beta-cell-specific T-cells , present in the circulation , need to cross the endothelium to access the pancreatic islets . Whether antigen-specific T-cell entry favors the entry of non-tissue specific T-cells is still a matter of debate . The entry of encephalitogenic T-cells into the brain has been shown to pave the way for non-CNS-specific T-cells ( Lees et al . , 2010; Ludowyk et al . , 1992 ) yet the latter subset remained in an inactive state . Together , the data presented in this study demonstrate for the first time that myelin enters the endosomal/lysosomal pathway when internalized by BECs , irrespective of their activation status . This observation is also different to findings on professional APCs such as dendritic cells , which mainly internalize antigens being in an immature state . The fact that BECs maintain to internalize and process exogenous antigens in an activated state is advantageous during infection-induced inflammation in the brain ( e . g . meningitis ) as it will facilitate the presence of antigen-specific effector T-cells to resolve the unwanted infection . However , this continuous facilitation of immune cell entry into the CNS is destructive in case of MS . Overall , our results demonstrate that BECs can take up and process myelin particles in a time-dependent manner . Although the focus of the present study was to examine whether antigen-presentation by BECs contributes to transmigration of myelin-reactive T cells , it can be speculated that uptake of myelin , consisting of large particles , by brain endothelial cells predominantly occurs via phagocytosis . BECs have been shown to use different endocytosis mechanisms to internalize particles , which is dependent on the size and composition of the particle ( Faille et al . , 2012; Falcone et al . , 2006; Georgieva et al . , 2011 ) . Furthermore , the upregulation of MHC-II expression under inflammatory conditions reinforces the idea of a non-professional antigen presenting cell role . Although we do not provide direct evidence for processing and presentation of internalized myelin , our data strongly suggest that myelin-derived antigens can be presented by brain endothelial cells in MHC-II to antigen-specific T cell subsets , aiding in the diapedesis of these cells in an MHC-II dependent fashion . These results demonstrate that the brain endothelium is an active contributor to disease pathogenesis . Furthermore , these findings have major implications in neuro-inflammatory disorders such as MS , since increased immune cell trafficking has a detrimental effect in disease progression . Therapies directed at antigen processing and presentation by BECs could be effective to dampen unwanted immune cell infiltration in MS .
The human brain endothelial cell ( BEC ) line hCMEC/D3 ( Weksler et al . , 2005 ) was kindly provided by Dr PO Couraud ( Institut Cochin , Universite Paris Descartes , Paris , France ) . BECs were grown in EBM−2 medium supplemented with hEGF , hydrocortisone , GA-1000 , FBS , VEGF , hFGF-B , R3-IGF-1 , ascorbic acid and 2 . 5% fetal calf serum ( Lonza , Basel , Switzerland ) . For antigen internalization experiments , resting or 24 hr rhTNFα activated ( 5 ng/ml , Peprotech , UK ) BECs were seeded in collagen-coated plates and when confluent , incubated with 10 µg/ml labeled myelin ( myelin−555 ) for 4 hr or 24 hr . Subsequently , cells were extensively washed with PBS to remove external myelin and fluorescence intensity was measured using a FACS Calibur flow cytometer ( Becton and Dickinson , San Jose , CA ) . The following antibodies were used to detect the presence of MHC and costimulatory molecules on resting or TNFα activated BECs: FITC-conjugated anti-HLA-ABC ( clone DX-17 ) and -VCAM−1 ( clone STA ) ; PE-conjugated anti-HLA-DR ( clone G46-6 ) ; -CD80 ( clone L307 . 4 ) ; -CD86 ( clone 2331 ) . Binding of unconjugated anti-CD40 ( clone TRAP-1 ) was detected using goat-anti-mouse IgG1-A488 ( Life Technologies ) . All antibodies were obtained from BD Pharmingen , except anti-VCAM which was obtained from eBiosciences . Confluent BECs were seeded in 6-well plates ( Corning , Amsterdam , The Netherlands ) and stimulated with 5 ng/ml rhTNFα for 24 hr . 10 µg/ml of fluorescent-labeled human myelin was added to BECs for 4 hr or 24 hr . Cells were then extensively washed with ice-cold PBS , detached with trypsin and fixated with 4% formaldehyde . Cells were then permeabilized with 0 . 05% saponin for 30 min at RT and subsequently blocked with 10% goat serum in PBS/BSA . Cells were labeled with EEA1-FITC ( BD Bioscience ) , LAMP1 ( BD Pharmingen ) and goat anti-mouse Alexa 488 ( Molecular Probes , Eugene , OR ) . Cells were analyzed on the ImageStream X100 ( Amnis-Merck Millipore ) imaging flow cytometer as previously described ( García-Vallejo et al . , 2015 ) . A minimum of 15 , 000 cells were acquired per sample . Internalization and co-localization scores were calculated as previously described ( García-Vallejo et al . , 2015 ) . Briefly , cells were acquired on the basis of their area . Analysis was performed with single cells after compensation ( with a minimum of 5000 cells ) . For standard acquisition , the 488 nm laser line ( for EEA-1 and LAMP-1 ) was set at 10 mW and the 642 nm laser line ( for myelin ) was set at 5 mW . Firstly , a mask was designed based on the surface of BECs in the brightfield image . This mask was then eroded to exclude the cell membrane . Finally , the resulting mask was applied to the fluorescence channel . The internalization score was then calculated on this mask using the Internalization feature provided in the Ideas v6 . 0 software ( Amnis-Merck Millipore ) . Internalization can be interpreted as a log-scaled ratio of the intensity of the intracellular space versus the intensity of the entire cell . Cells that have internalized antigen typically have positive scores , while cells that show the antigen still on the membrane have negative scores . Cells with scores around 0 have similar amounts of antigen on the membrane and in intracellular compartments . Co-localization is calculated using the bright detail similarity R3 feature in the Ideas software . This feature corresponds to the logarithmic transformation of Pearson’s correlation coefficient of the localized bright spots with a radius of 3 pixels or less within the whole cell area in the two input images . Myelin particle counts were calculated using the peak mask in combination with the spot count feature as previously described ( García-Vallejo et al . , 2014 ) . Confluent BECs were seeded in 8-well Ibidi slides ( Ibidi , GmbH , Munchen , Germany ) and incubated with 10 µg/ml Atto 633 labeled myelin for 24 hr . Subsequently , cells were extensively washed with PBS and fixated with 4% formaldehyde . Non-specific binding was blocked with 5% goat serum in PBS/BSA containing 0 . 3% Triton-X100 . Cells were labeled with rabbit anti-EEA1 ( Cell Signaling ) or rabbit anti-LAMP1 ( Cell Signaling ) . Antibodies were visualized after 1 hr incubation with goat anti-rabbit Alexa488 ( Molecular Probes ) . Finally , sections were stained with Hoechst ( molecular Probes , Invitrogen ) to visualize cellular nuclei and with phalloidin rhodamine to visualize F-actin ( Molecular Probes , Invitrogen ) . Sections were mounted with mounting medium . Co-localization was analyzed using a Confocal Laser Scanning Microscope ( Leica DMI 6000 , SP8 , Leica , Mannheim , Germany ) ; images were acquired using LCS software ( version 2 . 61 , Leica ) . Primary mBECs were isolated from brains of C57BL/6 mice as described previously ( Coisne et al . , 2005 ) . Brains were harvested and superficial blood vessel , meninges and cerebellum were removed . Brains were homogenized in isolation medium ( HBSS supplemented with 10 mM HEPES and 0 . 1% BSA ) in a potter and centrifuged . The pellet was resuspended in 15% dextran ( 70 kDa ) and centrifuged at 3000 g for 25 min . Subsequently , the pellet was resuspended in 0 , 2% collagenase/dispase with 10 μg DNase in culture medium ( DMEM supplemented with 20% FCS , 1% amino acids , 2% sodium pyruvate and 50 µg/ml gentamycin ) and incubated for 30 min in a 37°C waterbath . After washing , the obtained fragments of blood vessels were seeded in collagen-coated dishes in culture medium containing puromycin to avoid contamination with pericytes . After 24 hr of culture , medium was supplemented with 1 ng/ml FGF . At the end of culture , endothelial purity was checked by qPCR for CD31 ( endothelial ) , GFAP ( astrocytes ) , and PDGF-receptor beta ( pericytes ) as described before ( Reijerkerk et al . , 2013 ) and cultures were found to be consisting of 95% endothelial cells . Single cells suspensions of spleens and lymph nodes from 2D2 Tg mice ( generous gift from L . Berod , TWINCORE Institute , Hannover , Germany ) were depleted of erythrocytes using ACK lysis buffer . Subsequently , CD4+ T-cells were enriched using the mouse CD4+ T-cell enrichment kit ( eBiosciences ) according to manufacturer’s instructions; stained with anti-CD4-PE and CD62L-APC antibodies and naive CD4+CD62Lhigh T cells were sorted using a MoFlow ( DakoCytomation , Glostrup , Denmark ) . Naive T-cells ( 5 × 104 ) were incubated with MOG35–55/LPS loaded BMDCs ( 1 × 104 ) to promote Th1 differentiation . Incubation of naive CD4+ T-cells with MOG-loaded BMDCs in the presence of PGN ( 10 µg/ml ) promoted Th17 differentiation . Two days later , 10 U/ml rmIL−2 ( Invitrogen , Bleijswijk , The Netherlands ) was added to the Th1 promoting cultures and another three days later T-cells were harvested and used in functional assays . Messenger RNA was isolated from mBECs using the TRIzol method ( Life Technologies , Bleiswijk , the Netherlands ) and cDNA was synthesized with the Reverse Transcription System kit ( Promega , Leiden , the Netherlands ) . The following primer sequences were used: IFN-γ FWD: TACTACCTTCTTCAGCAACAGC , IFN-γ REV: AATCAGCAGCGACTCCTTTTC , IL−10-FWD: GGCGCTGTCATCGATTTCTC; IL−10 REV: ATGGCCTTGTAGACACCTTGG , T-bet FWD: CAGGGAACCGCTTATATG , T-bet REV: CTGGCTCTCCATCATTCA , RORγT FWD: GGAGCAGAGCTTAAACCCCC; RORγT REV: TCCCAGATGACTTGTCCCCA , GAPDH FWD: GACAACTCATCAAGATTGTCAGCA; GAPDH REV: TTCATGAGCCCTTCCACAATG . Oligonucleotides were synthesized by Invitrogen ( Bleiswijk , the Netherlands ) . Quantitative PCR ( qPCR ) reactions were performed in an ABI7900HT sequence detection system using the SYBR Green method ( Applied Biosystems , New York , USA ) . Expression levels were normalized to GAPDH expression levels . Ex-vivo isolated mBECs were seeded on collagen-coated 5 µm pore size Costar transwells ( Corning , Amsterdam , The Netherlands ) for 5–7 days . mBECs were loaded with 72 . 5 µg/ml myelin , 10 µg/ml MOG35–55 or 10 µg/ml OVA in the presence of 25 ng/ml TNFα for 24 hr . Cells were thoroughly washed and 1 × 105 Th1 or Th17 were added per transwell . Anti-mouse MHC-II blocking antibody ( #16-5321-81 , eBioscience ) was added at 5 µg/ml per transwell , 1 hr prior to addition of T cells . After 3 hr T-cells were recovered from the lower well and 20 , 000 beads ( Beckman Coulter , USA ) were added to each sample . Samples were analyzed by flow cytometry on a FACScalibur ( BD , San Jose , USA ) and by gating and counting 5000 beads , the number of migrated T-cells was determined . Statistical analysis was performed using GraphPad Prism software ( v5 . 01 GraphPad Software , La Jolla , CA ) using either unpaired Student t test or one-way ANOVA followed by posthoc Bonferroni correction . | The blood vessels in the brain help to control the entry of nutrients , cells and waste products into and out of the brain . In doing so , they create a protective barrier between the blood and the brain known as the blood-brain barrier . However , this barrier loses its protective function in individuals with multiple sclerosis or other disorders that affect the brain . Multiple sclerosis patients develop inflammation and their immune cells become able to enter the brain . These immune cells may then attack layers of insulation called myelin that surround nerve cells . Myelin helps nerve cells to work properly so the loss of this insulation can lead to tissue damage and cognitive problems . When immune cells called T cells enter the brain they can become primed to recognize myelin and attack it in the same way that they would attack viruses or bacteria . However , it is not clear precisely how these T cells develop the ability to cross the blood-brain barrier and attack myelin . Now , Lopes Pinheiro et al . show that “endothelial” cells in the blood-brain barrier are able to present fragments of myelin to T cells , which enables the T cells to identify myelin and move into the brain . First , the blood-brain barrier cells absorb and break down proteins in the myelin , and then they present fragments of these proteins on their surfaces with the help of protein clusters called major histocompatibility complexes ( MHCs ) . Other protein fragments that can also activate T cells in other parts of the body did not affect the blood-brain barrier when they were presented by MHCs , which suggests that the effect could be specific to myelin proteins . The experiments also show that it is possible to stop T cells from crossing the blood-brain barrier by preventing them from interacting with myelin fragments presented by MHCs . This suggests that therapies that interfere with the ability of blood-brain barrier cells to break down myelin proteins and present them to T cells might help to protect the brains of patients with multiple sclerosis . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience",
"immunology",
"and",
"inflammation"
] | 2016 | Internalization and presentation of myelin antigens by the brain endothelium guides antigen-specific T cell migration |
Nervous system maps are of critical importance for understanding how nervous systems develop and function . We systematically map here all cholinergic neuron types in the male and hermaphrodite C . elegans nervous system . We find that acetylcholine ( ACh ) is the most broadly used neurotransmitter and we analyze its usage relative to other neurotransmitters within the context of the entire connectome and within specific network motifs embedded in the connectome . We reveal several dynamic aspects of cholinergic neurotransmitter identity , including a sexually dimorphic glutamatergic to cholinergic neurotransmitter switch in a sex-shared interneuron . An expression pattern analysis of ACh-gated anion channels furthermore suggests that ACh may also operate very broadly as an inhibitory neurotransmitter . As a first application of this comprehensive neurotransmitter map , we identify transcriptional regulatory mechanisms that control cholinergic neurotransmitter identity and cholinergic circuit assembly .
Nervous system maps that describe a wide range of distinct structural and molecular parameters are essential for an understanding of nervous system development and function . Tremendous efforts have been and are being made to map connectomes ( Bargmann and Marder , 2013; Plaza et al . , 2014 ) . Connectomes now exist for small anatomic regions of mouse and fly brains ( Helmstaedter et al . , 2013; Kasthuri et al . , 2015; Takemura et al . , 2013 ) , but the only complete , system-wide connectome remains that of the nematode Caenorhabditis elegans , both in its hermaphroditic and male form ( Albertson and Thomson , 1976; Jarrell et al . , 2012; White et al . , 1986 ) . However , these anatomical maps are incomplete without the elucidation of chemical maps that describe the synaptically released neurotransmitters through which anatomically connected neurons communicate with one another . But even in C . elegans , let alone other organisms , there have so far been only limited efforts to precisely map neurotransmitter identities on a system-wide level with single-neuron resolution . In C . elegans , a combination of direct staining methods and expression analysis of neurotransmitter-specific enzymes and transporters have defined the probably complete complement of GABAergic , glutamatergic and aminergic neurotransmitter systems . Specifically , out of the 118 anatomically distinct neuron classes in the hermaphrodite ( amounting to a total of 302 neurons ) , six classes ( 26 neurons ) are GABAergic ( McIntire et al . , 1993 ) , 38 are glutamatergic ( 78 neurons ) ( Serrano-Saiz et al . , 2013 ) and 13 ( 26 neurons ) are aminergic ( i . e . serotonergic , dopaminergic , etc . ; Chase and Koelle , 2007 ) . One prominent neurotransmitter system – the cholinergic system – has not been completely mapped . Antibody staining against the vesicular acetylcholine ( ACh ) transporter , VAChT ( encoded by unc-17 ) and the ACh-synthesizing choline acetyltransferase ChAT ( encoded by cha-1 ) revealed the cholinergic identity of a number of neurons in the nervous system ( Alfonso et al . , 1993; Duerr et al . , 2008 ) . However , due to the synaptic localization of the VAChT and ChAT proteins , expression could only be unambiguously assigned to about one dozen neuron classes , mostly in the ventral nerve cord and a few isolated head and tail neurons ( see Table 1 for a summary of previous studies on cholinergic neuron identity ) . The authors of these previous studies explicitly noted that many additional VAChT/ChAT-expressing neuron classes await identification ( Duerr et al . , 2008 ) . Ensuing studies using reporter genes that capture cis-regulatory elements of parts of the unc-17/VAChT locus identified the cholinergic identity of a few additional neuron classes ( Table 1 ) , but the extent to which ACh is used in the nervous system has remained unclear . In the male nervous system , composed of 23 additional neuron classes , neurotransmitter identities are even less well defined ( not just ACh , but other systems as well ) . Here , we map the usage of ACh in both the hermaphrodite and male nervous systems . We show that ACh is the most broadly used neurotransmitter in the C . elegans nervous system , employed by more than half of all neurons . 10 . 7554/eLife . 12432 . 003Table 1 . Cholinergic neurons in the hermaphrodite . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 003Neuron typeNeuron classVAChT/ChAT1ChT 2AChE 3Co-transmitterPrevious ID 4Sensory neuron ( 9 classes ) ADF L/R ++ ++ Serotonin no ALN L/R ++ ++ yes 5 ASJ L/R + ++ no AWB L/R ++ ++ no IL2 D/V L/R +++ +++ ace-3/4 yes 6 PLN L/R ++ ++ yes 5 URA D/V L/R +++ ++ ace-3/4 yes 6 URB L/R ++ + ace-3/4 yes 6 URX L/R ++ ++ ace-3/4 no Interneuron ( 19 classes ) AIA L/R +++ ++ ace-3/4 yes 7 AIN L/R ++ +++ yes 6 AIY L/R +++ +++ yes 7 AVA L/R +++ ++ ace-2 no AVB L/R +++ ++ ace-2 no AVD L/R +++ ++ ace-2 no AVE L/R +++ ++ ace-2 no AVG + no DVA +++ +++ ace-2; ace-3/4 no 8 PVC L/R +++ + yes 5 PVN L/R ++ no PVP L/R +++ ++ yes 5 RIB L/R ( + ) * +++ no RIF L/R ++ ++ no RIH +++ ++ ace-2; ace-3/4 Serotonin no RIR ++ ++ no RIV L/R ++ ++ ace-3/4 no SAA D/V L/R ++ no SDQ L/R ++ ++ yes 5 Motor neuron ( 17 classes ) AS1-11 ++ ++ ace-2 yes 5 DA1-9 ++ ++ ace-2 yes 5 DB1-7 ++ ++ ace-2 yes 5 HSN L/R ++ Serotonin yes 5 PDA ++ ++ ace-2; ace-3/4 no PDB ++ ++ no RMD D/V L/R +++ +++ ace-3/4 yes 5 RMF L/R ++ ++ no RMH L/R ++ ++ no SAB D V L/R ++ ++ yes 9 SIA D/V L/R +++ ++ ace-3/4 no SIB D/V L/R +++ ++ no SMB D/V L/R +++ ++ no SMD D/V L/R +++ +++ ace-3/4 no 10 Motor neuron ( 17 classes ) VA1-12 ++ ++ ace-2 yes 5 VB1-11 ++ ++ ace-2 yes 5 VC1-3 VC6 ++ ++ yes 5 VC4-5 ++ Serotonin yes 5 Pharyngeal Polymodal ( 7 classes ) I1 L/R ++ no I3 ++ no MC L/R ++ yes 11 M1 ++ no M2 L/R ++ no M4 +++ +++ ace-2 no M5 +++ +++ no unc-17 ( + ) : 52 classes , 159 neurons See the legend to Figure 2A and Table 2 for notes on neuron classification . Data for the male nervous system is shown in Table 5 . '+' indicate relative expression levels . See Figure 1 for images . *Expression of cho-1 in the RIB neurons is strong but unc-17 expression is , at best , very dim . 1Gray shading indicates unc-17/cha-1 ( VAChT/ChAT ) expression as assessed by fosmid reporter and antibody staining . 2Gray shading indicates cho-1 ( ChT ) expression as assessed by fosmid reporters . 3Gray shading indicates reporters expression of one of the C . elegans ace ( AChE ) genes . 4Previously identified as a cholinergic neuron: 'yes' – see indicated references . 'no' - newly identified in this study . Only published data is considered , personal communications in Rand and Nonet ( 1997 ) were not taken into consideration . 5Duerr et al . ( 2008 ) . 6Zhang et al . ( 2014 ) . 7Altun-Gultekin et al . ( 2001 ) . 8Previously proposed to be DVC ( Duerr et al . , 2008 ) but based on position and markers reassigned to DVA . 9Zhao and Nonet ( 2000 ) . 10Based on our identification as SMB as cholinergic , Kim et al . ( 2015 ) demonstrated that lim-4 controls SMB cholinergic identity ( see also Table 6 ) . 11Raizen et al . ( 1995 ) . The tremendous benefits of a neurotransmitter map include the ability to precisely dissect and understand neuronal circuit function . For example , knowledge of the cholinergic identity of the AIY interneuron ( Altun-Gultekin et al . , 2001 ) helped to define the two distinct behavioral outputs of AIY , one controlled via an ACh-mediated activation of the RIB interneuron and another controlled by ACh-mediated inhibition of the AIZ interneuron , via an ACh-gated chloride channel ( Li et al . , 2014 ) . The cholinergic neurotransmitter map presented here will provide a resource to further functionally dissect circuit function in the C . elegans nervous system . Since neurotransmitter identity represents a key feature of a neuron , the knowledge of the cholinergic identity provides a resource for studying how a neuron adopts its specific fate during development . For example , the assignment of glutamatergic identity to a host of distinct C . elegans neurons has enabled us to define phylogenetically conserved regulatory features of glutamatergic neuron differentiation ( Serrano-Saiz et al . , 2013 ) . Moreover , the long-known cholinergic identity of ventral cord motor neurons provided an entry point to study how their terminal differentiation is controlled ( Kratsios et al . , 2011; 2015 ) . Previous studies describing the mechanism of cholinergic identity regulation have pointed to a modular control system in which neuron-type specific combinations of transcription factors turn on cholinergic pathway genes ( Altun-Gultekin et al . , 2001; Kratsios et al . , 2011; Zhang et al . , 2014 ) . Since previous studies only examined a relatively small number of neurons , the problem of cholinergic identity regulation has not yet encompassed a circuit level analysis . Through a genetic screen and a candidate gene approach we reveal common themes in the form of circuit-associated transcription factors that control the identity of all neurons within defined circuits or circuit-associated network motifs . Taken together , we anticipate that neurotransmitter maps like those provided here represent an invaluable resource for the C . elegans community that will serve as a high-resolution starting point for various types of behavioral and developmental analyses .
Cholinergic neurotransmitter identity is defined by the expression of the enzyme choline acetyltransferase ( ChAT; encoded by cha-1 in C . elegans ) and the vesicular ACh transporter ( VAChT; encoded by unc-17 in C . elegans ) ; see Figure 1A for a description of the cholinergic pathway genes . Co-expression of these two genes is ensured via their organization into an operon-like structure called the cholinergic locus ( Figure 1B ) . This operon-like organization is conserved from invertebrates to vertebrates ( Eiden , 1998 ) . Other possible diagnostic features of cholinergic neurons often used in vertebrates are the expression of the enzyme that breaks down ACh , acetylcholinesterase ( AChE/ace; four genes in C . elegans; [Arpagaus et al . , 1998] ) and the reuptake transporter of the breakdown product choline ( ChT; encoded by cho-1 in C . elegans [Okuda et al . , 2000] ) . Whether these genes are expressed in all cholinergic neurons and/or restricted to all cholinergic neurons is , however , unclear . 10 . 7554/eLife . 12432 . 004Figure 1 . Expression of cholinergic pathway genes in the adult C . elegans hermaphrodite . ( A ) Cholinergic pathway genes . Ch = choline; ACh = acetylcholine; ChAT = choline acetyltransferase; VAChT = vesicular ACh transporter , AChE = ACh esterase , ChT = choline transporter . ( B ) Fosmid reporters used in this study . The unc-17 fosmid reporter was kindly provided by the TransgeneOme project ( Sarov et al . , 2012 ) . It was previously reported that the expression of unc-17/VAChT and cha-1/ChAT overlap completely ( Mathews et al . , 2015 ) . ( C ) unc-17 and cho-1 fosmid reporter expression in an L4 hermaphrodite . The fluorescent reporter inserted into the cho-1 locus is targeted to the nucleus ( see Materials and methods ) , while the fluorescent reporter inserted into the unc-17 locus is fused directly to the unc-17 gene ( resulting in cytoplasmic localization ) . ( D , E ) unc-17 and cho-1 fosmid reporter expression in head ( D ) , retrovesicular ganglion and tail ganglia ( E ) . In ( E ) bottom panels , neurons are labeled with a green pan-neuronal marker , ric-19 . Transgenes: otIs576 = unc-17 fosmid reporter; otIs544 = cho-1 fosmid reporter , otIs380 = ric-19 reporter ( Stefanakis et al . , 2015 ) . ( F ) Immunofluorescent staining for endogenous UNC-17 protein of unc-104 ( e1265 ) animals that express the cho-1 fosmid reporter transgene otIs544 . ( G ) Co-labeling cholinergic ( cho-1/ChT-positive ) and glutamatergic ( eat-4/VGLUT-positive ) neurons illustrate no overlap in neurotransmitter ACh and Glu expression , and co-labeling with pan-neuronal marker rab-3 illustrates that most neurons now have a neurotransmitter assignment . Transgenes: otIs544 = cho-1 fosmid reporter , otIs388 = eat-4 fosmid reporter ( Serrano-Saiz et al . , 2013 ) , otIs355 = rab-3 reporter . ( H ) ace/AChE genes are expressed in a subset of cholinergic neurons and in non-cholinergic neurons . ace-1 fosmid reporter expression in head neurons ( left panel ) . ace-2 fosmid reporter expression in head neurons together with cho-1 fosmid reporter ( middle panel ) . ace-3/4 reporter expression together with cho-1 fosmid reporter in head neurons ( right panel ) . Transgenes: otEx4435 = ace-1 fosmid reporter; otEx4431 = ace-2 fosmid reporter; fpIs1 = ace-3/4 transcriptional reporter . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 00410 . 7554/eLife . 12432 . 005Figure 1—figure supplement 1 . Neuronal cell identification . Neuronal identity was confirmed by crossing cho-1 fosmid reporter and/or unc-17 fosmid reporter with specific markers . ( A ) ADF and RIH were labeled by cho-1 ( otIs354 ) and cat-1 ( otIs625 ) . ( B ) ASJ and AWB were labeled by cho-1 ( otIs354 ) and DiI staining . ( C ) AVA , AVE and AVD were labeled by cho-1 ( otIs544 ) and nmr-1 ( akIs3 ) . ( D ) AVA , AVE and AVD were also labeled by cho-1 ( otIs544 ) and glr-1 ( hdIs30 ) . ( E ) AVB was labeled by cho-1 ( otIs544 ) and acr-15 ( wdEx290 ) . ( F ) AVB was also labeled by cho-1 ( otIs544 ) and sra-11 ( otIs123 ) . ( G ) AVB was not labeled by glr-1 ( hdIs30 ) as had been previously published ( Brockie et al . , 2001 ) . ( H ) AWA was labeled by odr-10 ( kyIs37 ) but did not show cho-1 ( otIs544 ) expression . ( I ) AWA was labeled by gpa-4 ( otEx6381 ) but did not show unc-17 ( otIs576 ) expression . ( J ) AVG and RIF were labeled by cho-1 ( otIs544 ) and odr-2 ( otEx4452 ) . ( K ) DVA was labeled by ser-2 ( otIs358 ) and cho-1 ( otIs544 ) . ( L ) PDA was labeled by cho-1 ( otIs544 ) and ace-3/4 ( fpIs1 ) . ( M ) ALN and PLN were labeled by cho-1 ( otIs544 ) and lad-2 ( otIs439 ) . ( N ) SMB and SMD were labeled by cho-1 ( otIs544 ) and lad-2 ( otIs439 ) . ( O ) SMD and RIV were labeled by cho-1 ( otIs544 ) and lad-2 ( otIs439 ) . ( P ) SIA and SIB were labeled by cho-1 ( otIs544 ) and ceh-24 ( ccIs4595 ) . ( Q ) URX , RIR and RIH were labeled by cho-1 ( otIs544 ) and unc-86 ( otIs337 ) . ( R ) VC4 and VC5 were labeled by cat-1 ( otIs221 ) but not by cho-1 ( otIs544 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 00510 . 7554/eLife . 12432 . 006Figure 1—figure supplement 2 . Neurotransmitter identity of pharyngeal neurons . ( A ) The different panels show the expression of unc-17 ( otIs576 ) and cho-1 ( otIs544 ) fosmids in the pharyngeal neurons in the anterior and posterior bulbs . Only M4 and M5 express both fosmids . Schematic for the cholinergic pharyngeal neurons is shown . ( B ) Expression of eat-4 ( otIs518 ) and cho-1 ( otIs344 ) fosmids in the pharyngeal neuron M5 . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 00610 . 7554/eLife . 12432 . 007Figure 1—figure supplement 3 . Expression of unc-17 and cho-1 fosmid reporters in the male tail . ( A ) The top panel shows the male pre-anal ganglion on a ventral view and the bottom panel shows the pre-anal ganglion and tail neurons in a lateral view . ( B ) Male tail ventral view where PVS , PVU and the male-specific neurons PVZ and HOB were labeled by cho-1 fosmid and ida-1::gfp reporter . Transgenes: otIs576 = unc-17 fosmid reporter; otIs544 = cho-1 fosmid reporter; inIs179 = ida-1 reporter . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 007 To define cholinergic neuron types , we generated transgenic lines expressing fosmid based reporters for the unc-17 and cha-1 locus , the cho-1 locus and several ace genes ( Figure 1B ) . Fosmids contain 30–40 kb genomic sequences , including genes upstream and downstream of the gene of interest and usually contain all cis-regulatory information involved in regulating expression of a specific gene . Differently colored fluorescent proteins were used to assess the relative overlap of these genes to one another ( Figure 1C–E ) . The fosmid lines that monitor cho-1 and ace-1/-2 expression are nuclear localized reporters , in which the fluorescent tag is separated from the respective genomic locus by an SL2 trans-splicing event and targeted to the nucleus ( see Materials and methods ) . The fosmid line for the unc-17 locus is , in contrast , a direct fusion of gfp to the unc-17 gene , thereby revealing the subcellular localization of unc-17 . Multiple lines for each reporter transgene were analyzed and no differences between lines were found ( for example , green and red fluorescent signals from a cho-1fosmid::mCherry transgenic line and an independent cho-1fosmid::yfp transgenic line perfectly overlap; data not shown ) . Preliminary neuron identifications were done based on cell position and axonal projections . These identifications were then confirmed for each neuron by crossing the unc-17 and/or cho-1 fosmid reporter strains with a differently colored reporter with a known , neuron type-specific expression pattern ( Figure 1—figure supplement 1; see also Materials and methods ) . Furthermore , we validated unc-17 fosmid reporter expression by immunofluorescent staining with an antibody generated against the UNC-17 protein ( Duerr et al . , 2008 ) . As previously noted , the punctate localization of UNC-17 protein , as detected with the UNC-17 anti-serum , limits the ability to reliably identify cells in the absence of markers ( Duerr et al . , 2008 ) . However , immunostaining for UNC-17 in combination with the nuclear localization of cho-1 fosmid reporter in an unc-104 mutant background ( UNC-104 is required for UNC-17 transport to synapses ) , allowed us to precisely define the complete set of cells that stain for endogenous UNC-17 protein . We found the overlap of UNC-17 antibody staining with cho-1 fosmid reporter expression to be the same as the overlap of unc-17 fosmid reporter expression with cho-1 fosmid reporter expression ( Figure 1F ) , thereby validating the reliability of fosmid reporter expression patterns . unc-17/VAChT expression defines cholinergic identity and is present in 52 of the 118 classes of adult hermaphroditic neurons , amounting to 159 out of 302 neurons ( Figure 1 , Figure 2A; Table 1; Figure 1—figure supplement 2 ) . Compared to all other neurotransmitter systems , this makes ACh the most abundantly employed neurotransmitter system in C . elegans ( Glu: 38 classes , GABA: 6 classes , Aminergic: 13 classes , six of which are exclusively aminergic; Figure 2A , Table 2 ) . The abundance of ACh usage is illustrated in an even more striking manner if one considers the C . elegans connectome ( White et al . , 1986 ) : 85% ( 100/118 ) of all neuron classes are innervated by a cholinergic neuron ( Table 3 ) . With one exception ( the highly unusual CAN neurons , which show very little synaptic connectivity with any other neuron ) , all neurons that do not receive cholinergic input are either themselves cholinergic neurons or innervate neurons that are cholinergic ( Table 3 ) . In other words , all but one neuron class in the C . elegans nervous system are either cholinergic , receive cholinergic input or innervate a cholinergic neuron . 10 . 7554/eLife . 12432 . 008Figure 2 . Distribution of neurotransmitters throughout the nervous system of the hermaphrodite . ( A ) Pie chart with numbers/distributions of cholinergic ( this study ) , glutamatergic ( Serrano-Saiz et al . , 2013 ) , GABAergic ( McIntire et al . , 1993 ) and aminergic ( Chase and Koelle , 2007 ) neurons ( including pharyngeal neurons ) . Inset: Pie charts of extrapharyngeal sensory , motor- and interneurons . Neurons that contain a classic fast transmitter plus an aminergic transmitter ( e . g . RIH ) are counted in the fast transmitter category . Classification of C . elegans neurons into sensory , inter- and motor neurons is complicated by the fact that a subset of sensory neurons are also motor neurons , i . e . synapse directly onto muscle ( we count those neurons here only as sensory neurons ) . Conversely , a large number of motor neurons also extensively synapse onto other motor neurons or interneurons and hence classify as 'interneuron' as well; these neurons are shown exclusively in the motor neuron category . A number of neurons that were originally assigned as 'interneurons' by John White and colleagues are now considered motor neurons ( because of the more recent identification of NMJs; e . g . SIA , SIB , SAB neurons ) , or are considered sensory neurons ( because of their position in connectivity diagrams or expression of molecular markers; e . g . URA , URB , URXY , URY ) . See Table 2 for a complete list of neurons and their neurotransmitter assignment . Lastly , we note that unpublished results from our lab demonstrate that at least two additional interneurons , not shown here , utilize GABA ( M . Gendrel and O . H . , unpubl . data ) . ( B ) Distance of sensory neurons to motor output ( processing depth ) of cholinergic and glutamatergic sensory neurons . ( C ) Location of neurons with different neurotransmitter identities in the head ganglia . ( D ) Neurotransmitter identity does not track with lineage history . Neurotransmitter identity is superimposed on the embryonic lineage diagram ( Sulston et al . , 1983 ) , with each color line indicating one neuron type with a defined identity . White lines indicate no known neurotransmitter identity , gray lines indicate non-neuronal cells . Lines with two colors illustrate co-transmitter identities . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 00810 . 7554/eLife . 12432 . 009Table 2 . Neurotransmitter map of the hermaphrodite nervous system . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 009Neuron classNeuronNeurotransmitterNotesADA ADAL Glu ADAR Glu ADE ADEL DA ADER DA ADF ADFL ACh & 5HT ADFR ACh & 5HT ADL ADLL Glu ADLR Glu AFD AFDL Glu AFDR Glu AIA AIAL ACh AIAR ACh AIB AIBL Glu AIBR Glu AIM AIML Glu & 5HT AIMR Glu & 5HT AIN AINL ACh AINR ACh AIY AIYL ACh AIYR ACh AIZ AIZL Glu AIZR Glu ALA ALA Unknown ( orphan ) Newly assigned as mechanosensory ( based on Sanders et al . , 2013 ) ALM ALML Glu ALMR Glu ALN ALNL ACh Classified as sensory because of expression of oxygen sensors ALNR ACh AQR AQR Glu AS AS1 ACh AS2 ACh AS3 ACh AS4 ACh AS5 ACh AS6 ACh AS7 ACh AS8 ACh AS9 ACh AS10 ACh AS11 ACh ASE ASEL Glu ASER Glu ASG ASGL Glu ASGR Glu ASH ASHL Glu ASHR Glu ASI ASIL Unknown ( orphan ) ASIR Unknown ( orphan ) ASJ ASJL ACh ASJR ACh ASK ASKL Glu ASKR Glu AUA AUAL Glu AUAR Glu AVA AVAL ACh AVAR ACh AVB AVBL ACh AVBR ACh AVD AVDL ACh AVDR ACh AVE AVEL ACh AVER ACh AVF AVFL Unknown ( orphan ) AVFR Unknown ( orphan ) AVG AVG ACh AVH AVHL Unknown ( orphan ) AVHR Unknown ( orphan ) AVJ AVJL Unknown ( orphan ) AVJR Unknown ( orphan ) AVK AVKL Unknown ( orphan ) AVKR Unknown ( orphan ) AVL AVL GABA AVM AVM Glu AWA AWAL Unknown ( orphan ) AWAR Unknown ( orphan ) AWB AWBL ACh AWBR ACh AWC AWCL Glu AWCR Glu BAG BAGL Glu BAGR Glu BDU BDUL Unknown ( orphan ) BDUR Unknown ( orphan ) CAN CANL unknown MA ( cat-1 ) CANR unknown MA ( cat-1 ) CEP CEPDL DA CEPDR DA CEPVL DA CEPVR DA DA DA1 ACh DA2 ACh DA3 ACh DA4 ACh DA5 ACh DA6 ACh DA7 ACh DA8 ACh DA9 ACh DB DB1/3 ACh DB2 ACh DB3/1 ACh DB4 ACh DB5 ACh DB6 ACh DB7 ACh DD DD1 GABA DD2 GABA DD3 GABA DD4 GABA DD5 GABA DD6 GABA DVA DVA ACh DVB DVB GABA DVC DVC Glu FLP FLPL Glu FLPR Glu HSN HSNL ACh & 5HT HSNR ACh & 5HT IL1 IL1DL Glu Also a clear motor neuron IL1DR Glu IL1L Glu IL1R Glu IL1VL Glu IL1VR Glu IL2 IL2DL ACh Also a clear motor neuron IL2DR ACh IL2L ACh IL2R ACh IL2VL ACh IL2VR ACh LUA LUAL Glu LUAR Glu OLL OLLL Glu OLLR Glu OLQ OLQDL Glu OLQDR Glu OLQVL Glu OLQVR Glu PDA PDA ACh PDB PDB ACh PDE PDEL DA PDER DA PHA PHAL Glu PHAR Glu PHB PHBL Glu PHBR Glu PHC PHCL Glu PHCR Glu PLM PLML Glu PLMR Glu PLN PLNL ACh PLNR ACh PQR PQR Glu PVC PVCL ACh PVCR ACh PVD PVDL Glu PVDR Glu PVM PVM Unknown ( orphan ) PVN PVNL ACh Only very few minor NMJs , more prominent neuron-neuron synapses PVNR ACh PVP PVPL ACh PVPR ACh PVQ PVQL Glu PVQR Glu PVR PVR Glu PVT PVT Unknown ( orphan ) PVW PVWL Unknown ( orphan ) PVWR Unknown ( orphan ) RIA RIAL Glu RIAR Glu RIB RIBL ACh RIBR ACh RIC RICL Octopamine RICR Octopamine RID RID Unknown ( orphan ) RIF RIFL ACh RIFR ACh RIG RIGL Glu RIGR Glu RIH RIH ACh & 5HT RIM RIML Glu & Tyramine RIMR Glu & Tyramine RIP RIPL Unknown ( orphan ) RIPR Unknown ( orphan ) RIR RIR ACh RIS RIS GABA RIV RIVL ACh Only very few minor NMJs , more prominent neuron-neuron synapses RIVR ACh RMD RMDDL ACh RMDDR ACh RMDL ACh RMDR ACh RMDVL ACh RMDVR ACh RME RMED GABA RMEL GABA RMER GABA RMEV GABA RMF RMFL ACh RMFR ACh RMG RMGL Unknown ( orphan ) RMGR Unknown ( orphan ) RMH RMHL ACh RMHR ACh SAA SAADL ACh SAADR ACh SAAVL ACh SAAVR ACh SAB SABD ACh Makes clear neuromuscular junctions SABVL ACh SABVR ACh SDQ SDQL ACh SDQR ACh SIA SIADL ACh Makes clear neuromuscular junctions SIADR ACh SIAVL ACh SIAVR ACh SIB SIBDL ACh Makes clear neuromuscular junctions SIBDR ACh SIBVL ACh SIBVR ACh SMB SMBDL ACh SMBDR ACh SMBVL ACh SMBVR ACh SMD SMDDL ACh SMDDR ACh SMDVL ACh SMDVR ACh URA URADL ACh Also a clear motor neuron URADR ACh URAVL ACh URAVR ACh URB URBL ACh URBR ACh URX URXL ACh URXR ACh URY URYDL Glu URYDR Glu URYVL Glu URYVR Glu VA VA1 ACh VA2 ACh VA3 ACh VA4 ACh VA5 ACh VA6 ACh VA7 ACh VA8 ACh VA9 ACh VA10 ACh VA11 ACh VA12 ACh VB VB1 ACh VB2 ACh VB3 ACh VB4 ACh VB5 ACh VB6 ACh VB7 ACh VB8 ACh VB9 ACh VB10 ACh VB11 ACh VC VC1 ACh VC2 ACh VC3 ACh VC4 ACh & 5HT VC5 ACh & 5HT VC6 ACh VD VD1 GABA VD2 GABA VD3 GABA VD4 GABA VD5 GABA VD6 GABA VD7 GABA VD8 GABA VD9 GABA VD10 GABA VD11 GABA VD12 GABA VD13 GABA Summary for extrapharyngeal neurons Sensory neuron: Sensory neuron: 38/104 classes ACh: 9 classes 87/282 total neurons Glu: 22 Motor neuron: GABA: 0 24/104 Aminergic: 3 ( all Dopa ) 118/282 Unknown: 4 ( ASI , AWA , PVM , ALA ) Interneuron Motor neuron: 42/104 ACh: 17 classes 77/282 Glu: 1 ( RIM ) GABA: 5 Aminergic: 0 Unknown: 1 ( RMG ) Interneuron: ACh: 19 classes Glu: 11 GABA: 1 ( RIS ) Aminergic: 2 ( CAN , RIC ) Unknown: 9 Pharyngeal neurons I1 I1L ACh Due to connectivity and rudimentary sensory endings , all polymodal I1R ACh I2 I2L Glu I2R Glu I3 I3 ACh I4 I4 Unknown ( orphan ) I5 I5 Glu & 5HT I6 I6 Unknown ( orphan ) M1 M1 ACh M2 M2L ACh M2R ACh M3 M3L Glu M3R Glu M4 M4 ACh M5 M5 ACh MC MCL ACh MCR ACh MI MI Glu NSM NSML 5HT NSMR 5HT 10 . 7554/eLife . 12432 . 010Table 3 . Neurons receiving cholinergic inputs . Includes pharyngeal neurons . Data from www . wormwiring . org . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 010Connectivity *Neuron class#Receiving ACh input Cholinergic neurons ADF , AIA , AIN , AIY , ALN , AS , ASJ , AVA , AVB , AVD , AVE , AWB , DA , DB , DVA , I3 , IL2 , M2 , M4 , PLN , PVC , PVN , PVP , RIB , RIF , RIH , RIR , RIV , RMD , RMF , RMH , SAA , SAB , SDQ , SIA , SIB , SMB , SMD , URA , URB , URX , VA , VB , VC 44 Non-cholinergic neurons ADA , ADE , ADL , AFD , AIB , AIM , AIZ , ALA , ALM , AQR , ASE , ASG , ASH , ASI , ASK , AUA , AVF , AVH , AVJ , AVK , AVL , ASA , AWC , BAG , BDU , CEP , DD , DVC , I2 , I4 , I5 , IL1 , LUA , M1 , M3 , MC , NSM , OLL , OLQ , PQR , PVQ , PVR , PVT , PVW , RIA , RIC , RID , RIG , RIM , RIP , RIR , RIS , RME , RMG , URY , VD 56 Receiving no ACh input Cholinergic neuron AVG , HSN , I1 , M5 , PDA , PDB 6 Innervate cholinergic neuron AVM , DVB , FLP , I6 , MI , PDE , PHA , PHB , PHC , PLM , PVD , PVM 12 Neither of the above CAN 1 There does not appear to be any change in neurotransmitter identities in the first larval stage versus the adult stage , with the obvious exception of postembryonically generated neurons ( mostly motor neurons ) . Expression of the cholinergic locus ( unc-17 and cha-1 ) commences in the 1 . 5-fold stage of embryogenesis and by the three-fold stage , expression is seen in all cholinergic neurons ( data not shown ) . cho-1/ChT expression extensively correlates with expression of unc-17/VAChT , both in terms of onset ( by threefold stage; data not shown ) and cellular specificity in the mature nervous system . In the hermaphrodite worm , all neurons that express cho-1 also express unc-17/VAChT ( even though expression of unc-17 may be very low in at least one class , RIB ) , while 11 out of the 52 unc-17 ( + ) classes do not express cho-1 ( half of these neuron classes are in the pharyngeal nervous system; Figure 1;Table 1; Figure 1—figure supplement 2 ) . In contrast , as a previous analysis of small reporter gene fusions already suggested ( Combes et al . , 2003 ) , expression of the acetylcholinesterase ( AChE ) -encoding ace genes does not correlate with unc-17 expression . First , only one third of all cholinergic neuron classes express an ace gene ( Table 1 ) ; and second , expression is observed in body wall muscle as well as in a few non-cholinergic neurons ( Figure 1H ) . Given that the diffusible ACE proteins are secreted into the synaptic cleft , it may not come as a surprise that their site of synthesis does not necessarily match the site of ACh synthesis and release . The situation is similar in vertebrates; the only vertebrate AChE gene is expressed in cholinergic neurons , but the overlap is not complete and expression can also be observed in non-cholinergic neurons ( Gwyn and Flumerfelt , 1971; Levey et al . , 1984; Reiss et al . , 1996 ) . A list of all C . elegans neurons with their presently assigned neurotransmitter identity is shown in Table 2 . There are no overlaps in usage of the main neurotransmitter systems glutamate , GABA and ACh in the core nervous system of the hermaphrodite , but within the pharynx , one single neuron , the motor neuron M5 , strongly expresses both cholinergic pathway genes unc-17 and cho-1 and , albeit very weakly , the glutamatergic marker eat-4/VGLUT ( Figure 1—figure supplement 2 ) . We visualized the general lack of overlap with a transgenic line that expresses three different , nuclear localized , fluorescent tags – one marking cholinergic neurons ( cho-1fosmid::mChOpti ) , one marking glutamatergic neurons ( eat-4/VGLUTfosmid::yfp ) and one marking all neurons ( rab-3prom::bfp strain ) ( Figure 1G;Video 1 ) . There are , however , some overlaps of cholinergic identity with aminergic identity in the core nervous system: the ADF , HSN , RIH and VC4/5 neurons are cholinergic , but also serotonergic ( Duerr et al . , 2001; Sze et al . , 2000 ) ; similarly , some glutamatergic neurons are also aminergic ( Serrano-Saiz et al . , 2013 ) . The case of the postembryonically generated , hermaphrodite-specific VC motor neurons is particularly notable because of the distinct identities of specific VC subtypes . All six VC neurons express unc-17/VAChT and are therefore cholinergic , but VC4 and VC5 are also serotonergic ( Duerr et al . , 1999 ) . Notably , the expression of serotonergic identity in VC4 and VC5 correlates with a failure to express cho-1/ChT , which is only expressed in VC1 , 2 , 3 and VC6 ( Figure 1—figure supplement 1 , Figure 5C ) . VC4 and VC5 innervate vulval muscles and some aspects of their identity ( namely expression of the unc-4 gene in VC4/5 , but not VC1 , 2 , 3 , 6 ) are controlled by signals from vulval tissues ( Zheng et al . , 2013 ) . We find that elimination of this vulval signal , or genetic elimination of the target muscle of the VC4/5 neurons ( vulval muscle ) , does not impinge on the absence of cho-1 expression in VC4/5 ( data not shown ) . 10 . 7554/eLife . 12432 . 011Video 1 . Cholinergic and glutamatergic head neurons . Confocal image stack of a transgenic worm expressing cho-1::mChopti ( otIs544 ) and eat-4::yfp ( otIs388 ) fosmid reporter gene constructs in the head . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 01110 . 7554/eLife . 12432 . 012Figure 3 . Neurotransmitter distribution in nervous system-wide circuit diagrams . ( A , B ) Circuit diagrams , taken from White et al . ( 1986 ) , with neurotransmitter identities added in colors , as indicated . Panel A shows what White et al . called the “Circuitry associated with motoneurons in the nerve ring” and panel B shows the “Circuitry associated with the motoneurons of the ventral cord” . ( C ) A visualization of the C . elegans connectome that reflects signal flow through the network as well as the closeness of neurons in the network , as previously proposed and described ( Varshney et al . , 2011 ) . Coordinates from the diagram were kindly provided by Lav Varshney . The vertical axis represents the signal flow depth of the network , i . e . the number of synapses from sensory to motor neurons . The horizontal axis represents connectivity closeness . We superimpose here neurotransmitter identity onto this network diagram , illustrating some network cluster enriched for ACh usage ( shaded gray ) . ( D ) A graphic representation that focuses on processing depth , illustrating whether a neurotransmitter is used more frequently in upper ( sensory ) or lower ( motor ) layers of the network . ( E ) Network motifs enriched in the C . elegans connectome and their neurotransmitter usage . Colors indicate if the neurons in this position are enriched for the usage of Glu or ACh . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 012 ACh is used by sensory neurons , interneurons and motor neurons . Of the 45 extrapharyngeal cholinergic neuron classes , 9 are sensory neurons , 19 are interneurons and 17 are motor neurons ( Figure 2A; Table 2; we only consider extrapharyngeal neurons because most pharyngeal neurons are polymodal , i . e . have sensory , inter- and motor neuron features; Albertson and Thomson , 1976; D . H . H . unpubl . data ) . Compared to other transmitter systems , motor neurons have a preference for employing ACh ( 17/24 extrapharyngeal motor neuron classes are cholinergic; Figure 2A; Table 2 ) . In contrast , sensory neurons are predominantly glutamatergic ( 22/38 use Glu ) , but there is nevertheless an appreciable number of cholinergic sensory neurons ( 9/38 extrapharyngeal sensory neuron classes use ACh; Table 2 ) . Intriguingly , most cholinergic sensory neurons have very shallow processing depth , i . e . are closely connected to the motor system ( Figure 2B ) . Two ( IL2 and URA ) directly synapse onto muscle ( i . e . are sensory-motor neurons ) , another four ( ALN , PLN , ADF and URB ) synapse directly onto motor neurons , while another two ( URX and AWB ) synapse onto cholinergic command interneurons that innervate motor neurons . The latter two cases are the only cases in the entire C . elegans nervous system where a multi-neuron pathway from sensory , via inter- to motor neurons is entirely made of exclusively one neurotransmitter system . In comparison , glutamatergic sensory neurons do not display such a narrow processing depth ( Figure 2B ) . The predominance of ACh as a neurotransmitter does not solely stem from its widespread usage in motor neurons . ACh is also the most broadly used neurotransmitter of interneurons ( 19 classes; compared to 11 glutamatergic; Figure 2A; Table 1 ) . In comparison to ACh and glutamate ( Glu ) , the neurotransmitter GABA is only very sparsely used by interneurons ( presently only 1 class; [McIntire et al . , 1993] ) ; this is no reflection of a paucity of inhibitory neurotransmission in C . elegans since both ACh and Glu can act as inhibitory neurotransmitter through the gating of postsynaptic chloride channels ( see below ) . The most notable set of interneurons to which we assigned a cholinergic neurotransmitter identity are the command interneurons , which are well-characterized central integrators of information flow in the nervous system that directly synapse onto motor neurons ( Chalfie et al . , 1985; Von Stetina et al . , 2006 ) . Their neurotransmitter identity was previously not known and we verified their cholinergic identity through a number of different co-stains ( summarized in Figure 1—figure supplement 1; see Materials and methods ) . Expression of the cho-1 fosmid reporter overlapped with expression of the glutamate receptors glr-1 and nmr-1 in the AVA , AVE , AVD and PVC command interneurons . To confirm the cholinergic identity of the AVB command interneuron , we crossed the cho-1 fosmid reporter with sra-11 and acr-15 reporters . Overlap of cho-1 with these two reporters allowed us to confirm that AVB expresses cholinergic identity genes . All command interneurons showed expression of the unc-17 fosmid reporter . Apart from assigning cholinergic neurotransmitter identity to different types of neurons ( sensory vs . inter vs . motor neurons ) , we examined whether cholinergic neurotransmitter identity correlates with other intrinsic neuronal features . We find that the adoption of cholinergic neurotransmitter identity does not correlate with position of the neuron within the nervous system , as shown in Figure 2C , with the notable exception of cholinergic motor neurons in the ventral head ganglion and along the ventral nerve cord . There is no correlation between the adoption of cholinergic identity and developmental history of the neurons . We arrived at this conclusion by mapping neurotransmitter identity onto the entire lineage diagram and not detecting any obvious lineage clusters of cells that uniquely employ ACh ( or any other neurotransmitter; Figure 2D ) . With the identification of the complete set of cholinergic neurons , and with the consideration of previously identified glutamatergic , GABAergic and monoaminergic neurons , a neurotransmitter identity can now be assigned to ~90% of all neuron classes ( 102/118 ) and total neurons ( 275/302; Table 2 ) . While some of the remaining orphan neurons ( e . g . the prominent olfactory neuron AWA ) contain small synaptic vesicles that are indicative of the usage of an as yet uncharacterized neurotransmitter system , about half of the remaining 16 'orphan' neuron classes display , according to John White’s EM analysis , a notable paucity or even absence of synaptic vesicles and/or are predominated by dark staining vesicles ( e . g . AVF , AVH , AVJ , RIP ) ( White et al . , 1986 ) , suggesting that these neurons either signal mostly via electrical synapses or via neuropeptides . The assignment of neurotransmitter identity to ~90% of neurons prompted us to take a system-wide view of neurotransmitter usage . We started by examining neurotransmitter usage within a number of specific circuitries described by John White and colleagues , including circuitries associated with amphid sensory neurons , with head motor neurons and with motor neurons in the ventral nerve cord ( White et al . , 1986 ) . While some circuitries show a mixed usage of different neurotransmitter systems ( one example shown in Figure 3A ) , the circuitry associated with the motor neurons of the ventral nerve cord show the striking feature of being mainly cholinergic ( with the exception of the GABAergic DD/VD motor neurons; Figure 3B ) . That is , not only do most motor neurons ( 'MNs' ) ( SAB , DA , DB , VA , VB , AS ) employ ACh , but all neurons that innervate these neurons ( and that are also strongly interconnected among each other ) are also cholinergic . This includes the command interneurons AVA , AVB , AVE , AVD , PVC , as well as the DVA interneuron , which is also closely associated with the motor circuit ( Figure 3B ) . Due to the extent of their interconnectedness , this group of six interneuron classes has previously been defined as a 'rich club' of neurons ( Towlson et al . , 2013 ) . The adoption of cholinergic identity within an entire functional circuit prompts the immediate question whether activity of the circuit plays a role in the expression of cholinergic genes . However , we find that genetic silencing of the C . elegans nervous system , achieved through elimination of the snb-1/synaptobrevin gene , has no impact on the expression of cholinergic identity markers in arrested L1 larvae ( data not shown ) . Taking a broader view we mapped neurotransmitter identity on a wiring diagram that reflects signal flow through the network as well as connectivity closeness of neurons in the network , as suggested by Varshney and colleagues ( Varshney et al . , 2011 ) ( Figure 3C ) . We also examined the parameter of 'processing depth' in isolation , as had been done previously ( Varshney et al . , 2011 ) . We considered the distance of each neuron from sensory input to motor output , assigned this relative position the parameter 'processing depth' and then the portion of neurons that use each neurotransmitter at each processing depth ( Figure 3D ) . Both types of representations quantify and effectively visualize what the identity of many of the cholinergic neurons already suggested: compared to other neurotransmitter systems ( particularly Glu ) ACh is enriched , but not exclusively located to lower levels of information processing . Another notable feature of this presentation is that it visualizes the connectivity closeness of distinct clusters of cholinergic motor neurons ( shaded in gray in Figure 3C ) ; these neurons are the head sensory-motor neurons , head motor neurons and the above-mentioned ventral nerve cord ( VNC ) motor neuron circuitry . We considered neurotransmitter usage within the several types of recurring network motifs , composed of three or four neurons , which have been described to be enriched within the C . elegans connectome , such as feedforward motifs of three neurons ( Milo et al . , 2002 ) . Our goal was to examine whether the usage of ACh ( or any other neurotransmitter system ) is biased for certain positions of a neuron within these motifs . Using previously described approaches ( Milo et al . , 2002 ) ( see Materials and methods ) , we identified five 3-neuron motifs and fourteen 4-neuron motifs that are significantly enriched in the C . elegans connectome using the latest connectivity dataset ( Figure 3E ) . We found either ACh or Glu to be enriched in specific positions in all but one of these motifs . ACh was enriched at a specific position in 11 out of these 19 motifs . Generally , there is a strong trend of ACh being more frequently used at the downstream end of the signaling flow within specific motifs , while Glu tends to be located at upstream positions within motifs ( Figure 3E ) , which is consistent with the processing depth analysis described above ( Figure 3D ) . In one 3-neuron motif , previously termed a 'regulated mutual motif' ( Milo et al . , 2004 ) , each one of the interconnected neurons is enriched for a specific neurotransmitter and we examined this motif more closely for reasons that will become evident in later sections of this paper . The general architecture of this motif is defined by one neuron ( 'A' ) innervating two reciprocally connected neurons ( 'B' and 'C'; Figure 3E ) . 224 occurrences of this motif can be found in the C . elegans hermaphroditic connectome . This motif is significantly enriched for the presence of cholinergic neurons in either position 2 or 3 , or in both . Notably , position 1 is significantly enriched for Glu usage . In 146 out of the 224 motif occurrences , ACh is used by both neuron '2' and '3' ( listed in Table 4 ) , and 134 of these 146 motifs break down into a number of two striking types . In the first type , reciprocally connected command interneurons are either innervated by a sensory neuron or by an interneuron ( Table 4 ) . In virtually all of these cases , the innervating sensory neurons are glutamatergic . In many cases , the reciprocally connected command interneurons are neurons that control different directions of movement ( forward vs . reverse; Table 4 ) . In the second type , the SMD or RMD head motor neurons are reciprocally connected and innervated again either by mostly glutamatergic sensory neurons or by interneurons ( Table 4 ) . 10 . 7554/eLife . 12432 . 013Table 4 . Occurences of the 'Regulated Mutual' network motif . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 013Type 1: Sensory>command interneuronsType 2: Interneurons>command interneuronsType 3: Sensory neurons>head motor neuronsType 4: Interneurons>head motor neuronsType 5: Egg laying circuitMiscellaneousSN CI CI opp . IN CI CI opp . SN hMN hMN IN hMN hMN ADEL AVAL AVAR ADAL AVAL AVAR CEPVL RMDDL RMDVR RIAL RMDDL RMDVR AIML AVFL AVFR PHBL VA12 AVAL ADER AVAL AVAR ADAL AVAR AVBL yes IL1DL RMDDR RMDVL RIAL RMDDR RMDVL AIMR RIFR HSNR PHCL VA12 AVAL ADER AVAR AVDR ADAL AVAR AVBR yes IL1DR RMDDL RMDVR RIAR RMDDL RMDVR AIMR AVFL AVFR VA12 PVCL PVCR ADER AVAR AVER ADAR AVAR AVBL yes IL1L RMDDL RMDVR RIAR RMDDR RMDVL AIMR AVFL HSNR AVEL DA01 AS01 ADLL AVAL AVAR ADAR AVAR AVBR yes IL1L RMDL RMDR RICR SMDDL SMDVR AIMR AVFR HSNL AVER DA01 AS01 ADLL AVAL AVDL ADAR AVAR AVDR IL1R RMDDR RMDVL RICR SMDDR SMDVL HSNL AVFL HSNR AVHL ADFR AWBR ADLL AVAR AVBL yes ALA AVAR AVER IL1R RMDL RMDR RIML SMDDR SMDVL AWAR ADFR AWBR ADLL AVAR AVDR AUAR AVAR AVER IL1VL RMDDL RMDVR RIMR RMDL RMDR CEPVR IL2VR URAVR ADLR AVAR AVBL AVBR AVAL AVDL IL1VR RMDDR RMDVL RIMR SMDDL SMDVR ADLR AVAR AVBR AVDL AVAR AVDR IL2L RMDL RMDR RIS RMDL RMDR ADLR AVAR AVDR AVEL AVAL AVAR OLLL SMDDL SMDVR RIVR SMDDL SMDVR ADLR AVAR PVCL yes AVER AVAL AVDL OLLR SMDDL SMDVR RMGR RMDL RMDR ALML PVCL PVCR yes AVER AVDL AVEL URYDL RMDDR RMDVL AQR AVAL AVAR AVG AVAL AVAR URYDR RMDDL RMDVR unc-42 unc-42 AQR AVAL AVDL AVG AVAR AVBL yes URYDR SMDDL SMDVR AQR AVAL PVCR yes AVG AVAR AVBR yes URYVL RMDDL RMDVR AQR AVAR AVBL yes AVG AVAR AVDR URYVR RMDDR RMDVL AQR AVAR AVBR yes AVJR AVAR AVBL yes AQR AVAR PVCR yes AVJR AVAR AVDR unc-42 unc-42 ASHL AVAL AVDL AVJR AVAR AVER ASHR AVAR AVBR yes AVJR AVAR PVCL yes ASHR AVAR AVER AVJR AVAR PVCR yes ASHR AVAR PVCL AVJR PVCL PVCR yes AVM PVCL PVCR BDUR AVAL PVCL yes BAGL AVAR AVER DVA AVAL PVCL yes FLPL AVAL AVAR DVC AVAL AVAR FLPL AVAL AVDL LUAR AVAL AVDL FLPL AVAL PVCR yes LUAR AVAL PVCR yes FLPL AVAR AVBL yes PVCR AVDL AVEL FLPL AVAR AVBR yes PVNL AVAL AVDL FLPL AVAR AVDR PVNL AVAL PVCL yes FLPL AVAR PVCR yes PVPL AVAL AVAR FLPR AVAL AVAR PVPL AVAL PVCL yes FLPR AVAL AVDL PVPL AVAL PVCR yes FLPR AVAR AVBL yes PVPL AVAR AVBL yes FLPR AVAR AVBR yes PVPL AVAR AVBR yes FLPR AVAR AVDR PVPL AVAR AVDR FLPR AVAR AVER PVPL AVAR PVCL yes FLPR AVDL AVEL PVPL AVAR PVCR yes PHBL AVAL AVAR PVPL PVCL PVCR PHBL AVAL AVDL PVPR AVAR AVBR yes PHBL AVAL PVCL yes PVPR AVAR PVCL yes PHBL AVAR PVCL yes PVPR AVAR PVCR yes PHBR AVAL AVAR PVPR PVCL PVCR PHBR AVAL AVDL RIBR AVAR AVER PHBR AVAL PVCL yes RICL AVAL AVAR PHBR AVAL PVCR yes RICR AVAL AVAR PHBR AVAR AVDR SDQL AVAL AVAR PHBR AVAR PVCL yes SDQL AVAL AVDL PHBR AVAR PVCR yes PHBR PVCL PVCR unc-3 unc-3 PHCL AVAL PVCL yes PQR AVAL AVAR PQR AVAL AVDL unc-3 unc-3 Yellow: Glu , red: ACh , green: Aminergic , blue: GABA . opp . : command interneurons control opposite drives ( forward/reverse ) . SN: sensory neuron , IN: interneuron , CI: command interneuron , hMN: head motor neuron . Black bar: Transcription factor controlling cholinergic identity . Note that most interconnected neurons are controlled by the same transcription factor . The C . elegans genome encodes not only conventional , excitatory ACh-gated cation channels , but also inhibitory ACh-gated anion channels ( Hobert , 2013; Putrenko et al . , 2005 ) . Based on the synaptic connectivity diagram and the knowledge of the identity of all cholinergic neurons , it is therefore possible to predict potential inhibitory cholinergic transmission by examining which neurons express an ACh-gated anion channel . The C . elegans genome encodes at least four ACh-gated anion channels , acc-1 through acc-4 , two of which ( acc-1 and acc-2 ) were electrophysiologically validated to be inhibitory receptors , while the function of two others ( acc-3 and acc-4 ) is less clear ( Putrenko et al . , 2005 ) . We examined their expression pattern using available but previously uncharacterized fosmid-based reporter constructs ( Sarov et al . , 2012 ) . An acc-3 fosmid reporter showed no appreciable expression throughout the animal , whereas acc-1 and acc-2 fosmid reporters show very restricted and non-overlapping expression in the adult nervous system ( Figure 4 ) . The acc-1 fosmid reporter is expressed in a subset of cholinergic neurons , including cholinergic neurons in the ventral nerve cord , the retrovesicular ganglion and a few head neurons ( including the SMD , RMD motor neurons , the AVA and AVE command interneurons and the SAA neurons ) . A small number of glutamatergic neurons also express acc-1 ( including the pharyngeal neurons MI and M3 , the PLM neurons and an unidentified neuronal pair in the lateral ganglion ) . The acc-2 fosmid reporter is expressed in a distinct , small set of glutamatergic neurons ( RIA , RIG , PHA , AIZ ) and cholinergic neurons ( URX , RIH ) . We also found that the acc-2 fosmid reporter is strongly expressed in the newly identified male-specific MCM neurons . 10 . 7554/eLife . 12432 . 014Figure 4 . Expression pattern of ACh-gated chloride channels . Expression pattern of acc fosmid reporters in L4 stage animals are shown . Transgenes: otEx6374 = acc-1 fosmid reporter; otEx6375 = acc-2 fosmid reporter; otEx6376 = acc-4 fosmid reporter; otIs545 = cho-1 fosmid reporter; otIs518 = eat-4 fosmid reporter . Besides the neurons shown here , acc-1 and acc-2 are expressed in a small number of additional neurons ( not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 014 The acc-4 fosmid reporter showed the most striking expression pattern . As assessed by coexpression with cho-1 , the acc-4 fosmid reporter is expressed almost exclusively in almost all of the 52 classes of cholinergic neurons ( Figure 4 ) . The only cholinergic neuron classes not expressing acc-4 are the ASJ and RIB neurons and the only acc-4-expressing non-cholinergic neurons are the AVF neurons . If acc-4 indeed is able to operate as inhibitory receptor ( as suggested by its sequence ) , this expression data indicates that most cholinergic neurons can be silenced by presynaptically released ACh . In line with this prediction , more than half of all cholinergic neurons are innervated by cholinergic neurons . Among those neurons are the cholinergic command interneurons . This is particularly intriguing in light of laser ablation , electrophysiological and modeling data which indicate that specific command interneuron classes inhibit each others activity ( Rakowski et al . , 2013; Roberts et al . , 2016 ) . Another notable case of likely cross-inhibitory cholinergic connection is between members of two distinct head motor neuron classes ( RMD and SMD classes ) . Notably , both the cross-inhibitory command interneurons and cross-inhibitory head motor neurons are parts of the above-described 'regulated mutual' network motif in which inter-connected cholinergic neurons are innervated by the same upstream neuron ( Figure 3E ) . Regulated mutual motifs with negative interactions can operate as toggle switches that commit to one specific drive ( forward movement ) while inhibiting the alternative ( reversal ) drive . However , it is important to keep in mind that a number of cholinergic neurons ( including the command interneurons , but also VNC MNs ) are also known to express excitatory ACh-gated ion channels ( acr genes; www . wormbase . org ) , indicating that cholinergic input into these neurons may be complex . We furthermore note that a substantial number of cholinergic neurons that express acc-4 are not innervated by cholinergic neurons ( as predicted by the connectome ) , raising the intriguing possibility that ACC-4 may act as an inhibitory autoreceptor on cholinergic neurons . In the context of gene expression networks , negative autoregulation can confer a number of useful functions , including speed-up of circuit responses and noise reduction ( Hart and Alon , 2013 ) . A substantiation of this hypothesis will require a determination of the localization of ACC-4 protein as well as additional subunits with which ACC-4 must act to constitute an inhibitory receptor ( Putrenko et al . , 2005 ) , a feat beyond the scope of this present study . The C . elegans male contains 91 sex-specific neurons , defining 24 classes , most of them located in the tail . We find that 16 out of these 24 classes are cholinergic ( Figure 5A , Figure 1—figure supplement 3 , Table 5 ) . These cholinergic neurons include the only male-specific head neurons ( the CEM sensory neurons ) and an additional , male-specific class of motor neurons in the ventral nerve cord , the CA neurons . The three key themes observed in the hermaphrodite nervous system also apply to the male-specific neurons: ( 1 ) ACh is the most broadly used neurotransmitter in the male nervous system; ( 2 ) ACh is used in sensory , inter- and motor neurons of the male-specific nervous system; ( 3 ) the male-specific sensory neurons that are cholinergic are all in close proximity to the motor circuitry: most of them directly innervate muscle ( i . e . are sensory/motor neurons; PCB , PCC , SPC; several ray neurons ) while all others ( HOB , SPV ) innervate motor neurons . Like in the pharyngeal nervous system , we found neurons labeled by two conventional fast transmitters markers– the PVV neurons and the R6A neurons express unc-17/VAChT and eat-4/VGLUT ( data not shown ) . ACh/Glu cotransmission has been observed in some central synapses in the vertebrate central nervous system ( Nishimaru et al . , 2005; Ren et al . , 2011 ) . 10 . 7554/eLife . 12432 . 015Figure 5 . Sexual and temporal dynamics of cholinergic identity . ( A ) Male-specific CEM neurons are cholinergic , but turn on cho-1 ( otIs544 ) and unc-17 ( otIs576 ) only in late L4 . In the top panels CEM neurons are labeled by the pkd-2 reporter ( bxIs14 ) . See Figure 1—figure supplement 3 and Table 5 for a list of all male-specific cholinergic neurons . ( B ) Hermaphrodite-specific HSN neurons turn on the cholinergic marker unc-17 and pan-neuronal rab-3 also in late L4 . HSN neurons are labeled by a nuclear localized unc-86 fosmid reporter ( otIs337 ) . At L4 and later stages , unc-17 fosmid expression ( otIs576 ) becomes apparent in both soma and axon ( top panels ) . The expression of the pan-neuronal marker rab-3 ( otIs355 ) is also first observed in late L4 ( bottom panels ) . ( C ) Hermaphrodite-specific VC neurons turn on unc-17 and cho-1 only in late L4 ( note that cho-1 is NOT in VC4/5 ) ; this is later than the onset of the same genes in VA and VB neurons ( VA , VB and VC neurons are labeled with the HOX gene lin-39 ) . Transgenes: wgIs18 = lin-39 fosmid reporter; otIs544 = cho-1 fosmid reporter . ( D ) Sexually dimorphic neurotransmitter identity of a sex-shared neuron class . The AIM neuron expresses cho-1 ( and unc-17; not shown ) in adult males , but expresses eat-4/VGLUT instead in hermaphrodites Transgenes: otIs354 = cho-1 fosmid reporter; otIs518 = eat-4 fosmid reporter . ( E ) Sexually dimorphic neurotransmitter switch . Until the L3 stage , both male and hermaphrodite AIM neurons are glutamatergic ( express eat-4/VGLUT ) . While hermaphrodites continue to express eat-4 , males downregulate eat-4 and turn on cho-1 ( and unc-17; not shown ) . ( F ) The neurotransmitter switch is cell-autonomously controlled by the sex-determination pathway . In the upper panels , the masculinizing fem-3 gene is force-expressed in the AIM neurons ( with the eat-4prom11 driver ) in otherwise hermaphroditic animals; in the lower panels , the masculinizing tra-2 intracellular domain ( 'tra-2ic' ) is expressed in AIM neurons of the male . Quantification is provided on the right . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 01510 . 7554/eLife . 12432 . 016Table 5 . Male-specific cholinergic neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 016Neuron typeNeuron classunc-17 fosmid expressioncho-1 fosmid expressionCo-transmitterPrevious IDSensory neuron ( 7 classes ) CEM D/V L/R ++ ++ no R1A , R2A , R3A , R4A , R6A ++ ++ yes 1 SPC L/R ++ ++ yes 2 SPV L/R ++ ++ yes 3 HOB +++ ++ no PCB L/R ++ ++ yes 2 PCC L/R ++ ++ no Interneuron ( 6 classes ) DVE ++ no DVF ++no PDC ++ ++ PDC or PGA are also serotonergic 4 no PGA ++ PDC or PGA are also serotonergic 4 no PVY +++ ++ yes3 PVX +++ ++ yes3 Motor neuron ( 3 classes ) PVZ +++ ++ no PVV +++ ++ Glutamatergic6 no CA1-9* ++ ++ no7 *CA7-9 do not express cho-1 and have lower levels of unc-17 than CA1-6 . 1Koo et al . ( 2011 ) . 2Garcia et al . ( 2001 ) . 3LeBoeuf et al . ( 2014 ) . 4Loer and Kenyon ( 1993 ) . 5Sherlekar et al . ( 2013 ) . 6Our unpublished data . 7Rand and Nonet cite unpublished observations of cholinergic identity of four CA neurons ( Rand and Nonet , 1997 ) . We observe expression of unc-17 in all nine CA neurons ( albeit lower in CA7-9 ) . Most of the male-specific neurons are generated postembryonically from embryonically generated blast cells that divide during larval stages . One notable exception is the male-specific head sensory neuron class CEM . The two pairs of CEM neurons are generated in the embryo in both sexes , but are removed specifically in the hermaphrodite through programmed cell death ( Sulston et al . , 1983 ) . We examined the onset of cholinergic differentiation of these neurons in males and found that they only start expressing cholinergic identity features at the L4 larval stage ( Figure 5A ) . Hence , even though generated in the embryo , long before sexual maturation , neurotransmitter identity of CEM male-specific neurons only becomes established during overt sexual maturation in late larval stages . The same applies to the two classes of hermaphrodite-specific cholinergic neurons , the HSN and VC neurons . HSN is born embryonically , and VCs are born in the first larval stage , yet onset of cholinergic pathway genes is only observed in late L4 larval stages ( Figure 5B–C ) . The late onset of neurotransmitter expression in the VC neurons is particularly notable if one compares the onset of cholinergic marker expression in the VC neurons with other cholinergic motor neurons born at the same time , namely the VA , VB and AS-type neurons . In these neurons , the onset of cholinergic marker expression is observed already in late L1 stage animals , contrasting the late L4 onset in the VC neurons ( Figure 5C ) . Other than the CEM neurons , there are no sex-specific neurons located in the head of the worm . We were therefore surprised to note a pair of neurons , located next to the cholinergic AIY interneurons in the ventral head ganglion that expressed cholinergic markers only in males , but not hermaphrodites ( Figure 5D ) . This neuron pair is the AIM neuron pair , previously implicated in olfactory memory formation ( Lakhina et al . , 2015 ) and mate searching behavior ( Barrios et al . , 2012 ) . In hermaphrodites , the AIM neurons are glutamatergic , expressing the vesicular glutamate transporter eat-4/VGLUT ( Figure 5D ) . In males , the AIM neurons also initially express eat-4/VGLUT , but only until the L3 stage . During the L4 stage eat-4/VGLUT expression becomes downregulated and unc-17/VAChT and cho-1/ChT expression becomes induced ( Figure 5E ) . We assessed whether the neurotransmitter switch of the AIM neurons is programmed in a cell autonomous manner . To this end , we generated sexually mosaic animals in which we masculinized AIM in otherwise hermaphroditic animals and we feminized AIM in otherwise male animals , using previously described strategies ( Lee and Portman , 2007; Mowrey et al . , 2014; White and Jorgensen , 2012; White et al . , 2007 ) . Specifically , masculinization was achieved by degrading the global regulator of hermaphroditic cellular identity , TRA-1 , by ectopic expression of FEM-3 in specific hermaphroditic cells; FEM-3 is normally functioning in males to globally degrade TRA-1 . Feminization is achieved by preventing FEM-3 downregulation of TRA-1 in male cells through ectopic expression of the intracellular domain of TRA-2 ( TRA-2ic ) , which normally acts in hermaphrodites to inhibit FEM-3 . FEM-3 or TRA-2ic were expressed under a fragment of the eat-4 locus , which is exclusively expressed in the AIM neurons in the head ganglia of the worm ( E . S . and O . H . , unpubl . ) . We found that masculinization of the AIM neurons ( 'eat-4prom11::fem-3' ) in otherwise hermaphroditic animals results in downregulation of eat-4/VGLUT and upregulation of cho-1 expression ( Figure 5F ) . Conversely , feminization of AIM in male animals results in sustained eat-4 expression and no induction of cho-1 expression ( Figure 5F ) . These results demonstrate that the neurotransmitter switch is programmed cell autonomously . Neurotransmitter maps can serve many different purposes . One of their applications relates to nervous system development . Since the neurotransmitter identity of a neuron defines a critical identity feature of any specific neuron type , a neurotransmitter map provides an entry point to study the molecular mechanisms by which neuronal identity is acquired . Previous work from our lab has defined transcription factors that control cholinergic identity in a small number of sensory , inter- and motor neurons . Specifically , we have reported that the POU homeobox gene unc-86 controls the cholinergic identity of three cholinergic sensory neurons ( IL2 , URA , URB ) ( Zhang et al . , 2014 ) , that the LIM homeobox gene ttx-3 controls the cholinergic identity of two cholinergic interneurons ( AIY , AIA ) ( Altun-Gultekin et al . , 2001 ) and that the COE ( Collier/Olf/EBF ) -type Zn-finger factor unc-3 controls cholinergic identity of most motor neuron classes in the VNC as well as the SAB head motor neurons ( Kratsios et al . , 2011 , 2015 ) . We sought to extend this analysis to other neuron classes , with the specific question in mind whether broad themes of neurotransmitter identity control may be revealed through the establishment of a comprehensive 'regulatory map' . To identify transcriptional regulators , we examined candidate factors known to be expressed in specific neurons and also conducted genetic screens using gfp-based identity markers of cholinergic neurons ( see Materials and methods ) . Our analysis resulted in the identification of a total of 7 regulators that control the identity of 20 of the 52 cholinergic neuron types ( Table 6; Figure 6 ) . 10 . 7554/eLife . 12432 . 017Figure 6 . Regulatory factors affecting cholinergic identity . We examined 20 animals for each genotype and for every mutant strain the described phenotype was observed in >80% of animals . ( A ) The LIM homeobox transcription factor lim-4 is required for unc-17 fosmid reporter expression ( left panel ) and cho-1 fosmid reporter expression ( right panel ) in AWB and SMB neurons . AWB neurons were visualized by DiI staining in the unc-17 fosmid reporter expressing strain . AWB and SMB show no fosmid reporter expression in the lim-4 mutant . ( B ) The Otx-type homeobox transcription factor ceh-14 is required for unc-17 and cho-1 fosmid reporter expression in PVC and unc-17 fosmid reporter expression in PVN . PVC neurons show a decrease in unc-17 and cho-1 fosmid reporter expression in the ceh-14 mutant compared to wild type . PVN neurons show no unc-17 fosmid reporter expression in the ceh-14 mutant . Note that PVN does not express cho-1 fosmid reporter in wild type animals . ( C ) The homeobox transcription factors unc-30 and lin-11 are required for normal expression of the unc-17 and cho-1 fosmid reporters . Cholinergic identity genes are downregulated in PVP neurons starting at L1 ( top panels ) and continuing until the L4/adult stage ( bottom panels ) in unc-30 and lin-11 mutant strains compared to wild type . ( D ) The homeobox transcription factor unc-42 is required for unc-17 and cho-1 fosmid reporter expression in RIV , SMD , RMD and SIB . ( E ) The POU homeobox transcription factor unc-86 is required for unc-17 and cho-1 fosmid reporter expression in RIH . ( F ) A wild type male is shown in the top panel for reference . unc-86 ( middle panel ) is also required for unc-17 and cho-1 fosmid reporter expression in URX and in the CEM male-specific neurons . In the absence of unc-86 the AIM neurons did not show expression of unc-17 and cho-1 fosmid reporters in the L4/adult male . The LIM homeobox transcription factor ceh-14 is required for the AIM neurons to express unc-17 and cho-1 fosmid reporters in the L4/adult male ( bottom panel ) . Transgenes: otIs576 = unc-17 fosmid reporter; otIs544 = cho-1 fosmid reporter . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 01710 . 7554/eLife . 12432 . 018Figure 6—figure supplement 1 . Continuous expression of transcription factors fosmid reporters in cholinergic neurons . AIM and PVC were labeled by cho-1 and ceh-14 fosmid reporters . PVN was labeled by ceh-14 fosmid reporter but it did not express cho-1 ( see Table 1 ) . ADF and PVP were labeled by cho-1 and lin-11 fosmid reporters . URX , RIR and RIH were labeled by by cho-1 and unc-86 fosmid reporters . PVP was labeled by cho-1 and unc-30 fosmid reporters . AVA , AVE , AVEs , RIV , RMD , SAA , SIB and SMD were labeled by cho-1 and unc-42 fosmid reporters . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 01810 . 7554/eLife . 12432 . 019Table 6 . Newly identified transcriptional regulators of cholinergic identity . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 019Gene*DNA binding domainNeuron classEffect on identity featuresOther neurotransmitter identities affected ( neuron class ) Cholinergic identity**Other identity features**unc-17 cha-1 cho-1 unc-3 ( EBF ) Zn finger PDA yesyesyes PDB yesyesn . d . DVA yesyesnoPVC yesyesnoAVA yes yesnoAVB yesyesnoAVD yesyesnoAVE yesyesnoPVN yesn . a . n . d . unc-42 ( Prd-type ) Homeodomain RIV yesyes n . d . Glu ( ASH ) 7 RMD yesyes yes SMD yesyesn . d . SIB yes yesn . d . AVA no no yes1AVD no no yes1AVE no no yes1lim-4 ( Lhx6/8 ) Homeodomain AWB yesyesyes2 SMB yesyesyes3RIV nonon . d . lin-11 ( Lhx1 ) Homeodomain ADF nonono Glu ( ASG , ADL ) 7 PVP yes yes yes4unc-30 ( Pitx ) Homeodomain PVP yes yes yes4GABA ( DD , VD ) 9 RIH no no n . d . unc-86 ( Brn3 ) Homeodomain CEM ( male ) yes yes yes5Glu ( ALM , PLM , AIM , AIZ , AQR , PQR , PVR ) 7 URX yes yes yes6AIM ( male ) yes yes yes7RIH yes yes yes8ceh-14 ( Lhx3/4 ) Homeodomain AIM ( male ) yes yes yes Glu ( AFD , DVC , PHA , PHB , PHC ) 7 PVN yes yes n . d . PVC yes yes yes *Vertebrate orthologs in parenthesis . All neuron classes listed express the respective transcription factor tested . **'yes' = expression is downregulated or completely absent; 'no' = no readily observable effect; 'n . d . ' = not determined; 'n . a . = not applicable because gene is not expressed in this cell . For primary data see Figure 6 , and Figure 8 . For data on 'other markers' ( ≥2 markers tested ) , see individual footnotes ( this data is partly our own data , partly previously reported data ) . . Previously identified regulators of cholinergic identity are: unc-3 in A- , B-type , AS and SAB motor neurons , unc-86 in IL2 , URA , URB , cfi-1 in IL2 , URA , ttx-3 in AIY and AIA and ceh-10 in AIY ( Altun-Gultekin et al . , 2001; Kratsios et al . , 2011 , 2015; Wenick and Hobert , 2004; Zhang et al . , 2014 ) . 1Baran et al . ( 1999 ) ; Brockie et al . ( 2001 ) . 2Alqadah et al . ( 2015 ) ; Sagasti et al . ( 1999 ) . 3Kim et al . ( 2015 ) . 4Hutter ( 2003 ) . 5Shaham and Bargmann ( 2002 ) . 6Qin and Powell-Coffman ( 2004 ) . 7Serrano-Saiz et al . ( 2013 ) . 8Sze et al . ( 2002 ) . 9McIntire et al . ( 1993 ) In line with a similar observation that we made upon analysis of glutamatergic neuron identity control ( Serrano-Saiz et al . , 2013 ) , we observed a striking preponderance of homeodomain containing proteins in the transcription factors that we newly identified as cholinergic identity regulators . Specifically , we found that the three LIM homeobox genes lim-4 , lin-11 and ceh-14 control cholinergic identity of six distinct cholinergic neuron types , including sensory neurons ( lim-4 in AWB ) , interneurons ( lin-11 in PVP; ceh-14 in AIM , PVC ) and motor neurons ( ceh-14 in PVN , lim-4 in SMB; Table 6; Figure 6 ) . However , we find that lim-7 , the C . elegans homolog of vertebrate Islet , which specifies cholinergic identity in the spinal cord and forebrain in mice ( Cho et al . , 2014 ) , is not required to specify cholinergic identity in C . elegans ( as assessed by normal cho-1 expression throughout the nervous system in lim-7 null mutants; data not shown ) . Therefore , while the usage of LIM-type homeobox genes in controlling cholinergic neurotransmitter identity appears to be conserved from C . elegans to vertebrates , different family members appear to execute this function in different species and cell types . Moreover , we found that the Pitx-type homeobox gene unc-30 controls cholinergic identity of the PVP interneurons ( in conjunction with lin-11 ) and that the POU homeobox gene unc-86 controls cholinergic identity of the URX , RIH and male-specific CEM neurons . unc-86 , in conjunction with ceh-14 , is also required for the AIM neurons to adopt their cholinergic identity in males; both factors also control glutamatergic identity of the AIM neurons in hermaphrodites ( and males till the third larval stage; Figure 6 ) . All of the above-mentioned transcription factors are continuously expressed throughout the life of these neurons ( Figure 6—figure supplement 1 ) , suggesting that these factors not only initiate but also maintain cholinergic identity . From a EMS-induced genetic mutant screen that we conducted for regulators of RMD neuron identity ( see Materials and methods ) , we uncovered unc-42 , a Prox-type homeobox gene as a regulator of cholinergic gene expression in RMD motor neurons . We also found that unc-42 affects cholinergic identity of four additional , distinct types of cholinergic head neurons , most of them motor neurons ( Figure 6 ) . unc-42 is continuously expressed in all these postmitotic neuron types ( Figure 6—figure supplement 1 ) . The only exception to the homeobox theme is what appears to be the most remarkable regulator of cholinergic identity , the phylogenetically conserved COE-type unc-3 transcription factor . In addition to the previously reported impact of unc-3 on cholinergic ventral cord motor neuron identity ( SAB- , A- , B- , AS-type MNs ) , we found that unc-3 regulates expression of the cholinergic identity genes cho-1 and unc-17 in all command interneurons ( Figure 7 , Table 6 ) . Moreover , the PDA , PDB , PVN and DVA tail neurons also require unc-3 for their normal expression of cholinergic identity genes ( Figure 7 , Table 6 ) . DVA is particularly notable here because like the command interneurons , the DVA neuron also takes a central role in the overall C . elegans connectivity network ( Varshney et al . , 2011 ) ( Figure 3B ) and this central location is paralleled by the dependence of these neurons on unc-3 activity . The expression pattern of unc-3 had not previously been reported in most of these neurons . Using a fosmid reporter and a gfp reporter inserted into the unc-3 locus through CRISPR-Cas9 , we confirmed expression of unc-3 in all these cholinergic neuron types , including the command interneurons ( Figure 7A ) . 10 . 7554/eLife . 12432 . 020Figure 7 . unc-3 is a circuit-associated transcription factor . ( A ) Expression pattern of an unc-3 fosmid-based reporter ( otIs591 ) . Overlap with a cho-1 fosmid-based reporter ( otIs544 ) is shown in all panels . The upper panels are the same as the lower , but a Nomarski image has been added for orientation purposes . unc-3 expression was also detected in PDA , PDB and PVP in the pre-anal ganglion ( data not shown ) . ( B ) The expression of the unc-17 and cho-1 fosmid reporters is downregulated in command interneurons ( AVA , AVB , AVD , AVE , PVC ) and the tail neuron DVA in unc-3 mutant animals ( identical results were obtained using two unc-3 alleles , e151 generates a premature STOP and n3435 is a deletion allele ) . Quantification is shown on the right . Twenty animals were analyzed at the fourth larval stage ( L4 ) per genotype . Note that the effect of unc-3 on unc-17 expression in the command interneurons ( this figure ) is not as fully penetrant as it is in VNC motor neurons ( Kratsios et al . , 2011 ) . ( C ) Gap junctions that command interneurons make are visualized with gfp tagging the innexin protein UNC-7 , as previously described ( Starich et al . , 2009 ) ( transgene: iwIs47 ) . Dotted white lines delineate the location of the VNC . A significant decrease in the number of the UNC-7::GFP puncta was observed in the VNC of unc-3 ( n3435 ) mutant animals ( quantification shown on the right with average values and standard deviation ) . A student’s t test was performed . ***p value <0 . 0001 . ( D ) Reconstruction of the chemical synapse connectivity of the AVA command interneurons in a wild type and an unc-3 ( e151/MnH205 ) mutant animal . Less synaptic input onto AVA neurons and output from the AVA neurons was observed in the unc-3 mutant animal . This is not merely an effect of axonal process misplacement since in unc-3 mutants , AVA processes still run adjacent to the processes of the neurons it normally makes synaptic contacts to . More than 600 electron micrographs were reconstructed per genotype . In square brackets , the location ( number of electron micrograph ) for each chemical synapse is shown , and the number of consecutive micrographs in which a synapse was detected is also shown in parenthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 02010 . 7554/eLife . 12432 . 021Figure 7—figure supplement 1 . UNC-3 has no effect on glutamate receptor expression in command interneurons . ( A ) The expression of PDA identity markers exp-1 , ace-3/4 , and cog-1 is affected in unc-3 mutant animals . Quantification is provided on the right . For cog-1prom::gfp , n = 25 for wild type and unc-3 ( e151 ) . For ace-3/-4prom::gfp and exp-1prom::gfp , n = 20 for wild type and unc-3 ( e151 ) . Fisher’s exact test was performed . **p value <0 . 01; ***p value < 0 . 0001 . ( B ) The expression of multiple glutamate receptor genes ( nmr-1 , nmr-2 , glr-1 , glr-2 , glr-4 , glr-5 ) is unaffected in command interneurons ( AVA , AVB , AVD , AVE , PVC ) of unc-3 null animals . Similarly , the expression of the ACh receptor subunit encoding gene acr-15 is not affected in the AVA and AVB neurons of unc-3 mutants . Quantification is provided on the right . Number of animals examined = 20 animals per reporter gene per genotype . Moreover , the expression of flp-18 and rig-3 ( AVA markers ) , as well as opt-3 ( AVE marker ) is not affected in unc-3 mutants ( N = 20 , data not shown ) . In addition , the expression of several identity genes ( glr-5 , glutamate receptor; twk-16 , potassium channel; nlp-12 , neuropeptide; zig-5 , immunoglobulin superfamily gene; ser-2 , serotonin receptor ) for the DVA interneuron is unaffected in unc-3 mutants ( data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 021 Apart from the preponderance of homeobox genes , another striking theme we found is the employment of the same transcription factor in completely different cellular contexts , apparently a reflection of the operation of transcription factors in distinct combinations . For example , unc-86 controls cholinergic identity in the IL2 sensory neurons and the unrelated AIM interneurons ( in the male ) . In these different cellular contexts unc-86 cooperates with distinct cofactors , cfi-1 in IL2 ( Zhang et al . , 2014 ) and ceh-14 in AIM ( Figure 6G ) . The need for specific combinations of transcription factors to drive a specific identity program explains why we find that a factor that is expressed in multiple cholinergic neuron types does not necessarily regulate cholinergic identity in all neuron types in which it is expressed ( Table 6 ) . For example , lim-4 which is expressed in the cholinergic AWB , SMB and RIV neurons controls cholinergic identity in AWB and SMB ( Figure 6 ) , but not in RIV . This is likely because the cofactors that work together with lim-4 in AWB and/or SMB may not be expressed in RIV . Likewise , ceh-10 forms a heterodimer with ttx-3 in AIY to control its cholinergic identity ( Altun-Gultekin et al . , 2001; Wenick and Hobert , 2004 ) , but it is not required for cholinergic identity of the AIN neurons which express ceh-10 , but not ttx-3 ( data not shown ) . Transcription factors that we find to control cholinergic neurotransmitter identity are also employed in the control of other neurotransmitter identities ( see Table 6 ) , likely in the context of distinct transcription factors combinations . For example , the Pitx-type homeobox gene unc-30 controls cholinergic identity of the PVP neurons , apparently in conjunction with lin-11 ( this study ) , but also controls the GABAergic identity of D-type VNC MNs , likely in conjunction with an as yet unidentified factor ( Jin et al . , 1994 ) . Likewise , the LIM homeobox gene ceh-14 , which controls cholinergic identity of the PVN and PVC neurons ( Table 6 ) , likely together with unc-3 ( this study ) , controls glutamatergic identity of various amphid and phasmid sensory neurons in which ceh-14 operates independently of unc-3 ( Serrano-Saiz et al . , 2013 ) . Lastly , we note that loss of two of the transcription factors that we examined , unc-3 and ceh-14 , results in derepression of cholinergic identity features in normally non-cholinergic neurons ( data not shown ) . In unc-3 mutants , two cells in the dorsal ganglion ectopically express cholinergic features; these are probably the RID neuron and its sister cell . In ceh-14 mutants at least one pair of tail neurons ectopically expresses cholinergic markers . Most of the transcriptional regulators that we defined here control not only cholinergic identity in the respective neuron classes , but also control other identity features . For example , we find that loss of unc-3 affects multiple aspects of PDA motor neuron identity ( expression of the exp-1 ligand gated ion channel , ace-3/4 cholinesterase , cog-1 homeobox gene ) and loss of unc-42 affects metabotropic glutamate receptor ( mgl-1 ) expression in the RMD neurons . Apart from affecting cholinergic identity , loss of ceh-14 affects neuropeptide ( flp-10 ) expression , as well as the serotonergic co-transmitter identity of the AIM neurons and it affects expression of the ionotropic glutamate receptors nmr-1 and glr-1 in the PVC command interneuron ( Figure 8 ) . In addition , lin-11 and unc-30 were previously found to control many terminal identity features of the PVP ( Hutter , 2003 ) and these two factors also control cholinergic identity of PVP ( Figure 6E ) . Similarly , lim-4 controls cholinergic identity of the AWB neurons but also several other identity features ( Alqadah et al . , 2015; Sagasti et al . , 1999 ) . The coupling of adopting cholinergic identity control with the adoption of other identity features has been observed in previously described regulators of cholinergic identity: unc-3 for VNC MNs ( Kratsios et al . , 2015 2011 ) , unc-86 for IL2 ( Zhang et al . , 2014 ) , ttx-3 for AIY and AIA ( Altun-Gultekin et al . , 2001; Zhang et al . , 2014 ) , and also in the context of neurons with distinct neurotransmitter identities ( e . g . [Flames and Hobert , 2009; Serrano-Saiz et al . , 2013] ) . 10 . 7554/eLife . 12432 . 022Figure 8 . Coupling of cholinergic identity with other identity features . ( A ) An mgl-1 reporter transgene does not show expression in RMD neurons in the absence of unc-42 . ( B ) In the absence of ceh-14 the AIM neurons do not show eat-4 fosmid reporter and flp-10 reporter expression . 5-HT staining is not detectable in AIM neurons in the ceh-14 mutant . In the absence of ceh-14 the PVC neurons do not show nmr-1 or glr-1 reporter expression . Number of animals examined = 20 animals per reporter gene per genotype . ( C ) The expression of PDA identity markers exp-1 , ace-3/4 , and cog-1 is lost in unc-3 mutant animals . For cog-1prom::gfp , 25 of 25 wild-type and 1 of 25 unc-3 ( e151 ) animals showed cog-1prom::gfp expression in PDA . For ace-3/-4prom::gfp , 20 of 20 wild-type and 0 of 20 unc-3 ( e151 ) animals showed ace-3/-4prom::gfp expression in PDA . For exp-1prom::gfp , 20 of 20 wild-type and 11 of 20 unc-3 ( e151 ) animals showed exp-1prom::gfp expression in PDA . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 022 However , we also noted a number of striking exceptions to the coupling of neurotransmitter identity with other terminal identity features . The serotonergic identity of the hermaphrodite specific motor neurons HSN is controlled by the unc-86 POU homeobox gene ( Sze et al . , 2002 ) , but unc-86 does not affect unc-17/VAChT expression in HSN ( data not shown ) . The most striking example for a separation of neurotransmitter identity from other identity features is observed in relation to the function of the unc-3 gene . We had previously shown that in all motor neurons in which unc-3 is expressed ( SAB head motor neurons , A- B- and AS-type VNC MNs ) , unc-3 not only controls neurotransmitter identity , but also a multitude of other terminal molecular markers ( Kratsios et al . , 2015 , 2011 ) . In contrast , the activity of unc-3 in the AVA , AVB , AVD , AVE and PVC command interneurons and DVA interneuron appears to be restricted to select subfeatures of these neurons . We arrived at this conclusion by analyzing the expression of more than ten additional identity markers of these unc-3-expressing neurons ( including glutamate receptors , neuropeptides and ion channels ) . Not a single one besides the cholinergic reporter genes is affected in the worms lacking unc-3 ( Figure 7—figure supplement 1; Table 7 ) . Within a subset of these neurons , namely the command interneurons AVA , AVD and AVE , three transcription factors , unc-42 ( homeobox ) , fax-1 ( nuclear hormone receptor ) and cfi-1 ( ARID-type ) , have been shown to control subsets of these unc-3-independent terminal identity markers ( Table 8 ) . The observation of a piece-meal regulation of distinct terminal identity features by a number of distinct transcription factors , each acting in a highly cell-type and target gene-specific manner ( Table 8 ) , represents a remarkable departure from the commonly observed theme of co-regulation of multiple identity features by the same set of transcription factors ( Alqadah et al . , 2015; Cinar et al . , 2005; Duggan and Chalfie , 1995; Etchberger et al . , 2009; Guillermin et al . , 2011; Hobert , 2011; Kratsios et al . , 2011; Serrano-Saiz et al . , 2013; Wenick and Hobert , 2004; Zhang et al . , 2014 ) . 10 . 7554/eLife . 12432 . 023Table 7 . unc-3 affects the differentiation of a broad set of cholinergic neuron types . nmr and glr genes encode glutamate receptors and expression of neither is affected by unc-3; many of them are instead regulated by either the unc-42 , fax-1 or cfi-1 , as summarized in Table 8 . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 023 unc-3 ( + ) neuronsCholinergic identity in unc-3 ( - ) animals 1 Other identity features that are NOT affected in unc-3 ( - ) animals 1 Other identity feature that are affected in unc-3 ( - ) animalsINTERCommand inter- neurons AVA unc-17 , cho-1 affected nmr-1 , nmr-2 , glr-1 , glr-2 , glr-4 , glr-5 , acr-15 , rig-3 , flp-18 AVB unc-17 , cho-1 affected acr-15 AVD unc-17 , cho-1 affected nmr-1 , nmr-2 , glr-1 , glr-2 , glr-5 AVE unc-17 , cho-1 affected nmr-1 , nmr-2 , glr-1 , glr-2 , glr-5 , opt-3 PVC unc-17 , cho-1 affected nmr-1 , nmr-2 , glr-1 , glr-2 , glr-5 Other inter- neurons DVA unc-17 , cho-1 affected glr-4 , glr-5 , twk-16 , nlp-12 , zig-5 , ser-2 SAA unc-17 NOT affected2 PVP unc-17 , cho-1 NOT affected MOTOR Head MNs SAB unc-17 , cho-1 affected3 8/8 markers tested3 VNC MNs A , B , AS unc-17 , cho-1 affected3 29/34 markers tested3 Tail MNs PDA unc-17 , cho-1 affected exp-1 , ace-3/4 , cog-1 1PDB unc-17 , cho-1 affected PVN unc-17 affected2 1See Figure 7 , Figure 7—figure supplement 1 2Cho-1 not expressed in these neurons . 3As previously reported by Kratsios et al . ( 2011 ) , ( 2015 ) . 10 . 7554/eLife . 12432 . 024Table 8 . Transcription factors affecting command interneuron differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 024TF AVA AVB AVD AVE PVC unc-3 unc-42 fax-1 cfi-1 ceh-14 ACh unc-17 & cho-1 unc-3 effect unc-3 effect unc-3 effect unc-3 effect unc-3 effect unc-42 NO effect unc-42 NO effect unc-42 NO effect ceh-14 effect Other nmr-1 ( GluR ) unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-42 NO effect unc-42 NO effect unc-42 NO effect ceh-14 effect fax-1 effect fax-1 NO effect fax-1 effect cfi-1 effect cfi-1 effect nmr-2 ( GluR ) unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-42 NO effect unc-42 NO effect unc-42 NO effect fax-1 effect fax-1 NO effect fax-1 effect glr-1 ( GluR ) unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-42 effect unc-42 NO effect unc-42 effect unc-42 effect ceh-14 effect fax-1 no effect fax-1 NO effect fax-1 no effect fax-1 no effect cfi-1 effect cfi-1 effect glr-2 ( GluR ) unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-42 NO effect unc-42 NOeffect unc-42 NOeffect fax-1 NO effect fax-1 NO effect fax-1 NO effect glr-4 ( GluR ) unc-3 NO effect unc-42 effect fax-1 no effect glr-5 ( GluR ) unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-3 NO effect unc-42 effect unc-42 NOeffect unc-42 effect unc-42 effect fax-1 no effect fax-1 NO effect fax-1 no effect fax-1 no effect opt-3 unc-3 NO effect unc-42 effect fax-1 effect rig-3 ( IgSF ) unc-3 no effect flp-18 ( FMRF ) unc-3 no effect Gray shading: gene normally expressed in this cell . 'Effect' ( red ) indicate that respective reporter gene fails to be expressed in the respective mutant background in the indicated cells , 'no effect' ( green ) indicates the opposite . unc- 42 , cfi-1 and fax-1 data on non-ACh marker from Wightman et al . ( 2005 ) Shaham and Bargmann ( 2002 ) and Brockie et al . ( 2001 ) Moving beyond a cell- and gene-centric consideration of regulatory factors , we asked whether there are any overarching , circuit-based themes of cholinergic identity control . Specifically , since every transcription factor that we identified here to control cholinergic neurotransmitter identity exerts its effect on more than one neuron type , we asked whether neurons whose neurotransmitter identity is controlled by the same regulatory factor are part of synaptically connected circuits . Such an observation would suggest that the respective transcription factor may define and coordinate the activity of entire circuits and perhaps may also define aspects of circuit assembly . We indeed found several examples of transcription factors that control the identity of synaptically connected neurons . The most striking example is the ventral nerve cord motor circuit which is composed of a multitude of interconnected motor neurons ( six different classes ) and a highly interconnected 'rich club' of interneurons ( also six different classes ) ( White et al . , 1986; Towlson et al . , 2013 ) . As noted above , the entire ventral nerve cord motor circuit uses ACh ( except DD/VD; shown again in Figure 9A ) . Strikingly , unc-3 is expressed and required for the adoption of cholinergic identity in all neurons in this circuit ( schematized in Figure 9A; data in Figure 7B ) . unc-3 is therefore a circuit-associated transcription factor that is selectively associated with this circuit ( unc-3 is expressed only in a few neurons outside this circuit ) and that defines a critical feature of the circuit , namely the ability of neurons in the circuit to communicate among each other . We furthermore note that the regulated mutual 3-neuron network motif mentioned above ( Figure 4E ) frequently occurs in the unc-3-dependent motor circuit , with the mutually connected neurons being unc-3-dependent command interneurons that receive inputs either from glutamatergic neurons outside the circuit or from cholinergic , and also unc-3-dependent neurons within the circuit ( Figure 9A; Table 4 ) . 10 . 7554/eLife . 12432 . 025Figure 9 . Circuit-associated transcription factors . ( A ) Ventral cord motor circuit as shown in White et al . ( 1986 ) , but now superimposed with neurotransmitter identity and expression pattern of the unc-3 transcription factor . unc-3 controls the cholinergic identity of every single neuron in this circuit . Next to the circuit diagram , a number of different regulated mutual 3-neuron networks motifs are shown . These motifs are either embedded in the circuit and provide a connection to neurons located outside the circuit ( e . g . glutamatergic sensory neurons ) . In all cases unc-3 controls cholinergic identity of the mutually connected command interneurons ( 'CI' ) and in those cases where the mutually connected neurons receive cholinergic interneuron input , unc-3 controls the identity of the entire microcircuit . ( B ) unc-42 controls the cholinergic identity of interconnected head motor neurons , and glutamatergic signaling between ASH sensory neurons ( whose glutamatergic identity is controlled by unc-42 ( Serrano-Saiz et al . , 2013 ) and cross-connected command interneurons in which unc-42 controls glutamate receptor expression ( Brockie et al . , 2001 ) ( shown in Table 4 ) . Red boxes indicate the neurons affected by the indicated transcription factor . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 025 To investigate whether neurotransmitter identity is the only parameter of the circuit that is disrupted in unc-3 mutants , we examined connectivity between neurons in the VNC MN circuit . In our previous analysis of unc-3 function , we had identified neuromuscular junction defects , i . e . disorganized or absent synapses from VNC MNs onto body wall muscle ( Kratsios et al . , 2015 ) , but upstream layers of the motor circuit ( i . e . connections of command interneurons to MNs and connections among command interneurons ) had not been examined . unc-3-expressing command interneurons make prominent electrical synapses to other command interneurons and to motor neurons and these synapses can be visualized through gfp-tagging of a gap junction component that connects command interneurons and motor neurons , the innexin unc-7 ( Figure 7C ) , which is expressed in command interneurons ( Starich et al . , 2009 ) . UNC-7::GFP puncta , visualized with a translational reporter are severely reduced in unc-3 mutants ( Figure 7C ) . Expression of a transcriptional , fosmid-based unc-7 reporter is unaffected in unc-3 mutants ( data not shown ) , leading us to conclude that unc-3 affects electrical synapse formation at a step independent of regulation of innexin expression . To examine chemical synapses within neurons of the motor circuit , we reconstructed the chemical synapse connectivity of the AVA command interneuron in unc-3 null mutants using serial analysis of electron micrographs . We reconstructed a defined part of the anterior ventral nerve cord between two different motor neurons ( AS1 and AS3 ) . In this region , AVA makes prominent chemical synapses onto MNs and other command interneurons and it receives several synaptic inputs ( Figure 7D ) . In unc-3 null mutants , we found connectivity defects on all levels: AVA receives less chemical synaptic input from within the motor circuit ( i . e . from other command interneurons ) and it makes less chemical synapses onto other motor neurons and onto other command interneurons ( Figure 7D ) . There is also an overall disorganization of the placement of axonal processes in the VNC of unc-3 mutants ( data not shown ) ; however , AVA still neighbors the command interneurons that it normally connects to , indicating that the connectivity defects are not a secondary consequence of placement defects . Non-cholinergic synaptic inputs from sensory neurons into the motor circuit appear not to be affected by unc-3 . We arrived at this conclusion by examining the synaptic connections of the glutamatergic PHB neuron to the AVA interneuron , normally made in the pre-anal ganglion ( White et al . , 1986 ) . This synaptic connection can be visualized using a GFP reconstitution system ( 'GRASP'; ( Park et al . , 2011 ) . We find these synaptic GFP signals to be unaffected in unc-3 mutants ( data not shown ) . Remarkably , the pan-circuit control of cholinergic neurotransmitter identity by unc-3 is mediated via a single UNC-3 binding site ( 'COE motif' ) controlling neurotransmitter pathway genes . Its deletion in the context of the cho-1/ChT fosmid-based reporter eliminates expression not only in the ventral nerve cord motor neurons , but also in all other unc-3 dependent cholinergic neurons , not just within the motor neuron circuit , but also outside the circuit ( Figure 10A–C ) . On the other hand , a 280 bp region from the cho-1 and a 250 bp region from the unc-17 locus that contain the COE motif are not sufficient to drive expression in all unc-3-dependent inter- and motor neurons of the motor circuit , but only drives expression in motor neurons ( Figure 10D , E ) . This finding suggests that unc-3 may cooperate with distinct cofactors in distinct neuron types . 10 . 7554/eLife . 12432 . 026Figure 10 . A single UNC-3 binding site is required for cho-1 expression in all distinct unc-3-dependent cholinergic neuron types . ( A ) Schematic showing of the cho-1 locus and the location of the UNC-3 binding site ( COE motif ) relative to ATG for the fosmid reporters and 280bp promoter fusion . ( B , C ) A cho-1 fosmid reporter ( ~28 kb ) that contains an intact COE motif shows expression in all cholinergic neurons including the ventral nerve cord ( VNC ) motor neurons ( MNs ) , the command interneurons ( AVA , AVB , AVD , AVE , PVC ) , and the interneuron DVA . Mutation of the COE motif in the context of this cho-1 fosmid-based reporter results in selective loss of reporter gene expression in VNC MNs residing at the retrovesicular ganglion and all command interneurons ( only AVA and AVE head interneurons are shown ) . ( B ) cho-1 fosmid reporter versus cho-1_COEmut fosmid reporter in an adult head . ( C ) cho-1 fosmid reporter versus cho-1_COEmut fosmid reporter in an adult tail . Reporter gene expression is also lost in tail neurons DVA and PVC . The transgenic line rab-3prom::rfp drives reporter gene expression in the entire nervous system and was used in the background to facilitate neuronal identification . ( D ) A short fragment ( 280 bp ) of the cho-1 cis-regulatory region containing the COE motif is sufficient to drive reporter gene expression only in VNC MNs . This fragment does not show expression in command interneurons located at the head and tail of the animal . ( E ) A short fragment ( 250 bp ) of the unc-17 cis-regulatory region containing the COE motif is sufficient to drive reporter gene expression only in VNC MNs . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 026 Taken together , unc-3 activity is required not only for the expression of proper neurotransmitter identity , but also for synaptic connectivity throughout the VNC motor neuron circuit , not just in motor neurons but also in command interneurons . However , as mentioned above , unc-3 is not required to control the expression of other identity features of command interneurons , such as the many types of distinct glutamate receptors expressed by the command interneurons ( Figure 7—figure supplement 1 ) . unc-3 may not be the only transcription factor whose activity is required for the function and assembly of an entire circuit . On a micro-circuit level , we note that the homeobox gene unc-42 is , like unc-3 , also frequently employed in the context of the 'regulated mutual' 3-neuron network motif described above . This motif is predominantly found either ( a ) in the context of the innervation of cross-connected command interneurons or ( b ) the context of cross-connected head motor neurons ( SMDs and RMDs; Figure 9B; Table 4 ) . unc-42 has functions in both of these motifs . In the case of the cross-connected head motor neurons ( RMDs , SMDs ) and the RIV interneuron that innervates these cross-connected neurons , unc-42 specifies the cholinergic identity of all of these neurons ( and other signaling input to these neurons , exemplified by the above-mentioned regulation of the metabotropic Glu receptor mgl-1 by unc-42 ) . In the cross-connected command interneurons , unc-42 does not affect their cholinergic identity , but it does affect the expression of multiple ionotropic Glu receptors ( GluRs ) expressed in the command interneurons ( Brockie et al . , 2001 ) . Intriguingly , in a number of motif occurrences , the cross-connected command interneurons are innervated by the glutamatergic ASH sensory neurons , which sense a number of repulsive cues ( Table 4 ) ( Kaplan and Horvitz , 1993 ) . We previously found that the glutamatergic identity of the ASH neurons is controlled by unc-42 ( Serrano-Saiz et al . , 2013 ) . unc-42 therefore controls and apparently coordinates the expression of presynaptic neurotransmitter identity and postsynaptic receptor expression in a repulsive reflex circuit ( Figure 9B ) . Notably , the above-mentioned unc-42-dependent 'regulated mutual' 3 neuron network motifs are connected to one another , as illustrated in Figure 9B . Mutually connected , unc-42-dependent head motor neurons are coupled by electrical synapses ( Figure 9B ) . Moreover , the unc-42-dependent head motor neurons are connected to the unc-42-dependent ASH>command interneuron motif . This connection is made by the glutamatergic AIB interneurons; strikingly , their glutamatergic identity is also controlled by unc-42 ( E . S . and O . H . , unpubl . data ) . Taken together , a network of interconnected neurons in the head of the worm that employ distinct neurotransmitter systems all require unc-42 , either for the acquisition of their neurotransmitter identity or for the ability to receive neurotransmitter signals ( unc-42-dependent GluR expression in command interneurons ) .
The C . elegans mutant strains used in this study were: unc-104 ( e1265 ) ; lim-4 ( ky403 ) ; ceh-14 ( ch3 ) ; lin-11 ( n389 ) ; unc-30 ( e191 ) ; unc-42 ( e419 ) ; unc-86 ( n846 ) ; unc-3 ( e151 ) ; unc-3 ( n3435 ) . The unc-17 , acc-1 , acc-2 , acc-3 and acc-4 fosmid reporter constructs were kindly provided by the TransgeneOme project ( Sarov et al . , 2012 ) . gfp is fused directly to the respective loci ( 'translational reporters' ) . The unc-17 fosmid DNA was injected at 15 ng/μl into N2 worms together with lin-44::yfp as a co-injection marker . The cho-1 fosmid reporter construct was generated using λ-Red-mediated recombineering in bacteria as previously described ( Tursun et al . , 2009 ) . For the cho-1 fosmid reporter , either an SL2 spliced , nuclear- localized mChOpti::H2B sequence was engineered right after the stop codon of the locus ( resulting transgene: otIs544 ) or SL2 spliced , nuclear-localized yfp::H2B sequence was engineered at the same position , as previously reported ( Stefanakis et al . , 2015 ) ( resulting transgene: otIs354 ) . For the 'transcriptional' , fosmid-based unc-7 reporter an sl2::H2B::yfp cassette was inserted at the 3’ end of the locus . The acc-1 , -2 , -3 , and -4 and the cho-1 fosmid DNA were injected at 15 ng/μl into a pha-1 ( e2123 ) mutant strain with pBX as co-injection marker ( Granato et al . , 1994 ) . The following reporter strains were generated for this study: unc-17 fosmid reporter ( otIs576 ) , cho-1 fosmid reporter ( otIs544 ) , acc-1 fosmid reporter ( otEx6374 ) , acc-2 fosmid reporter ( otEx6375 ) , acc-4 fosmid reporter ( otEx6376 ) . For the mutation of the COE motif in the context of the cho-1 fosmid-based reporter construct , two nucleotides ( wild-type COE motif: aaaacggtctccagggagagaaa; mutated COE motif: aaaacggtctggagggagagaaa ) that are critical for UNC-3 binding were mutated as previously described in Stefanakis et al . , 2015 . The following additional , and previously described neuronal markers were used in the study: ric-19::gfp ( otIs380 ) , eat-4fosmid::sl2::yfp::H2B ( otIs388 ) , eat-4fosmid::sl2::mChOpti::H2B ( otIs518 ) , ace-3/4::gfp ( fpIs1 ) , rab-3::bfp ( otIs355 ) , pkd-2::gfp ( bxIs14 ) , unc-86fosmid::yfp ( otIs337 ) , rab-3::rfp ( otIs355 ) , lin-39fosmid::gfp ( wgIs18 ) , opt-3::gfp ( gvEx173 ) , flp-18::TagRFP ( otEx6491 ) , rig-3::gfp ( otEx239 ) . Additional transgenes used for cell identifications are listed in Table 9 . 10 . 7554/eLife . 12432 . 027Table 9 . Molecular markers for cell identification . The respective markers were crossed with cho-1 or unc-17 fosmid reporters to validate cell identification . DOI: http://dx . doi . org/10 . 7554/eLife . 12432 . 027NeuronMolecular markerHermaphrodite ADFL/R cat-1::GFP ( otIs625 ) 1AIA L/R ttx-3 fosmid::GFP ( wgIs68 ) AIN L/R ttx-3 fosmid::GFP ( wgIs68 ) AIY L/R ttx-3 fosmid::GFP ( wgIs68 ) ALN L/R unc-86 fosmid::YFP ( otIs337 ) ; lad-2::GFP ( otIs439 ) AS1-11 unc-3 fosmid::GFP ( otIs591 ) ASJ L/R DiI/DiO staining AVA L/R glr-1::DsRed ( hdIs30 ) ; nmr-1::GFP ( akIs3 ) AVB L/R acr-15::GFP ( wdEx290 ) ; sra-11::GFP ( otIs123 ) AVD L/R glr-1::DsRed ( hdIs30 ) ; nmr-1::GFP ( akIs3 ) AVE L/R glr-1::DsRed ( hdIs30 ) ; nmr-1::GFP ( akIs3 ) AVG odr-2::DsRed ( otEx4452 ) ; unc-6 fosmid::GFP ( otEx6370 ) AWB L/R DiI/DiO staining DA1-9 unc-3 fosmid::GFP ( otIs591 ) ; acr-2::GFP ( juIs14 ) DB1-7 unc-3 fosmid::GFP ( otIs591 ) ; acr-2::GFP ( juIs14 ) DVA ser-2::GFP ( otIs358 ) HSN L/R unc-86 fosmid::YFP ( otIs337 ) IL2 D/V L/R ( x6 ) unc-86 fosmid::YFP ( otIs337 ) PDA unc-3 fosmid::GFP ( otIs591 ) ; ace-3/4::GFP ( fpIs1 ) PDB unc-3 fosmid::GFP ( otIs591 ) PLN L/R unc-86 fosmid::YFP ( otIs337 ) ; lad-2::GFP ( otIs439 ) PVC L/R nmr-1::GFP ( akIs3 ) PVN L/R 2PVP L/R lin-11 fosmid::GFP ( wgIs62 ) ; unc-30 fosmid::GFP ( wgIs395 ) RIB L/R 2RIF L/R odr-2::DsRed ( otEx4452 ) ; unc-6 fosmid::GFP ( otEx6370 ) RIH cat-1::GFP ( otIs625 ) RIR8 unc-86 fosmid::YFP ( otIs337 ) RIV L/R unc-42 fosmid::GFP ( wgIs173 ) ; lad-2::GFP ( otIs439 ) RMD D/V L/R ( x6 ) glr-1::DsRed ( hdIs30 ) RMF L/R 2RMH L/R 2SAA D/V L/R ( x4 ) lim-4::GFP ( mgIs19 ) ; lad-2::GFP ( otIs439 ) SAB D V L/R ( x3 ) unc-4::GFP ( vsIs45 ) SDQ L/R unc-86 fosmid::YFP ( otIs337 ) ; lad-2::GFP ( otIs439 ) SIA D/V L/R ( x4 ) ceh-24::GFP ( ccIs4595 ) SIB D/V L/R ( x4 ) ceh-24::GFP ( ccIs4595 ) SMB D/V L/R ( x4 ) lim-4::GFP ( mgIs19 ) ; lad-2::GFP ( otIs439 ) SMD D/V L/R ( x4 ) lad-2::GFP ( otIs439 ) URA D/V L/R ( x4 ) unc-86 fosmid::YFP ( otIs337 ) URB L/R unc-86 fosmid::YFP ( otIs337 ) URX L/R flp-10::GFP ( otIs92 ) ; unc-86 fosmid::YFP ( otIs337 ) VA1-12 unc-3 fosmid::GFP ( otIs591 ) ; acr-2::GFP ( juIs14 ) VB1-11 unc-3 fosmid::GFP ( otIs591 ) ; acr-2::GFP ( juIs14 ) VC1-6 lin-11::GFP ( nIs106 ) ; ida-1::GFP ( inIs179 ) Pharyngeal I1 L/R 3I3 ( L ) 3M1 ( R ) 3M2 L/R 3M4 ( L ) 3M5 ( L ) 3MC L/R 3M CEM D/V L/R ( x4 ) pkd-2::GFP ( bxIs14 ) CA1-9 ida-1 ( inIs179 ) DVE , DVF 2HOB ida-1 ( inIs179 ) PCB , PCC , SPC 4PDC , PGA 2PVV 2PVX , PVY 2PVZ ida-1 ( inIs179 ) R1A , R2A , R3A , R4A , R6A 2SPV 21Excluded AWA due to lack of overlap of cho-1 fosmid reporter with odr-10::gfp and unc-17 fosmid reporter gfp reporter with gpa-4::mCherry . See Figure 1—figure supplement 1 . 2Identified based on position and axonal projections because of the lack of available markers . 3Pharyngeal neurons identified based on axonal projections which are visible with the unc-17 fosmid reporter . 4Garcia et al . ( 2001 ) . FEM-3 and TRA-2ic were expressed under the control of a fragment of the eat-4 locus ( between -2680 and -2155bp from ATG ) . The plasmids were injected in him-5 ( e1490 ) at 50 ng/ul . Two lines expressing FEM-3 were then crossed with otIs388 ( eat-4fosmid::sl2::yfp::H2B ) and otIs354 ( cho-1fosmid::sl2::yfp::H2B ) independently generating the following strains: OH13753 [otIs388; otEx6377 ( eat-4prom11::fem-3::sl2::tagRFP;unc-122::GFP ) ] , OH13802 [otIs388; otEx6378 ( eat-4prom11::fem-3::sl2::tagRFP;unc-122::GFP ) ] , OH13805 [otIs354; otEx6377 ( eat-4prom11::fem-3::sl2::tagRFP;unc-122::GFP ) ] and OH13806 [otIs354; otEx6378 ( eat-4prom11::fem-3::sl2::tagRFP;unc-122::GFP ) ] . Similarly , two lines expressing TRA-2ic were crossed with otIs388 and otIs354 generating the following strains: OH13803 [otIs388; otEx6379 ( eat-4prom11::tra-2ic::sl2::tagRFP; unc-122::GFP ) ] , OH13804 [otIs388; otEx6380 ( eat-4prom11::tra-2ic::sl2::tagRFP; unc-122::GFP ) ] , OH13807 [otIs354; otEx6379 ( eat-4prom11::tra-2ic::sl2::tagRFP; unc-122::GFP ) ] and OH13808 [otIs354; otEx6380 ( eat-4prom11::tra-2ic::sl2::tagRFP; unc-122::GFP ) ] . eat-4 and cho-1 expression was analyzed at 1 day adult animals with and without the array . eat-4 expression in AIM was normalized by its expression in the RIGL/R neurons , while for cho-1 the expression in AIM was normalized by its expression in AIYL/R neurons . Immunofluorescence for UNC-17 was performed as described earlier ( Duerr et al . , 2008 ) using a an unc-104 ( e1265 ) mutant strain carrying the cho-1 fosmid reporter ( otIs544 ) . Worms were fixed using methanol/acetone and stained with a rabbit anti-UNC-17 serum diluted 1/100 ( gift from James Rand ) . Immunofluorescence for serotonin was performed using a tube fixation protocol as described earlier ( Serrano-Saiz et al . , 2013 ) . The anti-5HT S-5545 antibody was used at 1/100 and purchased from Sigma Aldrich . Worms were anesthetized using 100 mM of sodium azide ( NaN3 ) and mounted on 5% agarose on glass slides . All images ( except Figure 8 and Figure 7—figure supplement 1 ) were acquired using a Zeiss confocal microscope ( LSM880 ) . Several z-stack images ( each ~0 . 4 μm thick ) were acquired with the ZEN software . Representative images are shown following orthogonal projection of 2–10 z-stacks . Images shown Figure 8 and Figure 7—figure supplement 1 were taken using an automated fluorescence microscope ( Zeiss , AXIO Imager Z1 Stand ) . Acquisition of several z-stack images ( each ~1 μm thick ) was performed with the Micro-Manager software ( Version 3 . 1 ) . Representative images are shown following max-projection of 2–10 z-stacks using the maximum intensity projection type . Image reconstruction was performed using ImageJ software ( Schneider et al . , 2012 ) . For quantification of UNC-7::GFP puncta shown in Figure 7C , images were acquired and z-stack were generated as described above . Manual counting of the UNC-7::GFP puncta was performed using the cell counter plug-in of the ImageJ software . For the quantification of eat-4 and cho-1 expression in AIM for the analysis shown in Figure 5F , images were acquired using a Zeiss confocal microscope ( LSM880 ) and the fluorescence intensity mean was obtained with the ZEN software tool . For results shown in Figures 5E–F and Figure 7C statistical analysis was performed using the Student’s t-test ( tail 2 , type 2 ) . Values are expressed as mean ± standard deviation ( s . d . ) or standard error of the mean ( sem ) as indicated in each figure legend . For results shown in Figure 7 and Figure 7—figure supplement 1 we performed Fisher’s exact test ( two-tailed ) . Wild-type and unc-3 mutant animals were reconstructed in the anterior region of the ventral cord in order to determine neuron morphology and synaptic circuitry . The reconstructions were made from electron micrographs of serial sections as described in White et al . ( 1986 ) . The regions reconstructed were ~150 µm in length and included ~1800 serial sections . We reconstructed the region of the ventral nerve cord that roughly includes the region from AS01 to AS03 motor neurons . Every third section was photographed and printed . All the processes of neurons with cell bodies in the region reconstructed were followed . The neurons were identified by characteristic synaptic or morphological features together with the relative position of their cell bodies in the sequence of cell bodies in the ventral cord ( White et al . , 1986 ) . The two reconstructed animals were the wild-type N2U and the unc-3 ( e151 ) allele in trans to a covering deficiency ( mnDf5 ) . This strain was generated by crossing unc-3 ( e151 ) the strain SP266 mnDp1 ( X;V ) / V; mnDf5 X . The otIs341 ( mgl-1::gfp ) transgenic strain was used to identify mutants affecting the identity of the cholinergic RMDD/V motor neurons . A conventional semi-clonal EMS screen identified the ot712 mutation , which was found to be closely linked to the transgene vsIs33V also present in the strain background . ot712 animals are uncoordinated and unc-42 maps on LGV . Complementation tests between ot712 and two alleles of unc-42 ( e419 and e270 ) confirmed that ot712 is an allele of unc-42 . Sanger sequencing reveals that ot712 harbors a late nonsense mutation ( W181>Stop ) in exon 6 . | To better understand the nervous system—the most complex of all the body’s organs—scientists have begun to painstakingly map its many features . These maps can then be used as a basis for understanding how the nervous system develops and works . Researchers have mapped the connections – called synapses – between all the nerve cells in the nervous system of a simple worm called Caenorhabditis elegans . Cells communicate by releasing chemicals called neurotransmitters across the synapses , but it is not fully known which types of neurotransmitters are released across each of the synapses in C . elegans . Now , Pereira et al . have mapped all worm nerve cells that use a neurotransmitter called acetylcholine by fluorescently marking proteins that synthesize and transport the neurotransmitter . This map revealed that 52 of the 118 types of nerve cells in the worm use acetylcholine , making it the most widely used neurotransmitter . This information was then combined with the findings of previous work that investigated which nerve cells release some other types of neurotransmitters . The combined data mean that it is now known which neurotransmitter is used for signaling by over 90% of the nerve cells in C . elegans . Using the map , Pereira et al . found that some neurons release different neurotransmitters in the different sexes of the worm . Additionally , the experiments revealed a set of proteins that cause the nerve cells to produce acetylcholine . Some of these proteins affect the fates of connected nerve cells . Overall , this information will allow scientists to more precisely manipulate specific cells or groups of cells in the worm nervous system to investigate how the nervous system develops and is regulated . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2015 | A cellular and regulatory map of the cholinergic nervous system of C. elegans |
Human lung adenocarcinoma exhibits a propensity for de-differentiation , complicating diagnosis and treatment , and predicting poorer patient survival . In genetically engineered mouse models of lung cancer , expression of the BRAFV600E oncoprotein kinase initiates the growth of benign tumors retaining characteristics of their cell of origin , AT2 pneumocytes . Cooperating alterations that activate PI3’-lipid signaling promote progression of BRAFV600E-driven benign tumors to malignant adenocarcinoma . However , the mechanism ( s ) by which this cooperation occurs remains unclear . To address this , we generated mice carrying a conditional BrafCAT allele in which CRE-mediated recombination leads to co-expression of BRAFV600E and tdTomato . We demonstrate that co-expression of BRAFV600E and PIK3CAH1047R in AT2 pneumocytes leads to rapid cell de-differentiation , without decreased expression of the transcription factors NKX2-1 , FOXA1 , or FOXA2 . Instead , we propose a novel role for PGC1α in maintaining AT2 pneumocyte identity . These findings provide insight into how these pathways may cooperate in the pathogenesis of human lung adenocarcinoma .
Non-small cell lung cancer ( NSCLC ) is the leading cause of cancer-related death , with lung adenocarcinoma ( LUAD ) being the most common NSCLC subtype ( Siegel et al . , 2016 ) . Due to the morbidity and mortality associated with LUAD , there is an urgent need to better characterize how key genetic drivers contribute to the pathogenesis of this disease . To that end , since the original discovery of KRAS mutations in human lung cancer cells ( Capon et al . , 1983 ) , it has emerged that ~75% of LUADs display mutational activation of key components of receptor tyrosine kinase ( RTK ) signaling that , in turn , promote activation of RAS and its key downstream effectors: the RAF→MEK→ERK→MAP kinase ( MAPK ) and the PI3’-lipid pathways ( Cancer Genome Atlas Research Network , 2014; Heist and Engelman , 2012 ) . Moreover , mutational activation of RTKs or downstream signaling proteins ( e . g . EGFR/ERBB1 , ALK , ROS1 , NTRK , BRAF ) serve as predictive biomarkers for the clinical deployment of FDA-approved inhibitors of these oncoprotein kinases for the treatment of genetically-defined subsets of lung cancer ( Drilon et al . , 2018; Hyman et al . , 2015; Rosell et al . , 2012; Scagliotti et al . , 2010; Shaw et al . , 2013; Shaw et al . , 2014 ) . Mutational activation of BRAF occurs in ~8% of LUAD , with the most common single mutation ( BRAFT1799A ) encoding the BRAFV600E oncoprotein kinase ( Cancer Genome Atlas Research Network , 2014 ) . To model BRAFV600E driven cancers , we previously described BrafCA mice carrying a CRE-activated allele of Braf that expresses normal BRAF prior to CRE-mediated recombination , after which BRAFV637E ( orthologous to human BRAFV600E and for simplicity henceforth referred to as BRAFV600E ) , is expressed from the endogenous chromosomal locus ( Dankort et al . , 2007 ) . This mouse has proven useful in modeling many cancer types in which BRAFV600E is implicated as a driver oncoprotein ( Charles et al . , 2011; Dankort et al . , 2009; Sakamoto et al . , 2017; Trejo et al . , 2013; Wang et al . , 2012 ) . Taken together , these studies indicate that BRAF mutation serves as a foundational initiating event for tumorigenesis in many target tissues . However , the progression of benign tumors initiated by BRAFV600E expression into malignant cancer invariably requires additional events such as silencing of tumor suppressors ( e . g . INK4A-ARF , TP53 , PTEN , CDX2 ) or activation of cooperating oncogenes ( PIK3CA , CTNNB1 , c-MYC ) ( Charles et al . , 2014; Dankort et al . , 2007; Dankort et al . , 2009; Huillard et al . , 2012; Juan et al . , 2014; Sakamoto et al . , 2017; Trejo et al . , 2013; Tsao et al . , 2004; Yu et al . , 2009 ) . In mouse models of lung carcinogenesis , there are key similarities between the early stages of tumorigenesis observed in response to expression of either the KRASG12D or BRAFV600E oncoproteins ( Dankort et al . , 2007; Trejo et al . , 2012 ) . While tumors initiated by BRAFV600E remain as benign adenomas with certain features of senescence ( Dankort et al . , 2007; Jackson et al . , 2001 ) , a proportion of KRASG12D initiated lung tumors progress to frank adenocarcinomas within six months , most likely due to the ability of KRASG12D to activate the PI3’-kinase signaling pathway ( Castellano et al . , 2013; Murillo et al . , 2018; Rodriguez-Viciana et al . , 1994; Vivanco and Sawyers , 2002; Yuan and Cantley , 2008 ) . Consistent with this hypothesis , co-expression of BRAFV600E and PIK3CAH1047R , a mutationally-activated form of PI3’-kinase-α ( PI3Kα ) , in AT2 pneumocytes leads to rapid growth of lung tumors , many of which display progression to frank malignancy bearing various hallmarks of the cognate human disease ( Kinross et al . , 2012; Trejo et al . , 2013 ) . Thus , these genetically manipulated mice provide a unique opportunity to genetically and biochemically separate and analyze the effects of activation of these two critical downstream arms of RTK→RAS signaling individually or in combination . Whereas the original BrafCA mouse allowed insights into cancer initiation , progression and therapy , there remain many questions that this mouse is inadequately configured to address . For example , it is not trivial to identify and isolate pure populations of tumor cells without significant stromal contamination , particularly in contexts when tumor cells are rare , such as in the earliest stages of tumorigenesis , or in the context of minimal residual disease following pathway-targeted inhibition of BRAFV600E signaling ( Dankort et al . , 2007 ) . To address these issues we and others have used mice carrying CRE-activated alleles that express fluorescent proteins , such as the mT-mG allele in which the activity of CRE recombinase silences the expression of tdTomato and elicits expression of EGFP ( Muzumdar et al . , 2007 ) . However , this approach is confounded by the observation that not all cells expressing the desired oncoprotein also express EGFP and vice versa . In order to unequivocally identify BRAFV600E expressing cells we have generated BrafCAT mice carrying a new CRE-activated Braf allele . Like the original BrafCA allele , BrafCAT encodes normal BRAF prior to CRE-mediated recombination , after which the recombined allele expresses a bicistronic BrafT1910A-P2A-tdTomato mRNA encoding both BRAFV600E and the red fluorophore tdTomato . Moreover , here we report the use of BrafCAT mice to explore the cooperation of oncogenic BRAFV600E and PI3KαH1047R in lung carcinogenesis in greater mechanistic detail . In brief , BRAFV600E-driven lung tumors maintain expression of markers of AT2 identity , including the known regulators of AT2 identity , NKX2-1 , FOXA1 , and FOXA2 ( Bruno et al . , 1995; Camolotto et al . , 2018; DeFelice et al . , 2003; Hamvas et al . , 2013; Lazzaro et al . , 1991; Minoo et al . , 1995; Snyder et al . , 2013; Stahlman et al . , 1996; Winslow et al . , 2011; Yuan et al . , 2000 ) . By contrast , co-expression of BRAFV600E and PI3KαH1047R leads to development of lung tumors that show variable and widespread loss of expression of markers of AT2 pneumocyte terminal differentiation including the well-characterized surfactant proteins , SFTPA , SFTPB , and SFTPC . Notably reduced expression of AT2 markers begins early in tumor development and occurs despite sustained NKX2-1 , FOXA1 , and FOXA2 expression in tumor cells . Hence , these data shed light on the mechanisms by which pathways that cooperate in lung tumorigenesis also cooperate to influence the differentiation state of tumor cells . Indeed , these findings bear similarity to observations in human lung adenocarcinomas in which poorly differentiated and metastatic cancers often show loss of expression of functional markers of lung identity despite maintaining expression of NKX2-1 ( Yatabe et al . , 2002 ) . Consequently , our results may shed light on our understanding of human lung cancer progression and how normal lung epithelial cells may lose their differentiation status following activation of cooperating oncogenic pathways .
To generate a reporter of BRAFV600E oncoprotein expression , we linked its expression to the expression of the red fluorophore , tdTomato . To accomplish this , we made use of the design of the original BrafCA allele , in which the modified exon 18 and the remainder of the Braf allele is not transcribed prior to the action of CRE recombinase ( Dankort et al . , 2007 ) due to the insertion of a triple polyadenylation/mRNA transcription termination signal from SV40 ( Figure 1A ) ( Srinivas et al . , 2001 ) . Consequently , we designed a targeting vector containing the final coding exon ( 22 ) of mouse Braf in which the stop codon was removed , followed by sequences encoding: 1 . an in-frame glycine-serine-glycine-P2A self-cleaving peptide; 2 . sequences encoding a membrane-tethered tdTomato-CAAX protein and; 3 . a PGK-PURO selectable marker flanked by Frt sites for subsequent removal by FLP recombinase . Following electroporation of this construct into 2H1 BrafCA/+ ES cells , from which the original BrafCA mice were generated , 288 puromycin resistant clones were selected and screened by PCR for homologous recombination of the construct into the distal end of the BrafCA allele ( Dankort et al . , 2007 ) . However , because the targeted ES cells are heterozygous for both normal Braf and the genetically manipulated BrafCA allele , and because there was no way to direct homologous recombination of the targeting vector to the previously targeted BrafCA allele , we expected to target both homologues . Because BRAF is expressed in ES cells , we reasoned that modification of the normal allele would lead to ES cells with constitutive tdTomato-CAAX expression . Indeed , ~50% of PCR positive ES cell clones displayed constitutive membrane associated red-fluorescence and were used to generate BrafTOM mice , in which tdTomato serves as a marker for any cells expressing normal BRAF ( van Veen et al . , 2016 ) . By contrast , homologous recombination of the targeting vector into the BrafCA allele should give rise to ES cells that do not express tdTomato due to the strong transcriptional termination signal . However , upon the addition of a cell permeable TAT-CRE protein to these cells , they should initiate the expression of both BRAFV600E and tdTomato ( Figure 1B ) . Hence , this in vitro strategy allowed us to both identify appropriately targeted BrafCAT/+ ES cells and also indicated the appropriate functioning of the BrafCAT allele in response to CRE-mediated recombination prior to the generation of mice . BrafCAT mice were generated from an appropriately targeted ES cell clone ( 1E6 ) . To compare and contrast lung tumorigenesis following CRE-mediated recombination of the BrafCAT versus the original BrafCA allele , mice of the appropriate genotype were infected with 107 pfu of Ad-CMV-CRE and analyzed at 8 weeks post-initiation ( p . i . ) . We observed similar lung tumor formation in BrafCAT versus BrafCA mice ( Figure 1C ) with the only discernable difference being the red fluorescence of lung tumors arising in the BrafCAT mice ( Figure 1C , D and Figure 1—figure supplement 1A ) . In embryonic fibroblasts ( MEFs ) derived from BrafCAT mice , tdTomato fluorescence was detected by flow cytometry within 24 hr after expression of CRE and plateaued by 96 hr ( Figure 1—figure supplement 1B ) . Following CRE-mediated recombination of BrafCAT in both MEFs and mouse lung cells ( Figure 2B ) , the total amount of fluorescence from the BrafCAT allele was modest , likely due to being driven by the endogenous Braf promoter . However , tdTomato expressing cells were readily differentiated from autofluorescence by using a channel with no fluorophore ( FITC ) as a marker of autofluorescence , as has been previously described ( Dane et al . , 2006 ) . Together , these data indicate that the BrafCAT allele functions analogously to the BrafCA allele for the development of benign lung tumors following CRE-mediated initiation of BRAFV600E expression , and that cells expressing the BRAFV600E-P2A-tdTomato mRNA are readily identified by flow cytometry within a short time frame . To address mechanism ( s ) of cooperation between BRAFV600E and PI3’-lipid signaling in lung cancer progression , lung tumorigenesis was initiated in BrafCAT or BrafCAT; Pik3calat-H1047R ( Pik3caHR hereafter ) mice ( Figure 2A ) and analyzed at 2 , 6 or 12 weeks p . i . ( Figure 2B , for detailed gating strategy see Figure 2—figure supplement 1A ) . Importantly , to initiate oncoprotein expression solely in AT2 pneumocytes , we utilized Ad5-SpC-CRE , which restricts expression of CRE recombinase to Sftpc expressing cells ( Sutherland et al . , 2014 ) . tdTomato expressing tumor cells were detectable by flow cytometry in both BrafCAT and BrafCAT; Pik3caHR mice as early as two weeks p . i . ( Figure 2B ) . To identify alterations in mRNA expression that might explain how PI3KαH1047R promotes malignant transformation of lung tumors initiated by BRAFV600E , we performed RNA-Seq analysis of flow sorted tdTomato+ lung tumor cells driven either by BRAFV600E alone or the combination of BRAFV600E plus PI3KαH1047R . To gain a broad view of pathways and processes differing in these two tumor genotypes , we used Gene Set Enrichment Analyses ( GSEA ) ( Figure 2C ) on , samples from all time points ( for GSEA analyses separated by week see Figure 2—figure supplement 1B ) . Comparing ‘hallmark’ gene sets ( Broad Institute MSigDB: Hallmarks ) , and consistent with the engineered characteristics of the lung tumor cells , GSEA revealed that PI3K→AKT→MTOR signaling ( Figure 2C and D ) and epithelial→mesenchymal transition ( Figure 2C and E ) related genes were significantly elevated in BRAFV600E/PI3KαH1047R-driven lung tumors compared to BRAFV600E-driven tumors . To examine differentiation state we constructed gene sets comprised of the 100 most specific described mRNA markers of the different cell types expressed in the distal lung epithelium ( Han et al . , 2018; Treutlein et al . , 2014 ) namely alveolar type 1 ( AT1 ) and type 2 ( AT2 ) pneumocytes , as well as club and ciliated cells , hereafter referred to as AT1-100 , AT2-100 , club-100 , and ciliated-100 , respectively . Despite these gene sets representing related cell types of the distal lung epithelium , there is only modest overlap between their members , ranging from 2 to 12 of the 100 genes . GSEA using these gene sets demonstrated that , compared to BRAFV600E expressing tumor cells , BRAFV600E/PI3KαH1047R expressing tumor cells showed a significant decrease ( adj . p=0 . 01 ) in expression of transcripts associated with AT2 cell identity ( Figure 2C and F ) . This effect encompassed nearly all AT2 markers including the classical markers , Sftpa/b/c , and newly described markers detected in many different gene classes including Lcn2 ( lipid transporter ) , Bex2 ( transcription factor ) , and Dlk1 ( encoding a non-canonical NOTCH ligand ) . Together these data suggest that PI3KαH1047R signaling promotes reduced expression of markers of AT2 differentiation . Notably , when examined in parallel with 50 hallmark gene sets and markers of other cell types in the distal lung epithelium , the loss of expression of AT2 mRNAs was the strongest effect observed in association with PI3KαH1047R expression ( NES: −2 . 71 , Figure 2C and F ) . Decreased AT2 marker expression was statistically significant as early as 2 weeks ( Figure 3A and Figure 2—figure supplement 1B ) and was also observed at 6 weeks ( Figure 3B and Figure 2—figure supplement 1B ) and 12 weeks ( Figure 3C and Figure 2—figure supplement 1B ) p . i . Hence the effects of PI3KαH1047R on expression of AT2 differentiation markers initiates more rapidly than has been reported in tumors driven by KRASG12D ( Desai et al . , 2014 ) , KRASG12D/TP53Null ( Winslow et al . , 2011 ) , KRASG12D/CTNNB1Δex3 ( Pacheco-Pinedo et al . , 2011 ) , BRAFV600E/TP53Null ( Garnett et al . , 2017; Shai et al . , 2015 ) , or BRAFV600E/CTNNB1Δex3 ( Juan et al . , 2014 ) . In addition to reduced AT2 mRNA marker expression , PI3KαH1047R expression elicited a marked increase in AT1 marker expression when comparing all time points ( Figure 2C ) . By contrast to the changes observed in AT2 marker expression , this change was most significant at earlier time points and was no longer observed to be significant by 12 weeks p . i . ( Figure 2—figure supplement 1B ) . Together , these results suggest a capability of PI3’-lipid signaling to influence AT2 pneumocyte identity . Consistent with the mRNA expression data , Ad5-Sftpc-CRE initiated BRAFV600E/ PI3KαH1047R-driven lung tumors displayed decreased expression of surfactant proteins A and C ( SFTPA and SFTPC ) and Lysozyme , all of which are AT2 pneumocyte markers ( Figure 3B , D and F ) . We next built a pipeline in CellProfiler , which allows for quantification of immunofluorescence images with single cell resolution in thousands of cells algorithmically ( Figure 3—figure supplement 1A–H ) . By this means we noted significant reductions of SFTPA , SFTPC , and LYZ expression at 12 weeks p . i . ( Figure 3C , E and G . Wilcoxon p=0 . 0001 , . 02 , . 02 respectively , data from Figure 3—source data 1 , 2 and 3 ) . It has previously been suggested that a population of cells at the bronchioalveolar junction co-expressing SFTPC and club cell antigen ( CCA ) has properties of bronchio-alveolar stem cells ( BASCs ) ( Kim et al . , 2005 ) , and also that ERK1/2 signaling tone allows for expansion of club cell derived tumors ( Cicchini et al . , 2017 ) . Thus , a hypothesis that might explain our observations is that expression of PI3KαH1047R allows for enhanced expansion of BASC-derived tumors , which express lower levels of AT2 marker genes . We reject this hypothesis based on three lines of evidence . First , there did not appear to be two classes of BRAFV600E/PIK3CAH1047R-induced tumors with respect to AT2 marker expression . Instead , within each tumor we observed variegated loss of AT2 marker expression ( Figure 3B , D , F insets ) . Second , to test whether CCA/SFTPC double positive cells might be the source of emergence of a separate tumor type , we co-stained our AT2 marker panel with antisera to detect expression of club cell antigen ( CCA ) . While CCA staining was readily detectable in airways , it was not observed throughout the BRAFV600E/PI3KαH1047R-driven tumors , including those cells that have reduced AT2 marker expression ( Figure 3B , D and F ) . Finally , we observed tumors arising predominantly within alveolar spaces , a pattern characteristic of AT2-derived tumors , whereas club cell derived tumors tend to arise predominantly at bronchioalveolar duct junctions ( Cicchini et al . , 2017 ) . Taken together , these data suggest that BRAFV600E/PI3KαH1047R-driven lung tumors arise from AT2 cells but rapidly lose their differentiated identity under the influence of PI3’-lipid signaling . Extensive research demonstrates that the NKX2-1 and FOXA1/2 transcription factors pattern and maintain the differentiated identity of normal lung cells ( Bruno et al . , 1995; Camolotto et al . , 2018; DeFelice et al . , 2003; Hamvas et al . , 2013; Lazzaro et al . , 1991; Minoo et al . , 1995; Snyder et al . , 2013; Stahlman et al . , 1996; Yuan et al . , 2000 ) . As these transcription factors also display decreased expression in some models of LUAD ( Juan et al . , 2014; Snyder et al . , 2013; Winslow et al . , 2011 ) , we next examined if changes in bulk NKX2-1 or FOXA1/2 expression in BRAFV600E/PIK3CAH1047R-driven lung tumors might explain the observed alterations in AT2 marker gene expression . Analysis of RNA-Seq data showed no consistent decrease of NKX2-1 or FOXA1/2 mRNA expression in BRAFV600E/PIK3CAH1047R-driven lung tumors . As this was initially surprising , we sought to further verify that AT2 identity was lost independent of NKX2-1 expression levels or localization in BRAFV600E/PIK3CAH1047R-driven lung tumors . Dual color immunofluorescence staining demonstrated that the observed loss of AT2 identity is not associated with a decrease or change in nuclear localization of NKX2-1 at either 2 or 12 weeks p . i . ( Figure 4A–B ) . Quantification of immunostaining supports this observation , with no significant alteration in nuclear NKX2-1 staining at 2 weeks , and only a slight but significant increase in nuclear NKX2-1 staining observed when comparing 12 week p . i . BRAFV600E/PIK3CAH1047R-driven tumors to paired BRAFV600E-driven tumors ( Figure 4C , data from Figure 4—source data 1 ) . Quantification of the same tumors confirmed a significant decrease of SFTPA staining in BRAFV600E/PIK3CAH1047R-driven tumors first observed at 2 weeks and persisting at 12 weeks p . i . ( Figure 4D ) . Since we performed our quantification with single cell resolution we were able to compare NKX2-1 and SFTPA immunofluorescence staining on a cell-by-cell basis ( Figure 4E–H ) . BRAFV600E-driven tumors show association of NKX2-1 and SFTPA staining consistent across time points ( Spearman Rho = 0 . 23- . 27 , Figure 4E ) . BRAFV600E/PIK3CAH1047R-driven tumors by contrast initially show almost no association between levels of NKX2-1 and SFTPA ( Spearman Rho = 0 . 07 , Figure 4F ) , but by 12 weeks p . i . the association of these two markers has increased markedly ( Spearman Rho = 0 . 40 , Figure 4F ) . Because the reduction of SFTPA precedes the observed association of NKX2-1 and SFTPA , we conclude that other factors must explain the rapid reduction in SFTPA expression . Dividing tumor cells into classes representing NKX2−1 ± and SFTPA +/- ( for definitions , see Materials and methods ) shows significant effects ( chi-squared p<0 . 001 ) of tumor genotype on co-expression of NKX2-1 and SFTPA ( Figure 4G–H ) . At both 2 and 12 weeks p . i . , the largest proportional increase driven by PIK3CAH1047R expression is seen in NKX2-1+/SFTPA- tumor cells ( Figure 4G–H ) , implying that at both early and late time points , decreased expression of NKX2-1 cannot explain the observed decrease in SFTPA expression . Similarly , neither the expression of FOXA1 ( Figure 4—figure supplement 1A ) nor FOXA2 ( Figure 4—figure supplement 1B ) at 12 weeks p . i . correlated with decreased SFTPA expression as assessed by immunostaining . Nor is the phosphorylation status of NKX2-1 at a critical ERK targeted residue ( pS327 ) associated with loss of SFTPA expression ( Figure 4—figure supplement 1C ) . Together these data suggest that the decreased expression of markers of AT2 cell differentiation that we observed upon co-activation of PI3’-lipid signaling in BRAFV600E driven tumors is largely independent of the expression levels of these known regulators of AT2 cell identity . We also note that at later time points , repression of NKX2-1 , FOXA1 , or FOXA2 protein expression in scattered cells may serve to augment the dedifferentiation phenotype that we observed beginning at 2 weeks p . i . To test the generality of our results , we examined KRASG12D ( Figure 4—figure supplement 1D ) and KRASG12D/PIK3CAH1047R-driven lung tumors ( Figure 4—figure supplement 1E ) at 16 weeks p . i . , a time at which mutationally-activated PI3KαH1047R strongly enhances KRASG12D-driven lung tumorigenesis ( Green et al . , 2015 ) . Indeed , in this model , we observed a similar and significant decrease in SFTPA expression comparing KRASG12D/PIK3CAH1047R- to KRASG12D-driven tumors ( Figure 4—figure supplement 1D , E , G , data from Figure 4—source data 2 ) , which did not correspond to reduced NKX2-1 expression ( Figure 4—figure supplement 1E , F ) . Intriguingly , although mutationally-activated KRASG12D is reported to activate PI3’-lipid signaling in lung tumors ( Castellano et al . , 2013; Engelman et al . , 2008; Gupta et al . , 2007b; Molina-Arcas et al . , 2013; Murillo et al . , 2018; Rodriguez-Viciana et al . , 1994 ) , we did not observe extensive loss of SFTPA immunoreactivity in these tumors ( Figure 4—figure supplement 1D ) . Both KRASG12D/PIK3CAH1047R-driven tumors and KRASG12D-driven tumors showed similarly modest association of SFTPA and NKX2-1 expression ( Figure 4—figure supplement 1H–I ) . Taken together these data suggest that either there exist additional factors that regulate AT2 pneumocyte identity independently of the NKX2-1/FOXA1/FOXA2 regulatory axis , or that these well-known regulators of pneumocyte identity require the presence of one or more additional factor ( s ) for their transcriptional function . To explore the increase in AT1 marker expression observed in BRAFV600E/PI3KαH1047R-driven tumors in more detail , we immunostained both early and late tumors for the expression of the AT1 marker AQP5 . At 2 weeks p . i . we observed a striking increase in AQP5 expression when comparing BRAFV600E/PI3KαH1047R-driven tumors to BRAFV600E-driven tumors ( Figure 5A and B , data from Figure 5—source data 1 ) . Interestingly , at 12 weeks p . i . the difference in AQP5 immunoreactivity is no longer significant between BRAFV600E/PI3KαH1047R and BRAFV600E-driven tumors , mirroring our mRNA expression profiling results ( Figure 2 and Figure 2—figure supplement 1b ) . The pattern of AQP5 expression likely explains this finding , as modest AQP5 staining is seen throughout BRAFV600E-driven tumors as previously reported ( Trejo et al . , 2013 ) , whereas BRAFV600E/PI3KαH1047R-driven tumors show strong AQP5 in some areas and essentially absent AQP5 in other areas ( Figure 5A ) . Co-immunostaining of AQP5 and LYZ showed similar patterns in which BRAFV600E-driven tumors display widespread expression of both of AT1 and AT2 markers ( Figure 5C ) . By contrast , BRAFV600E/PI3KαH1047R-driven tumors show some areas that are double positive for both AQP5 and LYZ ( Figure 5C , cyan arrowheads ) , some areas with only expression of AQP5 ( Figure 5C , green arrowheads ) , some areas with only expression of LYZ ( Figure 5C , red arrowheads ) , and some areas where neither is expressed ( Figure 5C , yellow arrowheads ) . Quantification of these data shows correlation ( Spearman Rho = 0 . 54 ) between AQP5 and LYZ in BRAFV600E-driven tumors ( Figure 5D ) , but lower correlation ( Spearman Rho = 0 . 13 ) between these markers in BRAFV600E/PI3KαH1047R-driven tumors ( Figure 5E ) . Comparing these two tumor types directly shows a significant effect of genotype on staining distribution ( Figure 5F ) , with BRAFV600E/PI3KαH1047R-driven tumors showing a strong decrease in AQP5+/LYZ+ double positive cells , with corresponding increases in each of the remaining classes of cells ( AQP5+/LYZ-; AQP5-/LYZ+; and AQP5-/LYZ- ) . Co-expression of AT1 or AT2 markers in BRAFV600E-driven tumors is reminiscent of what is observed in bipotent progenitor cells , which are reported to give rise to both AT1 and AT2 cells ( Desai et al . , 2014 ) . To search for additional indicators of a bipotent progenitor like state induced by BRAFV600E→MEK→ERK signaling , we analyzed transmission electron micrographs of BRAFV600E-driven lung tumor sections from suitably manipulated BrafCA mice 11 weeks p . i . Normal AT2 cells display a cuboidal morphology and contain many surfactant rich lamellar bodies ( Figure 5—figure supplement 1A , cyan ) . BRAFV600E-driven tumor cells displayed gross morphological similarities to AT2 cells ( Figure 5—figure supplement 1B ) , but in a subset of tumor cells large vacuoles , not seen in normal AT2 cells , were observed ( Figure 5—figure supplement 1B , purple ) . Enhanced magnification of these structures demonstrated a rough chrysanthemum like pattern of electron dense material ( Figure 5—figure supplement 1C ) characteristic of glycogen storage ( Revel , 1960 ) . As glycogen storage vacuoles are another hallmark of bipotent progenitor cells ( Desai et al . , 2014 ) , we propose that BRAFV600E→MEK→ERK signaling alone drives AT2 pneumocytes toward this fate . As BRAFV600E/PI3KαH1047R-driven tumors show some cells with co-expression of AT1 and AT2 markers , and many cells with reductions of either or both of these marker classes , we believe the coincident activation of ERK1/2 plus PI3’-lipid signaling promotes more profound de-differentiation of AT2 cells . To identify novel candidate transcription factors that may participate in the establishment or maintenance of AT2 pneumocyte identity and function ( Figure 6A ) , we took a three-step approach . First , we performed whole transcriptome correlation network analysis using Weighted Gene Correlation Network Analysis ( WGCNA ) ( Langfelder and Horvath , 2008 ) to identify potentially relevant gene modules ( Figure 6B ) . The majority of the AT2-100 and AT1-100 mRNAs fell into a single cluster ( Cluster 2 – Dark blue Figure 6B , C ) . Next , to identify candidate regulators within cluster 2 , we filtered the 2852 mRNAs in this cluster to select for those with demonstrated roles in transcriptional regulation . Finally , we filtered these selected transcription factors for differential expression in BRAFV600E- vs BRAFV600E/PI3KαH1047R-driven lung tumors using a promiscuous cutoff of adjP <0 . 2 . The intersection of these three methods left a single candidate , the nuclear receptor co-activator , PGC1α ( Figure 6D ) . PGC1α is a known transcriptional regulator , clusters with the majority of AT1 and AT2 genes , and its mRNA is significantly decreased in BRAFV600E/PI3KαH1047R-driven tumors compared to BRAFV600E-driven tumors ( Figure 6E ) . The control of PPARGC1A levels by PI3’-lipid signaling may also be true in human lung tumors: those bearing mutations in either PIK3CA , AKT1 , or PTEN have significantly reduced PPARGC1A mRNA expression compared to tumors bearing none of these mutations ( Figure 6—figure supplement 1A ) . Finally , we sought to determine if PGC1α expression correlates with maintenance of lung identity on a cell-by-cell basis within BRAFV600E/PI3KαH1047R-driven tumors in mice . Immunostaining revealed that cells with decreased expression of SFTPA lack nuclear PGC1α ( Figure 6F , red arrows ) . Conversely , tumor cells with detectable nuclear localization of PGC1α display readily detectable levels of SFTPA ( Figure 6F , green arrows ) . To directly test if PGC1α regulates AT2 identity , we crossed a floxed , conditional null allele of Pgc1α ( Ppargc1alox/lox ) to BrafCAT mice and generated littermate cohorts of BrafCAT; Ppargc1alox/lox , and BrafCAT; Ppargc1alox/+ mice . Lung tumorigenesis was initiated in these mice using 106 pfu Ad5-Sftpc-CRE ( Figure 7A ) with lungs harvested 12 weeks p . i . for isolation and analysis of tdTomato+ tumor cells by RNA sequencing . As expected , BRAFV600E/PGC1αNull-driven lung tumor cells showed decreasd expression of mRNAs encoding proteins involved in oxidative phosphorylation ( Figure 7A ) . BRAFV600E/PGC1αNull-driven tumors also displayed a widespread decrease in markers of AT2 differentiation status as compared to BRAFV600E-driven tumors that retain PGC1α expression ( Figure 7A and Figure 7—figure supplement 1 ) . Interestingly , silencing of PGC1α recapitulated some other aspects of PI3Kα activation , including increasing markers of EMT , but some noticeable differences were observed in mRNA profiles including significant increases in ciliated and club cell markers ( Figure 7A ) . LYZ and SFTPA immunoreactivity in BRAFV600E and BRAFV600E/PGC1αNull-driven tumors validated the decrease in AT2 marker expression in the absence of PGC1α expression ( Figure 7B–C , data from Figure 7—source data 1 ) . We next sought to test if the role of PGC1α in AT2 pneumocyte identity maintenance could be through direct action on the promoters of AT2 pneumocyte specific genes . As PGC1α generally co-activates nuclear receptors such as PPARγ , we sought to discover its potential binding partner relevant to AT2 pneumocyte gene regulation . To this end we performed a motif discovery analysis using Multiple Em for Motif Elicitation ( MEME ) ( Bailey and Elkan , 1994; Bailey et al . , 2009 ) algorithm and scanning the promoter regions of the AT2-100 . The most enriched novel motif ( Figure 7—figure supplement 1B–C ) bears significant similarity to the known binding motif of the nuclear receptor , NR5A2 ( JASPAR: MA0505 . 1 ) , also known as Liver Receptor Homolog ( LRH ) 1 ( Gupta et al . , 2007a ) . To functionally test whether PGC1α can act at promoter sequences to regulate key markers of AT2 identity , we generated reporter constructs in which ~ 5 kb upstream of the transcription start site ( TSS ) of the genes encoding surfactant proteins A , B , or C ( Sftpa/b/c ) was inserted into a luciferase reporter plasmid . Transfection of individual expression constructs for NR5A2 , PGC1α , NKX2-1 , or FOXA1 showed modest promoter induction of up to four fold over mock-transfected cells ( Figure 7D ) . However , co-transfection of these four factors together showed a dramatic 25–100 fold induction of the SFTPA , SFTPB , and SFTPC promoters . Importantly , the absence of PGC1α severely hampered the ability of NKX2-1 and FOXA1 to transactivate the surfactant A and B promoters , suggesting a functional role of PGC1α in the transcriptional transactivation of AT2 promoters . Based on the discovery that NR5A2 and PGC1α can cooperate with NKX2-1 and FOXA1 to transactivate AT2 pneumocyte promoters , we hypothesized that these proteins may exist in a biochemical complex . To test this directly , we used the immortalized mouse lung epithelial line MLE-12 ( Wikenheiser et al . , 1993 ) . We first verified that these cells indeed express the pertinent proteins and maintain surfactant expression ( Figure 7—figure supplement 1D ) . We next performed co-immunoprecipitation assays from MLE-12 cell extracts . While magnetic beads conjugated to mouse IgG did not enrich eluates for NKX2-1 , magnetic beads conjugated to a mouse monoclonal antibody directed against PGC1α readily co-immunoprecipitated NKX2-1 ( Figure 7E ) . To confirm this result , we showed that , while magnetic beads conjugated to rabbit IgG did not enrich eluates for PGC1α , magnetic beads conjugated to a rabbit monoclonal antibody directed against NKX2-1 also co-immunoprecipitated PGC1α ( Figure 7F ) . Thus , PGC1α and NKX2-1 appear to co-exist in a complex in cells that express genes encoding AT2 expressed surfactant proteins .
Genetically engineered mouse models have proven to be invaluable tools that complement the significant advances being made in the genetic and biochemical characterization of the initiation , progression and maintenance of human cancers . As genome sequencing efforts catalog clinically actionable mutations and their correlations with the cancer’s observed phenotypes , directly testing how these mutations affect disease initiation , progression , and response to novel therapeutics becomes a high priority . Here we describe a new GEM model that has allowed a deeper dissection of the stages of BRAFV600E-driven lung cancer , a disease that kills ~4000 patients per year in the U . S . A . ( Siegel et al . , 2016 ) . To engineer this mouse we employed a strategy in which the previously targeted CRE-activated BrafCA allele was re-targeted to allow for the expression of both BRAFV600E and tdTomato from a single bicistronic mRNA following CRE-mediated recombination . Consequently , all BRAFV600E expressing cells are predicted to be red fluorescent by virtue of co-expression of tdTomato . This is an alternative approach to the more widely employed method of using a fluorophore expressed in trans from a different promoter ( e . g . Rosa26 ) the expression of which is co-activated by CRE recombinase ( Livet et al . , 2007; Madisen et al . , 2010; Muzumdar et al . , 2007; Snippert et al . , 2010 ) . The approach described here has both advantages and disadvantages compared to the use of a fluorophore expressed in trans approach . The main disadvantage of our approach appears to be the modest fluorescence emission of tdTomato when expressed downstream of a P2A element from the mutationally activated BrafT1910A mRNA . Many LSL-XFP alleles have been designed for very high levels of expression by the incorporation of enhancers and strong promoters , whereas our model is driven by the endogenous Braf enhancer/promoter sequences . Hence , although the level of tdTomato fluorescence in lung tumors arising in BrafCAT mice is readily detected by flow cytometry , detection of tdTomato expression by immunohistochemistry in FFPE tissue sections is problematic . However , we have developed protocols that allow for detection of native tdTomato fluorescence in frozen sections ( Figure 1D ) or by a ligation proximity assay in FFPE sections ( Daphne Pringle , unpublished ) . By contrast , the use of a fluorophore expressed in trans approach inherently relies on the simultaneous activity of CRE at distinct regions of the genome . Thus it is possible for the proto-oncogenic BrafCA locus to be subject to CRE-mediated recombination without recombination of the reporter , or vice-versa . While this is acceptable for some applications , it adds confounding noise that may be amplified under conditions when tumor cells have been specifically depleted by the application of pathway targeted or immunomodulatory therapy that has negligible effect on normal cells . Our approach avoids this noise entirely and results in a level of fluorescence that is directly correlated with expression of BRAFV600E on a cell-by-cell basis ( van Veen et al . , 2016 ) . Importantly , our approach is especially compatible with the emerging technology of single cell RNA sequencing ( scRNA-Seq ) , in which it is desirable to isolate single BRAFV600E oncoprotein kinase expressing cells for analysis . Our studies were aided greatly by recent characterizations of the gene expression changes accompanying development of the mouse distal lung epithelium ( Treutlein et al . , 2014 ) and KRASG12D-mediated oncogenic transformation of AT2 cells ( Desai et al . , 2014 ) . Using scRNA-Seq , the authors reconstructed the lineage hierarchies of AT1 , AT2 , club and ciliated cells , as well as provided a new list of markers useful in identifying these major cell types of the distal lung epithelium . We used these marker lists extensively and without bias or selection for most of our analyses . Importantly , we constructed gene sets for the GSEA analyses that showed widespread decreased AT2 marker expression in BRAFV600E/PIK3CAH1047R-driven lung tumors . Analysis of the full complement of AT2 markers paints a dramatic picture; nearly every AT2 identity marker is diminished in expression in BRAFV600E/PIK3CAH1047R-driven tumors as compared to BRAFV600E-driven tumors , as early as two weeks p . i . Importantly we noted a similar early and consistent repression of PGC1α expression in BRAFV600E/PIK3CAH1047R-driven tumors . Here our data support previous findings that indicate that PGC1α expression is repressed by the PI3K→AKT signaling axis via three insulin response sequences found in the PGC1α promoter shown to bind to the AKT-regulated FOXO1 transcription factor ( Daitoku et al . , 2003; Kemper et al . , 2014 ) . It has also been suggested that BRAFV600E signaling can suppress PGC1α expression via the MITF transcription factor in melanoma , further demonstrating the complicated interplay between growth factor signaling and oxidative phosphorylation ( Haq et al . , 2013 ) , though this study did not examine any potential connection of PGC1α to tumor cell differentiation status . PIK3CA mutation is found in ~4% of human LUAD ( Campbell et al . , 2016 ) that , although quite rare , still represents a significant patient population due to the high prevalence of lung cancer in our society . Previous studies have demonstrated that the ability of mutationally-activated KRASG12D to bind and activate PI3Kα is critical for its tumor promoting activities , as well as tumor maintenance ( Castellano et al . , 2013; Gupta et al . , 2007b; Molina-Arcas et al . , 2013; Murillo et al . , 2018 ) . Despite this , activation of PI3’-lipid signaling remains rate limiting with respect to tumor initiation and growth in KRASG12D-driven lung tumors as evidenced by the strong cooperation between KRASG12D and PIK3CAH1047R in promoting lung tumorigenesis in a GEM model ( Green et al . , 2015 ) . This cooperation likely reflects the fact that the initial expression of KRASG12D is a relatively weak activator of both PI3’-lipid signaling and the RAF→MEK→ERK pathways ( Cicchini et al . , 2017 ) . Indeed , whereas KRASG12D derived lung cancer cells show little phosphorylation of AKT at a key activating amino acid ( S473 ) , KRASG12D/PIK3CAH1047R derived cells show strong pS473-AKT phosphorylation ( Green et al . , 2015 ) . Our results suggest that the lack of PI3’-lipid signaling in KRASG12D-driven lung tumors is not only limiting in tumor growth , but in propensity for AT2 pneumocyte de-differentiation . Intriguingly , recent studies have shown that stromally derived IGF-1 promotes a cancer stem cell like phenotype in Kras mutated cell lines ( Chen et al . , 2014 ) , via the PI3K→AKT signaling axis . It will be interesting to examine in future studies how various tumor genotypes and tumor microenvironments converge upon the initiation and evolution of LUAD de-differentiation . Lung adenocarcinoma differentiation status , as judged by pathological criteria , remains a critically important prognostic factor in predicting patient survival ( Yoshizawa et al . , 2011 ) . However , only in recent years have we begun to understand the genetic aberrations that can directly promote loss of differentiation status . The protein most directly demonstrated to influence both lung adenocarcinoma differentiation status and progression is NKX2-1 , a homeodomain transcription factor . Indeed , in the KrasLSL-G12D GEM model of KRASG12D-driven lung adenocarcinoma , NKX2-1 expression is diminished or silenced in the most poorly differentiated tumors . Moreover , shRNA-mediated inhibition of NKX2-1 expression is reported to enhance the metastatic potential of KRASG12D/TP53Null-driven lung cancer cells ( Winslow et al . , 2011 ) . Interestingly , concomitant expression of KRASG12D with genetic silencing of NKX2-1 expression has qualitatively different effects , promoting trans-differentiation of tumor cells into a gastric fate resembling de-differentiated mucinous adenocarcinoma ( Snyder et al . , 2013 ) , an effect that requires the activity of FOXA1/FOXA2 ( Camolotto et al . , 2018 ) . Intriguingly , while even haploinsufficiency of NKX2-1 promoted the appearance of mucinous adenocarcinoma in the KrasLSL-G12D GEM model of lung adenocarcinoma , the same was not true in a lung adenocarcinoma model driven by expression of a mutationally-activated form of the EGF receptor ( Maeda et al . , 2012 ) . In this model , haploinsufficiency of NKX2-1 slowed tumor progression rather than enhancing it . Importantly , loss-of-function mutations or silencing of NKX2-1 is a relatively infrequent event in lung adenocarcinoma . Instead NKX2-1 is often found as the most significantly focally amplified locus ( Campbell et al . , 2016 ) . Further , in human NSCLC cell lines in which NKX2-1 is amplified , RNAi-mediated inhibition of NKX2-1 expression elicited decreased cell division and apoptosis ( Kwei et al . , 2008 ) . Hence , the role of NKX2-1 in LUAD progression is thus simultaneously critically important and also complicated . In this case , GEM models provide an ideal system in which to study the contribution of individual mutations to each step of tumor initiation and progression without the complications of mutagen-induced genome hypermutation as is common in KRAS-mutated human lung cancer cells ( Chalmers et al . , 2017 ) . While a wealth of literature has shown the direct effect of NKX2-1 on AT2 promoters ( Bruno et al . , 1995 ) , and the importance of NKX2-1 in normal lung development ( DeFelice et al . , 2003 ) , it has also been shown that transcriptional co-activators are critical for NKX2-1 function ( Cassel et al . , 2002; Di Palma et al . , 2003; Park et al . , 2004; Yi et al . , 2002 ) . Here we have shown that the binding motif of the nuclear receptor , NR5A2 is highly enriched in the promoters of AT2 specific genes . We also demonstrate that NR5A2 , and its known co-factor PGC1α ( Yazawa et al . , 2010 ) , can potently enhance the activity of NKX2-1 at the promoters for surfactant proteins A and B . Interestingly , while NR5A2 and PGC1α can activate the promoter of SFTPC alone , the added presence of NKX2-1 and FOXA1 does not further co-activate this promoter . It may be the case that PGC1α and NKX2-1 act independently at this promoter , or it may be that there are additional or alternative transcriptional regulators not present in 293 T cells which , when present , allow PGC1α and NKX2-1 to cooperate . In vivo , in BRAFV600E/PIK3CAH1047R-driven tumors , we observed an early decrease of PGC1α mRNA expression , and importantly , a correlation between decreased PGC1α and SFTPA on a cell-by-cell basis within tumors . Combined with functional data in GEM models in which PGC1α expression was genetically silenced , these data argue that mutational-activation of PI3’-lipid signaling in BRAFV600E-driven LUAD leads to diminished PGC1α expression , and that this reduced expression compromises the ability of NKX2-1/FOXA1 to maintain AT2 pneumocyte identity ( Figure 7—figure supplement 2A ) . It is important to note that while genetic silencing of PGC1α recapitulated some aspects of PI3K activation in BRAFV600E-driven driven lung tumors , there were interesting differences , including an increase in markers of club and ciliated cell identity , and no significant effect on AT1 marker expression . It is therefore highly likely that amongst the pleiotropic effects of PI3’-lipid signaling , inhibition of PGC1α-mediated signaling represents one of many important effector pathways . This novel mechanism of lung identity regulation is made more important by the observation that it can be induced by mutational-activation of PI3Kα in both KRASG12D- and BRAFV600E-driven driven lung adenocarcinomas . Since the best characterized role of PGC1α is in the regulation of mitochondrial biogenesis , we were initially surprised to discover the cooperativity that PGC1α shows in the regulation of AT2 cell identity . However , it has recently emerged that crippling mitochondrial function via loss of the critical pyruvate transporter , MPC1 , potently drives cells into a de-differentiated stem cell fate in drosophila and mouse intestinal cells ( Schell et al . , 2017 ) . These studies , and the data presented here , suggest exciting future studies to examine the role of PGC1α to act as a pleiotropic effector of tumor cell growth and differentiation state downstream of PI3’-lipid signaling .
Further information and requests for reagents should be directed to and will be fulfilled by the Lead Contact , Martin McMahon ( martin . mcmahon@hci . utah . edu ) . Individual animals , tumors , and different cell lines comprise independent biological replicates . Repeated testing on the same cell line is considered technical replication . All mouse work was done with the approval of either the University of California IACUC under approval #AN089594 or the University of Utah IACUC under approval #15-11014 . Mice were housed in microisolator cages on ventilated racks in AAALAC accredited vivaria . Mice were housed in groups , as possible , and were provided bedding enrichment . Animals were provided standard laboratory rodent chow or Capecchi’s breeder diet when appropriate . Cages were supplied with water via a lixit system built into the housing rack or with a water bottle placed in the microisolator cage . Institute husbandry staff performed twice daily health checks . For breeding purposes , FVB/N and C57BL/6J mice were obtained from the Jackson Laboratory . In all animal experiments , animals were age matched to within 4 weeks of one another , all being between 3 to 4 months old . A power analysis was conducted in R to determine the number of animals to include based on RNA-Seq data of Sftpa from a previous experiment ( power . t . test ( delta = 2895 . 4 , sd = 918 , sig . level = 0 . 05 , power = 0 . 8 ) ) indicating 3 . 4 mice per group would be sufficient to see a 2-fold difference . This was rounded up and four mice were initiated for each time point and genotype . Equal numbers of male and female mice were selected for each time point and genotype . Mice were then randomized within sex and genotype and assigned to groups for harvest at different time points such that at each time point , 2 females and two males were euthanized for analysis . All mice used in these experiments were completely drug naïve and had never undergone other experimental procedures . Mice were on a mixed background of C57BL/6 , 129 , and FVB . Tumor initiation was performed by experimenters blinded to genotype . MLE-12 immortalized lung cells were newly obtained from ATCC and therefore assumed to be of the correct identity and to be free of mycoplasma contamination . MLE-12 cells were cultured in HITES medium supplemented with 2% FBS ( HyClone ) and penicillin/streptomycin . 293 T cells were from a frozen lab stock , whose identity was confirmed by STR profiling at the Huntsman Cancer Institute DNA sequencing core , and whose mycoplasma contamination status was confirmed to be negative by PCR . 293 T cells were maintained in DMEM ( Life Technologies ) supplemented with 10% FBS ( HyClone ) and supplemented with 5 mM glutamine and penicillin/streptomycin . All PCR steps were performed with CloneAmp ( Clontech ) high-fidelity polymerase premix and all newly constructed vectors were verified by Sanger sequencing . The targeting vector used to produce the BRafCAT mouse strain was built by the cloning of three fragments into the dual selection targeting vector pDTA-TK . Two fragments , comprising 4 . 8 kb and 3 . 8 kb targeting homology arms were amplified from a C57BL/6J BAC containing the mouse BRaf locus . A third fragment comprising the entirety of the genetically engineered module was assembled by gene synthesis ( Genewiz ) . The three fragments were cloned into the AGEI site of pDTA-TK using In-Fusion ( Clontech ) . New luciferase reporter constructs for the promoters of mouse surfactant proteins A , B , and C were created by amplification of mouse genomic DNA from a tail biopsy . Primer blast ( NCBI ) was used to create primers which captured 4500–5500 base pairs of promoter sequence beginning with the bases found immediately before the first annotated transcribed exon ( ensembl . org ) . The TCF/LEF sites found in the M50 Super Top Flash construct were removed by digestion with KPNI and XHOI and In-Fusion was used to clone the SFTPA , B , or C promoters in their place . The FUW-Kate plasmid was constructed by gene synthesis ( IDT ) of sequence encoding the mKate2 red fluorophore with ends compatible for In-Fusion cloning into the EcoRI and BamHI sites of FUW . The BRafCAT targeting construct was linearized by restriction enzyme digestion with the rare cutting enzyme I-CeuI and electroporated into 2H1 BrafCA/+ ES cells , which were selected for construct integration using puromycin . Three 96-well plates of resistant clones were screened via PCR using one primer specific to the targeting construct and one primer found in the mouse genome , outside of the construct homology arms , such that a 4 . 8 kb product would be the result of homologous targeting construct integration , whereas integration by NHEJ would yield no product . Cell permeant TAT-CRE was added directly to culture media at a final concentration of 1 uM to test for functionality and CRE dependence of the fluorophore . Purified TAT-CRE was obtained from Excellgen ( Rockville , MD ) . Mice were euthanized for analysis of lung tumor cell fluorescence at 2 , 6 , or 12 weeks p . i . Mice destined for lung harvest at either 2 or 6 weeks were initiated with 107 pfu of Ad5-Sftpc-CRE ( Fasbender et al . , 1998; Sutherland et al . , 2014 ) . Mice destined for lung harvest at 12 weeks were initiated with a lower titer ( 106 pfu ) of Ad5-Sftpc-CRE to avoid encountering premature endpoints due to tumor burden . At euthanasia , mice were perfused by first cutting the vena cava caudalis underneath the liver and then injection of DEPC treated PBS into the right ventricle of the heart until lungs turned white ( Arlt et al . , 2012 ) . The cranial , medial and caudal lobes of the right lung were first taken and placed into ice cold PBS and placed on ice . A new syringe and needle containing 10% neutral buffered formalin ( NBF ) was inserted into the larynx of mice and 10 ml was slowly infused to initiate fixation of the remaining lung lobes . The two remaining partially fixed lung lobes were placed into 25 ml of NBF and incubated for 24 hr before being processed into paraffin and sectioned . Tumor bearing lungs were minced using fine scissors in a 0 . 5 mg/ml solution of Liberase TM ( Roche ) . Minced tissue was incubated in a 37°C degree water bath for 15 min before being dissociated by pipetting up and down with a 1 ml pipette tip . Red blood cell lysis was performed by the addition of BD PharmLyse to 1x concentration and samples were incubated for an additional 10 min . Samples were passed through a 100 um filter fitted on a 50 ml conical tube . Filters were rinsed with 9 ml of ice cold Hanks Balanced Salt Solution ( HBSS ) and flow through was pelleted by centrifugation . From this point on , cells were kept on ice until lysis . Pellets were resuspended in 10 ml ice cold HBSS and passed through a 70 um filter affixed to the same 50 ml conical tube . Finally , dissociated cells were resuspended in 1 ml HBSS and passed through 35 um cell strainer cap into 5 mL round bottom polystyrene tubes . FACS was performed on a Becton-Dickinson ARIA III fitted with 100 uM nozzle , using gates as shown in the figures . tdTomato positive cells were sorted into RLT lysis buffer , homogenized and stored at −80°C until RNA purification . Library Construction and Sequencing cDNA libraries for RNA-Sequencing experiments were produced by the High Throughput Genomics Core at the Huntsman Cancer Center . RNA was purified using the Qiagen RNeasy micro system . RNA integrity was assayed using the Agilent TapeStation 2200 and High Sensitivity RNA ScreenTapes . RNA data was included in final analyses if RINe values were above 6 . No other exclusion of samples was done . For all samples , libraries were prepared using the Nugen Ovation Ultralow Library System V2 . Libraries were sequenced using an Illumina HiSeq 2000 device with 50 single end cycles and v4 chemistry . For RNA sequencing analysis , RNA transcript abundance was estimated using Salmon with default parameters on the main instance of the Galaxy webserver ( https://usegalaxy . org ) . Differentially expressed genes were then determined by use of DESeq2 using default parameters on the main instance of the Galaxy webserver ( https://usegalaxy . org ) , resulting in the datasets Figure 2—source data 1 , Figure 7—source data 1 . Gene set enrichment analysis was next performed using the R-package ‘fgsea’ with default parameters using the scripts provided as Figure 2—source code 1 , Figure 3—source code 1 , Figure 7—source code 1 . Gene correlation network analysis was performed in the R-package ‘WGCNA’ with the following parameters: To decrease noise , genes were filtered for minimal expression ( R-norm >40 ) , leaving 14207 genes to be clustered . These genes were clustered in a single block with a soft-thresholding power of ‘3’ as recommended in the WGCNA documentation based on the scale free topology fit index of our data . WGCNA R script provided as Figure 6—source code 1 , data provided as Figure 6—source data 1 . For motif discovery , the promoter regions from the top 100 most specific AT2 marker genes were defined as 5 kb upstream of the transcriptional start site , and downloaded from biomart ( https://www . ensembl . org/biomart ) . This list of sequences was filtered by repeat masker to remove low complexity DNA ( repeatmasker . org ) . The filtered list was analyzed for enriched novel motifs using MEME ( http://meme-suite . org ) with default settings less the following parameters: Find 25 motifs of width between 6 and 25 nucleotides . Novel motifs were matched to known transcription factor motifs from human and mouse ( HOCOMOCO v11 full ) using TomTom ( http://meme-suite . org ) with default settings . To determine if PI3’-lipid signaling strength affects Ppargc1a transcript levels in human tumors , lung adenocarcinoma data were downloaded from the NCI Genomic Data Commons Data Portal and segregated into those cases predicted to have strong or weak activation of PI3’-lipid signaling ( Figure 6—figure supplement 1—source data 1 ) . The groups were defined as such: ‘Strong activation’ due to mutation in at least one of the following genes: Pik3ca , Pten , Pik3r1 , and Akt1 , ‘Unknown PI3K activation status’ due to none of these mutations being detectable . FPKM values for Ppargc1a were then compared between these two groups . This resulted in n = 211 control samples and n = 19 mutant samples with predicted strong activation of PI3K signaling . All R scripts written for this study are available at GitHub ( van Veen , 2019; copy archived at https://github . com/elifesciences-publications/vanveen-elife ) . Harvested tissues were processed and embedded in paraffin , and sectioned at 4 uM . After deparaffinization in Citra-Clear ( Stat Lab , McKinney , TX ) , sections were re-hydrated in an ethanol series and antigens were unmasked using heated incubation in Tris-EDTA-SDS ( Syrbu and Cohen , 2011 ) . Sections were blocked for non-specific interaction in 10% Normal Donkey Serum in PBS and antibody staining was performed using the following primary antibody concentrations: PGC1α ( AB3242 ) 1:50 . NKX2-1 ( AB76013 ) 1:250 . Phospho-S327-NKX2-1 ( 13608 ) 1:250 . FOXA1 ( AB23738 ) 1:250 . FOXA2 ( D56D6 ) 1:250 . SFTPA ( SC-7699 ) 1:250 . Lysozyme ( AB108508 ) 1:250 . SFTPC ( SC-7705 ) 1:250 . CCA ( SC-9772 ) 1:1000 . AQP5 ( SC-9890 ) 1:50 . Primary antibody incubation was performed overnight at four degrees . After washing , alexa-488 and alexa-594 conjugated donkey anti mouse and donkey anti rabbit secondary antibodies were diluted 1:250 , and incubated on sections for 2 hr at room temperature . Stained sections were counterstained in DAPI and mounted in fluoromount G . For overview see Figure 3—figure supplement 1 . Fluorescent imaging in Figure 6d was performed on a Zeiss Apotome . Fluorescent imaging in Figure 5 was performed on a Leica DM1000 . All other fluorescent imaging was performed on a Nikon Ti-E inverted microscope employing a high sensitivity Andor Clara CCD camera . All images being compared in figures and in quantification were captured at exactly the same parameters for light and exposure . Acquisition settings were set such that pixel intensities were below saturation within regions of interest . When image intensity scales were adjusted for clarity , the intensity scales of images compared were set at exactly the same input and output levels . Intensity scales were never modified before quantitation . Images of tumor bearing lungs were imported into NIH ImageJ , where individual tumors were traced , and matched TIFF files were exported for each available fluorescent channel . A custom pipeline was built in CellProfiler to identify tumor cells based first upon identifying tumor nuclei using NKX2-1 immunoreactivity , when available , or DAPI staining when NKX2-1 immunostaining had not been performed . Tumor cells were defined based on propagation from identified nuclei , and tumor cytoplasm was defined as tumor cell minus tumor nucleus . For tumors analyzed in Kras based models , tumors were more diffuse and intermingled with surrounding parenchyma and so identifying tumor cells based on propagation from nuclei led to poor performance , and so tumor cells were defined as a three pixel ring around the tumor nucleus . Measurements were taken from pertinent TIFF files within individual nuclear , cellular , and cytoplasmic objects . Data were exported into comma separated value files and imported into R Studio using a custom script . For the purposes of graphing , individual tumor cell points were graphed using a custom script employing the R function ggplot ( ) . For all measurements , median fluorescence within each cellular object was the primary data output . For the purposes of quantification , tumor cells were not treated as independent , but whole tumor averages were considered ( mean of median fluorescence values ) , and each tumor was treated as independent . Fluorescence was not assumed to fit a normal distribution , and as such two factor comparisons were done using Wilcoxon Rank Sum test to generate p values using the R function wilcoxon . test ( ) . For comparisons of more than two conditions , one way ANOVA was performed using aov ( ) followed by Tukey’s honest significant difference using TukeyHSD ( ) . Chi-Squared tests were performed in R with the function chisq . test ( ) . When quadrants were drawn defining ‘negative’ and ‘positive’ staining: BRAFV600E driven tumors were noted to have relatively uniform positive expression of markers studied , and so ‘negative’ was defined as any tumor cell with median fluorescence less than one standard deviation below the mean of median fluorescence in BRAFV600E driven tumor cells . For all imaging and quantification , images were captured from 2 to 4 separate animals bearing tumors per group . Specific scripts and data for figures as follows: Figure 3: CellProfiler pipeline ( Figure 3—source code 3 ) used with raw images to produce Figure 3—source data 1 , 2 and 3 . Figure 3—source data 1 , 2 and 3 then used with Figure 3—source code 2 to perform statistics and produce graphs . Figure 4: CellProfiler pipeline ( Figure 4—source code 2 and Figure 4—source code 3 ) used with raw images to produce Figure 4—source data 1 and Figure 4—source data 2 . Figure 4—source data 1 and 2 then used with Figure 4—source code 1 to perform statistics and produce graphs . Figure 5: Cellprofiler pipeline ( Figure 5—source code 2 ) used with raw images to produce Figure 5—source data 1 . Figure 5—source data 1 then used with Figure 5—source code 1 to perform statistics and produce graphs . Figure 7: CellProfiler pipeline ( Figure 3—source code 3 ) used with raw images to produce Figure 7—source data 1 . Figure 7—source data 1 then used with Figure 7—source code 2 to perform statistics and produce graphs . Using Fugene 6 , HEK293 cells were co-transfected with luciferase reporter constructs , candidate transcriptional regulators , a fluorescent reporter construct to measure transfection efficiency ( FUW-Kate ) , and empty vector ( pUC19 ) to standardize total amount of DNA transfected across conditions to 100 ng/well of 96 well plate . Transfection efficiency was measured by fluorescence on an Incucyte ZOOM automated microscopy system ( https://www . essenbioscience . com/ ) . Luciferase production was measured with the Pierce Firefly Luc One-Step Glow Assay Kit , normalized to transfection efficiency , and represented as fold change over cells transfected with no candidate transcriptional regulators . All assays were performed in triplicate . Data provided as Figure 7—source data 3 . For immunoblot analysis of protein expression in MLE-12 cells , cells were lysed on ice in RIPA buffer supplemented with Thermo Halt protease inhibitor complex . Lysates were separated by acrylamide electrophoresis on pre-cast Novex 4–12% gradient Bis-Tris gels and transferred onto PVDF membranes using an Invitrogen iblot two transfer device . After blocking of non-specific interactions using Li-Cor blocking reagent , membranes were incubated in PBS containing antibodies at the following concentrations: SFTPC: ( sc-13979 ) 1:1000; SFTPA: ( sc-13977 ) , 1:1000; NKX2-1: ( ab76013 ) , 1:1000; FOXA1: ( 58613 ) , 1:2500; PGC1α: ( ab3242 ) , 1:500; NR5A2 ( ab153944 ) , 1:1000 . Protocol was repeated twice for a total of n = 3 with equivalent results . For co-immunoprecipitation experiments , 10 ug of the following antibodies were bound to Dynabeads magnetic Protein G beads ( Ms anti-PGC1α ( 1F3 . 9 ) , Rb anti-NKX2-1 ( AB76013 ) , Normal Rabbit IgG ( CST#2729 ) , Mouse IgG1 ( CST#5415 ) ) for 15 min at room temperature . All following steps were performed at four degrees centigrade . MLE-12 cells were lysed by 15 passages through a 25 gauge needle in the following buffer: 20 mM Tris HCl pH 8 , 137 mM NaCl , 0 . 1% ( v/v ) Nonidet-P40 . Lysates were centrifuged for 5 min at 1000xg to remove insoluble fraction . Cleared lysates were divided and incubated with either target antibody or IgG bound magnetic beads for 30 min before proceeding with washing and elution steps following product protocol . Eluates were separated by acrylamide electrophoresis on pre-cast Novex 4–12% gradient Bis-Tris gels and transferred onto PVDF membranes using an Invitrogen iblot two transfer device . After blocking of non-specific interactions using Li-Cor blocking reagent , membranes were incubated in PBS containing antibodies at the following concentrations: NKX2-1: ( ab76013 ) , 1:1000; PGC1α: ( ab3242 ) , 1:500 . Protocol was repeated twice for a total of n = 3 with equivalent results . | Cancers appear when changes in the genetic information of a cell , also called mutations , allow it to multiply uncontrollably . The disease we know as “lung cancer” kills more people than any other cancer , but this term actually refers to different types of tumors that appear because of various mutations that happen in different kinds of lung cells . To complicate matters further , as lung cancer cells become more aggressive , they can stop appearing and behaving like the type of lung cell they came from . Yet , knowing the exact origin of the cancer is key , since it determines which treatment will work best to stop the disease in its tracks . Despite these differences , many lung cancer cells contain mutations that over-activate two molecular cascades called the MAP kinase and the PI3’-kinase pathways . Under normal conditions , these signaling pathways relay external messages to the inside of the cell , where they help cells multiply . Two separate mutations can respectively over-stimulate either the MAP kinase or the PI3’-kinase pathway , but it was unclear how these could work together to start and maintain aggressive lung tumors . Another unanswered question was how these cancer cells lose the characteristics of the healthy cells they came from . To address these issues , van Veen et al . genetically engineered mice that carry a mutation which activates the MAP kinase pathway . The lung cells with this genetic change also made a red fluorescent protein that marked cancer cells , so that these could be separated from the rest of the lung and analyzed . This revealed that cells with only the MAP kinase mutation turned into small and benign tumors that began in lung cells , known as “type 2” cells . The PI3’-kinase mutation alone could not even start a tumor . However , together the mutations made tumors much more aggressive . Cells that carried both mutations also stopped producing proteins normally made by type 2 cells , therefore causing the cells to lose their original identity . The mice created by van Veen et al . could help to understand how lung cancers develop in these animals and also in human lung cancer patients . Ultimately , this information could be used to design new cancer treatments , especially since both the MAP kinase and PI3’-kinase pathways contain many proteins that can be targeted with drugs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2019 | Mutationally-activated PI3’-kinase-α promotes de-differentiation of lung tumors initiated by the BRAFV600E oncoprotein kinase |
Recent developments in human neuroimaging make it possible to non-invasively measure neural activity from different cortical layers . This can potentially reveal not only which brain areas are engaged by a task , but also how . Specifically , bottom-up and top-down responses are associated with distinct laminar profiles . Here , we measured lamina-resolved fMRI responses during a visual task designed to induce concurrent bottom-up and top-down modulations via orthogonal manipulations of stimulus contrast and feature-based attention . BOLD responses were modulated by both stimulus contrast ( bottom-up ) and by engaging feature-based attention ( top-down ) . Crucially , these effects operated at different cortical depths: Bottom-up modulations were strongest in the middle cortical layer and weaker in deep and superficial layers , while top-down modulations were strongest in the superficial layers . As such , we demonstrate that laminar activity profiles can discriminate between concurrent top-down and bottom-up processing , and are diagnostic of how a brain region is activated .
Using ‘ultra-high field’ MRI systems of 7T and above , it has become possible to non-invasively measure fMRI responses at lamina-resolved spatial resolutions in humans ( Dumoulin et al . , 2018; Koopmans et al . , 2011; Polimeni et al . , 2010 ) . This has allowed researchers to ask new questions about the functional organization of the human brain , and examine communication between brain areas in more detail than previously possible ( Kuehn and Sereno , 2018 ) . One important promise of laminar fMRI is its potential ability to distinguish between bottom-up and top-down BOLD responses . While these are spatially amalgamated at standard imaging resolutions ( Lawrence et al . , 2017; Self et al . , 2017 ) , they are expected to be expressed at different cortical depths . Bottom-up connections between brain areas are known to target the granular layer 4 , at middle cortical depths , while top-down connections target deeper and superficial layers but largely avoid layer 4 ( Anderson and Martin , 2009; Felleman and Van Essen , 1991; Rockland and Pandya , 1979 ) . It should therefore be possible to tease apart the bottom-up and top-down contributions to a stimulus-driven BOLD response by examining that response across cortical depth . Previous laminar fMRI studies suggest that this is indeed the case . For example , stimulus-driven responses in visual cortex have been shown to be strongest at middle depths ( Koopmans et al . , 2010 ) , while top-down signals embodying contextual inference , prediction , attention and working memory operate at deep and/or superficial , but not middle , cortical depths ( Klein et al . , 2018; Kok et al . , 2016; Lawrence et al . , 2018; Muckli et al . , 2015; Scheeringa et al . , 2016 ) . Similar results for top-down influences have also been reported for auditory ( De Martino et al . , 2015 ) and motor cortex ( Huber et al . , 2017 ) . Whilst these results are encouraging , these studies have typically measured top-down signals in the absence of a bottom-up response . The rationale for this choice is clear , as bottom-up drive could affect all cortical layers due to quick communication between layers ( Self et al . , 2013 ) and blurring from spatial hemodynamics ( Uğurbil et al . , 2003; Uludağ and Blinder , 2018; Yacoub et al . , 2005 ) could obscure layer-specific top-down effects . However , being constrained to measuring top-down responses in isolation limits the potential power of laminar fMRI experiments for exploring brain function . If the overall BOLD response to a stimulus could be separated into its bottom-up , stimulus-driven component and a top-down , modulatory component , this would open the door for increasingly complex task design in laminar fMRI experiments . Here we measured lamina-resolved fMRI responses from human participants as they viewed visual stimuli and were required to attend to a specific stimulus feature ( orientation ) . Our stimulus paradigm was designed to elicit concurrent bottom-up and top-down modulations of the stimulus-driven response through orthogonal manipulations of stimulus contrast ( bottom-up ) and feature-based attention ( top-down ) , both of which are known to influence early visual cortex responses ( Boynton et al . , 1999; Himmelberg and Wade , 2019; Kamitani and Tong , 2005; Martinez-Trujillo and Treue , 2004; Saenz et al . , 2002; Treue and Martínez Trujillo , 1999 ) . We predicted that response modulations driven by attention would operate at different cortical depths to those driven by changes in stimulus contrast . Specifically , contrast modulations were expected to be largest at middle cortical depths , as increases in contrast should be associated with stronger bottom-up input to the granular layer ( Hubel and Wiesel , 1972; Rockland and Pandya , 1979 ) . Top-down influences are generally expected to be strongest in the deep and/or superficial cortical depths ( agranular layers , Lawrence et al . , 2017 ) . However , previous research into laminar effects of attention have been varied . Studies by van Kerkoerle et al . ( 2017 ) and Klein et al . ( 2018 ) report largely agranular effects of attention in V1 while others report effects in all layers of V1 ( Denfield et al . , 2018; Hembrook-Short et al . , 2017 ) . Further studies have also reported attention effects in all layers of V4 ( Nandy et al . , 2017 ) and the superficial layers of primary auditory cortex ( De Martino et al . , 2015 ) . Moreover , previous laminar attention studies have employed spatial or object-based attention , but the laminar circuits involved in feature-based attention have , to our knowledge , not yet been studied . It is therefore not clear whether we should expect feature-based attention to modulate responses in all layers or only agranular layers . Critically , both eventualities yield the prediction that modulations from feature-based attention should be more strongly expressed in agranular layers compared to those from stimulus contrast , which should only be strong in the granular layer . To preview , we found that fMRI responses in the early visual regions ( V1-V3 ) were strongly modulated by changes in stimulus contrast and feature-based attention , and that these effects were indeed expressed at different cortical depths . As predicted , attentional modulations were more strongly expressed in agranular layers , particularly the superficial layers , while stimulus contrast modulations were largest in the granular layer .
Subjects were able to focus their attention on one set of oriented bars within a plaid and accurately discriminate changes in bar width between stimuli . On average , subjects performed at 83 . 5% correct ( SD = 2 . 3 ) for low and 84 . 5% correct ( SD = 1 . 5 ) for high contrast stimuli . Task difficulty was controlled by separate staircases for high and low contrast stimuli to match task difficulty across contrast levels . Despite this , the numerically small difference in task performance was significant ( t [22]=2 . 52 , p=0 . 019 ) . To assess the effects of attention and stimulus contrast on brain responses , we divided visually active voxels within V1 , V2 and V3 into subpopulations with a strong preference for clockwise orientations over counter-clockwise or vice versa ( Albers et al . , 2018; see Materials and methods ) . It was expected that voxels would respond more strongly during blocks in which their preferred orientation was attended , and that all voxels would respond more strongly to higher contrast stimuli . As expected , BOLD responses in early visual cortex were modulated by both subjects’ attention towards a specific orientation and changes in stimulus contrast ( see Figure 2 ) . Responses to high contrast stimuli were significantly higher than low contrast stimuli across V1-V3 ( F [23 , 1]=35 . 57 , p=4 . 00e−6 ) . The size of this effect varied across areas ( F [30 . 2 , 1 . 3]=46 . 53 , p=1 . 77e−8 ) , being larger in V1 than V2 and V3 . Voxel responses were also higher when their preferred orientation was attended , compared to when the orthogonal orientation was attended ( F [23 , 1]=25 . 67 , p=4 . 00e−5 ) . This effect also varied across visual areas ( F [46 , 2]=4 . 91 , p=0 . 012 ) , being slightly smaller in V1 compared to V2 and V3 . Overall , therefore , our paradigm was successful in inducing strong modulations of stimulus-driven BOLD responses using bottom-up ( contrast ) and top-down ( feature-based attention ) task manipulations . Next , we determined whether the effects of feature-based attention and stimulus contrast on BOLD responses varied across cortical depth , and whether they did so differently from each other . To this end we computed separate BOLD time courses specific to three equal volume gray matter depth bins defining deep , middle and superficial cortex ( Lawrence et al . , 2018; van Mourik et al . , 2018a , see Materials and methods for more information ) . Depth-specific time courses were normalized to remove overall differences in signal intensity between layers ( Figure 3—figure supplements 3 and 4 ) . Note that this normalization was not critical to the results reported ( Figure 3—figure supplement 5 ) . Normalized depth-specific time courses were analyzed to compare the laminar profile of activity modulations resulting from top-down attention and bottom-up stimulus contrast . To get an overall picture of depth-specific modulations across the visual cortex , we first combined voxels from V1 , V2 and V3 for this analysis . Response modulations from both feature-based attention and stimulus contrast were clearly present in depth-specific time courses ( Figure 3A–D ) . In order to fairly compare laminar profiles across conditions , we used data from the same time points ( highlighted in Figure 3A & C ) , which comprised the peak of the BOLD response during a block of stimuli and during which both the effects of attention ( F [23 , 1]=19 . 95 , p=1 . 76e−4 ) and contrast ( F [23 , 1]=35 . 98 , p=4 . 00e−6 ) were significant . Within this time window , the effect of feature-based attention on neural responses was present at all cortical depths and was largest in the superficial layers ( Figure 3b ) . There was a trend of activity differences between layers induced by the attentional manipulation ( F [46 , 2]=2 . 82 , p=0 . 070 ) . Unpacking this , the attentional modulation was significantly stronger in the superficial layers compared to the middle ( t [23]=2 . 11 , p=0 . 046 ) and deep layers ( t [23]=2 . 15 , p=0 . 042 ) , while there was no significant difference in the strength of the attentional modulation between the deep and middle layers ( t [23]=0 . 36 , p=0 . 723 ) . Modulations from changes in stimulus contrast were organized quite differently , peaking at middle depths ( Figure 3d ) . Indeed , contrast modulations varied significantly across depth ( F [46 , 2]=8 . 43 , p=0 . 001 ) , being largest at middle compared to deep ( t [23]=3 . 79 , p=0 . 001 ) and superficial ( t [23]=3 . 56 , p=0 . 002 ) depths . Critically , the organization of contrast-related modulations across depth was significantly different to those caused by feature-based attention , as shown by a source ( bottom-up , top-down ) X layer ( deep , middle , superficial ) interaction ( F [46 , 2]=4 . 39 , p=0 . 018 ) . As such , the laminar profiles of responses modulations across the early visual cortex were dependent on whether those modulations were bottom-up or top-down in origin . We predicted that top-down effects were more likely to be expressed in agranular layers compared to bottom-up effects . To explicitly test for this , we computed a score that described whether experimental effects were more agranular or granular . This was done by averaging the effect of feature-based attention ( or contrast ) from the superficial and deep depth bins ( agranular ) and subtracting that from the middle bin ( granular ) . As such , a positive score indicates a largely agranular effect , while a negative score indicates a granular effect . As predicted , feature-based attention effects were more agranular compared to stimulus contrast ( Figure 3E ) . This difference was significant ( t [23]=3 . 11 , p=0 . 005 ) , and 20 of our 24 subjects showed an effect in this direction ( Figure 3F ) . Therefore , it appears that top-down contributions to response modulations were stronger in the agranular layers compared to bottom-up contributions , which were strongest in the granular layer . As can be seen from Figure 3B , the agranular profile of attention was driven by the fact that the attentional modulation was strongest in the superficial layers . We next explored how modulations from feature-based attention and stimulus contrast varied across cortical depth within visual areas , and potential differences in organization between areas . We estimated depth-specific effects of attention and contrast for V1 , V2 and V3 using the same methods applied to the three areas combined ( Materials and methods ) . Similar to our original analysis , variation in the effect of attention across depth over V1-V3 ( Figure 4A ) did not reach significance ( F [46 , 2]=2 . 54 , p=0 . 090 ) , and attention depth profiles were similar across areas ( F [69 . 67 , 3 . 03]=0 . 44 p=0 . 778 ) . The effect of contrast did vary across cortical depth ( F [36 . 28 , 1 . 58]=7 . 52 , p=0 . 004 ) peaking at middle depths ( Figure 4B ) , but this profile was not significantly different between the three areas ( F [92 , 4]=1 . 39 , p=0 . 244 . When directly contrasting these two modulatory factors , there was an overall , area independent , difference between feature-based attention and stimulus contrast that approached significance ( F [46 , 2]=2 . 74 , p=0 . 075 ) , but no significant differences between areas ( F [80 . 38 , 3 . 50]=1 . 00 , p=0 . 407 ) . We also computed scores describing how agranular or granular effects of attention and contrast were within V1 , V2 and V3 ( Figure 4C ) . In general , modulations from feature-based attention were more agranular compared to those from stimulus contrast ( F [23 , 1]=5 . 48 , p=0 . 028 ) , and this was consistent across visual areas ( F [46 , 2]=0 . 51 , p=0 . 607 ) . Overall , these results show highly similar behavior of the three early visual regions ( V1 , V2 , V3 ) that we examined , in terms of both their bottom-up and top-down laminar activation profiles .
We measured laminar fMRI responses from the human visual cortex during a visual task designed to induce bottom-up and top-down response modulations via orthogonal manipulations of stimulus contrast and feature-based attention . BOLD responses were strongly modulated by both feature-based attention and stimulus contrast , and these effects were expressed at different cortical depths . Effects of stimulus contrast were considerably larger at middle cortical depths compared to deep and superficial depths , while effects of feature-based attention were more even across depth , peaking in superficial cortex . Moreover , by comparing the strength of attention and contrast modulation in agranular versus the granular layers , we found that attention effects were expressed more strongly in the agranular layers ( specifically the superficial layers ) compared to effects from stimulus contrast , which were more granular . Our results show clear differences in how bottom-up and top-down aspects of perceptual processing affect brain responses across cortical depth and are consistent with the anatomical organization of feedforward and feedback connections between brain areas ( Rockland and Pandya , 1979 ) . To our knowledge , our study also provides the first report of how visual cortex responses are modulated by feature-based attention at the laminar level . Most importantly , we demonstrate that laminar fMRI methods can be used to examine both the bottom-up and top-down components of the overall BOLD response as they co-occur during the processing of a stimulus . Previous laminar fMRI studies have either measured depth-specific effects in the absence of a physical stimulus ( Kok et al . , 2016; Lawrence et al . , 2018; Muckli et al . , 2015 ) , or in the presence of a stimulus that was held constant ( De Martino et al . , 2015; Klein et al . , 2018 ) . By orthogonally manipulating stimulus contrast and feature-based attention , we have shown that top-down effects can be separated from concurrent bottom-up modulations driven by the stimulus . This opens the door for future studies to further examine the dynamic interactions between bottom-up and top-down processing that occur in the context of stimulus processing . Top-down modulations of the BOLD response were expressed at all cortical depths relatively evenly , slightly peaking in the superficial layers . This partly contrasts with our previous study , which observed top-down activation of V1 during visual working memory that was strong in the agranular layers , but much weaker in the middle layer ( Lawrence et al . , 2018 ) . The most obvious difference between the two studies is the presence of a physical stimulus during top-down modulation in this study , while there was no stimulus during working memory in our previous study . Each stimulus is expected to trigger a large response in the middle layer of V1 driven by bottom-up connections from the LGN ( Hubel and Wiesel , 1972 ) . Interestingly , influences of feature-based attention have been reported in the LGN before ( Ling et al . , 2015; Schneider , 2011 ) , suggesting that this bottom-up signal could carry attentional modulations , consistent with our data . Electrophysiological studies of laminar effects of attention report mixed results regarding the involvement of the granular layer of V1 in attention . van Kerkoerle et al . ( 2017 ) report increased spike rate and current sinks with attention that were largest in the agranular layers . In particular , Van Kerkoerle et al . report strong attentional modulations in the deep layers compared to the middle layer , which was not the case in our data . Other studies ( Denfield et al . , 2018; Hembrook-Short et al . , 2017 ) report attentional effects on spike rate in all layers of V1 . However , it should be noted that these studies utilized spatial attention as opposed to feature-based attention , making it unclear how comparable their results are to our study . More research is required to further elucidate the laminar circuits involved in different modes of attentional control . The relatively similar strength of attentional modulations across cortical layers highlights an important aspect of our task design . Taken in isolation , the laminar profile of feature-based attention we report could be viewed as difficult to interpret , as there are no obvious differences between cortical depths . Crucially , however , the comparison of this profile to one derived from a manipulation of stimulus contrast revealed clear differences in how the visual cortex is modulated depending on the source of the modulation . We encourage future laminar studies exploring top-down responses to also include a bottom-up manipulation as a point of comparison , as the laminar organization of a BOLD activity difference between conditions on its own can be challenging to interpret ( Self et al . , 2017 ) . It is possible that we found top-down effects were similar across cortical depth due the blurring of BOLD responses across depth bins from spatial hemodynamics . Our task involved repeated presentation of a series of visual stimuli , which is expected to cause large swathes of stimulus-related activity in V1 that starts in layer four and quickly spreads to other layers ( Self et al . , 2013 ) . This activity is in turn expected to be spatially blurred in the BOLD response by venous draining towards the pial surface , which smooths responses across cortical depth , causing stronger responses at superficial depths ( Uğurbil et al . , 2003; Uludağ and Blinder , 2018; Yacoub et al . , 2005 ) . It is therefore possible that repeated visual stimulation could have effectively washed out depth-specific responses , increasing the likelihood of experimental effects being uniform across depth . That said , any influence of spatial hemodynamics should be consistent across experimental conditions , and therefore accounted for in our calculation of bottom-up/top-down modulations via a subtraction of the responses to different contrast/attention conditions . Indeed , the strikingly distinct laminar profile of stimulus contrast effects that clearly peaked in the middle layers indicates that our analysis could account for the influence of hemodynamics . Nevertheless , accurate depth-estimates of BOLD responses continues to be the biggest challenge in laminar fMRI . Recent developments in modeling spatial aspects of the BOLD response for applying a spatial deconvolution to BOLD data ( Markuerkiaga et al . , 2016 , ISMRM , abstract; Marquardt et al . , 2018 ) and improved measurement protocols ( Huber et al . , 2017 ) could help to alleviate this issue . We show that modulations of stimulus-driven responses were similar across areas within the early visual cortex . For stimulus contrast , this is consistent with a purely stimulus-driven effect that changes response amplitude at early , subcortical levels and is inherited through the visual system via bottom-up connections targeting layer 4 ( Hubel and Wiesel , 1972; Rockland and Pandya , 1979 ) . With regards to attention , there is little work addressing laminar differences between visual brain areas . Nandy et al . ( 2017 ) report attentional modulations in all layers of V4 , consistent with our findings in extrastriate areas V2 and V3 , as well as V1 , but they do not provide a comparison to other brain areas . Buffalo et al . ( 2011 ) report attentional modulation of gamma and alpha oscillations in deep and superficial cortex that were similar in V1 , V2 and V4 . Though they did not measure from granular layer neurons , and thus cannot comment on whether attentional modulations occurred in all layers or only agranular layers , the similarity of results across visual brain areas appears consistent with our study . However , we again note that how these results compare to our own is unclear as these studies used spatial attention , not feature-based attention , as well as a variety of electrophysiological measurements with an unclear relation to the BOLD signal . For future studies , laminar fMRI is well suited to exploring laminar differences between brain areas as it affords simultaneous measurements over larger areas of cortex compared to electrophysiological methods . In conclusion , we have shown that fMRI responses in visual cortex are strongly modulated by changes in stimulus contrast and feature-based attention , and that these effects operate at different cortical depths . Top-down modulations from attention were overall stronger in agranular layers ( specifically the superficial layers ) compared to those from stimulus contrast , which were strongest in the granular layer . We have shown that , in a task where bottom-up and top-down influences are manipulated independently , the overall BOLD response can be separated into top-down and bottom-up components by examining how these effects are organized across depth . Future studies can use similar strategies to further explore the dynamic interactions between bottom-up and top-down processing that occur in perception and cognition .
Twenty-six healthy participants ( all right-handed , nine males , mean age 25 . 5 , age range 19–47 ) with normal or corrected-to-normal vision completed the experiment . This sample size ( N = 26 ) provided us with 80% power to detect one-sided experimental effects that had at least medium effect size ( Cohen’s d > 0 . 6 ) . All gave written informed consent and the study was approved by the local ethics committees ( CMO region Arnhem-Nijmegen , The Netherlands , and ethics committee of the University Duisburg-Essen , Germany , protocol CMO 2014/288 ) . Participants were reimbursed for their time at the rate of €10 per hour . All participants completed a 1 hr 3T fMRI retinotopic mapping session , a 1 hr psychophysics session , and a 1 hr 7T fMRI session for the main task . The experiment and analysis plan were preregistered on the Open Science Framework ( https://osf . io/46adc/ ) . Retinotopic mapping data were acquired and analyzed using identical methods to those reported in our previous study ( Lawrence et al . , 2018 ) . In brief , brain responses to rotating wedge and expanding ring checkerboard stimuli were acquired using a Siemens 3T Trio MRI system ( Siemens , Erlangen , Germany ) with a 32-channel head coil and a T2*-weighted gradient-echo EPI sequence ( TR 1500 ms , TE 40 ms , 68 slices , 2 mm isotropic voxels , multi-band acceleration factor 4 ) . One high resolution anatomical image was also acquired with a T1-weighted MP-RAGE sequence ( TR 2300 ms , TE 3 . 03 ms , 1 mm isotropic voxels , GRAPPA acceleration factor 2 ) . Anatomical data were automatically segmented into white matter , gray matter and CSF using FreeSurfer ( http://surfer . nmr . mgh . harvard . edu/ ) . Functional data were analyzed using the phase encoded approach in MrVista ( http://white . stanford . edu/software/ ) . Polar angle and eccentricity data were visualized on an inflated cortical surface and the boundaries of V1 , V2 and V3 were drawn manually using established criteria ( Engel et al . , 1994; Sereno et al . , 1995; Wandell et al . , 2007 ) . During the psychophysics session subjects completed the same visual task ( Figure 1 ) that was used in the 7T main task fMRI session . Plaid stimuli were programmed in MATLAB ( MathWorks , Natick , MA ) and presented using PsychToolbox ( Brainard , 1997 ) on a 24 inch BenQ XL2420T monitor ( http://www . benq . eu/product/monitor/ , resolution 1920 × 1080 , refresh rate 120 Hz ) . Plaids comprised orthogonally oriented sets of bars ( one set black , one set white ) , overlaid on top of each other . Areas of overlap between bars were made mid-gray ( the same as the background ) , to facilitate mental separation of the two component stimuli . Subjects viewed the stimuli from a chin rest mounted 70 cm from the display and were instructed to fixate on a central fixation dot ( 0 . 5 degrees of visual angle across ) at all times . Plaids were presented centrally behind an annulus mask ( inner radius one degree , outer radius eight degrees , and had a spatial frequency of 1 cycle/degree and random phase . Stimulus edges were softened with a linear ramp that started 0 . 5 degrees from the edge of the mask . The task used a block design . Stimulus blocks were preceded by an attention cue that lasted 2 s , where attention was cued by the color of the fixation dot ( red = attend clockwise , green = attend counter-clockwise ) . The fixation dot remained red/green for the duration of the stimulus block . Stimulus blocks comprised a series of 8 plaid stimuli presented sequentially at a rate of 0 . 5 Hz ( 1 . 75 s on , 0 . 25 s off ) . Subjects’ task was to press one of two buttons indicating whether the bars in the attended orientation were thicker or thinner than they were in the previously presented stimulus . Subjects were instructed to attend , but not respond to , the first stimulus in each block ( as there was no preceding stimulus to compare to ) and to respond to all remaining stimuli within the block . Subjects were allowed to respond at any time during stimulus presentation or the inter-stimulus interval , but the trial was marked as incorrect if they did not respond before the next stimulus was presented . Bar width for clockwise and counter-clockwise bars varied independently from each other , meaning attention had to be focused on the cued orientation in order to succeed at the task . Changes in bar width between stimuli were controlled using a QUEST ( Watson and Pelli , 1983 ) staircase function targeting 80% correct performance , which was updated after each individual trial . Bars within the first plaid presented in each block had a bar width of 0 . 2 degrees ± a random increment between 0 and 0 . 02 degrees . For the remaining stimuli bar width was equal to the width of the previously presented stimulus ±an increment decided by the staircase . For both sets of bars , the direction of width increments was pseudo-randomized such that they were positive for four stimuli in each block and negative for four stimuli , presented in a random order . Stimulus luminance polarities were held constant within blocks but randomized between blocks , ensuring that both positive and negative luminance polarities were presented the same number of times for each experimental condition . After a stimulus block , the fixation dot turned black and was presented for 1 s . This was followed by performance feedback presented as a mark out of 7 for correct trials in the previous block , presented for 1 s . A 2 s attention cue then preceded the onset of the next stimulus block . Subjects completed 24 blocks of the task , at which point the discrimination threshold for 80% correct performance was recorded for use in the main fMRI task . This process was performed once using high contrast stimuli ( 80% Michelson contrast ) , and once using low contrast stimuli ( 30% Michelson contrast ) , meaning separate thresholds were estimated for two contrast levels , which were used to match task difficulty across contrast levels in the fMRI experiment . fMRI data for the main experiment were acquired using a Siemens Magnetom 7T MRI system ( Siemens , Erlangen , Germany ) with a commercial RF head coil ( Nova Medical , Inc , Wilmington , MA , USA ) with one transmit ( TX ) and 32 receive ( RX ) channels and a gradient coil ( Type AS095 , Siemens Healthcare , Erlangen , Germany ) with 38 mT/m gradient strength and 200 mT/m/ms slew rate . Functional data were acquired with a T2*-weighted 3D gradient-echo EPI sequence ( Poser et al . , 2010; TR 3408 ms , TE 28 ms , 0 . 8 mm isotropic voxels , 16° flip angle , 192 × 192×38 . 4 mm FOV , GRAPPA acceleration factor 4 ) . Shimming was performed using the standard Siemens shimming procedure for 7T . Anatomical data were acquired with an MP2RAGE sequence ( Marques et al . , 2010; TR 5000 ms , TE 2 . 04 ms , voxel size 0 . 8 mm isotropic , 240 × 240 mm FOV , GRAPPA acceleration factor 2 ) yielding two inversion contrasts ( TI 900 ms , 4° flip angle and TI 3200 ms , 6° flip angle ) , which were combined to produce a T1-weighted image . We also acquired a T2-weighted HASTE scan that was used to identify the calcarine sulcus to aid functional slice positioning ( TR 3230 ms , TE 67 ms , seven coronal slices , 0 . 625 × 0 . 625×5 . 10 mm voxels ) . Stimuli were programmed and displayed using the same methods described for the psychophysics session onto a rear-projection screen using an EIKI ( EIKI , Rancho Santa Margarita , CA ) LC-X71 projector ( 1024 × 768 resolution , refresh rate 60 Hz ) , viewed via a mirror ( view distance ~130 cm ) . Each subject completed 3 runs of the main task . The task was identical to the psychophysics session , except blocks of high and low contrast stimuli were randomly interleaved rather than presented in separate sessions , and timings were adjusted to sync with volume acquisition: The attention cue preceding a stimulus block was presented for 1 . 04 s , stimulus blocks lasted 16 s , followed by 1 s of fixation , 0 . 5 s of feedback , and a 15 . 54 s inter-block interval with a black fixation dot to allow the BOLD response to return to baseline before the next block . Changes in bar width for high and low contrast blocks were controlled by separate staircases , which were given starting estimates equal to contrast-specific discrimination thresholds measured in the psychophysics session plus a 20% increment . Due to a problem with recording button responses in one session , the behavioral results reported in the Results section were calculated using data from the remaining 23 subjects . five volumes were acquired per stimulus block , and five volumes between blocks , with 16 blocks in a single 555 . 5 s run . This run time also includes three dummy volumes that were discarded from the start of each run to allow for signal stabilization . After the main task , subjects completed an orientation localizer scan that was used to measure voxel-wise orientation preference . Single sets of oriented bars ( i . e . , one stimulus component from the plaid stimuli presented in isolation ) were presented in an AoBo block design . Stimulus blocks were 13 . 6 s long ( 4 TR ) and separated by rest blocks of the same length . During a stimulus block bars were repeatedly presented with the same orientation at a rate of 2 Hz ( 250 ms on , 250 ms off ) . Stimuli were presented at 100% contrast , luminance polarity was reversed with each stimulus presentation , and phase was randomized . Stimulus blocks alternated between blocks of clockwise bars ( 45° ) and blocks of counter-clockwise bars ( 135° ) . A total of 16 stimulus blocks were presented in a 446 . 4 s run; the first three volumes were again discarded . During the scan subjects maintained fixation and pressed a button every time the fixation dot flashed white for 0 . 25 s ( 1 to 4 flashes per block ) . We used the same data processing pipeline as our previous study ( Lawrence et al . , 2018 ) . Functional volumes were cropped so that only the occipital lobe remained , and spatially realigned within and then between runs using SPM8 ( http://www . fil . ion . ucl . ac . uk/spm ) . Finally , data were highpass filtered using FEAT ( fMRI Expert Analysis Tool ) v6 . 00 ( https://fsl . fmrib . ox . ac . uk/fsl ) with a cut off of 55 s to remove low frequency scanner drift . 7T anatomical data were segmented into white matter , gray matter and CSF using FreeSurfer’s ( http://surfer . nmr . mgh . harvard . edu/ ) automated procedure . The white and gray matter surfaces were then aligned to the mean functional volume using a standard rigid body registration ( Greve and Fischl , 2009 ) followed by a recursive non-linear distortion correction that has been described previously ( Lawrence et al . , 2018; van Mourik et al . , 2018a ) . We defined orientation-selective masks in V1 , V2 and V3 using methods we have described previously ( Lawrence et al . , 2018 ) . Note that the voxel selection procedure described here was applied to data from the orientation localizer scan; a data set that was independent from the main task . In brief , a GLM was applied to functional localizer data using FEAT v6 . 00 ( https://fsl . fmrib . ox . ac . uk/fsl ) to identify voxels that responded to significantly to all stimuli presented in the localizer ( z > 2 . 3 , p<0 . 05 ) . For two subjects , there were very few voxels within the visual cortex that survived this cluster correction , indicating they had failed to remain alert for the duration of the experiment , and so we did not make any further use of their data . Next we contrasted responses to clockwise and counter-clockwise bars , and created masks of 1000 voxels per area , containing the 500 voxels with the most positive t values in this contrast ( prefer clockwise ) and the 500 with the most negative t values ( prefer counter-clockwise ) . This was done separately for V1 , V2 and V3 . In any cases where there were fewer than 500 voxels within an area that met the required criteria for being visually active and having an orientation preference , we used as many voxels as did fulfill the criteria . To ensure that our results did not depend on how many and which voxels we chose to include on our masks , and that the selection we ran a battery of control analyses using an array of different mask sizes ( Figure 3—figure supplement 1–9 ) . Although effect sizes varied across mask sizes , all produced effects in the same direction as our main analysis . Overall effects of feature-based attention and changes in stimulus contrast were quantified using a temporal GLM applied using FEAT v6 . 00 ( https://fsl . fmrib . ox . ac . uk/fsl ) on the preprocessed functional data . Each of the four experimental conditions ( attend clockwise high contrast , attend clockwise low contrast , attend counter-clockwise high contrast , attend counter-clockwise low contrast ) were modeled as separate regressors of interest and contrasted against baseline to estimate % signal changes associated with each condition . This was applied to orientation-selective masks from V1-V3 combined , and also to each area separately . % signal changes are shown in Figure 2 . Signal changes associated with attended and unattended orientations were calculated by averaging responses from clockwise preferring voxels to attend clockwise blocks and counter-clockwise preferring voxels to attend counter-clockwise blocks . Likewise , unattended responses were calculated by averaging responses from clockwise preferring voxels to attend counter-clockwise blocks and vice versa . Laminar-specific time courses were estimated using the open fMRI analysis toolbox ( van Mourik et al . , 2018b ) as we have described previously ( Lawrence et al . , 2018 ) . In brief , segmented cortical meshes were divided into five depth bins: white matter , three equivolume gray matter bins , and CSF . The proportion of overlap between each voxel within our orientation-selective masks and these five bins were estimated , creating a matrix of depth weights describing the laminar organization of a population of voxels . These weights were regressed against the functional data from the same voxels to produce a single time course for each depth bin representative of the average response across the population at that cortical depth . This process was applied separately to the clockwise and counter-clockwise preferring voxel populations from V1-V3 combined to examine overall laminar activity across the visual cortex ( Figure 3 ) . We also did the same for V1 , V2 and V3 separately to examine differences in laminar organization between areas ( Figure 4 ) . It is well established that gradient-echo BOLD suffers from a bias in signal strength whereby responses in superficial cortex are stronger than responses from deep cortex ( Koopmans et al . , 2010; Uğurbil et al . , 2003; Uludağ and Blinder , 2018; Yacoub et al . , 2005 ) . This bias can be seen clearly in our raw data ( see Figure 2—figure supplement 1; Figure 3—figure supplement 3 ) . We attempted to alleviate this issue by converting time courses specific to deep , middle and superficial cortical layers to z scores , normalizing differences in overall signal strength between layers . The z scoring was performed on layer-specific time courses from a single run of the main task , meaning it was performed within layers , across all experimental conditions and within runs . This procedure removed overall amplitude and variance differences between layers , while preserving within-layer differences between conditions . This had the effect of making overall signal changes between depth bins very similar ( Figure 3—figure supplement 4 ) , while preserving potential differences between depth bins that are due to experimental manipulations ( rather than large differences in overall signal change ) . Of note , none of our results critically depend on this normalization step ( Figure 3—figure supplement 5 ) , but it allowed us to interpret those results in the absence of large-scale response differences between layers that are present in the raw data . We analyzed time courses specific to deep , middle and superficial gray matter depth bins in the following way to quantify depth-specific effects of feature-based attention and stimulus contrast . Z scored , depth-specific time courses were split into segments of 10 volumes each , corresponding to one stimulus block ( five volumes ) followed by an inter-block interval ( five volumes ) . To examine effects of feature-based attention , we computed an average attended time course by averaging responses for each block from the voxels that preferred the cued orientation in that block ( i . e . prefer clockwise for attend clockwise blocks and prefer counter-clockwise for attend counter-clockwise blocks ) , and an average unattended time course by averaging responses from voxels that preferred the ignored orientation for each block ( i . e . prefer clockwise for attend counter-clockwise blocks and prefer counter-clockwise for attend clockwise blocks ) . To examine effects of stimulus contrast , we averaged responses from both populations of voxels , regardless of orientation preference , averaging across all high contrast blocks and low contrast blocks to produce separate average time courses for high and low contrast stimuli . This analysis procedure was performed separately on time courses from the three gray matter depth bins within each subject , and then a group average was calculated . Figure 3A&C show group average time courses for each experimental condition , averaged across gray matter bins . The strength of modulations from feature-based attention and stimulus contrast were quantified as the difference between condition-specific time courses during the peak of the stimulus driven response ( highlighted in Figure 3A&C ) , which are plotted for each depth bin in Figure 3B&D . Finally , we computed a score to describe the extent to which an effect of interest was expressed in the agranular or granular layers . This was achieved by averaging the effect of attention or stimulus contrast ( Figure 3B&D ) from the superficial and deep gray matter bins ( agranular ) and subtracting the middle bin ( granular ) . A positive score therefore indicates a mostly agranular effect , while a negative score indicates a granular effect . The procedure described here was applied first to voxels from all visual areas combined ( Figure 3 ) , and then V1 , V2 and V3 separately ( Figure 4 ) . Overall effects of feature-based attention and stimulus contrast ( Figure 2 ) were assessed using a visual area ( V1/V2/V3 ) x contrast ( high/low ) x attention ( attended/unattended ) repeated measures ANOVA . Note that , though we plot the results from V1-V3 combined in Figure 2 , the ANOVA was performed on data from the three areas separately so that it would incorporate differences between areas . The effects of feature-based attention and stimulus contrast in laminar-specific time courses from V1-V3 combined were quantified by examining the difference between attended and unattended ( or high and low contrast ) time courses during the peak of the stimulus-driven response during a block of stimuli ( highlighted in Figure 3A&C ) . These were assessed using separate condition ( attended/unattended or high/low contrast ) x time point ( 6 . 8/10 . 2/13 . 6/17 s ) repeated measures ANOVAs . These tests were performed on time courses averaged across depth bins that are plotted in Figure 3A&C . Depth-specific time courses were analyzed in the same way , and the difference between attended/unattended and high/low contrast for each depth are plotted in Figure 3B&D , respectively . We investigated whether these laminar profiles were different from each other using a modulation ( attention/contrast ) by depth ( deep/middle/superficial ) repeated measures ANOVA . A significant interaction ( see Results ) revealed the profiles were different from each other , so we examined them independently with one-way repeated measures ANOVAs ( levels: deep/middle/superficial ) . In the cases that the main effect of depth was significant ( i . e . , for stimulus contrast ) , differences between depths were examined with paired-samples t tests . Finally , the difference between agranular – granular scores for attention and stimulus contrast was assessed using a paired-samples t test . Laminar-specific effects of attention and stimulus contrast were also compared between visual areas ( Figure 4 ) . Differences between laminar profiles of attention and contrast and between areas were assessed with a visual area ( V1/V2/V3 ) x modulation ( attention/contrast ) x depth ( deep/middle/superficial ) repeated measures ANOVA . The modulation x depth interaction approached significance ( see Results ) , and we chose to examine the effects of attention and contrast using separate ANOVAs so that we might relate these results to those obtained from V1-V3 combined . As such , we examined whether the effects of attention and contrast varied across depth and visual area using separate visual area ( V1/V2/V3 ) x depth ( deep/middle/superficial ) repeated measures ANOVAs . Finally , differences in agranular – granular scores between conditions and areas were assessed using a modulation ( attention/contrast ) x visual area ( V1/V2/V3 ) repeated measures ANOVA . For all the ANOVAs we conducted , in cases where the assumption of sphericity was violated the degrees of freedom were adjusted using a Huynh-Feldt correction . | Recent advances in brain imaging have made it possible to map brain activity in areas of tissue less than a millimeter in size . This resolution offers particular advantages for studying the brain’s outer surface , the cortex . The cortex is traditionally divided into several layers , each containing different types and arrangements of neurons . New high-resolution machines can now visualize the activity in individual layers of cortex , and this can reveal whether the layers also have different roles . In humans , a large area in the cortex is devoted to vision . Our visual cortex receives sensory information that arrives from the eyes via the optic nerve . This is known as bottom-up processing . But what we see depends on more than just incoming sensory information: it also relies on where we focus our attention , and on our expectations about how things should look . Many optical illusions , for example , work because the brain attempts to decipher an ambiguous visual signal based on previous experiences . This use of existing knowledge to interpret sensory input is called top-down processing . Using high-resolution brain scanning , Lawrence et al . show that bottom-up and top-down processing occur in different layers of visual cortex . Healthy volunteers viewed a series of images while lying inside a brain scanner . Lawrence et al . changed the contrast of the images to alter the volunteers’ bottom-up processing: this affected activity in the middle layer of visual cortex . To adjust their top-down processing , the volunteers were asked to attend to different features of the images on different trials: these changes in attention had more effect in the layers on either side of the middle layer . This suggests that bottom-up processing occurs in the middle layer of visual cortex , whereas top-down processing takes place in the layers above and below . The findings by Lawrence et al . will help to better measure activity in cortical layers using modern brain imaging techniques . With further technological improvements , it may become possible to image each layer in the brain in more detail , in particular for other areas that support complex cognitive processes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Dissociable laminar profiles of concurrent bottom-up and top-down modulation in the human visual cortex |
Alternative splicing ( AS ) can critically affect gene function and disease , yet mapping splicing variations remains a challenge . Here , we propose a new approach to define and quantify mRNA splicing in units of local splicing variations ( LSVs ) . LSVs capture previously defined types of alternative splicing as well as more complex transcript variations . Building the first genome wide map of LSVs from twelve mouse tissues , we find complex LSVs constitute over 30% of tissue dependent transcript variations and affect specific protein families . We show the prevalence of complex LSVs is conserved in humans and identify hundreds of LSVs that are specific to brain subregions or altered in Alzheimer's patients . Amongst those are novel isoforms in the Camk2 family and a novel poison exon in Ptbp1 , a key splice factor in neurogenesis . We anticipate the approach presented here will advance the ability to relate tissue-specific splice variation to genetic variation , phenotype , and disease .
Production of distinct mRNA isoforms from the same locus has been shown to be common phenomena across metazoans ( Barbosa-Morais et al . , 2012; Merkin et al . , 2012 ) . Different isoforms may arise through the use of alternative transcription start and end sites , or through alternative processing of pre-mRNA . A key process is alternative splicing ( AS ) of pre-mRNA , where different subsets of pre-mRNA segments are removed while others are joined , or spliced together . The resulting differences between the mature mRNA isoforms can , in turn , encode different protein products , or affect mRNA stability , localization , and translation . Over 95% of human multiexon genes undergo AS , and disease associated genetic variants have been shown to frequently lead to splicing defects ( Cooper et al . , 2009; Pan et al . , 2008; Wang et al . , 2008 ) . These observations emphasize the need to accurately map and quantify splice variations . RNA-Seq technology has advanced the detection and quantitation of splice variants by producing millions of short sequence reads derived from the transcriptome . Despite constant technological advancement , the combination of limited coverage depth , experimental biases , and reads spanning only a small fraction of the variable parts of transcripts has left accurate mapping of transcriptome variations an open challenge ( Alamancos et al . , 2014 ) . Transcriptome variations have been traditionally studied either at the level of full gene isoforms or through the specification of alternative splicing 'events' . The latter have been categorized into several common types , such as intron retention , alternative 3’/5’ splice sites , or cassette exons . Importantly , while exact isoforms and their quantifications cannot be directly inferred from the short RNA-Seq reads , AS events can be detected via reads that span across spliced exons ( junction reads ) . Both AS events and full isoforms can be captured by a gene schematic or a splice graph ( Heber et al . , 2002 ) , where edges ( lines ) connect pre-mRNA segments spliced together in different transcripts ( Figure 1A , top ) . 10 . 7554/eLife . 11752 . 003Figure 1 . LSV formulation and prevalence . ( A ) LSVs can be represented as splice graph splits from a single source exon ( yellow ) or into a single target exon ( pink ) . LSV formulation captures previously defined , 'classical' , binary alternative splicing cases ( top ) as well as other variations ( bottom ) . An asterisk denotes complex variations involving more than two alternative junctions; dash line denotes redundant LSVs that are a subset of other LSVs ( see Materials and methods ) . ( B ) Example of a complex LSV in the Camk2g gene . The gene’s splice graph ( top ) includes known splice junctions from annotated transcripts ( red ) and novel junctions ( green ) detected from RNA-Seq data . The splice graph includes a complex LSV involving exons 14–17 ( middle ) . RT-PCR validation of the LSV in brainstem , cerebellum , hypothalamus , muscle , and adrenal is shown at the bottom . Several isoforms are preferentially included in brain and muscle . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 003 While useful , the previously defined AS types fail to capture the full complexity of spliceosome decisions . Specifically , AS types represent spliceosome decisions as strictly binary , involving only two exons or two splice sites in the same exon . The bottom panel in Figure 1A illustrates a few possible splicing variations that do not fit the previously defined AS types and can involve more than two alternative junctions . Figure 1B serves as a visual summary for both the potential and challenges in analyzing splicing variations . Combining known transcripts and RNA-Seq data results in the Camk2g splice graph shown ( Figure 1B , top ) . This splice graph includes novel , unannotated , splice junctions detected from junction spanning RNA-Seq reads ( green ) , as well as a complex case where exon 14 can be spliced to exons 15 , 16 , or 17 ( Figure 1B , middle ) . Quantification by RT-PCR in several mouse tissues validate the existence of these variations and also points to isoforms that are predominantly produced in brain subregions and in muscle ( Figure 1B , bottom ) . In order to achieve such results we need to have a computational framework that efficiently combines RNA-Seq with existing gene annotation and enables us to accurately detect , quantify , and visualize diverse splicing variations across different experimental conditions .
To address the shortcomings of previously defined AS types we suggest the formulation of local splicing variations , or LSVs . LSVs are defined and easily visualized as splits ( multiple edges ) in a splice graph where several edges either come into or from a single exon , termed the reference exon . A Single Source ( SS ) LSV ( Figure 1 , yellow ) corresponds to a reference exon spliced to several downstream RNA segments while single target ( ST ) LSV ( Figure 1 , pink ) corresponds to a reference exon spliced to upstream segments . The full specification of an LSV also includes the relative location of the exons and junctions ( see Material and methods ) . Figure 1A illustrates how this formulation captures previously defined AS types ( top panel ) as well as more complex cases ( bottom panel ) . Specifically , previously defined 'classical' AS events appear as special cases of binary graph splits ( e . g . , include or skip a cassette exon ) , while LSVs capture non-classical binary splits and splits involving more than two junctions . Such non-binary splits are termed complex LSVs . LSVs can also involve intron retention ( intronic LSVs ) or be comprised of only exons ( exonic LSVs ) . Moreover , the transcriptome variability captured by LSVs may be the result of not only spliceosome decisions but also of alternative transcription start or end positions . For example , the gene in Figure 1A bottom panel involves two alternative first exons so a relative change in the transcription start site usage can result in changes in downstream LSVs quantification . Importantly , LSV formulation allows the probing of transcriptome structure and complexity yet , unlike full transcripts , can still be quantified directly from junction spanning reads . In order to address the challenges involved in detection , quantification and visualization of LSVs we developed a new computational framework that we have termed Modeling Alternative Junction Inclusion Quantification ( MAJIQ ) . MAJIQ’s first step ( Figure 2A , top ) is to parse a known database of transcripts , given as a GFF3 annotation file , along with a set of mapped and aligned RNA-Seq experiments ( indexed BAM files ) . Unlike many methods that only analyze known isoforms , MAJIQ supplements known transcripts with 'reliable' edges derived from de novo junction spanning reads . Several filters can be applied to define which edges are considered reliable and which LSVs have enough reads to be later quantified ( see Material and methods ) . Similarly , LSVs whose edges are a subset of other LSVs , such as those denoted with dashed rectangles in Figure 1A , are removed to avoid redundancy ( see Material and methods ) . Next , MAJIQ can be executed to quantify LSVs either in a specific condition or to compare two experimental conditions , with or without replicates . LSV quantification in a specific condition is based on the marginal percent selected index ( PSI , denoted Ψ ) for each junction involved in the LSV , while comparison of experimental conditions is based on relative changes in PSI ( dPSI , ΔΨ ) . MAJIQ uses a combination of read rate modeling , Bayesian Ψ modeling , and bootstrapping to report posterior Ψ and ΔΨ distributions for each quantified LSV . The results of MAJIQ’s LSV detection and quantification can be interactively visualized with the package VOILA in a standard web browser ( Figure 2A bottom ) . 10 . 7554/eLife . 11752 . 004Figure 2 . LSV analysis using MAJIQ . ( A ) MAJIQ’s analysis pipeline . RNA-Seq reads are combined with an annotated transcriptome to create splice graphs and detect LSVs for each gene , then LSVs are quantified and compared between conditions . The visual output ( VOILA ) lists LSVs with violin plots representing estimates of percent inclusion index ( PSI , Ψ ) or changes in inclusion ( dPSI , ΔΨ ) . Two cases are illustrated , for a single source three way LSV ( orange ) , and a single target two way LSV ( pink ) . ( B ) Correspondence between E[Ψ] by MAJIQ and Ψ by RT-PCR . R is the correlation coefficient . Colors and shapes represent different experimental conditions: mouse cerebellum and liver ( dark and light orange diamonds , respectively ) ; human unstimulated and stimulated T-Cells ( dark and light purple dots , respectively ) . Total n = 208 . ( C ) Correspondence between E[ΔΨ] by MAJIQ and ΔΨ by RT-PCR , where |ΔΨRT|>0 . 2 . R is the correlation coefficient . Changes in inclusion were measured between liver and cerebellum mouse tissues ( diamonds , n = 45 ) ; stimulated and unstimulated T-Cells ( dots , n = 9 ) . ( D ) Reproducibility ratio ( RR ) of high confidence differentially included LSVs , i . e . LSVs for which P ( |ΔΨ|> 0 . 2 ) > 0 . 95 ) , when comparing RNA-Seq from two conditions . A differentially included LSV is considered replicated if it maintains a rank at least as high as N in biological replicates , where N is the set size . LSVs are ranked by E[ΔΨ] and filtered for overlap . Twelve replicate pairs from Keane et al . ( 2011 ) were used to compute the histogram’s std ( light blue ) . Other lines show MAJIQ’s RR with replicates ( thick blue ) , RR for AS events detected by rMATS w/wo replicates ( light and dark green ) , MISO ( red ) , and RR for LSVs using Naïve Bootstrapping ( orange ) . The inset bar chart shows the number of LSVs or AS events ( N ) derived by each method and used in the RR plots ( see Materials and methods for more details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 00410 . 7554/eLife . 11752 . 005Figure 2—figure supplement 1 . Quantifying PSI and dPSI accuracy . ( A ) Correspondence between E[Ψ] by MAJIQ ( top ) or MISO ( bottom ) and Ψ by RT-PCR across four different experimental conditions . ( B ) The same set of LSVs used to measure correspondence between E[ΔΨ] by MAJIQ ( top ) and MISO ( bottom ) and ΔΨ by RT-PCR . Changes in inclusion were measured between cerebellum and liver mouse tissues ( diamonds , right panel , n = 50 ) ; stimulated and unstimulated T-Cells ( dots , center panel , n = 57 ) . Setting a threshold of ΔΨRT =20% for a significant change MAJIQ has no false positives and fewer false negatives compared to MISO . ( C ) Histogram of Ψ reproducibility , computed as the absolute difference between biological replicates of hippocampus and liver ( R = E[Ψr1]-E[Ψr1] ) . Overall , 81 . 2% of the junctions in quantifiable LSVs were reproducible within 5% ( R ( Ψ ) < 5% . Average n = 8058 . Twelve replicate pairs were used to compute the histogram’s std ( light color ) . Inset graph: comparing MAJIQ and MISO reproducibility for paired junction ( ΔR = RMISO - RMAJIQ ) . Plot shows the cumulative distribution over ΔR>0 ( blue ) and ΔR<0 ( red ) and over the subset with significant difference ( ΔR >0 . 05 , dashed lines ) . Overall MAJIQ improved Ψ reproducibility for approximately two thirds of the LSVs ( P ( ΔR = RMISO- RMAJIQ ) > 0 = 61 . 7% ) and over two fold more showed a significant improvement ( P ( ΔR>0 . 05 ) = 21 . 2% ) , P ( ΔR< -0 . 05 ) = 10 . 1% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 00510 . 7554/eLife . 11752 . 006Figure 2—figure supplement 2 . Quantifying differential splicing reproducibility . ( A ) Effect of the threshold applied to the significance of splicing changes on the number of LSVs identified as changing ( N ) and the reproducibility ratio ( RR ) . Both MAJIQ and rMATS estimate ( P ( ΔΨ ) > α ) > β ) for an inclusion difference α with confidence level β . In the paper we used a strict β = 95% to control for false positives and a conservative α = 20% to call differentially spliced LSVs . Here , the results with a relaxed criteria of α = 15% ( left ) and α = 10% ( right ) are shown . The plots are otherwise identical to Figure 2D . ( B ) Breakup by coverage level ( x-axis ) of the high confidence differentially spliced LSVs depicted in Figure 2D . Y-axis denotes reproducibility ratio by RNA-Seq from biological replicates and the numbers at the top of each bar denote reproducibility by RT-PCR ( |ΔΨRT|>0 . 2 ) of a randomly chosen subset of LSVs from that bin . The overall reproducibility is represented by the far left bin . ( C ) Correlation between MAJIQ E[ΔΨ] and average ΔΨ by RT-PCR among 3 biologic replicates for the most changing junction in validated complex LSVs examined in this paper between various pairs of tissues ( n = 78 ) . Of the junctions predicted to change between tissues ( |E[ΔΨ]| > 20% ) , 55/56 validated ( 98 . 2% ) by RT-PCR with an average |ΔΨ| > 20% . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 006 We assessed MAJIQ’s quantification accuracy for both Ψ and ΔΨ using a combination of RNA-Seq from biological replicates and an extensive set of 208 RT-PCR validations . These experiments included two mouse tissues ( cerebellum and liver [Zhang et al . , 2014] ) , and a human Jurkat T cell line ( unstimulated and stimulated , [Cole et al . , 2015] ) . While accuracy depended on the dataset used , MAJIQ achieved an overall correlation of R = 0 . 8 and R = 0 . 95 for PSI and dPSI quantification by RT-PCR , comparing favorably to alternative methods on all datasets ( Figure 2B , C , Figure 2—figure supplement 1 ) . Next , we used biological replicates from the Mouse Genome Project ( Keane et al . , 2011 ) to assess reproducibility of differential splicing detection from RNA-Seq when comparing two experimental conditions . The reproducibility ratio ( RR , see Material and methods ) captures the fraction of top ranked differentially spliced LSVs that maintain their top ranking when analyzing another set of replicate experiments . Figure 2D shows MAJIQ compares favorably to other methods , including MISO ( Katz et al . , 2010 ) , rMATS ( Shen et al . , 2014 ) , and a bootstrapping approach ( Xiong et al . , 2015 ) adopted for LSV . While MISO and rMATS achieved a reproducibility ratio of 61–67% we found the bootstrapping approach ( N . B . ) suffered from particularly high variance , which degraded reproducibility of LSVs ranking . In comparison , MAJIQ achieved a mean RR=77% when comparing two pairs of experiments and improving to RR=86% when the experiments compared had replicates . Notably , detection power was also improved . Defining differentially spliced LSVs as those for which P ( |ΔΨ|>0 . 2 ) > 0 . 95 , the number of detected LSVs ( N ) , after removing LSVs overlap ( see Materials and methods ) , was on average 400 for pairwise and 447 for group comparisons , compared to 240 and 260 respectively by rMATS . The improvement in both detection and reproducibility of differentially spliced LSVs ( N , RR ) was robust to the statistical threshold used to define N ( Figure 2—figure supplement 2A ) and when we removed MAJIQ’s de-novo junction detection the number of LSVs dropped as expected but reproducibility remained high ( N = 337 , RR= 87% , data not shown ) . Importantly , this result also indicated that including de-novo junctions increased the number of differentially spliced LSVs that could be detected by over 30% ( 337 vs . 447 ) , while retaining equivalent reproducibility . Defining differential splicing reproducibility by RT-PCR as LSVs for which |ΔΨRT|>20% resulted in 95% reproducibility . The higher reproducibility by RT-PCR can be expected given the lower experimental variability compared to RNA-Seq . Notably , the LSVs tested by RT-PCR were selected to cover a wide spectrum of read depth . We found that while higher coverage allowed more differential LSVs to be detected and steadily increased reproducibility by RNA-Seq , MAJIQ’s reproducibility by RT-PCR was stable across read coverage depth , pointing to the robustness of the method ( Figure 2—figure supplement 2B ) . Finally , we note that the above RT-PCR evaluation concentrated on binary LSVs to allow comparison to currently available methods , but we observed similar accuracy for the quantification of complex LSVs ( Figure 2—figure supplement 2C ) . To assess the significance of LSVs formulation we estimated LSVs prevalence in several metazoans , ranging from lizard to human ( Figure 3 ) . Naturally , this analysis is affected by how well a species transcriptome is annotated , and how permissive the database used is . In human for example , complex LSVs constitute 20 . 6% to 33 . 7% of the LSVs in annotated transcripts by RefSeq and Ensembl respectively , but only 1 . 86% in opossum’s Ensembl annotation ( Figure 3A , B ) . Next , we expanded the set of annotated transcripts with novel junctions detected from RNA-Seq junction spanning reads . Limiting our analysis to only 5–6 similar tissues in all species and conservative junction detection still increased the total number of LSVs in human by 11% and the fraction of complex LSVs from 33 . 7% to 37 . 1% ( Figure 3A ) . In species not as well annotated the effect of adding RNA-Seq data was more dramatic , jumping in opossum for example from 1 , 610 to 10 , 228 LSVs , of which 10% were complex . In summary , while LSV analysis across species was confounded by read coverage and transcriptome annotation we find that non-classical and complex LSVs make up a substantial fraction of observed transcriptome variations . Such complex LSVs are likely to be removed , undetected , or mislabeled by algorithms that only quantify binary AS events from previously annotated transcripts . 10 . 7554/eLife . 11752 . 007Figure 3 . LSV prevalence across diverse metazoans . ( A ) Number of LSVs ( top ) and fraction of complex LSVs ( bottom ) when using Ensembl annotated transcripts only ( grey ) or combining it with RNA-Seq from 5–6 similar tissues ( red ) . Mouse* is the dataset from Zhang et al . ( 2014 ) . ( B ) Number of LSVs ( top ) and fraction of complex LSVs ( bottom ) when using RefSeq ( orange ) and Ensembl ( blue ) . The RNA-Seq dataset is the same as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 007 Given the clear impact of the RNA-Seq dataset and transcriptome annotation , we chose to focus our genome wide analysis on a recent mouse dataset . This allowed us to analyze 12 tissues with an average of over 120M reads per sample , produced by a single lab ( Zhang et al . , 2014 ) . This data included three brain subregions , eight samples per tissue , and matching RNA for RT-PCR validations , leading to a total of 100 , 512 LSVs detected . First , we used this data to assess the usage of LSVs across tissues . In order to minimize LSVs that result from false junctions identified by the mapper we only included junctions with multiple uniquely mapped staggered reads across multiple biological replicates ( see Material and methods ) . Next , we tested the maximal inclusion level of the second , third , or the least used junction in an LSV across twelve mouse tissues . We detected a switch behavior where a different junction becomes dominant at 50% inclusion or more in approximately 5% of the classical binary LSVs ( Figure 4A , grey ) , compared to 12% for the second most used junction in complex LSVs ( Figure 4A , light green ) . Setting a conservative threshold of Ψ > 10% to denote splice junctions that are less likely to be splicing noise or database errors we find that for the classical binary LSVs approximately 32% , or 9 , 516 pass that threshold , compared to 57% and 19% of the complex LSVs that pass that threshold for the second and third most used junction respectively . These correspond to a total of 6 , 338 and 2 , 112 LSVs in our datasets , pointing to the importance of complex LSVs in transcriptome analysis . Even when testing for the least used junction in complex LSVs ( e . g . the ninth in a nine junction LSV ) , we still find almost 10% pass the 10% inclusion threshold ( Figure 4A , dark green ) . Finally , for intronic LSVs we find almost 11 , 000 cases where an intron is retained at least 50% in one tissue , and 3 , 844 cases where the intron is almost always retained with Ψ > 99% ( Figure 4—figure supplement 1D ) . This observation of widespread intron retention ( IR ) , especially in brain tissues , is in line with a recent study across many more tissues and cell lines ( Braunschweig et al . , 2014 ) , though our overall estimate of IR prevalence is more conservative . 10 . 7554/eLife . 11752 . 008Figure 4 . Genome wide view of exonic LSVs across twelve mouse tissues . ( A ) Cumulative distribution ( CDF ) for maximal junction inclusion ( PSI ) across tissues . Plot includes the least used junction in binary LSV ( grey ) , the second , third and least used junction in complex LSVs ( light , medium , dark green ) . Dashed vertical line denotes 10% inclusion . ( B ) Histogram of the most common exonic LSV types . ( C ) Histogram of the number of exons , junctions , 3’ and 5’ splice sites in all identified LSV . ( D ) Histogram of which 3’ ( left ) or 5’ ( right ) splice site are found to be dominant across all tissues and all LSVs . X-axis denotes the order of the splice site . Dominance is defined as E[Ψ] > 0 . 6 . Cases with no dominant junction are represented by the bars on the far left . ( E ) The fraction of complex LSVs ( green , top right ) from the total number ( purple , bottom left ) of differentially spliced LSVs ( |E[ΔΨ]| >0 . 2 ) between pairs of tissues . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 00810 . 7554/eLife . 11752 . 009Figure 4—source data 1 . dPSI values for all pairs of tissues . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 00910 . 7554/eLife . 11752 . 010Figure 4—figure supplement 1 . Intronic LSV detection and quantification . ( A ) Effect of intronic average coverage on the number of detected introns in LSVs . ( B ) Histogram of mean Ψ reproducibility as in Figure 2—figure supplement 1C but for intronic LSV . Ψ Reproducibility is computed as the absolute difference between biological replicates ( R = |E[Ψr1]-E[Ψr1]| ) . Twelve replicate pairs were used to compute each histogram’s mean . The histograms’ std was too small to be plotted clearly . Colors correspond to different thresholds on average intronic coverage . Numbers in the legend represent average number of introns quantified in experiment pairs . Based on the tradeoff between reproducibility and overall detection shown in ( A ) subsequent evaluations and figures were executed using average intronic coverage threshold of 0 . 5 . ( C ) Histogram of the most common intronic LSV types . Only non-redundant LSVs are included ( See Materials and methods ) . ( D ) Cumulative distribution function for the fraction of the introns in intronic LSVs as a function of the minimal intronic Ψ observed across the twelve mouse tissues . Vertical dashed line corresponds to Ψ=10% . ( E ) Bottom left ( purple ) : Each entry A ( i , j ) is the number of intron containing LSVs where the intron is differentially spliced between tissue i and tissue j . Upper right ( green ) : Each entry A ( i , j ) is the fraction of complex LSVs from the LSVs listed in the matching bottom left rectangle entry A ( j , i ) . Diagonal ( red ) : Each entry A ( i , i ) is the total number of unique differentially spliced intron containing LSVs in tissue i compared to all other tissues where the intron is differentially spliced . ( F ) Bottom left ( purple ) : Each entry A ( i , j ) is the number of intron containing LSVs where an exonic junction ( and not the intron ) is differentially spliced between tissue i and tissue j . Note that in this case the upper right ( green ) triangle that gives the fraction of complex LSVs from the LSVs listed in the matching bottom left rectangle is by definition 100% and is therefore not shown . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 010 Commonly occurring network substructures , or network motifs , have garnered much research attention in diverse fields ( Milo et al . , 2002 ) . Gene splice graphs can also be thought of as networks with exons as nodes and spliced junctions as edges . In this interpretation , LSVs can be thought of as small network motifs and used to shed light on the transcriptome complexity and commonly reoccurring sub-structure . Comparing the frequency of exonic LSV types ( Figure 4B ) we find that the more common non classical LSVs involve 3 to 5 exons , combine exon skipping with an alternative 3’/5’ splice site , or involve alternative transcript start/end at the LSV’s reference exon . In contrast , intronic LSVs are much less diverse , with classical intron retention making 68% of the cases ( Figure 4—figure supplement 1C ) . Figure 4C shows that for exonic LSVs 14% involve more than 2 exons , 30% of the single source and 20% and of the single target LSVs involve a reference exon with two or more 5’/3’ splice sites , respectively . Overall , complex ( non-binary ) LSVs comprise 36 . 2% of the transcriptome variations detected in the data and 27 . 5% of the variations deemed quantifiable ( see Materials and methods ) , yet spliceosome decisions still appear localized , with few LSVs involving more than 6 exons or junctions . When analyzing LSVs usage , we found that the biochemical 'proximity rule' , by which the splice site nearest to the reference exon is preferred ( Reed and Maniatis , 1986 ) , is commonly not reflected at the genomic level . Defining 'dominant' junctions as those included at least 60% , we found proximal junctions appear dominant in approximately two thirds of the cases involving binary LSVs ( Figure 4D ) while more complex LSV tend to have more evenly distributed inclusion levels with no dominant junction ( Figure 4D , left bars ) . This more evenly distributed usage of exons and junctions in complex LSVs further supports possible functionality of multiple isoforms . Figure 4E gives a genome wide view of the exonic LSVs that exhibit significant splicing changes ( |E[ΔΨ]|> 20% ) between mouse tissues . In line with previous reports ( Barash et al . , 2010; Barbosa-Morais et al . , 2012 ) , we find clear clusters for brain and muscle tissues ( average of 875 and 657 changing LSVs , respectively ) , a weaker cluster for digestive tissues ( liver , kidney ) with an average of 501 changing LSVs , and lung as a unique signal ( 549 changing LSVs ) . Brain regions have a higher average of 927 ( Cerebellum ) to 840 ( brainstem ) changing LSVs compared to non-brain tissues . The number of LSVs changing between brain subregions varies between 36% and 57% of those changing between CNS and non-CNS tissues , with hypothalamus standing out as more similar to the two other CNS tissues ( average of 937 and 343 changing LSVs when compared to non brain and other brain sub-regions , respectively ) . Overall , we find that complex LSVs make up almost 47% of the differentially spliced LSVs , a fold enrichment of 1 . 7 compared to their relative proportion of 27 . 5% in the quantifiable set ( P < 2 . 3 x10-278 , binomial test ) . Given the above result of complex LSV enrichment in tissue dependent splicing variations we decided to test whether this enrichment holds in other datasets that involve developmental stages , splice factor knockdowns , and disease . We performed a meta analysis of 31 mouse datasets that involve a total of 243 RNA-Seq experiments covering a variety of tissues , cell lines , developmental stages , and knockdowns of key splicing factors . To this set we also added a human dataset comparing Alzheimer’s disease and healthy brain samples ( Figure 5A and below ) . We found the median fraction of complex LSV in these datasets was 0 . 309 and their median fold enrichment in differentially spliced LSVs was 1 . 63 , a significant enrichment in 30/32 of the datasets ( 1 . 6x10-322 < p-val < 1x10-3 , Bonferroni corrected binomial test , see Figure 5A , and Figure 5—source data1 ) . This consistent overrepresentation of complex LSVs among differentially spliced LSVs across a variety of contexts further suggests that complex LSVs are an important aspect of regulated alternative splicing . 10 . 7554/eLife . 11752 . 011Figure 5 . Meta analysis of complex LSVs . ( A ) Fold enrichment ( green dots ) of complex LSVs calculated by comparing the fraction of complex LSVs among differentially spliced LSVs ( dark blue bars ) to their relative proportion ( light blue bars ) in 32 datasets . The corrected p-value column on the left measures significance of the fold enrichment ( binomial test , Bonferroni corrected p-value ) Medians are displayed for fold enrichment ( green line , 1 . 63 ) , fraction of complex LSVs among changing LSVs ( orange line , 0 . 52 ) , and fraction of complex LSVs among all detected LSVs ( red line , 0 . 31 ) . Human AD versus healthy brain data corresponds to the cohort from ( Bai et al . , 2013 ) . See Figure 5—source data 1 for more information . ( B ) Empirical cumulative distribution function ( CDF ) of the maximal change of junction inclusion ( ΔΨ ) across all mouse datasets in Figure 5A . Only the LSVs detected in the twelve mouse tissues ( Figure 4 ) are included . The plot includes junctions in binary LSVs ( grey ) , and the second , third , and least changing junction in complex LSVs ( light , medium , dark green ) . Dashed vertical line denotes ΔΨ of 10% . ( C ) Per nucleotide average conservation score ( phastCons60 track ) in regions proximal to single source ( top ) and single target ( bottom ) LSVs that were differentially spliced between any pair of tissues shown in Figure 4 . The average is plotted for the subsets of complex ( green ) LSVs and binary ( grey ) LSVs as well as around a randomly selected set of constitutively spliced junctions ( red , see Materials and methods for details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01110 . 7554/eLife . 11752 . 012Figure 5—source data 1 . LSV enrichment meta analysis table . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01210 . 7554/eLife . 11752 . 013Figure 5—figure supplement 1 . Empirical cumulative distribution function ( CDF ) of the maximal junction inclusion ( E[Ψ] ) across all mouse datasets in Figure 5A . Only the LSVs detected in the twelve mouse tissues ( Figure 4 ) are included . This plot is equivalent to the ΔΨ plot in Figure 5B and includes junctions in binary LSVs ( grey ) , as well as the second , third , and least included junction in complex LSVs ( light , medium , dark green ) . Dashed vertical line denotes 10% inclusion . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 013 Next , we asked how does the inclusion of junctions change across these datasets . For this , we took a conservative approach monitoring only the LSVs that have been already identified in normal tissues used to build the genome wide view of LSVs ( Figure 4 ) . Figure 5B shows over 20% of all complex LSVs detected in more than one sample had the third most differentially included junction exhibit |ΔΨ|> 10% , corresponding to 2 , 236 LSVs . Strikingly , these additional experimental contexts showed that over 39% of all complex LSVs detected in our normal tissue set had their third most included junction with Ψ > 10% , corresponding to 4 , 201 LSVs ( Figure 5—figure supplement 1 ) . Finally , we plotted the conservation level around constitutive exons and differentially spliced LSVs shown in Figure 4 that are either binary or complex ( Figure 5C ) . Inline with previous reports , we found tissue regulated splicing involves significantly higher conservation in the intron proximal to the variable exonic segments , a region known to include cis elements to which tissue specific splice factors bind . However , we also found that differentially spliced complex LSVs exhibited significantly higher conservation levels in these regions compared to their binary counterparts . This finding may be the result of the more complex splicing changes that need to be controlled or tighter control associated with complex LSVs specific function . In summary , these different lines of evidence all support the functional relevance and utility of accurately mapping and quantifying complex splicing variations in genome wide studies . The observed evolutionary pressure to conserve intronic segments around tissue dependent LSV raises the questions what are the functional consequences of LSVs and whether complex LSVs are functionally distinct from classical binary ones . To probe possible function we mapped exons in LSVs into their matching protein domains ( see Material and methods ) . We then grouped LSV junctions based on whether they were part of binary or complex LSVs and whether they were differentially included across tissues . In line with previous works ( Ellis et al . , 2012 ) , we find that binary LSVs , such as cassette exons , which are also differentially included across tissues , more frequently affect low-complexity , disordered regions when compared to non-changing binary LSVs ( p<1x10-4 , corrected Fisher’s exact test ) . Interestingly , differentially included complex LSVs are similarly enriched for such low-complexity regions ( p<1x10-4 ) , but also show enrichment for specific protein families ( e . g . spectrin/filamin ) and domains ( e . g . RNA recognition motifs ) when compared to non-changing complex LSVs . These families and domains are largely distinct from those enriched in binary LSVs ( e . g . WW domains or coiled coils ) . The complete list of enriched protein features can be found in Supplementary file 1 . Overall , this analysis suggests that regulated alternative splicing of both binary and complex LSVs can affect protein interactions via unstructured protein regions , or affect the inclusion of distinct protein domains in specific families . To further demonstrate the power of MAJIQ and our LSV based approach we validated a set of complex LSVs that exhibit tissue and brain region dependent splicing patterns . Surprisingly , this analysis revealed a previously uncharacterized , brain-specific exon in the gene encoding PTBP1 , an extremely well studied splicing factor critical to neural development ( Keppetipola et al . , 2012 ) ( Figure 6A , Figure 6—figure supplement 1A ) . While this novel exon remained undetected when running cufflinks ( Trapnell et al . , 2010 ) on this dataset ( data not shown ) , expression of this novel exon as part of a complex LSV was supported by RT-PCR from cerebellum and adrenal tissues ( Figure 6B , top ) with good concordance with MAJIQ’s PSI quantification ( Figure 6B , bottom ) . Products including exon 14 were also weakly detected by RT-PCR of brainstem and hypothalamus-derived RNA , but not from any of the other eight tissues tested ( Figure 6—figure supplement 2 ) . Together these data strongly support exon 14 as brain-specific . 10 . 7554/eLife . 11752 . 014Figure 6 . Identification of a novel , brain-specific , PTC-introducing , developmentally-regulated exon in Ptbp1 . ( A ) Top: Splice graph representation of a complex target LSV containing a previously unannotated , PTC-introducing exon in Ptbp1 ( exon 14 , green ) . Stop signs indicate multiple conserved premature termination codons . Bottom: UCSC Genome Browser tracks of RNA-seq reads from adrenal ( red ) and cerebellum ( blue ) , and conserved Rbfox binding sites ( [U]GCAUG ) found within the bounds of this LSV . ( B ) Top panel: RT-PCR validation of RNA from replicate cerebellar and adrenal tissues with isoforms illustrated on the left . Asterisk denotes a background band that migrates non-specifically . Bottom panel: E[Ψ] violin plots of MAJIQ quantification for the colored junctions in ( A ) . Matching isoforms are indicated on the left . ( C ) Top: RNA-seq reads from mouse cortices ( Yan et al . , 2015 ) . Developmental time points indicated on the right with exons colored as in ( A ) . Bottom: Ψ violin plots for the PTC-introducing exon 14 across brain development . ( D ) Top panel: Top regulatory motifs predicted by AVISPA to influence the neuronal-specific splicing of exon 14 . Stacked bars represent the normalized feature effect ( NFE ) for each motif . Colors indicate the contribution of the corresponding motif in the region indicated in the inset . ( E ) MAJIQ Ψ quantification of the LSV shown in ( A ) , using RNA-seq from one month old wild type whole brain ( left ) and nestin-specific Rbfox1 KO littermates ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01410 . 7554/eLife . 11752 . 015Figure 6—figure supplement 1 . Novel exon and PTCs in Ptbp1 are conserved , independent from known PTC event , and regulated by Rbfox1 and 2 . ( A ) PTBP1 domain structure ( top ) and splice graph from cerebellum data ( bottom ) highlighting approximate locations of known alternatively spliced linker region ( dark grey ) encoded by exon 13 ( exon 9 in the literature ) , the novel PTC introducing exon 14 ( red stop sign in protein , green exon in splice graph ) , and the known PTC upon exclusion of exon 16 ( exon 11 in the literature ) . ( B ) UCSC genome browser view with sequence alignment and placental mammalian conservation . Novel exon 14 is highlighted in blue and boxed regions correspond to conservation of 3’ and 5’ splice sites and the in frame PTCs . ( C ) Locations of additional primers with RT-PCR from replicate cerebellum RNA . ( D ) UCSC genome browser view showing conserved Rbfox binding sites ( [U]GCAUG ) and brain RNA-seq reads from wild type one month old mice ( top ) and Rbfox1 KO littermates ( bottom ) corresponding to experiments quantified in Figure 6E . Exon 14 location is highlighted in blue . ( E ) MAJIQ Ψ quantification of junctions as illustrated in ( C ) from one month old wild type mice ( top ) and Rbfox2 KO littermates ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01510 . 7554/eLife . 11752 . 016Figure 6—figure supplement 2 . RT-PCR validation of complex Ptbp1 LSV across 11 mouse tissues . ( A ) Representation of Ptbp1 target LSV analyzed with primers indicated by arrows . ( B ) RT-PCR from replicates across tissues indicated with isoforms indicated on the left . ( C ) Representative RT-PCR from tissues indicated with isoforms indicated on the left . [Bstm: brainstem; Hyp: hypothalamus; Cer: cerebellum; Adr: adrenal gland; Kid: kidney; Hrt: heart; Mus: muscle; Bfat: brown adipose; Wfat: white adipose; Liv: liver]DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 016 Interestingly , Ptbp1 exon 14 shows conservation of splice sites between mouse and human and inserts multiple premature termination codons ( PTCs ) in both species , as well as in other mammals , before RMMs 3–4 of PTBP1 ( Figure 6—figure supplement 1A , B ) , suggesting that mRNAs including this exon are likely targets of nonsense-mediated decay ( NMD ) . Regulated alternative splicing that introduce PTCs is a common theme among numerous splicing factors ( Ni et al . , 2007 ) and exclusion of Ptbp1 exon 16 ( exon 11 in the literature ) has already been identified and shown to induce NMD ( Figure 6—figure supplement 1A ) ( Wollerton et al . , 2004 ) . Remarkably , exclusion of exon 16 is barely detectable in the brain regions examined and inclusion of exon 14 is not associated with this event ( Figure 6—figure supplement 1C ) . Together , this suggests that these splicing events are independent mechanisms to control Ptbp1 expression and that inclusion of novel exon 14 plays a larger role in the brain regions examined , with 26% of the Ptbp1 transcripts in the cerebellum containing PTCs . Embryonic down regulation of Ptbp1 by miR-124 is crucial at the onset of neurogenesis ( Makeyev et al . , 2007 ) and leads a change in splicing programs ( Boutz et al . , 2007; Keppetipola et al . , 2012 ) , but cannot account for additional postnatal down regulation of this protein ( Boutz et al . , 2007; Zheng et al . , 2012 ) . Remarkably , MAJIQ analysis of RNA-seq data from mouse cortices across development ( Yan et al . , 2015 ) reveals clear developmental regulation of exon 14 with a dramatic increase in inclusion from P15 through adulthood ( Figure 6C ) . Taken together , this complex LSV offers a novel mechanism for postnatal neuronal reduction in Ptbp1 . To identify putative regulators of novel exon 14 , we used AVISPA ( Barash et al . , 2013 ) , a web tool that utilizes splicing code models to suggest motifs important for tissue-specific splicing , and identified the [U]GCAUG binding motif of the Rbfox family as important for neuronal splicing outcome ( Figure 6D ) . AVISPA’s map of regulatory motifs pointed to a number of Rbfox binding sites downstream of exon 14 ( Figure 6A ) . These motifs , perfectly conserved between mouse and human , suggested enhancement of inclusion by the Rbfox family ( Lovci et al . , 2013 ) . Consistent with this regulatory hypothesis , MAJIQ analysis of RNA-seq data from one month old nestin-specific Rbfox1 KO mice revealed a marked decrease in inclusion of exon 14 from ~16% in wild type mice to nearly undetectable in the KO ( Figure 6E; Figure 6—figure supplement 1D ) and similar decreased inclusion was observed upon Rbfox2 KO ( Lovci et al . , 2013 ) ( Figure 6—figure supplement 1E ) . Together these data demonstrate the power of MAJIQ , in combination with the VOILA and AVISPA analysis tools , in identifying previously uncharacterized isoforms and understanding the regulation of biologically important transcript variation . Several of the brain specific LSVs we detected were found in genes encoding calcium/calmodulin-dependent protein kinase II ( CAMK2 ) subunits which regulate functions in the brain such as neurotransmitter synthesis and release , cellular transport , neurite extension , synaptic plasticity , learning and memory ( Griffith , 2004 ) . We focused on Camk2d and Camk2g as these exhibit complex changes and were expressed in nearly all tissues examined ( Figure 4—source data 1 ) . Figure 1B and Figure 7—figure supplement 1B show MAJIQ’s analysis and matching RT-PCR validation of a Camk2g LSV containing three exons across five tissues . Figure 7 shows similar verification for another complex LSV but in Camk2d . In both cases , exon inclusion creates consensus NLS motifs ( KKRK ) , which localize these subunits to the nucleus ( Braun and Schulman , 1995 ) . For Camk2g the NLS motif is contained in exon 15 whose inclusion levels are highest in the brain , particularly in the brainstem ( Figure 1B , Figure 7—figure supplement 1B ) . 10 . 7554/eLife . 11752 . 017Figure 7 . Camk2d LSV exhibits complex developmental dynamics and is misregulated in Alzheimer’s disease . ( A ) Representation of complex source LSV in Camk2d with matching RT-PCR validation in five tissues ( brainstem , cerebellum , hypothalamus , heart , and adrenal ) . Colored arcs represent the junctions quantified by MAJIQ for this LSV while dashed arcs correspond to junctions in the RNA-seq data that are not part of the quantified LSV . Violin plots on the bottom display Ψ quantifications ( x-axis ) for each of the colored junctions ( y-axis ) across the five tissues with appropriate isoforms from the gel on the right . Isoforms with known tissue-specific splicing patterns are labeled as in the literature ( B ) Line graphs of MAJIQ E[Ψ] quantification ( y-axis ) of junctions as in ( A ) across time points ( x-axis ) through cortex development ( top ) and heart development ( bottom ) . Points represent mean Ψ and error bars represent one standard deviation in E[Ψ] . ( C ) ΔΨ quantification comparing changes between control and Alzheimer’s patient brains of the homologous junctions illustrated in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01710 . 7554/eLife . 11752 . 018Figure 7—figure supplement 1 . Complex and de novo LSVs in Camk2g are developmentally regulated and dysregulated in Alzheimer’s disease . ( A ) Splice graph representation of Camk2g in the cerebellum . Green junctions and exons denote de novo detection from RNA-seq data . Numbers represent number of raw reads across the junction . ( B ) Representation of complex source LSV in Camk2g ( top ) with matching RT-PCR validation in five tissues ( brainstem , cerebellum , hypothalamus , heart , and adrenal , middle ) . Colored arcs represent the junctions quantified by MAJIQ for this LSV while dashed arcs correspond to junctions in the RNA-seq data , but not directly quantified by the LSV . Violin plots on the bottom display E[Ψ] quantifications ( x-axis ) for each of the colored junctions ( y-axis ) across the five tissues with appropriate isoforms from the gel on the right . ( C ) Line graphs of MAJIQ Ψ quantification ( y-axis ) of junctions as in ( B ) across time points ( x-axis ) through cortex development ( top ) and heart development ( bottom ) . Points represent mean Ψ and error bars represent one standard deviation . ( D ) Representation of de novo exon 13 detected in mouse ( top ) and MAJIQ Ψ across mouse cortex development , points represent mean Ψ and error bars represent one standard deviation ( bottom ) . ( E ) VOILA ΔΨ visualization of LSV from ( D ) that is conserved in human showing E[Ψ] values ( stacked bar chart , sides ) and E[ΔΨ] ( center ) for each junction between control and Alzheimer’s disease brains . ( F ) Top regulatory motifs predicted by AVISPA to influence the CNS splicing patterns of exon 13 . Stacked bars represent the normalized feature effect ( NFE ) for each motif as in ( Barash et al . , 2013 ) . Colors indicate the contribution of the corresponding motif in the region indicated in the inset . ( G ) VOILA ΔΨ visualization LSV from ( B ) showing E[Ψ] values ( stacked bar chart , sides ) and E[ΔΨ] ( center ) between wild type and Rbfox1 ( top ) or Rbfox2 ( bottom ) KO mice . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01810 . 7554/eLife . 11752 . 019Figure 7—figure supplement 2 . LSV in Camk2a is developmentally regulated oppositely in the brain and heart . ( A ) Splice graph representation of Camk2a across three tissues indicated . Dashed box indicates region containing cassette exon that inserts a consensus NLS . ( B ) MAJIQ E[ΔΨ] between cerebellum and muscle for inclusion ( green ) and exclusion ( blue ) isoforms . Red junction corresponds to an alternative 5’ss not highly used in any tissue . ( C ) Line graphs of MAJIQ Ψ quantification ( y-axis ) of junctions as in ( B ) across time points ( x-axis ) through cortex development ( left ) and heart development ( right ) . Points represent mean Ψ and error bars represent one standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 01910 . 7554/eLife . 11752 . 020Figure 7—figure supplement 3 . Developmentally controlled , complex LSV in Camk2b is regulated by Ptbp2 . ( A ) VOILA thumbnail representation of complex target LSV in Camk2b detected from cortex development data including a alternative transcription start in exon 21 ( green junction ) that is not highly utilized and NAGNAG alternative 3’ splice sites of reference exon 22 . ( B ) Line graphs of MAJIQ Ψ quantification ( y-axis ) of junctions as in ( A ) across time points ( x-axis ) through cortex development show known increase in exon 20 inclusion through development , coupled with a novel switch from proximal NAG 3’ss ( red ) to almost exclusive use of distal NAG 3’ss ( orange ) by adulthood . Points represent mean Ψ and error bars represent one standard deviation . ( C ) UCSC genome browser view of mapped reads from cortex of embryonic 16 . 5 mouse ( top , purple ) or postnatal 21-month mouse ( bottom , blue ) . Dashed box highlights nucleotides corresponding to conserved NAGNAG alternative 3’ss that is developmentally regulated . ( D ) Top regulatory motifs predicted by AVISPA to influence the CNS splicing patterns of exon 20 . Stacked bars represent the normalized feature effect ( NFE ) for each motif . Colors indicate the contribution of the corresponding motif in the region indicated in the inset . ( E ) Violin plots representing MAJIQ Ψ for wild type E18 . 5 mice ( top ) and Ptbp2 KO littermates ( bottom ) shows embryonic Ptbp2 represses adult specific inclusion of exon 20 , as previously reported ( Li et al . , 2014 ) , in addition to the switch in NAGNAG 3’ splice site use . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 02010 . 7554/eLife . 11752 . 021Figure 7—figure supplement 4 . Analysis of CAMK2D , CAMK2D , and CLTA LSVs in an independent Alzheimer’s cohort . ( A ) Boxplot showing distribution of E[Ψ] values and all E[Ψ] values ( dots ) for the most changing junction in the CAMK2D event examined in Figure 7 from a larger , independent cohort of normal and AD patients in the given brain sub regions . Two-tailed rank sum p-values are shown . ( B ) Same as ( A ) but for CAMK2G event examined in Figure 7—figure supplement 1 . ( C ) Same as ( A ) but for CLTA event examined in Figure 7—figure supplement 6 . Total samples analyzed for frontal pole normal and AD are 58 and 62; superior temporal gyrus normal and AD , 37 and 50; parahippocampal gyrus normal and AD , 33 and 45 . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 02110 . 7554/eLife . 11752 . 022Figure 7—figure supplement 5 . Complex alternative end of Alzheimer’s-associated Klc1 . ( A ) Splice graph representation of a complex alternative end LSV of Klc1 . Dark grey represents a 26 nt alternative 5’ss of exon 13 . ( B ) Top panel: RT-PCR validation with RNA from replicate cerebellar and adrenal tissues with isoforms illustrated on the left . Dark outlined isoforms are those that include the 26 nt alternative 5’ss of exon 13 . Bottom panel: PSI violin plots of MAJIQ quantification of junctions as colored in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 02210 . 7554/eLife . 11752 . 023Figure 7—figure supplement 6 . Clta splicing is developmentally regulated and dysregulated in Alzheimer’s Disease . ( A ) Splice graph for Clta and representation of target LSV . ( B ) Top panel: RT-PCR validation with RNA from replicate tissues with isoforms illustrated on the left . Bottom panel: PSI violin plots of MAJIQ quantification of junctions as colored in ( A ) . ( C ) Line graphs of MAJIQ Ψ quantification ( y-axis ) of junctions as in ( A ) across time points ( x-axis ) through cortex development . Points represent mean Ψ and error bars represent one standard deviation . ( D ) ΔΨ quantification comparing changes between control and Alzheimer’s patient brains of the homologous junctions illustrated in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 02310 . 7554/eLife . 11752 . 024Figure 7—figure supplement 7 . Eif4g3 splicing shows brain subregion-specificity and a novel exon in muscle . ( A ) Representation of complex source LSV in Eif4g3 . Red junction , red portion of exon 10 correspond to a novel alternative 3’ss detected in the brain . Purple junction , purple portion of exon 12 , and dashed exon 13 correspond to Ensembl annotated tandem cassette exons with no support in any experiment . Larger , pink portion of exon 12 corresponds to 120 nt , unannotated exon that is included with exon 11 in muscle . ( B ) Top panel: RT-PCR validation with RNA from replicate tissues with isoforms and expected product sizes illustrated on the left . Bottom panel: PSI violin plots of MAJIQ quantification of junctions as colored in ( A ) . ( C ) Line graphs of MAJIQ Ψ quantification ( y-axis ) of junctions as in ( A ) across time points ( x-axis ) through cortex development . Points represent mean Ψ and error bars represent one standard deviation . ( D ) UCSC genome browser view of the bounds of this LSV with mapped reads from representative muscle sample . Inset shows zoomed area of dashed lines corresponding to the location of the novel bounds of exon 11 . ( E ) RT-PCR from replicate muscle RNA using an additional primer set from exon 10 to 14 . DOI: http://dx . doi . org/10 . 7554/eLife . 11752 . 024 Several other important aspects of Camk2d splicing are accurately captured by MAJIQ . These include near 100% skipping of exons 21 through 23 in all non brain or muscle tissues ( known in the literature as isoform C or Camk2δC , ( Xu et al . , 2005 ) ) , high relative inclusion of NLS containing exon 21 in heart ( isoform B or Camk2δB ) , and high levels of isoform A ( Camk2δA ) , which includes exons 22 and 23 , in the brain regions examined ( Figure 7A ) . This result is consistent with previous reports of Camk2d splicing patterns and isoform A being neuronal-specific ( Xu et al . , 2005 ) . Importantly though , MAJIQ also detects isolated inclusion of exon 23 in the heart ( Figure 7A , green junction ) , which is supported by both the RT-PCR experiment and analysis of an independent dataset across heart development ( see below ) . Previous studies focused on splicing regulation of Camk2d in the heart used junction spanning primers that preclude detection of this highly utilized splicing choice ( Xu et al . , 2005; Ye et al . , 2015 ) . Because CAMK2 has been implicated in neurodevelopment and is proposed to be critical for postnatal heart development ( Xu et al . , 2005 ) , we next looked for developmental changes in LSVs by analyzing RNA-seq data derived from mouse cortices ( Yan et al . , 2015 ) and hearts ( Giudice et al . , 2014 ) at different time points . In the brain there is a switch in the splicing of Camk2d between the C and the A isoforms , reaching over 80% use of the A isoform by postnatal day 15 , corresponding to a time of intense synaptogenesis and plasticity ( Licatalosi et al . , 2012 ) ( Figure 7B , top ) . In the heart we see a more modest decrease in isoform C and increase in exon 23 only during postnatal heart development ( Figure 7B , bottom , compare purple with green ) , consistent with results from RT-PCR from eight week old mice ( Figure 7A ) . Notably , other CAMK2 subunits also displayed developmental dynamics in both tissues , such as inclusion of NLS containing exons in Camk2g and Camk2a ( Figure 7—figure supplement 1C and 2 ) , an unannotated mouse cassette exon in Camk2g regulated by the Rbfox family ( Figure 7—figure supplement 1D–G ) , and a complex LSV in the variable domain of Camk2b that affects autophosphorylation and is regulated by Ptbp2 ( Li et al . , 2014 ) ( Figure 7—figure supplement 3 ) . Given the suggested role of calcium signaling in neurodegeneration ( Marambaud et al . , 2009 ) and CAMK2 implication in Alzheimer’s disease ( AD ) ( Steiner et al . , 1990 ) , we also analyzed RNA-seq data from three control brains and compared them to three AD brains ( Bai et al . , 2013 ) . Strikingly , in CAMK2D we observe a marked decrease of ~38% of the neuronal specific isoform of the complex , developmentally-regulated mouse LSV we validated above , with reciprocal increase in the all exclusion , isoform C in AD brains ( Figure 7C ) . We also observe changes in a CAMK2G LSV that corresponds to an unannotated mouse exon ( Figure 7—figure supplement 1D , E ) . Importantly , these exons are perfectly conserved between mouse and human at the amino acid level , further suggesting physiologic importance of the novel splicing variations detected by MAJIQ . Finally , we validated that the observed CAMK2 splicing changes in AD brains can be reproduced in a second independent study . We used data from the AMP-AD Target Discovery Consortium ( doi:10 . 7303/syn2580853 ) involving a larger cohort of 157 samples from AD patient’s brains and 128 control samples , across three different brain sub regions ( Figure 7—figure supplement 4 ) . Overall , we detected approximately 200 LSVs that are reproducibly differentially spliced between AD and normal brains ( see Methods ) and enriched in GO terms such as cytoskeleton , GTPase regulator activity , and synapse organization ( data not shown ) . This set constitutes approximately 12% of the changing LSVs detected in the original dataset , a fraction that grows to 21% but only 164 LSVs if stricter filtering is applied to both datasets ( data not shown ) . This relatively low percentage of reproducible changes across the two datasets can be at least partially attributed to the small number of samples in the original study combined with an average of 1 . 8 fold lower coverage in the second , larger dataset . Notably though , among the reproducible set of differentially spliced LSVs 79 are complex , a significant , 1 . 2-fold enrichment compared to their relative proportion among all LSVs detected ( p=0 . 04 , binomial test ) . While the validation and experimental follow up on these LSVs is beyond the scope of this paper these results and the related CAMK2 analysis demonstrate the usefulness of our combined approach for LSV detection , quantification , and visualization for disease studies . Overall , our analysis of CAMK2 is in line with previous studies but also detects additional isoforms and exons that are conserved , developmentally regulated , and dysregulated in AD , making for a more accurate picture of CAMK2 splicing patterns . Additional complex LSVs we validated and analyzed include brain specific isoforms of the kinesin light chain Klc1 , recently shown to be an amyloid-beta accumulation modifier ( Morihara et al . , 2014 ) ( Figure 7—figure supplement 5 ) ; the clathrin light chain Clta , which displays developmental dynamics and dysregulation in both Alzheimer's disease cohorts ( Figure 7—figure supplement 6 , Figure 7—figure supplement 4 ) ; and the translation initiation factor scaffold Eif4g3 , which has high inclusion of a cassette microexon specifically in cerebellum and a novel , muscle-specific exon ( Figure 7—figure supplement 7 ) .
The work presented here spans a wide spectrum of topics from a new formulation of transcriptome variations in units of local splicing variations ( LSVs ) ; through algorithms for detecting , quantification and visualization of LSVs; a genome wide map of LSVs; analysis of the prevalence and functional significance of complex LSVs; to validation of several complex LSVs that affect protein domains in developmentally regulated genes with key roles in neurogenesis or other brain functions . For the latter , we also demonstrated dysregulation in Alzheimer’s disease using two independent datasets . The new formulation of LSVs sheds light on what has thus far been mostly a 'dark side' of the transcriptome and RNA-Seq based studies , i . e . complex splicing variations . Several previous works aimed to address the apparent representational gap between full transcripts and the classical binary AS events . For example , ( Nagasaki et al . , 2006 ) developed an efficient bit array representation for the various exonic segments that make up different gene isoforms , and ( Sammeth et al . , 2008 ) suggested an elaborate notational system that allowed them to catalogue all the splicing variations in a given transcriptome , comparing the frequencies of different AS types across 12 metazoa . More recently , ( Pervouchine et al . , 2013 ) developed bam2ssj , a package implementing a general intron centric approach to estimate AS from RNA-Seq data that can capture non classical AS variations . bam2ssj gives a BAM-file–processing pipeline that counts junction reads to compute the ratio of inclusion levels either from the 5’ or the 3’ end of an intron , denoted Ψ5and Ψ3 . A different , graph based , approach was taken by ( Hu et al . , 2013 ) where a splice graph is divided into subunits termed alternative splicing modules ( ASMs ) . ASMs are hierarchically structured , each capturing all the possible paths along a splice graph between specific start ( ‘single entry’ ) and end ( ‘single exit’ ) points . The matching algorithm , DiffSplice , then aims to identify cases of differential transcription of ASMs between two experimental conditions . All of these works differ substantially in the formulation of splicing variation , the underlying algorithms , and visualization approach , yet all share the effort to capture non classical AS types . In comparison , MAJIQ offers a unique approach that spans formulation , detection , quantification and visualization of splicing variations . Unlike ASMs , LSVs can be inferred directly from junction spanning reads and result in quantitative PSI and dPSI estimates , while MAJIQ’s probabilistic model offers significant accuracy boost for PSI and dPSI estimates compared to alternative methods . The importance of LSVs formulation is manifested in how common complex LSVs are in diverse metazoans , making up at least a third of observed LSVs in human and mouse . Complex LSVs are also enriched for regulated splicing when analyzing over thirty datasets across different tissues , developmental stages , splice factor knockdowns and neurodegenerative disease . In addition , LSV formulation can be used to investigate substructures of the transcriptome . We found that the biochemically-based proximity rule is commonly overcome at the genomic level and that complex LSVs are less likely to have a dominant splice junction . As for LSVs possible function , our results indicate that tissue dependent binary and complex LSVs both tend to occur in unstructured regions known to affect protein-protein interactions , as well as in specific yet distinct protein domains and families . In order to benefit from the new LSV formulation matching software is needed . The software we developed , MAJIQ , is LSV focused and compares favorably with available tools on AS quantification based both on RNA-Seq from biological replicates and on a compendium of over 200 RT-PCR experiments . Unlike many tools , MAJIQ supplements annotated transcriptomes with novel splice junctions , while VOILA allows the resulting LSVs to be interactively visualized within standard web browsers . Thus , MAJIQ and VOILA offer a compelling LSV centered addition to tools such as MISO ( Katz et al . , 2010 ) , rMATS ( Shen et al . , 2014 ) and cuffdiff ( Trapnell et al . , 2013 ) that allow users to quantify whole isoforms relative abundance , alternative polyadenylation , or differential expression . Immediate applications of the novel LSV framework and the MAJIQ software cover a wide spectrum . Examples include improved disease studies where transcriptome variations play a role , enhancing predictive models for splicing and for the effect of genetic variants , studying the regulatory underpinning of complex LSVs , and examining their evolutionary history . At the most basic level , our results illustrate the potential for novel discoveries in reanalyzing previously published data with the new LSV based methods . We anticipate the framework and resources provided here will form the basis of many additional new discoveries in diverse fields .
All RNA-Seq was mapped using STAR ( Dobin et al . , 2013 ) . STAR was run with alignSJoverhangMin 8 . We created the STAR genome based on mm10 or hg19 , with an in-house junction DB containing all possible junctions within each gene . An LSV ( local splice variation ) is defined as a split in a splice graph into or from a single exon , termed the reference exon . Single Source LSV ( SS-LSV ) correspond to splits from a reference exon to multiple 3’ splice sites in downstream exons , single target LSV ( ST-LSV ) correspond to multiple 5’ splice sites spliced to an upstream reference exon . The reference exon may include multiple 3’ splice sites ( ST-LSV ) or 5’ splice sites ( SS-LSV ) . An LSV type is defined by the reference exon type ( SS , ST ) and the set of junctions it includes . Each junction is defined by the splice site ID in the reference exon , and the splice site ID in its target/source exon . Under the above formulation some SS-LSV and ST-LSV may include exactly the same set of edges or one LSV may contain a subset of another LSV’s edges . For example the SS-LSV from exon 4 and the ST-LSV into exon 5 in Figure 1A bottom are comprised of exactly the same edges , while the ST-LSV into exon 2 is a subset of the SS-LSV from exon 1 . Such cases are easily detected and removed from further analysis to avoid redundancy . It is important to note that under the LSV formulation classical cassette exons correspond to two distinct LSVs , a single source and a single target . These LSVs are not redundant as they correspond to different lines of experimental evidence; one from junction reads connecting the alternative middle exon with the upstream exon ( SS-LSV ) and one connecting the middle exon to the downstream exon ( ST-LSV ) . The separate quantification for such LSVs , combined with the joint visualization using VOILA ( see below ) , help distinguish between cases where the two LSVs give similar PSI or dPSI quantifications and cases where they disagree . A case of possible disagreement is illustrated in the last three exons of Figure 1A , where an alternative transcription start site and a third junction going into the last exon may lead to different PSI values . The above definition gives a one-to-one mapping between a local splice graph split and an LSV type . Given a set of LSVs we can compute a distribution over their types or group several types together to detect a distribution over specific LSV features . We note that unlike the analysis of network motifs in Milo et al . ( 2002 ) , we do not compare LSVs to random connections in a network as the null hypothesis , but rather to a sequential network where all exons are connected via a single path . Thus , we compute a distribution over relevant statistics such as the number of junctions in the reference exon , the total number of junctions or the total number of exons in the LSV ( Figure 4 ) . MAJIQ is comprised of two main components , a builder and a quantifier . The builder analyzes a given set of RNA-Seq experiments and a transcriptome database to detect LSVs ( either known or de-novo ) and create a splice graph for each gene in the database . The quantifier subsequently estimates PSI or dPSI for LSVs detected by the builder . VOILA creates HTML5 based visualization of gene splice graphs , LSVs , PSI and dPSI estimates . It uses two types of input files: a binary file output from MAJIQ builder summarizing gene splice graphs , and another binary file from MAJIQ quantifier summarizing LSV PSI/dPSI quantifications . The HTML5 lists splice graphs and associated LSVs according to user defined filters . Distributions over PSI or dPSI are represented using violin plots and each splice graph and LSV is also linked to the UCSC genome browser to allow comparison to raw reads or other track information . Interactive filters allow users to select which types of LSVs to display while a table view allows users to sort and search LSVs . The VOILA splice graphs , LSVs cartoons and violin plots are shown in Figure 6 , 7 and their respective supplementary figures . The original VOILA plots used for these figures can be found at: majiq . biociphers . org . More information regarding VOILA usage and parameters can be found in the software's user guide , available at majiq . biociphers . org . PSI reproducibility by RNA-Seq from biological replicates was evaluated using the following procedure . First , MAJIQ Builder was executed to detect the union set of LSVs in a set of biological replicates of hippocampus and liver from Keane et al . ( 2011 ) . To avoid redundancy and enable comparison to other methods only a single junction from binary LSVs were included in downstream analysis . Next , for each replicate pair the difference in LSV quantification for each LSV was computed as R ( ΨMAJIQ ) = E[Ψr1]-E[Ψr1] . LSVs that were only detected in one of the replicates were discarded . The same set of LSVs were fed into MISO using the MAJIQ Builder GFF3 output file and the same procedure was executed to compute R ( ΨMISO ) . This procedure was repeated 6 times to compute the mean and standard errors for the empirical R ( Ψ ) PDF shown in Figure 2—figure supplement 1C . The empirical PDF and standard error for the difference in reproducibility ΔR = RMISO- RMAJIQ ( Figure 2—figure supplement 1C inset graph ) were computed by a similar procedure . PSI reproducibility by RT-PCR was evaluated using the following procedure . For the data from Zhang et al . ( 2014 ) , we first selected LSVs that were estimated by MAJIQ to be differentially spliced with high confidence ( P ( ΔΨ >0 . 2 ) > 0 . 95 ) ) when using three samples from cerebellum and liver . This allowed us to also assess dPSI reproducibility for a wide range of dPSI values ( see below ) . Next , for each LSV the total number of reads starting at positions within all the LSV’s junctions in each replicate were summed together for each tissue . Then , the LSVs were binned by the average total read coverage in the two tissues . Bins were defined to be: 10–30 , 30–40 , 40–80 , 80–200 , and above 200 reads . From each such bin , a set of LSVs was randomly selected for RT-PCR validation . Each RT-PCR was executed in triplicates ( see below ) . Finally , the average PSI by RT-PCR and the expected PSI by either MAJIQ or MISO were used to produce Figure 2 , and Figure 2—figure supplement 1 . MISO was executed with default parameters . For the stimulated and unstimulated T-Cell dataset , we collected a compendium of historical RT-PCR quantifications for previously annotated cassette exons . These experiments were executed by different Lynch lab members across several years and pre selected for specific studies regardless of dPSI or RNA-Seq coverage level . The vast majority of these cassette exons did not exhibit differential splicing between stimulated and unstimulated cells and some lacked triplicates . This set of previously annotated cassette exons was mapped to MAJIQ’s LVS and then quantified using RNA-Seq from Cole et al . ( 2015 ) ( Figure 2 , Figure 2—figure supplement 1 , circle shaped points ) . dPSI reproducibility by RNA-Seq from biological replicates was evaluated using the following procedure . First , the MAJIQ Builder was executed for all replicates of hippocampus and liver experiments from Keane et al . ( 2011 ) , yielding the union of all LSVs in these experiments . Next , for each liver and hippocampus pair of experiments , all quantifiable LSVs were ranked according to their E[ΔΨ] and the set of N LSVs with significant splicing changes at high confidence was defined as LSVs for which P ( ΔΨ >0 . 2 ) > 0 . 95 . This threshold was selected to be conservative , but see Figure 2—figure supplement 2A for more relaxed thresholds . This process was then repeated in another pair of experiments and the relative rank of the original set of N LSVs was recorded . The reproducibility ratio RR ( nN ) of any ranked LSVs subset n∈1…N was defined as the fraction n*N where n* is the subset of the first n ranked LSVs that were in the N best ranked LSVs by the replicate experiments . Similar to the IRD statistic used to assess reproducibility of Chip-Seq peak calling ( Li et al . , 2011 ) , a perfect RR ( n ) graph follows the diagonal line . Unlike IRD though , the RR statistic is invariant to small or even complete perturbations in the relative rank of the top ranked LSVs . Intuitively , this means that the RR will remain the same as long as the same subset n* makes the best N cutoff . It is important to note that the RR value can vary greatly , affected by biological , experimental , and technical factors . Nonetheless , one can use the RR to assess reproducibility in specific settings , or compare dPSI reproducibility by different algorithms under the same experimental setup . An inherent challenge in comparing MAJIQ to other methods is that MAJIQ quantifies LSVs while other methods quantify the classical AS event types . One complication as a result of that is that while redundant LSVs are removed ( see above ) different LSVs may still partially overlap . A good example for that are cassette exons . In the LSV formulation a differentially included cassette exon may have two LSVs that support it , corresponding to different lines of experimental evidence ( junction reads from the up and downstream exons ) but other methods/tools will only count this exon as a single event . This in turn may bias both the reproducibility ratio ( RR ) and detection power ( N ) in favor of MAJIQ . In our experiments , when we ignored such possible overlap of LSVs the reproducibility ratio remained the same but the number of differentially spliced LSVs detected was significantly higher ( RR=86% , N = 752 , data not shown ) . In order to avoid such a bias in favor of MAJIQ we implemented a conservative approach where the ranked LSVs are filtered so that no LSV contained overlapping exons with another LSV . We note this is a conservative filter as there may be complex LSVs that involve multiple differentially spliced exons , or cases where the same exone involves different variations ( e . g . skipping the exon but also alternative 3’ or 5’ splice sites ) . In such cases only a single LSV would pass that filter while the methods we compared to would still be able to retain separate AS events for those . dPSI reproducibility for MISO was evaluated by the following procedure . First , we followed MISO’s ( Katz et al . , 2010 ) guidelines for performing exon-centeric analysis ( i . e . AS events ) rather than whole transcripts analysis . For this , we used the set of alternative events for the mm10 mouse genome provided by MISO . We indexed the GFF3 file and ran MISO with default parameters on the same data pairs of experiments described above to compute expected dPSI ( E[ΔΨMISO] ) . Finally , we ranked LSVs by decreasing expected dPSI and computed the reproducibility ratio ( RR ) as described above . As MISO does not supply a statistical criteria for selecting the number of events ( N ) from its ranked list , we used the number produced by rMATS . Changing N to the higher number of LSVs detected by MAJIQ degraded MISO’s performance ( data not shown ) . dPSI reproducibility for rMATS was evaluated by the following procedure . We ran rMATS ( Shen et al . , 2014 ) with replicates ( groups ) and without them ( pairs ) , using ENSEMBL annotation file in GTF format . rMATS estimates differential expression for each one of the classic alternative splicing events it identifies from the annotation file ( exon skipping , 5 and 3 prime splice site donor/acceptor , mutually exclusive exons and intron retention ) . We used the default parameters except for the cutoff employed to compute the FDR associated to each AS event quantification , which was set to 0 . 2 ( see Figure 2—figure supplement 2A for the impact in reproducibility of different cutoffs ) . Lastly , we extracted the RR for confident changing AS events identified by rMATS ( FDR < 0 . 05 , P ( ΔΨ >0 . 2 ) as reported in the rMATS output file ) . dPSI reproducibility for the Naive Bootstrapping approach used in Xiong et al . ( 2015 ) for cassette exons was adopted for LSVs using the following procedure . First , we implemented the bootstrapping over junction positions described in Xiong et al . ( 2015 ) , with the same beta prior to avoid zero read counts . These samples gave an empirical distribution over possible PSI values and these were subsequently used to estimate the expected PSI . Similar to MISO , the Naive Bootstrapping approach does not assume a joint prior so that the expected dPSI estimates are simply the difference in the expected PSI in each experiment . The resulting expected dPSI was then used to rank the LSVs , filter them for possible overlap of exons , and compute RR as described above . dPSI accuracy by RT-PCR was evaluated by the same procedure as that described above for PSI . ΔΨRT was then computed as the difference between the average of each triplicate set of experiments in cerebellum and liver or the difference between previously recorded measurements in the Lynch Lab for the stimulated vs . unstimulated T-Cells . dPSI reproducibility by RT-PCR was defined as cases for which ΔΨRT >20% . This definition allowed assessing false positives and false negatives ( Figure 2—figure supplement 1B , Figure 2—figure supplement 2B ) . In order to construct the LSV junctions and protein features ( PF ) table we first built the union set of LSVs detected from Zhang et al . ( 2014 ) . We used ENSEMBL RESTful services [http://www . ncbi . nlm . nih . gov/pubmed/25236461] to retrieve PF along with their genomic coordinates associated with transcripts containing LSVs . Next , we annotated each LSV junction by PF that overlap its reference exon and its target/source exon , discarding junctions in non-coding regions . Because the overlap of a LSV junction region and a protein feature can be partial , we considered a PF to be associated with a junction when there was at least a 20% overlap . Lastly , we labeled as changing junctions those that had an estimated delta psi greater than 20% in any two tissues . We assessed relative enrichment of PFs using the following procedure when comparing groups of LSV junctions such as changing vs . unchanging , or binary vs . complex . For each PF we computed the p-value by Fisher’s Exact Test ( FET ) for its distribution between the two junction groups compared . To correct for multiple hypotheses testing while accounting for the high correlation between some PF we applied a permutation based testing procedure ( Column M ) . Specifically , we shuffled the labels ( e . g . changing , unchanging ) but controlled for the LSV origin of each junction . Thus , junctions from the same LSV were randomly switched with junctions from an LSV with the same number of junctions . This procedure guarantees that the number of labeled junctions remains the same , helps control for correlation between PF of junctions in the same LSV and for the distribution of LSV types . We repeated this process 10000 times and then calculate an empirical corrected FET p-value . The results from this analysis are included in Supplementary file 1 . RNA-Seq data for lizard and chicken was downloaded from Barbosa-Morais et al . ( 2012 ) ; opossum and chimp datasets were downloaded from Brawand et al . ( 2011 ) . RNA-Seq for human was downloaded from Illumina’s Body Atlas 2 . 0 ( NCBI GSE30611 ) . Transcriptomes were downloaded from Ensembl for lizard ( genome assembly AnoCar2 . 0 ) , chicken ( assembly Galgal4 ) , opossum ( assembly monDom5 ) , chimp ( assembly Pan_troglodytes-2 . 1 . 3 ) , mouse ( assembly GRCm38 . p3 ) and human ( assembly GRCh38 . p2 ) . For mouse and human RefSeq transcriptome annotations were used for comparison ( Figure 3 ) . The latest genome builds annotated in RefSeq were used , GRCm38/mm10 for mouse and GRCh37/hg19 for human . In order to assess the prevalence and potential enrichment of complex LSVs across additional datasets beyond the 12 mouse tissues , we analyzed a number of additional datasets shown in Figure 5A and Figure 5—source data 1 . All raw data was downloaded from SRA and mapped using STAR as described above . In a select number of older or low-coverage experiments , mapped reads from replicates were pooled together before analyzing with MAJIQ ( see 'Notes on processing' , Figure 5—source data 1 ) . MAJIQ dPSI was run for each comparison ( e . g . , tissue pairs , pairwise developmental time points , control versus altered splice factor expression ) . LSVs were considered differentially spliced if E[ΔΨ] > 20% . For datasets with multiple conditions ( e . g . 12 tissues , or multiple developmental timepoints ) , the union of differentially spliced LSVs and all detected LSVs between all pairwise comparisons was considered . The enrichment of complex LSVs in the differentially spliced group compared to their relative proportion among all detected LSVs in each dataset was evaluated using a binomial test , with a Bonferroni correction for the number of datasets used . All counts , SRA and GEO data accession numbers , and PubMed IDs for each study are detailed in Figure 5—source data 1 . To assess the distributions of PSI and dPSI across all datasets in Figure 5B and Figure 5—figure supplement 1 , we considered only the LSVs and junctions detected across the 12 mouse tissues and required exact matches to these junctions in the additional datasets in order to consider those PSI or dPSI values in the analysis . This conservative approach ensured we only monitored 'natural' LSVs and no LSVs that are unique to a specific cell line or KD experiment . In order to validate splicing changes in AD identified for the complex LSVs examined in this study ( CAMK2D , CAMK2G , and CLTA ) we took all differentially spliced LSVs we identified from the 3 healthy and 3 AD brains ( Bai et al . , 2013 ) and looked for similar changes in an independent , larger cohort . We used data from the Mount Sinai Brain Bank ( MSBB ) RNA sequencing study ( ID: syn3157743 , accessed at https://www . synapse . org/# ! Synapse:syn3157743 ) We focused on samples that came from healthy brains and definite AD brains , based on CERAD Neuropathology Criteria given , across the following brain regions: frontal pole ( healthy: n=58 , AD: n=62 ) ; superior temporal gyrus ( healthy: n=37 , AD: n=50 ) ; parahippocampal gyrus ( healthy: n=33 , AD: n=45 ) . Because overall coverage was lower in these datasets compared to the original cohort , which affects the ability to detect intron retention ( data not shown ) , we ran MAJIQ Builder on both datasets with a high threshold for IR detection ( --min_intronic_cov 1000 ) in order to only compare exonic LSVs . Additionally , to account for heterogeneity in the data and to save computational time we considered PSI values for each patient separately , as opposed to running all possible pairwise dPSI comparisons . An LSV that was changing in the first cohort was considered validated if in the MSBB cohort the distribution of PSI values for the most changing junction was significantly different between healthy and AD individuals in at least one brain subregion ( p<0 . 05 , two-tailed rank sum test ) with a difference in the median PSI of > 10% in the same direction as the original cohort . This lead to 199 LSVs in 145 genes changing in both cohorts . Finally , DAVID was used to find enriched GO terms among these genes with shared differentially spliced LSVs between the two cohorts using default parameters ( Huang et al . , 2008 ) . The conservation plots in Figure 5C were generated using the union of the changing LSV from the 66 pairwise tissue comparisons shown in Figure 4E . For each such LSV we extracted the phastCons60 conservation scores for vertebrates ( Siepel et al . , 2005 ) for the first 50 positions in each exon and the first 300 intronic positions proximal to each exon . It is not immediately clear which of the variable regions in a complex LSV ( left hand plot for the single source LSVs , right hand plot for the single target LSVs in Figure 5C ) should be included in such conservation analysis . Previous work focused on binary cases of cassette exons and compared constitutive exons to the regions around the alternative exon , which tend to be more conserved . Since the focus of this analysis was on conservation of possible regulation in LSV units we chose to apply the max function for each position in such variable LSV regions . To partially correct for the possible bias for high scores that the max operation may introduce we also applied it to the binary LSVs and to randomly selected sets of K constitutive exons , where the size K is sampled based on the distribution of number of exons in LSVs ( Figure 4C ) . Overall we sampled 5000 such sets for the constitutive regions plot . Finally , the lines in Figure 5C were smoothed using a 5 bases sliding window . Total RNA was extracted from mouse tissues as described previously ( Zhang et al . , 2014 ) . For each tissue three samples corresponding to RNA from circadian times 31 , 41 , and 53 were used for validation . For additional validations we used total RNA extracted from a clonal Jurkat T cell line ( JSL1 , described in detail previously [Lynch and Weiss , 2000] ) cultured in RPMI medium supplemented with 5% heat-inactivated fetal bovine serum ( unstimulated ) or the same growth medium supplemented with the phorbol ester PMA ( Sigma-Aldrich , St . Louis , MO ) at a concentration of 20 ng/mL ( stimulated ) . Stable identify of this clonal line is continuously monitored by assessing hallmark changes in splicing induced by PMA ( Cole et al . , 2015; Martinez et al . , 2012; Shankarling et al . , 2013 ) . Low cycle reverse transcription-PCR ( RT-PCR ) was performed on 0 . 5 micrograms of RNA as described previously in detail ( Rothrock et al . , 2003 ) using sequence specific primers . Gels were quantified by densitometry with the use of a Typhoon PhosphorImager ( Amersham Biosciences , UK ) . Primers and expected size of products for all events are given in Supplementary file 2 . For cassette exon LSVs percent spliced in was calculated as the percent of isoforms including the alternative exon over the total inclusion and exclusion isoforms . For complex LSVs , each band present on the gel was quantified . The percent selected index ( PSI ) for each junction of an LSV was calculated as the isoform ( s ) including that junction over the total isoforms present . For example , for the Camk2g source LSV ( Figure 1C ) the percent usage of the red junction that goes from the reference source exon 14 to exon 15 corresponds to the sum of the bands corresponding to the 214 nt isoform that includes exon 15 alone and the 256 nt isoform that includes both exons 15 and 16 . MAJIQ and VOILA are available for download at majiq . biociphers . org | Genes contain coded instructions to build other molecules that are collectively referred to as gene products . Building these products requires the gene’s instructions to be copied into a molecule of RNA in a process called transcription . Over 90% of human genes undergo a process by which different segments of the transcribed RNA molecule are either removed or retained . This process , termed alternative splicing , results in a single gene encoding different gene products that can perform in different ways . Alternative splicing can also mean that gene products vary between different cells , tissues and individuals . Some of these variations can be harmful and lead to disease . However , it is difficult with current methods to accurately identify variations in gene products that are due to alternative splicing and see how these products differ between groups of people , such as patients and healthy controls . Vaquero-Garcia , Barrera , Gazzara et al . have now developed new methods to define , measure and visualize the variations in RNA gene products . First , splicing variations were catalogued across a range of species from lizards to humans , which revealed that some fairly complicated variations were much more common than previously appreciated . These complex variations had not been studied much before , but the new methods showed that they make up a third of the variations in the RNA products copied from human genes . Vaquero-Garcia , Barrera , Gazzara et al . then showed that the new methods are more accurate and sensitive than previous methods , and can be used to discover splicing variations that were previously unknown . For example , applying the new methods to data collected in other studies revealed variations in genes that are important for brain development and activity . Further analysis then showed that these variations were also altered in brain samples from patients with Alzheimer disease . The new methods developed by Vaquero-Garcia , Barrera , Gazzara et al . can now shed new light on gene product variations , especially the more complex ones that have not been studied before . The next challenge is to use these tools to better understand the regulation and purpose of splicing variants and how they can contribute to diseases in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"computational",
"and",
"systems",
"biology"
] | 2016 | A new view of transcriptome complexity and regulation through the lens of local splicing variations |
Specific cell shapes are fundamental to the organization and function of multicellular organisms . Fibroblast Growth Factor ( FGF ) signaling induces the elongation of lens fiber cells during vertebrate lens development . Nonetheless , exactly how this extracellular FGF signal is transmitted to the cytoskeletal network has previously not been determined . Here , we show that the Crk family of adaptor proteins , Crk and Crkl , are required for mouse lens morphogenesis but not differentiation . Genetic ablation and epistasis experiments demonstrated that Crk and Crkl play overlapping roles downstream of FGF signaling in order to regulate lens fiber cell elongation . Upon FGF stimulation , Crk proteins were found to interact with Frs2 , Shp2 and Grb2 . The loss of Crk proteins was partially compensated for by the activation of Ras and Rac signaling . These results reveal that Crk proteins are important partners of the Frs2/Shp2/Grb2 complex in mediating FGF signaling , specifically promoting cell shape changes .
During the development of complex multicellular organisms , changes in epithelial cell morphology are essential for the tissue-specific cell differentiation patterning that leads to the subsequent formation of functional organs ( Settleman and Baum , 2008 ) . This is particularly clear when considering the formation of the ocular lens , which has served as a model system to delineate many developmental pathways ( Cvekl and Zhang , 2017; Gunhaga , 2011 ) . The development of the mouse lens begins at embryonic day 9 . 5 when the optic vesicle extends toward the presumptive lens ectoderm , inducing the latter to thicken into a cuboidal layer of epithelial cells commonly referred to as the lens placode . At E10 . 5 , the cells making up the lens placode undergo apical constriction to form the lens pit , which eventually closes at its anterior surface to form the lens vesicle ( Chauhan et al . , 2011 ) . After the newly differentiated primary fiber cells extend anteriorly from the posterior region of the lens vesicle , the anterior epithelial cells begin to migrate posteriorly towards the equatorial region of the lens , at which point they begin to differentiate into secondary fiber cells that continue to occupy the space within the lens interior . This process is accompanied by up to a 1000-fold increase in the length of the newly formed secondary fiber cells , which coordinate with the primary fiber cells to organize an elegant concave pattern that maintains the structural integrity and transparency of the mature lens ( Bassnett , 2005; McAvoy et al . , 1999; Sue Menko , 2002 ) . Previously proposed models that have been considered for the mechanism by which lens fiber cells elongate consist of microtubule reorganization , increased cell volume and actin dynamics ( Audette et al . , 2017 ) . Microtubules are prominent components of the cytoskeleton lining the plasma membrane of lens fiber cells . They are oriented longitudinally along the lens with their minus end towards the anterior pole and their plus ends facing the posterior ( Byers and Porter , 1964 ) . which is consistent with the presence of microtubule organizing centers at the apical ends of lens fiber cells ( Lo et al . , 2003 ) . However , Beebe et al reported that Nocodazole inhibition of microtubule polymerization failed to disrupt fiber cell elongation in chick lens explants ( Beebe et al . , 1979 ) . Instead , they proposed that the expansion of cell volume was the key mechanism by which fiber cells elongated to form the functional mature lens . However , utilizing precise measurement techniques , it was later found that there was no apparent increase in lens volume in vivo at the onset of fiber cell elongation ( Bassnett , 2005 ) . The differentiation of lens epithelial cells into lens fiber cells is also associated with the assembly of actin filaments beneath the cortical membrane ( Weber and Menko , 2006 ) . When this cortical actin structure was inhibited by cytochalasin D in lens explants , both the differentiation and elongation of lens cells were blocked . Nevertheless , neither the disruption of the cell adhesion molecules N-cadherin or β1-integrin nor the ablation of the actin regulators Rac1 and Rho completely prevented the lengthening of lens fiber cells ( Logan et al . , 2017; Maddala et al . , 2011 , 2015; Pathania et al . , 2016; Pontoriero et al . , 2009 ) . Consequently , how the actin cytoskeleton is controlled during fiber cell elongation has remained an open question . Fibroblast Growth Factor ( FGF ) signaling is known to play an important role in lens cell differentiation and elongation . Transgenic expression of human FGF-1 using an αA-crystallin promoter was found to induce lens epithelial cells to acquire elongated shapes and fiber cell characteristics ( Robinson et al . , 1995 ) . In explant cultures , FGF was also found to promote lens fiber cell differentiation and elongation in a dose dependent manner ( Lovicu and McAvoy , 2001; McAvoy and Chamberlain , 1989 ) . Conversely , genetic ablation of FGF receptors leads to a complete loss of lens cell differentiation and elongation ( Zhao et al . , 2008 ) . Several proteins have been implicated in the direct engagement with these active FGF receptors , including Frs2/3 , Grb14 , Shb , PLCγ and Crk ( Brewer et al . , 2015; Klint and Claesson-Welsh , 1999 ) . Of particular interests are Crk and the related protein Crkl , which are mammalian homologs of the viral Crk oncogene that prossess the ability to promote the tyrosine phosphorylation of cellular proteins ( Feller , 2001 ) . Lacking intrinsic tyrosine kinase activity , the Crk family of proteins act as adaptors that transduce signals from upstream phosphotyrosine-containing proteins to downstream SH3-interacting partners ( Birge et al . , 2009 ) . Biochemical studies have shown that FGF2-stimulated endothelial cell proliferation is dependent on the binding of Crk to the phosphorylated tyrosine residue 463 in FGFR1 ( Larsson et al . , 1999 ) . In line with this finding , Crk null mice display some of the cardiovascular and cranial features of Noonan syndrome , which is caused by aberrant Ras-MAPK signaling ( Park et al . , 2006; Roberts et al . , 2007; Schubbert et al . , 2006; Tartaglia et al . , 2001; Tartaglia et al . , 2007 ) . Crkl was also identified as a component of an FGF8-induced feed forward loop , resulting in anchorage-independent cell growth ( Seo et al . , 2009 ) . Consistent with this , the human CRKL gene lies within the chromosome 22q11 deletion region that causes DiGeorge syndrome , which shares the pharyngeal and cardiac defects seen in Fgf8-deficient mice ( Moon et al . , 2006 ) . Despite these findings implicating Crk and Crkl in FGF signaling , a recent study has shown that mutating their putative Y463 binding site in Fgfr1 did not produce any observable phenotype in transgenic mice ( Brewer et al . , 2015 ) . Therefore , the potential role and mechanism of Crk proteins in FGF signaling remain uncertain . In this study , we showed that the lens specific knockout of Crk and Crkl disrupted lens fiber cell elongation without affecting differentiation , suggesting that lens cell morphogenesis can be uncoupled from differentiation during development . FGF loss- and gain-of-function experiments demonstrated that Crk proteins act downstream of FGF signaling to enhance ERK phosphorylation . Contrary to the previous belief that Crk proteins directly bind to the Fgfr , we found that mutating the purported Crk docking site on Fgfr1 failed to perturb lens development or Crk phosphorylation . Instead , our data showed that Crkl was recruited to the Frs2/Shp2/Grb2 complex after FGF stimulation . Crk/Crkl deficient animals phenocopied Rac1 but not Rap1 mutants , and activation of Rac1 and Ras signaling partially reversed the observed lens elongation defects caused by the deletion of Crk and Crkl . These results show that the Crk family of adaptor proteins are essential partners of the Frs2/Shp2/Grb2 complex that forms during FGF signaling , and are specifically required for stimulating the actin reorganization that is necessary for the morphological shaping of lens cells .
We observed that Crk and Crkl proteins displayed a restricted localization pattern in the lens . At E10 . 5 , Crk and Crkl were predominantly confined to the apical side of the lens vesicle ( Figure 1A , arrows ) , away from the basal side where integrins interact with the basement membrane ( Figure 1A , dotted lines ) . By contrast , Crk and Crkl exhibited a more diffuse pattern at E12 . 5 when the posterior lens vesicle cells gave rise to the primary lens fibers ( Figure 1A ) . However , by E14 . 5 , Crk and Crkl were specifically enriched in the transitional zone where the lens epithelial cells begin to differentiate and elongate into the secondary lens fiber cells ( Figure 1A , arrowheads ) . Using an antibody that recognizes the phosphorylated forms of both of these proteins , we were able to observe that the phosphorylation of Crk and Crkl also mainly occurs in the transition zone of the lens at this stage of development ( Figure 1B , arrowheads ) . These results suggest that Crk activity is under dynamic regulation as the lens cells undergo successive morphological changes during development . We next ablated Crk genes using Pax6Le-Cre , also known as Le-Cre , which is initially active in the lens placode and later in the lens epithelium ( Ashery-Padan et al . , 2000 ) . As expected , this resulted in the loss of both Crk/Crkl and pCrk/pCrkl in the Pax6Le-Cre;Crkflox/flox;Crklflox/flox ( CrkCKO ) lens after E10 . 5 ( Figure 1A and B , dotted line ) . Although deletion of either Crk or Crkl alone did not perturb lens development ( Figure 1—figure supplement 1A–C ) , the CrkCKO lens displayed a reduction in lens size , rotation of the lens epithelial layer , and disorganization of the lens fiber cells at E14 . 5 ( Figure 1C , arrow ) . Using F-actin staining to better delineate individual cell shapes , we observed a significant reduction in the length of the lens fiber cells ( Figure 1C ) . In addition to these fully penetrant lens phenotypes , the neural retina was often observed to aberrantly protrude toward the diminished lens . This observation is consistent with the known role of the properly developed lens in the correct placement of the retina ( Figure 1A , asterisks ) ( Ashery-Padan et al . , 2000 ) . During lens development , the transcription factors Pax6 and Prox1 are of vital importance for the normal differentiation of the transparent lens . Pax6 controls lens induction and fate determination while Prox1 regulates fiber cell differentiation and crystallin expression ( Ashery-Padan et al . , 2000; Audette et al . , 2016 ) . Interestingly , notwithstanding the severe morphological defects including the frequent ventral rotation of the lens epithelial layer , Prox1 , Pax6 and multiple forms of crystallin ( α , β , γ ) were still expressed ( Figure 2A and B ) . In addition , the polarity of the lens fiber cells was also preserved as evidenced by the localization of Zo-1 and β1 integrin to the apical and basal sides of the lens , respectively ( Figure 2C , arrowheads ) . However , there were both a reduction in cell proliferation and an increase in apoptosis in the lens epithelial cells as shown by Ki67 and TUNEL staining ( Figure 2D and E , arrowheads ) , which likely accounted for the diminished lens epithelial layer ( Figure 2A and B , arrowheads ) . Collectively , these data show that Crk proteins are dispensable for lens differentiation and polarity , but are essential for proliferation , survival and elongation of the cells that make up the structurally mature lens . To investigate the role of Crk proteins in FGF signaling , we took a three-prong approach , combining in vitro , in vivo and ex vivo experiments . First , we performed loss-of-function experiments to examine whether FGF signaling is required for Crk protein activity . Consistent with previous studies showing that FGF signaling is essential for early lens development ( Garcia et al . , 2011 ) , we observed that genetic ablation of Fgfr1/2 in Pax6Le-Cre;Fgfr1flox/flox;Fgfr2floxflox ( FgfrCKO ) mutants disrupted lens vesicle invagination at E10 . 5 ( Figure 3A ) . Although the mutant lens cells still expressed Pax6 , the lens specific marker αA-crystallin was not induced . Importantly , whereas the control embryos displayed the phosphorylation of Erk and Crk in the invaginating lens vesicle , these staining patterns were absent in FgfrCKO lens cells ( Figure 3A ) . To further corroborate these results in vitro , we isolated primary mouse embryonic fibroblast ( MEF ) cells from Fgfr1flox/flox;Fgfr2floxflox embryos and infected them with a Cre-expressing adenovirus to ablate Fgfr1 and Fgfr2 . Unlike wild type controls , these mutant MEF cells failed to elevate the level of pCrk/Crkl and pErk upon FGF stimulation , demonstrating that both Crk and Erk proteins are under the tight regulation of FGF signaling ( Figure 3B ) . To probe the relationship between Crk and Erk , we next infected Crkflox/flox;Crklfloxflox MEF cells with the Cre adenovirus to deplete Crk proteins ( Figure 3C ) . As a result , both the intensity and duration of FGF-stimulated Erk phosphorylation were down regulated , suggesting that Crk proteins modify the quantity of FGF-ERK signaling by elevating and prolonging ERK activation . Consistent with this , we noticed that Erk phosphorylation was prominent in the elongating primary lens fiber cells at E12 . 5 and in the transitional zone of the lens at E14 . 5 ( Figure 3D ) . In the CrkCKO mutant lens , however , pERK staining was significantly reduced ( Figure 3D ) . Together , this data shows that Crk proteins regulate FGF-induced Erk activation in the developing lens . The second approach we took to probe the role of Crk genes in FGF signaling was based on gain-of-function experiments . We utilized two Fgf3 transgenes that are driven by the αA-crystallin promoter to target the lens ( Robinson et al . , 1998 ) , which led to an anterior expansion of pErk staining ( Figure 4C ) . The Fgf3OVE393A line displayed excessive elongation of the lens fiber cells that protruded through the corneal epithelium , while the Fgf3OVE391 line showed a more modest enlargement of the lens ( Figure 4A ) . After crossing these mice with Crk mutants , however , lens abnormalities in both lines were suppressed and ectopic phospho-Erk staining was abrogated . The length of the fiber cells was reduced to the same size as those seen in the CrkCKO mutants , demonstrating that Crk genes were necessary for the induction of fiber cell elongation by FGF ( Figure 4B ) . Fgf3 overexpression also led to premature exiting of the cell cycle as indicated by the loss of the cell proliferation marker Ki67 ( Figure 4C , arrowheads ) and an increase in expression of the cell cycle inhibitor p57 ( Figure 4C , arrows ) . As a result , the anterior lens epithelial cells differentiated prematurely to express the fiber cell markers Jag1 and C-Maf at the expense of the epithelial cell markers E-cadherin and Foxe3 . Interestingly , the cell cycle and differentiation abnormalities were not rescued after crossing the Fgf3 transgenic mice with CrkCKO mutants . These genetic epistasis experiments further highlighted that the specific functionality of Crk genes is in mediating FGF signaling for lens cell elongation but not differentiation . In the third approach , we used the mouse lens explant system to test the direct role of Crk proteins in FGF-induced fiber cell elongation ( Korol et al . , 2014 ) . In this assay , the explant lens isolated from P3 mouse embryos carrying Crkflox/flox;Crklfloxflox alleles was infected with a Cre-expressing adenovirus to achieve acute genetic ablation . This avoids any potential complication that may stem from defects in early lens development or other compartments of the eye . Using a ROSAmTmG mouse line that switches the reporter expression from membrane TdTomato to membrane GFP upon Cre mediated recombination , we showed that a 2 day incubation period with the virus was sufficient to induce genetic changes in all lens epithelial cells ( Figure 4D ) . In control explants , β-catenin staining revealed a robust elongation of the lens epithelial cells after FGF2 exposure . By contrast , the lens epithelial cells in Crkflox/flox;Crklfloxflox explants treated with the Cre adenovirus retained the epithelial specific hexagonal shape without any obvious signs of elongation . Taken together , these three lines of evidence established that Crk proteins are essential mediators of FGF signaling whose specific function relates to the fiber cell elongation that occurs during lens development . Previous studies have suggested that Crk proteins bind directly to the phosphorylated tyrosine-463 residue on Fgfr1 to mediate downstream signaling ( Larsson et al . , 1999 ) . Surprisingly , mice carrying a Y463F mutation in Fgfr1 ( Fgfr1Crk ) lacked any observable phenotypes and were reported to be both viable and fertile ( Brewer et al . , 2015 ) . The lack of abnormality in Fgfr1CRK mice prompted us to examine whether it was due to the compensatory functions of other Fgf receptors . We isolated MEF cells from Fgfr1Crk;Fgfr2flox mice and removed Fgfr2 by Cre-mediated recombination in vitro . Although we have shown above that genetic inactivation of Fgfr1 and Fgfr2 together was sufficient to abrogate FGF signaling in MEF cells ( Figure 3B ) , FGF was still able to induce phosphorylation of Crk/Crkl and Erk in Fgfr1CRK MEF cells after the deletion of Fgfr2 ( Figure 3—figure supplement 1A ) . We next investigated the requirement of the Y463 residue of Fgfr1 for Crk signaling in lens development . For this purpose , we genetically ablated all FGF receptors ( Fgfr2 , Fgfr3 and Fgfr4 ) with the exception of Fgfr1 in the mouse lens , resulting in only a modest reduction in lens size ( Figure 3—figure supplement 1B ) . Even in such a stringent genetic background , a homozygous Fgfr1Crk mutation did not further worsen the lens phenotype or disrupt pCrk staining . These results show that the putative Y463 Crk binding site on Fgfr1 is dispensable for FGF-dependent lens development and Crk signaling . Since our biochemical and genetic data did not support a functional role for the direct interaction between FGF receptors and the Crk protein , we next explored whether FGF signaling may engage Crk indirectly through intermediaries . Frs2 is a myristylated protein located at the plasma membrane that binds to Fgf receptors specifically at a juxtamembrane site . With multiple tyrosine residues phosphorylated by activated Fgf receptors , Frs2 acts as a nexus of FGF signaling by presenting easily accessible docking sites for the phosphotyrosine-binding proteins Shp2 and Grb2 , which in turn activate Ras-MAPK signaling . Taking advantage of a mutant Fgfr1 allele ( Fgfr1ΔFrs ) that lacks the Frs2 binding site ( Hoch and Soriano , 2006 ) , we showed that formation of the lens vesicle was indeed disrupted in Pax6Le-Cre;Fgfr1flox/ΔFrs;Fgfr2flox/flox embryos ( Figure 5A ) , which resembled the phenotype of FgfrCKO null mutants ( Figure 3A ) . Importantly , we observed that the loss of the Fgfr-Frs2 interaction also abrogated the phosphorylation of Crk proteins . This was further confirmed in vitro using Fgfr1flox/ΔFrs;Fgfr2floxflox MEF cells infected with a Cre adenovirus . The resulting Fgfr1-/ΔFrs;Fgfr2-/- MEF cells failed to elevate the levels of pERK or pCrk/Crkl in response to FGF2 stimulation ( Figure 5B ) , demonstrating that the Frs2-binding site on Fgfr1 is necessary for Crk signaling . The above results suggested that Frs2 may be the adaptor protein that recruits Crk to Fgf receptors . To test this idea , we first examined Crk phosphorylation in MEF cells that were mutated by the Cre-mediated deletion of Frs2 and Shp2 . As shown in Figure 5C , phosphorylation of Crk proteins induced by FGF was significantly reduced in Frs2 null MEF cells . Moreover , in MEF cells lacking the Frs2 interacting protein Shp2 , even the normally seen basal levels of pCrk were lost . We have previously showed that both Frs2 and its binding partner Shp2 have relatively slow turnover rates in vivo ( Li et al . , 2014 ) . As a result , the conditional inactivation of Frs2 or Shp2 alone exhibited only a modest lens phenotype . Combined deletion of Frs2 and Shp2 in Le-Cre;Frs2flox/floxs;Shp2floxflox ( Frs2CKO;Shp2CKO ) embryos , however , did block lens development at E14 . 5 . At this stage , the control lens displayed a significant accumulation of Shp2 protein in the transitional zone of the lens , which is also the area of maximum pErk and pCrk/Crkl staining ( Figure 5D , arrows ) . In contrast , neither Shp2 nor the pCrk/Crkl proteins were detectable in the Frs2CKO;Shp2CKO lens . Notably , an in vitro ablation of either Crk or Shp2 reduced the Erk phosphorylation that is normally induced by FGF ( Figure 3C and Figure 5D ) . Consistent with this , both CrkCKO and Shp2CKO embryos at E12 . 5 continued to display residual levels of pErk staining in the lens ( Figure 5E ) . In CrkCKO;Shp2CKO mutants , however , pErk staining was entirely abolished , suggesting that Shp2 and the Crk family of proteins act synergistically to regulate Erk signaling . Grb2 is another binding partner of Frs2 that is essential for the activation of Ras-MAPK signaling . In Grb2-depleted MEF cells , we observed a similar reduction in Crk and Erk signaling in response to FGF stimulation ( Figure 5F ) . The attenuation of Crk phosphorylation in Frs2 , Shp2 and Grb2 deficient cells raises the possibility that Crk proteins are recruited to the Fgf receptors via the Frs2-Shp2-Grb2 complex . To probe the physical interaction between these proteins , we transfected NIH-3T3 cells with a TAP-Crkl construct that encodes the Crkl protein conjugated to a tandem affinity purification ( TAP ) tag used for purification ( Hallock et al . , 2015 ) . When the concentration of FGF was increased , we observed a shift of Frs2 mobility in our immunoblot analysis that is representative of the phosphorylated form of Frs2 being generated as has been previously reported ( Figure 5G ) ( Kouhara et al . , 1997 ) . Consistent with this , there was also an increase in the levels of pErk present in the cell lysates . Interestingly , immunoprecipitated TAP-Crkl only pulled down the phosphorylated form of Frs2 induced by FGF stimulation , which was also accompanied by Shp2 and Grb2 . Overall these results indicate that the Frs2-Shp2-Grb2 complex is responsible for recruiting Crk proteins to the activated Fgf receptor . Crk proteins have been implicated in the activation of the small molecular GTPases Rac1 and Rap1 , which play important roles in cell adhesion , cytoskeletal rearrangement and cell shape changes ( Birge et al . , 2009; Feller , 2001 ) . Previous studies have reported that the conditional deletion of Rac1 in the lens resulted in impaired actin polymerization that lead to a morphologically impaired lens ( Maddala et al . , 2011 ) . Since Rac2 has also been reported to be expressed in the lens ( Rao et al . , 2004 ) , we decided to inactivate both Rac1 and Rac2 to compare their phonotype with that of the Crk mutants . Immunofluorescence confirmed the specific depletion of Rac1 in the Pax6Le-Cre;Rac1flox/flox;Rac2-/- ( RacCKO ) lens ( Figure 6A–B ) . Unlike the CrkCKO lens , however , the RacCKO lens did not display a decrease in pErk staining ( Figure 6C–D ) . Nonetheless , depletion of Rac proteins resulted in a lens fiber cell elongation defect , albeit milder than that seen in the CrkCKO lens ( Figure 6E–G ) . To further explore the relationship between Rac and Crk proteins , we utilized the R26-Rac1LSL-G12V allele that results in the expression of a constitutively active form of Rac1 ( Rac1CA ) after the Cre mediated excision of its STOP cassette ( Srinivasan et al . , 2009 ) . In the Pax6Le-Cre;Crkflox/flox;Crklfloxflox;R26-Rac1LSL-G12V ( CrkCKO;Rac1CA ) lens , there was a statistically significant increase in fiber cell length as compared to that of the CrkCKO lens ( Figure 6H and U ) . The partial rescue of the CrkCKO phenotype by Rac1 activation supports the idea that Rac1 is a downstream effector of Crk in the signaling pathways that lead to lens fiber cell elongation . Rap1 is also known to be targeted by Crk proteins via activation of the guanine nucleotide exchange factor ( GEF ) , C3G , and FGFR1 has previously been reported to activate Rap1 in endothelial cells ( Quilliam et al . , 2002; Yan et al . , 2008 ) . We reasoned that if Rap1 was a downstream effector of Crk proteins in FGF signaling , Rap1 and Crk deficient lenses should phenocopy each other as well . In agreement with a previous report ( Maddala et al . , 2015 ) , we found that the conditional knockout of two Rap1 genes ( Pax6Le-Cre;Rap1aflox/flox;Rap1bflox/flox ( Rap1CKO ) ) disrupted the epithelial polarity of the lens , leading to an ectopic expression of the epithelial-mesenchymal-transition ( EMT ) marker smooth muscle actin ( SMA ) in lens epithelium cells ( Figure 6I–L ) . Due to cell adhesion defects , the posterior lens fiber cells also failed to remain attached to the anterior epithelial cells as is represented by a noticeable gap in the anterior part of the lens . In contrast , neither of these phenotypes were observed in the CrkCKO mutant lens ( Figure 6M and N ) , which instead displayed a more severe lens fiber cell elongation defect than the Rap1CKO mutants . We also considered the possibility that Rap1 and Rac may be functionally redundant in the developing ocular lens . The combined deletion of these four genes in Le-Cre;Rac1flox/flox;Rac2-/-;Rap1aflox/flox;Rap1bflox/flox ( RacCKO;Rap1CKO ) embryos , however , did not produce any lens differentiation abnormalities or enhance the fiber cell elongation defects previously observed in the RacCKO mutants ( Figure 6O–U ) . Taken together , these results argue against the Rap1 family of proteins being downstream targets of Crk signaling during lens development . The above experiments showed that a constitutively active form of Rac1 resulted in only modest attenuation of the lens fiber cell elongation defects seen in the CrkCKO mutant , suggesting that there exist additional downstream effectors of the Crk proteins that are important in regulating fiber cell shape . As we have observed a significant downregulation of pErk in the CrkCKO mutant lens , one potential candidate is the Ras-MAPK signaling pathway . We subsequently crossed the CrkCKO mutant with a KrasLSL-G12D mouse , which harbors a Cre-inducible allele of the oncogenic G12D mutated Kras ( Tuveson et al . , 2004 ) . Because this mutant is expressed from the endogenous Kras locus , it is expected to activate Ras signaling at a normal physiological level as opposed to being overexpressed . Consistent with this , we observed a modest increase in pERK staining in the Pax6Le-Cre;Crkflox/flox;Crklflox/flox;KrasLSL-G12D ( CrkCKO;KrasG12D ) lens ( Figure 7A–C ) . Although TUNEL staining showed that the cell apoptosis defect was not rescued in the CrkCKO; KrasG12D lens , the number of proliferative Ki67 positive cells increased significantly , which can likely explain the extension of the lens epithelium as evidenced by E-cadherin staining ( Figure 7D–L ) . Importantly , the length of the lens fiber cells in the CrkCKO;KrasG12D lens increased to about 90% of that of the control lens , demonstrating that the activation of Ras signaling can largely rescue the fiber elongation defects caused by the loss of Crk proteins ( Figure 7M–P ) . This result provides strong genetic evidence that Ras signaling plays an essential role in Crk-mediated lens cell shape changes .
The ocular lens is derived from a single cell type that undergoes an orderly set of successive differentiation change that are represented by various cell shape alterations occurring along the developmental timeline . This makes it an excellent model for studying the cell signaling pathways involved in these morphological processes . Although FGF signaling is known to regulate both lens fiber cell differentiation and elongation , it is unclear whether these two functions can be separated at the mechanistic level . In this study , we successfully showed that the Crk family of adaptor proteins specifically mediate FGF signaling in the control of lens cell shape but not differentiation . Contrary to previous claims , we demonstrated that the putative Crk binding site on Fgfr1 is dispensable . Rather , the recruitment of Crk proteins to the Frs2-Shp2-Grb2 complex via the Frs2 docking site on Fgf receptors was observed . Our study further showed that the downstream Crk effectors involved in regulating the precise cytoarchitecture of the lens are primarily Ras and to a lesser extent Rac1 , but not Rap1 ( Figure 7Q ) . These results identify Crk proteins as the essential adapters that link FGF signaling to cytoskeletal dynamics . This study also demonstrated that Crk and Crkl play essential overlapping functions in embryonic lens development . Whereas lack of either Crk or Crkl did not disrupt lens development in the mouse embryos , combined deletion of Crk and Crkl led to a profound defect in lens fiber cell elongation . There is a substantial overlap between Crk and Crkl in many biological functions including the Reelin pathway ( Park and Curran , 2008 ) , neuromuscular synapse formation ( Hallock et al . , 2010 ) , migration of T cells to sites of inflammation ( Huang et al . , 2015 ) and podocyte morphogenesis ( George et al . , 2014 ) . These studies and ours confirm that single and double floxed mice for Crk and Crkl are crucial models for investigating the distinct and overlapping biological functions of these closely related proteins . Crk and Crkl are versatile adaptor proteins that can interact with a wide spectrum of signaling molecules ( Birge et al . , 2009 ) . Previous studies have shown that they can be recruited by molecular scaffolds such as Dab1 to transmit Reelin signaling or p130Cas and paxillin to participate in integrin signaling ( Nojima et al . , 1996; Petit et al . , 2000; Sekine et al . , 2012 ) . Following growth factor stimulation , Crk and Crkl are also known to bind directly to several receptor tyrosine kinases ( RTKs ) via their SH2 domains and become rapidly phosphorylated at their specifically targeted tyrosine residues ( Y221 in Crk and Y207 in Crkl ) ( Antoku and Mayer , 2009; Feller et al . , 1994 ) . Upon ligand stimulation , FGF receptors undergo autophosphorylation at multiple tyrosine residues , which serve as docking sites for downstream signaling proteins . Previous studies have suggested that Crk and Crkl can recognize a conserved phosphotyrosine residue in FGFRs ( pY463 in FGFR1 and pY466 in FGFR2 ) through their SH2 domains ( Larsson et al . , 1999 ) . By examining the Y463F mutant of Fgfr1 both in vitro and in vivo , we showed that this site was dispensable for FGF signaling with regards to the activation of Crk proteins . By contrast , we showed that Crk proteins interact with Frs2 and Shp2 proteins , and that their loss prevents FGF-induced Crk phosphorylation . This is reminiscent of Frs2’s function in response to nerve growth factor ( NGF ) to assemble a complex containing Crk , C3G , Rap1 and Braf in order to prolong MAPK signaling ( Kao et al . , 2001 ) . Interestingly , in that case , the binding of Crk requires the Shp2-docking site on Frs2 . We have also identified Grb2 as yet another interacting partner of Crkl and showed that Grb2 was required for FGF-induced phosphorylation of Crk and Crkl . Since both Frs2 and Shp2 interact with Grb2 , we would like to suggest that the assembly of the entire Frs2-Shp2-Grb2 complex is necessary for the recruitment of the Crk family of adaptor proteins to Fgf receptors . C3G , Sos and Dock1 were the first three GEFs identified to be directly associated with the SH3 domains of Crk ( Hasegawa et al . , 1996; Oda et al . , 1994; Tanaka et al . , 1994 ) . While Sos is primarily involved in the activation of Ras-MAPK signaling , C3G and Dock1 promote the exchange of GDP for GTP in the small GTPases Rap1 and Rac1 ( Birge et al . , 2009 ) . These active GTP-bound proteins subsequently regulate integrin signaling and actin polymerization ( Gloerich and Bos , 2011; Ridley , 2011 ) . We showed that the Crk knockout did not recapitulate the Rap1 mutant lens phenotype , ruling out Rap1 as a critical downstream effector of Crk proteins . Instead , activation of Rac1 and , more importantly , Ras effectively rescued the lens fiber elongation defect seen in Crk mutants . In addition to these results further clarifying the downstream targets of Crk in lens development , they also raise an interesting question regarding the nature of Ras signaling in this process . Previous studies have shown that the Frs2/Shp2 mediated Ras-MAPK pathway acts downstream of FGF signaling to regulate lens fiber cell differentiation ( Li et al . , 2014; Madakashira et al . , 2012; Upadhya et al . , 2013 ) , but in this study , our genetic evidence demonstrated that the Crk-mediated Ras signaling pathway only promoted fiber cell elongation and not differentiation . The uncoupling of lens fiber cell differentiation and elongation have been previously observed in lens explants , where pharmacological inhibition of ERK suppressed the morphological changes induced by FGF signaling while not preventing the expression of differentiation markers such as β-crystallin ( Lovicu and McAvoy , 2001 ) . We propose that these results reveal a biphasic function of Ras signaling , which promotes either differentiation or elongation of lens cells in a dosage dependent manner . In this model , Ras signaling induced directly by the Frs2/Shp2/Grb2 complex is sufficient to stimulate cell differentiation , with the additional elevation of Ras activity potentiated by the binding of Crk proteins to the signaling complex resulting in the necessary promotion of cell shape changes . In support of this idea , we note that the loss of Crk proteins did not completely squelch Ras-MAPK signaling , allowing cell differentiation to proceed in Crk mutant lenses . Therefore , in response to heightened FGF signaling , Crk proteins act as additional boosters to Ras activity that results in the specific promotion of cell shape changes .
Mice carrying Crkflox , Crklflox , Fgfr3flox , Frs2αflox , Rap1aflox , Rap1bflox and Shp2flox alleles were bred and genotyped as described ( Lin et al . , 2007; Pan et al . , 2008; Park and Curran , 2008; Su et al . , 2010; Zhang et al . , 2004 ) . Fgf3OVE391 and Fgf3OVE393A were from Dr . Michael Robinson ( Miami University , Oxford , OH ) , Fgfr1ΔFrs from Dr . Raj Ladher ( RIKEN Kobe Institute-Center for Developmental Biology , Kobe , Japan ) , Fgfr1Crk from Dr . Philipo Soriano ( Washington University Medical School , St Louis , MO ) , Fgfr2flox from Dr . David Ornitz ( Washington University Medical School , St Louis , MO ) , Fgfr3flox from Dr . Xin Sun ( University of California San Diego , La Jolla , CA ) , Fgfr4-/- from Dr . Chu-Xia Deng ( National Institute of Health , Bethesda , MD ) , Grb2flox from Dr . Lars Nitschke ( University of Erlangen-Nürnberg , Erlangen , Germany ) , Pax6Le-Cre ( Le-Cre ) from Drs . Ruth Ashery-Padan ( Tel Aviv University , Tel Aviv , Israel ) and Richard Lang ( Children's Hospital Research Foundation , Cincinnati , OH ) , Rac1flox and Rac2-/- from Dr . Feng-Chun Yang ( Indiana University School of Medicine , Indianapolis , IN ) ( Ackermann et al . , 2011; Ashery-Padan et al . , 2000; Brewer et al . , 2015; Glogauer et al . , 2003; Hoch and Soriano , 2006; Roberts et al . , 1999; Robinson et al . , 1998; Weinstein et al . , 1998; Yu et al . , 2003 ) . KrasLSL-G12D mice were obtained from the Mouse Models of Human Cancers Consortium ( MMHCC ) Repository at National Cancer Institute ( Tuveson et al . , 2004 ) . Fgfr1flox ( Stock No: 007671 ) , R26-Rac1LSL-G12V ( Stock No: 012361 ) and ROSAmTmG ( Stock No: 007676 ) mice were obtained from Jackson Laboratory ( Hoch and Soriano , 2006; Muzumdar et al . , 2007; Srinivasan et al . , 2009 ) . In all conditional knockout experiments , mice were maintained on a mixed genetic background and at least three animals were analyzed for each genotype ( Supplementary file 1 ) . We did not observe phenotypic variation in lens development among Pax6Le-Cre and Pax6Le-Cre;Crkfloxl+;Crklflox/+ mice , and thus Pax6Le-Cre only mice were used as controls . Mouse maintenance and experimentation was performed according to protocols approved by Columbia University Institutional Animal Care and Use Committee . Mouse embryos were fixed with 4% paraformaldehyde ( PFA ) in PBS overnight and paraffin or cryo embedded . The paraffin sections ( 10 µm ) were rehydrated and stained with hematoxylin and eosin ( H and E ) for histological analysis . Lens sizes were measured as previously described ( Cai et al . , 2011; Pan et al . , 2010 ) . TUNEL staining and immunostaining were performed on the cryosections ( 8 µm ) as previously described ( Carbe et al . , 2012; Carbe and Zhang , 2011 ) . For phospho-ERK and Shp2 staining , the signal was amplified using a Tyramide Signal Amplification kit ( TSA Plus System , PerkinElmer Life Sciences , Waltham , MA ) . Antibodies used were: anti-Shp2 ( Sc-280 ) , anti-C-maf ( sc-7866 ) , anti-Crkl ( Sc-319 ) , anti-Jag1 ( Sc-6011 ) ( all from Santa Cruz Biotechnology , Santa Cruz , CA ) , anti-pCrkl ( Tyr207 ) ( #3181 , also recognize pCrk ( Tyr221 ) and anti-pERK1/2 ( #4370 ) ( both from Cell Signaling Technology , Boston , MA ) , anti-P57 ( ab75974 , from Abcam , Boston , MA ) , anti-α-SMA ( #C6198 ) , anti-β-catenin ( 6F9 ) , anti-E-cadherin ( U3254 ) ( all from Sigma , St . Louis , MO ) , anti-Ki67 ( #550609 , BD Pharmingen , San Jose , CA ) , anti-Crk ( #610036 ) , anti-Rac1 ( #610651 ) ( both from BD Transduction Laboratory , Franklin Lakes , NJ ) , anti-Prox1 ( PRB-238C ) and anti-Pax6 ( PRB-278P ) ( both from Covance , San Diego , CA ) . Antibodies against α- , β- and γ-crystallins were kindly provided by Sam Zigler ( National Eye Institute ) . Cell proliferation and apoptosis were measured as the ratio of Ki67 or TUNEL-positive cells versus DAPI-positive cells , and analyzed by one-way ANOVA analysis . Alexa Fluor 488 Phalloidin ( A12379 , ThermoFisher ) was used to stain F-actin . Postnatal day 0 to day 3 Crkflox/flox;Crklfloxflox mice were sacrificed and eyes enucleated . Lenses were then dissected out in lens explant culture medium containing DMEM with 1% BSA ( BP1600 , Fisher Scientific ) and 1:100 dilution of Antibiotic-Antimycotic ( 15240062 , ThermoFisher ) . Lens capsules were torn open from the posterior before the lens epithelium was peeled off with forceps and pinned down onto a cell culture dish . To delete Crk and Crkl in the lens explants , 2 × 107 adenoviruses expressing Cre recombinase ( Ad5CMVCre-eGFP , Cat #: VVC-U of Iowa-1174 , Gene Transfer Vector Core , University of Iowa , IA ) were added to the culture of 4 explants for 8 hr one day after explant isolation . GFP-expressing adenoviruses ( Ad5CMVeGFP , Cat #: VVC-U of Iowa-4 ) were used as a control . To induce fiber cell differentiation and elongation , these explants were further cultured in the lens explant culture medium with 2 mg/ml heparin sodium ( H3393 , Sigma ) and 100 ng/ml recombinant murine FGF2 ( 124–02 , ScienCell , Carlsbad , CA ) for 4–5 days . 4 . 24–6 . 36 × 105 MEF cells infected with Ad5CMVCre were seeded in 60 mm dishes and serum starved ( 0 . 5% FBS in DMEM ) for 36–48 hr before being stimulated by 50 ng/ml FGF2 ( R and D Systems , Minneapolis , MN ) for 5 min at 37°C as previously described ( Li et al . , 2014 ) . After washing twice in cold PBS , MEF cells were lysed in 160 µl RIPA buffer ( 50 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl , 1% NP40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 1 µg/ml aprotinin , 1 µg/ml pepstatin , 10 mM sodium pyrophosphate , 1 mM PMSF , 0 . 2 mM Na3VO4 , 50 mM NaF ) . Proteins were visualized by infrared-based western blot analysis using an Odyssey SA scanner ( LICOR Biosciences , Lincoln , NE ) . The signal intensity was quantified using the Odyssey software . The antibodies used were mouse anti-phospho-ERK1/2 ( sc-7383 , Santa Cruz Biotechnology ) , anti-phospho-Crk ( tyr221 ) ( #3491 , Cell Signaling Technology ) , rabbit-anti-Crkl ( sc-319 , Santa Cruz biotechnology ) , mouse-anti-Crkl ( sc-365471 , Santa Cruz biotechnology ) , anti-Shp2 ( sc-280 , Santa Cruz biotechnology ) , anti-Grb2 ( sc-255 , Santa Cruz biotechnology ) , and anti-Frs2 ( sc-8318 , Santa Cruz biotechnology ) . NIH3T3 cells from the American Type Culture Collection were tested mycoplasma-free . They were plated at 1 × 106 cell per 10 cm tissue culture plate and transfected with 2 . 5 μg TAP-Crkl plasmid ( Hallock et al . , 2015 ) using lipofectamine 3000 ( ThermoFisher , Springfield Township , NJ ) . 36 hr after the transfection , cells were serum depleted overnight and then were treated with 100 ng/ml recombinant human FGF2 ( 104–02 , ScienCell ) for 5 min . Both the treatment group and control were lysed with 1 ml immunoprecipitation buffer ( 25 mM Tris-HCl , pH 7 . 4 , 1 mM EDTA , 150 mM NaCl , 1% NP-40 , 5% glycerol ) supplemented with Halt protease inhibitor cocktail ( ThermoFisher ) . Lysates were then incubated with streptavidin resins ( 240207 , Agilent , Santa Clara , CA ) to pull down TAP-Crkl , according to manufacturer’s recommendations . Resins were later washed in streptavidin binding buffer ( 10 mM Tris-HCl , pH 7 . 5 , 1 mM EDTA , 150 mM NaCl ) twice at 4°C . All pulled-down proteins were eluted in 1X Laemmli SDS sample buffer ( 1 . 5% SDS , 9% glycerol , 62 . 5 mM Tris-HCl , pH 6 . 8 , 0 . 00025% Bromophenol blue , 2% β-mercaptoethanol ) . Samples were denatured at 95°C for 5 min before being loaded onto SDS-PAGE gels . | As an embryo develops , its cells divide multiple times to transform into the specialized cell types that form our tissues and organs . To carry out specific roles , cells need to be of a certain shape . For example , in mammals , the cells that make up the main portion of the eye lens , develop into a fiber-like shape to be perfectly aligned with each other . This enables them to transmit light to the retina at the rear end of the eye . To do so , the lens cells increase over 1000 times in length with the help of a group of proteins called the Fibroblast Growth Factor , or FGF for short . The FGF pathway includes a network of interacting proteins that transmit signals to molecules inside the lens cells to control how they specialize and grow . However , until now it was not clear how it does this . Here , Zhang et al . used mouse lens-cells grown in the laboratory to investigate how FGF signaling causes cells to change their structure . The experiments revealed two related proteins called Crk and Crkl that linked the FGF pathway with another signaling system . When these two proteins were removed from the lens cells , the lens cells were still able to specialize , but could no longer grow in length . This suggests that these two processes are independent of each other . Moreover , Crk and Crkl helped the cells to change shape by increasing the amount of another group of proteins called Ras , which are known to both help cells to specialize and to regulate their shape . Zhang et al . discovered that the amount of Ras proteins determined whether cells specialized or modified their shape by changing the organization of proteins in the cell . Millions of children are born with cataracts , a disease caused when lens cells fail to shape properly . A better knowledge of FGF signaling may help to understand how cataracts develop and inspire future treatments . Moreover , the pathways identified in this study could also apply to other organs and diseases in which FGF signaling is active . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | Crk proteins transduce FGF signaling to promote lens fiber cell elongation |
The mechanisms by which mammalian cells recognize and epigenetically restrict viral DNA are not well defined . We used herpes simplex virus with bioorthogonally labeled genomes to detect host factors recruited to viral DNA shortly after its nuclear entry and found that the cellular IFI16 , PML , and ATRX proteins colocalized with viral DNA by 15 min post infection . HSV-1 infection of ATRX-depleted fibroblasts resulted in elevated viral mRNA and accelerated viral DNA accumulation . Despite the early association of ATRX with vDNA , we found that initial viral heterochromatin formation is ATRX-independent . However , viral heterochromatin stability required ATRX from 4 to 8 hr post infection . Inhibition of transcription blocked viral chromatin loss in ATRX-knockout cells; thus , ATRX is uniquely required for heterochromatin maintenance during chromatin stress . These results argue that the initial formation and the subsequent maintenance of viral heterochromatin are separable mechanisms , a concept that likely extrapolates to host cell chromatin and viral latency .
Recognition and restriction of foreign DNA is ubiquitous to all cells . It is imperative that cells prevent foreign DNA from expressing its genes or integrating into the cellular genome lest foreign gene products disrupt normal cellular function . Bacteria possess restriction endonucleases and CRISPR-Cas9 mechanisms to recognize and cleave foreign DNA ( Barrangou et al . , 2007; Tock and Dryden , 2005 ) . Eukaryotic cells , however , recognize foreign DNA and redirect existing chromatin machinery to epigenetically silence the DNA by modifying it with repressive heterochromatin . Epigenetic modification of chromatin , such as DNA methylation and covalent modification of histone tails , is critical for regulating gene expression to ensure proper cellular function and expression of regulated genes at only the appropriate times ( Kanherkar et al . , 2014 ) . Likewise , epigenetic modifications also allow for the effective silencing of foreign DNA that is often associated with invading pathogens , for example DNA viruses . DNA virus genomes are recognized by nuclear DNA sensors , such as interferon-inducible protein 16 ( IFI16 ) , and are subject to epigenetic regulation by host cell factors during both lytic and latent infection ( Knipe , 2015; Knipe et al . , 2013; Orzalli and Knipe , 2014 ) . Nuclear unintegrated retroviral DNAs are also associated with nucleosomes and silencing histone modifications ( Wang et al . , 2016 ) . Indeed , retroviral genomes are epigenetically silenced even after integration into the host genome ( Yao et al . , 2004 ) . Both boon and bane for DNA viruses , viral chromatin enables DNA viruses to establish and maintain persistent latent infections in which the virus is poised for reactivation , but it also impedes productive viral transcription and replication during lytic infection ( Knipe et al . , 2013 ) . Viruses must overcome host-cell silencing mechanisms for reactivation from latency or for lytic infections to proceed . The double-stranded DNA genome of herpes simplex virus ( HSV ) is not associated with histones within the virion ( Pignatti and Cassai , 1980 ) . However , during lytic infection , nucleosomes are rapidly assembled on the viral DNA upon its entry to nuclei of epithelial or fibroblast cells ( Cliffe and Knipe , 2008; Lee et al . , 2016; Oh and Fraser , 2008 ) . Newly formed viral chromatin is immediately associated with silencing histone tail modifications , specifically H3 lysine 9 trimethylation ( H3K9me3 ) and histone H3 lysine 27 trimethylation ( H3K27me3 ) , that peak in density within 1–2 hr post infection ( hpi ) ( Lee et al . , 2016 ) . The resulting viral heterochromatin acts as an epigenetic barrier to viral gene expression and replication . Consequently , viral gene expression occurs in the regulated cascade of immediate-early ( IE ) , early ( E ) , and late ( L ) viral gene products that drive viral transcription , viral DNA synthesis , and virion assembly , respectively ( Knipe and Cliffe , 2008 ) . HSV proteins VP16 and ICP0 , a viral E3 ubiquitin ligase , promote viral gene expression , and removal of viral heterochromatin ( Cliffe and Knipe , 2008; Lee et al . , 2016; Herrera and Triezenberg , 2004 ) . However , the cellular machinery that drives formation of viral heterochromatin and epigenetic silencing of viral gene expression during the early stages of DNA virus infections are still not well described . The SWI/SNF chromatin remodeler protein , α-thalassemia X-linked intellectual disability ( ATRX ) , has recently gained attention as a crucial epigenetic regulator of eukaryotic gene expression and steward of silenced heterochromatin . ATRX is also known to play a role in restriction of DNA viruses ( Lukashchuk and Everett , 2010 ) . ATRX and death domain-associated protein ( DAXX ) together form a histone chaperone complex specific for non-canonical histone variant 3 . 3 ( H3 . 3 ) ( Drané et al . , 2010; Lewis et al . , 2010 ) . This complex is critical for maintaining repressive heterochromatin at many repeat-rich regions , including telomeres ( Lovejoy et al . , 2012 ) , pericentric repeats ( Elsässer et al . , 2015 ) , and endogenous retroviruses ( Elsässer et al . , 2015 ) , and mutations in the ATRX gene are linked to a developmental disorder and several cancer types ( Lovejoy et al . , 2012 ) . ATRX and DAXX are also two of the core components of promyelocytic leukemia protein ( PML ) nuclear bodies ( PML-NBs ) , nuclear punctate structures that have been implicated in a range of cellular activities including the DNA-damage response , transcriptional regulation , and restriction of viral infection ( Chang et al . , 2018 ) . In this context , ATRX and DAXX have been shown to restrict gene expression from DNA viruses and integrated retrovirus DNA ( Lukashchuk and Everett , 2010; Schreiner et al . , 2013; Shalginskikh et al . , 2013; Tsai et al . , 2011; Woodhall et al . , 2006 ) . The ATRX/DAXX complex also appears to be important for maintaining DNA virus latency as depletion of either ATRX or DAXX was reported to induce EBV reactivation from latently infected cells ( Tsai et al . , 2011 ) . In response , DNA viruses have evolved a number of strategies for alleviating the effects of host restriction factors and epigenetic silencing . The Epstein-Barr virus ( EBV ) BNFR1 protein interacts with DAXX to displace ATRX from the complex , effectively reprogramming DAXX and resulting in the activation of early gene transcription ( Tsai et al . , 2011; Tsai et al . , 2014 ) , while degradation of DAXX is promoted by pp71 and E1B-55K during human cytomegalovirus ( HCMV ) , and adenovirus ( AdV ) infections , respectively ( Schreiner et al . , 2013; Saffert and Kalejta , 2006 ) . During HSV infection , proteasome-dependent degradation of PML , ATRX , and IFI16 is promoted by ICP0 ( 27–31 ) . Though ATRX and DAXX have proven to be important host restriction factors , the mechanism of their antiviral activity is less clear . While ATRX and DAXX have been shown to possess nucleosome deposition and remodeling activity in vitro ( Drané et al . , 2010; Lewis et al . , 2010 ) , in vivo studies have largely investigated the effects of ATRX or DAXX depletion at integrated reporter elements or viral genomes during late stage infection , after chromatin had already formed . For instance , DAXX has been reported to promote H3 . 3 incorporation on HCMV , AdV , and EBV genomes at 18 , 24 , and 72 hpi , respectively ( Schreiner et al . , 2013; Tsai et al . , 2014; Albright and Kalejta , 2016 ) . While the histone chaperones HIRA and ASF1A have been implicated in the initial loading of histones onto HSV DNA during the first few hours of infection , these reports also demonstrated that HIRA and ASF1A actually promote productive viral infection ( Oh et al . , 2012; Placek et al . , 2009 ) . Thus , the mechanisms underpinning de novo formation of restrictive heterochromatin on viral DNA remain unclear . Lytic HSV infection provides a temporal reference point that can be used to determine the sequential order of events and dissect the function of cellular chromatin remodelers with greater resolution . Therefore , we used HSV to investigate de novo viral heterochromatin formation and maintenance in the presence or absence of chromatin remodeler and host restriction factor ATRX . Here we report on the mechanisms of viral gene restriction mediated by ATRX during the first 8 hours of infection . We used HSV with bioorthogonally-tagged genomes to quantitatively track viral entry and host restriction factor recruitment as a function of time . Along with nuclear DNA sensor IFI16 , we detected ATRX and PML recruitment to viral DNA by 15 min post infection ( mpi ) , almost immediately upon nuclear entry . Although ATRX restricted the expression of viral mRNA , ATRX had no effect on the initial formation of viral heterochromatin assembly at 2 and 4 hpi . However , ATRX was required for the maintenance of viral heterochromatin between 4–8 hpi during challenges to chromatin stability . Our findings argue for a biphasic model of epigenetic regulation in which de novo assembly of heterochromatin on viral genomes is ATRX-independent but that ATRX is required for viral heterochromatin stability during chromatin stress , such as replication and transcription .
To investigate early interactions of HSV genomes with host proteins that could potentially initiate chromatinization of viral DNA , we generated nucleoside analog-labeled HSV-1 preparations that could be used for the detection of input viral DNA in infected cells . First , we prepared labeled HSV ( HSV-EdC ) stocks by infecting confluent human foreskin fibroblasts ( HFF ) with HSV-1 in the presence of the nucleoside analog 5-ethynyl-2′-deoxycytidine ( EdC , 0 . 5 µM ) . HSV-EdC DNA could be detected by fluorescence microscopy after a copper-catalyzed bioorthogonal cycloaddition ( click chemistry ) of biotin to EdC incorporated into viral DNA followed by incubation with a streptavidin-fluorophore probe ( Figure 1A; Figure 1—figure supplement 1B ) . HSV-EdC genomes were detected as punctate nuclear foci that colocalized with the HSV immediate-early protein ICP4 , considered a marker for viral DNA , at 2 hr post infection ( hpi ) ( Figure 1—figure supplement 2A ) . We also observed that HSV-EdC viral DNA ( vDNA ) colocalized with early replication compartments , as defined by immunofluorescence ( IF ) staining for the viral early protein ICP8 at 4 hpi ( Figure 1—figure supplement 2B ) . As recently reported ( Sekine et al . , 2017 ) , HSV-EdC genomes that colocalized with ICP8 underwent a morphological decompaction that could be prevented by treatment with viral DNA synthesis inhibitor phosphonoacetic acid ( PAA , 400 µg/ml ) ( Figure 1—figure supplement 2B ) . HSV-EdC demonstrated only a slight decrease in ICP8 protein levels at 4hpi but no differences in either ICP4 or ICP8 levels by 8hpi ( Figure 1—figure supplement 2C ) . Intracellular HSV-EdC foci were prevented by treatment with heparin ( 50 µg/mL ) , which blocks viral attachment and entry to cells ( Bender et al . , 2005 ) ( Figure 1—figure supplement 2D ) . Labeling of host cell chromatin during infection with EdC-labeled viral preparations was not observed . In total , these results validated the use of EdC labeling of HSV to detect input HSV DNA and showed that the course of infection with HSV-EdC virus was normal . To determine the kinetics of restrictive host protein recruitment to input viral DNA , we quantitatively mapped the spatiotemporal kinetics of incoming HSV-EdC genome colocalization with host restriction factors IFI16 , PML , and ATRX ( Figure 1A , C–E; Figure 1—figure supplement 3A , B ; Video 1 ) . To measure colocalization of viral DNA and restriction factor foci in host cell nuclei , we first defined nuclear fluorescent foci using a software bot ( Cicconet et al . , 2017 ) and then used additional customized features of the software to define foci as colocalizing when the centers were within a distance threshold of 5 pixels , ~350 nm ( see Materials and methods for detailed description of software ) . Using this method , we observed click chemistry labeled viral genomes in the nuclei of HFFs as early as 15 mpi ( Figure 1B ) . The percentage of infected cells increased steadily over the first 30 min of infection . A stable 2–3 genomes per infected nucleus were maintained from 50 mpi ( Figure 1—figure supplement 3C ) . ATRX colocalization with HSV DNA peaked between 40–100 mpi ( Figure 1E ) , with 80% of viral DNA colocalizing with ATRX . We observed IFI16 and ATRX colocalizing with viral DNA at similar frequencies by 15 mpi ( Figure 1C ) . IFI16 has been shown to form transient foci at the nuclear periphery in response to HSV infection ( Diner et al . , 2016; Everett , 2016 ) . The kinetics of IFI16 colocalization with viral DNA supported this hypothesis as our observations revealed an early peak of IFI16-vDNA colocalization ( 20–30 mpi ) that rapidly decreased after 30 mpi ( Figure 1C , E ) . Little to no IFI16-vDNA colocalization was observed from 60 mpi ( Figure 1E ) . These results indicated that , under our conditions , IFI16 and ATRX localize with HSV genomes almost immediately upon nuclear entry , and while the majority of IFI16 colocalization with input viral DNA occurs between 15 and 30 mpi , ATRX colocalization is stable between 15 and 100 mpi . PML and ATRX colocalization with viral DNA occurred with nearly identical kinetics ( Figure 1D ) . We attempted to analyze the kinetics of DAXX colocalization with HSV DNA . However , several antibodies , while providing a signal for immunoblotting , gave little or no signal in IF experiments . ATRX colocalization with viral DNA began to wane by 100 mpi ( Figure 1E ) . The HSV E3 ubiquitin ligase , ICP0 , which has been shown to promote degradation of both PML and ATRX ( Chelbi-Alix and de Thé , 1999; Jurak et al . , 2012 ) . Indeed , we observed that the presence of ATRX foci in infected nuclei generally diminished as the presence of ICP0 increased across the infected population ( Figure 1—figure supplement 3D ) . By 90 min post infection , ICP0 was observed to localize to ATRX foci that also colocalized with HSV genomes ( Figure 1—figure supplement 3E ) . Though the percentage of infected cells positive for ICP0 did not change between 90 and 120 min , the percentage of infected cells positive for ATRX foci decreased by more than half ( Figure 1—figure supplement 3D ) . This may reflect a period of time in which ICP0 drives the dissolution of PML-NBs; however , the ICP0 antibody exhibited a high background signal that rendered it difficult to perform precise colocalization studies that would be needed to further define the spatiotemporal relationship between ICP0 and ATRX . To test whether the colocalization of vDNA with ATRX or IFI16 was due to random events , we compared the distances of vDNA to restriction factors with the distances of random intranuclear points to vDNA . Frequency distribution analysis of pixel distances from HSV-EdC genomes to nearest-neighbor ( nn ) ATRX foci revealed a non-parametric distribution of distances that had a median distance of 2 . 83 pixels ( ~200 nm ) , within the optical resolution of the system ( Figure 1F ) . Similarly , vDNA-to-nnIFI16 foci had a mean median distance of 2 . 24 pixels ( ~150 nm ) ( Figure 1G ) . We used a custom software script to generate foci at randomly positioned x , y coordinates in numbers equal to the ATRX ( or IFI16 ) foci that fell within a given nucleus ( see Materials and methods for detailed description of software ) . We then compared the frequency distribution of vDNA-to-nnATRX foci ( or vDNA-to-nnIFI16 foci ) distances with the distribution of distances of vDNA to randomly generated x , y coordinates ( Figure 1—figure supplement 1A ) . The frequency distributions of distances from vDNA-to-nnATRX foci and vDNA-to-nnIFI16 foci versus the distribution of distances from vDNA-to-nnRandom points at 30mpi were highly significant ( p < 0 . 001 ) ( Figure 1F , G ) . These results argued that IFI16 and PML-NB proteins localize to viral DNA very early and are candidates for the DNA sensors that initially respond to nuclear entry of viral genomes . We next investigated if IFI16 affects ATRX colocalization with HSV-EdC genomes or vice versa . We reported that IFI16 and PML recruitment to viral genome complexes are independent ( Orzalli et al . , 2013 ) , while another study observed only a 10% reduction in cells showing PML recruitment to genome complexes when IFI16 was depleted ( Cuchet-Lourenço et al . , 2013 ) . We depleted HFFs of IFI16 using small interfering RNAs ( siRNA ) against IFI16 ( siIFI16 ) ( Figure 2A , C ) . Cells treated with siIFI16 and infected at a multiplicity of infection ( MOI ) of 5 with HSV-EdC showed no significant reduction in ATRX colocalization with viral genomes at 30mpi ( Figure 2A , D ) . Likewise , depletion of ATRX via siRNA ( siATRX ) did not change the colocalization frequency of IFI16 with labeled HSV genomes ( Figure 2B , E ) . Interestingly , treatment with siRNAs reduced the frequency of IFI16 foci formation in both siNon-Targeting ( siNT ) and siATRX treated HFFs ( Figure 2E ) , perhaps due to an innate immune response to transfected siRNA ( Whitehead et al . , 2011 ) . While ATRX recruitment to viral DNA appears to be independent of IFI16 , it has been reported that mutation or alteration of the ATRX interaction domain of DAXX abolished ATRX localization to PML-NBs and early replication compartments ( Lukashchuk and Everett , 2010 ) . Indeed , ATRX colocalization with vDNA and PML-NBs was lost in HFFs depleted for DAXX , resulting in diffuse nuclear ATRX staining ( Figure 2F , Figure 2—figure supplement 1A ) . These results demonstrated IFI16 is not required for ATRX colocalization to viral DNA , as recently reported for PML and IFI16 ( Alandijany et al . , 2018 ) , and argued that ATRX and IFI16 are in distinct DNA sensing pathways that simultaneously detect input viral DNA as it enters the nucleus . We next investigated the functional effects of ATRX on epigenetic silencing of viral DNA . Depletion of ATRX from HepaRG cells was reported to increase viral plaque formation and slightly increase viral early protein UL42 as detected by immunoblotting ( Lukashchuk and Everett , 2010 ) . Because ICP0 promotes the dissolution of PML-NBs and the proteasome-mediated degradation of ATRX , we performed the following studies using the HSV-1 7134 virus that is ICP0-null . ATRX colocalized with input viral DNA from 15 to 100 mpi; therefore , we investigated if ATRX was associating with the viral DNA . We performed chromatin immunoprecipitation followed by quantitative polymerase chain reaction ( ChIP-qPCR ) on chromatin from HFF cells infected with HSV 7134 virus . We found that ATRX could be detected at viral gene promoters for both ICP27 and ICP8 at levels higher than GAPDH by one hpi , and to significantly higher levels by 4 hpi ( Figure 3A ) . Detection of ATRX at viral gene promoters suggested that ATRX may play a role in epigenetically regulating viral gene expression by associating with viral DNA . We next measured viral gene expression in siATRX-treated HFFs infected with HSV 7134 . We harvested infected cells at 2 hr intervals from 2 to 8 hpi and measured viral transcripts by reverse transcription ( RT ) -qPCR ( Figure 3B–D ) . ATRX-depleted HFFs showed significant increases in transcripts from genes of all kinetic classes , with the most significant effects on expression occurring from IE ( ICP27 ) and L genes ( gB ) at 6 and 8 hpi ( Figure 3B , D ) , while E gene ICP8 expression was significantly elevated at 8hpi ( Figure 3C ) . In parallel with the above experiment , we tested the impact of viral DNA replication on ICP0-null HSV gene expression in HFFs depleted of ATRX . To accomplish this , we treated cells with a viral DNA polymerase inhibitor , PAA ( 400 µg/ml ) , from 1 hr prior to infection and maintained PAA throughout the experiment . While overall viral gene expression was reduced in the presence of PAA , depletion of ATRX still resulted in significant increases in ICP0-null gene expression from each gene of the three kinetic classes ( Figure 3B–D ) . The increased accumulation of viral mRNA upon ATRX depletion argued that ATRX plays a role in preventing transcription from viral genes , and the increase in viral gene expression with and without PAA demonstrated that ATRX restricts gene expression from both input and progeny viral DNA . To facilitate our functional studies of ATRX and DAXX , we used CRISPR-Cas9 mediated gene editing to establish an ATRX-knockout cell line ( ATRX-KO ) derived from hTERT immortalized human fibroblasts ( Albright and Kalejta , 2016; Bresnahan and Shenk , 2000 ) . We also established a control cell line ( Control ) in parallel that expresses Cas9 but no guide RNA , resulting in passage-matched ATRX-KO and Control cell lines ( Figure 4—figure supplement 1A ) . The immortalized fibroblasts were not permissive for single cell cloning; therefore , we used a population of ATRX-KO cells maintained under puromycin selection . ATRX-KO cells yielded significantly higher viral titers of ICP0-null virus than Control cells ( MOI 3 ) ( Figure 4—figure supplement 1B ) . Similar to our observations in siRNA treated cells , ICP27 gene expression from ICP0-null HSV was significantly greater in ATRX-KO cells than Control by 4 hpi , with both ICP8 and gB exhibiting significantly elevated expression levels by six hpi ( Figure 4—figure supplement 1C ) . DAXX has also been shown to reduce HSV UL42 protein levels during ICP0-null HSV infection ( Lukashchuk and Everett , 2010 ) ; however , the effects of double depletion of ATRX and DAXX on viral mRNA levels have yet to be investigated . We treated ATRX-KO and Control cells with siRNAs against DAXX , ATRX , or a non-targeting control ( Figure 4—figure supplement 1D ) . Control cells treated with siDAXX exhibited elevated expression of ICP27 transcripts and a slight increase in ICP8 transcript levels by 8 hpi ( Figure 4A ) . In contrast , siDAXX-treated ATRX-KO cells did not exhibit elevated expression of either ICP27 or ICP8 in comparison to siNT and siATRX treated ATRX-KO at 8 hpi ( Figure 4A ) . Slightly elevated viral gene expression from ATRX-KO cells treated with siATRX is likely due to increased depletion efficiency in the heterogeneous ATRX-KO population . These results argued that DAXX works in conjunction with ATRX to restrict viral gene expression . Interestingly , siDAXX treated ATRX-KO cells exhibited a significant decrease in gB transcripts at 8 hpi when compared to siATRX treated ATRX-KO cells ( Figure 4A ) . Similarly , Control cells treated with siDAXX did not exhibit an increase in gB expression at 8 hpi . These results indicated that DAXX restricts viral gene expression while in complex with ATRX , but after ATRX is degraded by the activity of ICP0 ( 30 ) , it may have a separate role in promoting gB and potentially other viral late gene expression or replication . We next investigated the effects of DAXX depletion in ATRX-KO cells on viral yield . ATRX-KO cells treated with siDAXX and infected with ICP0-null HSV at either MOI 0 . 1 or 3 exhibited no increase in viral yield in comparison to cells treated with siATRX and siATRX + siDAXX ( Figure 4B , Figure 4—figure supplement 1E ) . Treatment with siNT decreased viral yield in both ATRX-KO and Control cells compared to untreated samples , possibly due to the cellular response to siRNA transfection as discussed above ( Figure 4—figure supplement 1E ) . Unlike siDAXX treatment in ATRX-KO cells , siIFI16 and siATRX treatment resulted in a significant increase in viral yield over siATRX treatment alone in ATRX-KO cells ( Figure 4C ) . We recently reported similar additive effects on ICP0-null HSV yield in cells depleted for both DAXX and IFI16 ( Merkl et al . , 2018 ) . These results supported a model in which ATRX restricts HSV gene expression in coordination with DAXX , and this pathway is distinct from IFI16 sensing and restriction of viral DNA . ATRX and DAXX have been reported to promote H3 . 3 deposition on viral DNA promoters of AdV and HCMV ( Schreiner et al . , 2013; Albright and Kalejta , 2016; Newhart et al . , 2013 ) ; however , the role ( s ) that ATRX and DAXX play in chromatin deposition on HSV DNA remain unknown . To investigate if ATRX plays a role in the initial chromatin deposition on input viral DNA , we performed ChIP-qPCR . ATRX-KO and Control cells were infected with ICP0-null 7134 virus at an MOI of 3 , and cell lysates were harvested at 2 , 4 , and 8 hpi . Chromatin was immunoprecipitated with antibodies specific for total H3 , H3 . 3 , H3K9me3 , and H3K27me3 . We then quantified the recovered chromatin via qPCR . Surprisingly , we detected little to no difference in chromatin occupation at either the ICP27 or ICP8 gene promoter at 2 and 4 hpi for each of the chromatin markers tested in ATRX-KO and Control cells ( Figure 5A–D; Figure 5—figure supplement 1A-D ) . However , by 8 hpi , we observed lower levels of viral chromatin in ATRX-KO cells , with H3 , H3 . 3 , H3K9me3 , and H3K27me3 all exhibiting significantly decreased levels compared to Control cells at both ICP27 and ICP8 gene promoters ( Figure 5A–D; Figure 5—figure supplement 1A-D ) . The overall density of H3K9me3 modifications per H3 at 8 hpi in ATRX-KO cells was less than half of those in Control cells ( Figure 5E ) . This argued that loss of H3K9me3 occurred at a greater rate than the removal of H3 . Indeed , H3K9me3 levels were reduced by greater than 10-fold at the ICP27 promoter in ATRX-KO cells compared to Control . DNA replication was also enhanced in ATRX-KO cells , indicating an earlier onset of viral DNA synthesis ( Figure 5F ) . These findings were supported by the observation that when infected with an ICP0-positive virus , 7134R , there was no significant difference in viral gene expression between ATRX-KO and Control cells ( Figure 5—figure supplement 1F ) . This argued that the unique restrictive effects exerted by ATRX on ICP0-null HSV gene expression occurred later than 2 hpi , when ICP0 disrupted PML-NBs ( Figure 1—figure supplement 3D ) , possibly in response to chromatin stability challenges , such as replication and transcription . From these results we concluded that ATRX promotes maintenance of stable viral heterochromatin during infection and delays the onset of viral DNA replication but is not uniquely required for the de novo formation of chromatin on input viral genomes . To determine if the changes in heterochromatin composition on ICP0-null HSV genomes were dependent on viral DNA synthesis , we performed ChIP-qPCR in ATRX-KO and Control cells with or without viral DNA synthesis inhibitor . ATRX-KO and Control cells were pretreated with PAA for 1 hr prior to infection and maintained for the duration of the experiment . We infected cells with HSV 7134 , harvested at 4 and 8 hpi , and performed ChIP-qPCR as described above . Consistent with our previous results , untreated ATRX-KO cells exhibited reduced H3 , H3 . 3 . H3K9me3 , and H3K27me3 at both the ICP27 and ICP8 promoters compared to their Control cell counterparts ( Figure 6—figure supplement 1A-D and 2A-D ) . However , when viral DNA synthesis was inhibited with PAA ( Figure 6E ) , viral chromatin density was similar in all conditions at 4 hpi ( Figure 6A–D ) . Viral heterochromatin continued to accumulate from 4 to 8 hpi in PAA-treated Control cells while it remained near 4 hr levels in PAA treated ATRX-KO cells; however , removal of heterochromatin was largely blocked by PAA treatment ( Figure 6A–D ) . Interestingly , H3 . 3 increased in occupancy at the ICP8 promoter from 4 to 8 hpi in ATRX-KO cells treated with PAA suggesting that H3 . 3 deposition may be differentially regulated at the ICP8 promoter ( Figure 6—figure supplement 2C ) . Overall , these results argued that ATRX maintained virus genome-associated heterochromatin on input HSV DNA , promoted increased viral heterochromatin stability and density from 4 to 8 hpi , and that removal of heterochromatin was enhanced by viral DNA replication . Chromatin structure is dynamic by nature and requires active maintenance in response to stability challenges including replication and transcription ( Schwabish and Struhl , 2004; Nair et al . , 2017 ) . To assess the effects of transcription on viral chromatin composition , we infected Control and ATRX-KO cells with transcriptionally silent HSV d109 virus . ATRX-KO cells showed no difference in either H3 or H3K9me3 levels at the ICP8 gene promoter at both 4 and 8 hpi ( Figure 7A ) . To further investigate the effects of ATRX on chromatin during transcription , we inhibited transcription with CDK9 inhibitor flavopiridol . Cells were treated with flavopiridol ( 1 µM , 1 hr pretreatment ) ( Figure 7—figure supplement 2E ) and infected with HSV . As previously observed in untreated cells , H3 and H3K9me3 levels on HSV DNA were similar at 4 hpi in ATRX-KO and Control cells but lower by 8 hpi in ATRX-KO fibroblasts ( Figure 7—figure supplement 1A , B ) . However , treatment with flavopiridol blocked heterochromatin loss and resulted in H3 and H3K9me3 accumulation from 4 to 8 hpi in ATRX-KO cells ( Figure 7B , Figure 7—figure supplement 2A ) . Likewise , treatment with polymerase inhibitor α-amanitin ( 2 µg/ml , 16 hr pretreatment ) ( Figure 7—figure supplement 2F ) blocked removal of H3 and H3K9me3 in ATRX-KO cells and resulted in heterochromatin accumulation from 4 to 8 hpi ( Figure 7C , Figure 7—figure supplement 1C , D and 2B ) . We also infected Control and ATRX-KO cells with HSV 7134 and added the transcriptional inhibitor actinomycin-D ( actD , 5 µg/ml ) at 4 hpi . Addition of actD prior to infection blocked chromatin loading to HSV DNA ( not shown ) , so we added actD at 4 hpi to arrest transcription once viral chromatin had been established . Treatment with actD at 4 hpi stabilized chromatin resulting in no loss of H3 or H3K9me3 in ATRX-KO and Control cells between 4 and 8 hpi ( Figure 7D; Figure 7—figure supplement 2C ) . Taking these results together , we concluded that ATRX plays a role in a process that protects against chromatin destabilization during transcription . Because PML has recently been implicated in influencing ATRX/DAXX activity ( Delbarre et al . , 2017 ) , we next investigated if PML affects viral chromatin maintenance . We infected HFFs stably expressing a short hairpin against PML ( shPML ) ( Wagenknecht et al . , 2015 ) with HSV 7134 , and we observed that at 4 hpi , PML-depleted cells ( Figure 8A ) showed no significant differences from shControl cells for H3 but a slight decrease in H3K9me3 at the ICP27 and ICP8 viral gene promoters ( Figure 8B–D ) . However , by 8 hpi , we observed a significant reduction of chromatin in shPML cells that was similar to what was observed in ATRX-KO fibroblasts ( Figure 8B–D ) . Though PML may play a role in promoting early H3K9me3 density at viral gene promoters , these results indicated that PML , like ATRX , is not uniquely required for the formation of viral heterochromatin . However , ATRX and PML are both required for the maintenance of viral heterochromatin stability .
Our results revealed that incoming viral DNA is detected and epigenetically silenced by multiple host-cell restriction pathways . We detected IFI16 , PML , and ATRX colocalizing with viral DNA by 15 mpi , the earliest time point at which we were able to detect viral DNA in the nuclei of HFFs . This observation argued that these host restriction factors detect HSV almost immediately upon nuclear entry . IFI16 has been observed to form transient foci at the nuclear periphery during early HSV infection ( Diner et al . , 2016; Everett , 2016 ) , but there was no definite way to connect the IFI16 foci to incoming viral DNA without labeling of the viral DNA as presented here . IFI16 colocalization with viral DNA was temporally limited , suggesting an initial detection of viral DNA followed by a displacement by additional factors or modifications , perhaps by nucleosomes which are known to impede IFI16 oligomerization ( Stratmann et al . , 2015 ) . In contrast , we observed ATRX stably colocalizing with viral genomes until PML-NBs were disbursed by ICP0 around two hpi . We previously showed that depletion of IFI16 does not affect recruitment of PML to ICP4 , a marker for viral DNA ( Orzalli et al . , 2013 ) . Furthermore , during the preparation of this manuscript , it was reported that IFI16 does not influence PML recruitment to labeled HSV DNA ( Alandijany et al . , 2018 ) . The latter report supports our own observations that ATRX and IFI16 are independently recruited to incoming viral DNA . Providing additional support for this argument , we showed the restrictive effects of ATRX and IFI16 on ICP0-null HSV replication are additive . These results echo our recent findings regarding the additive restrictive effects of DAXX and IFI16 ( Merkl et al . , 2018 ) . Together , these results argue that IFI16 and PML-NB components represent distinct nuclear DNA sensing pathways that act in parallel to detect and restrict invading viral DNA almost immediately upon nuclear entry . Our results revealed that ATRX is not required for H3/H3 . 3 deposition on viral DNA for the first 4 hpi; however , we cannot rule out the possibility that other histone chaperones or assembly complexes may compensate for the loss of ATRX/DAXX histone chaperone activity , as HIRA does for depletion of CAF-1 ( 52 ) . In fact , CAF-1 has been reported to compensate for loss of Daxx in murine cells ( Drané et al . , 2010 ) , and NAP1 can assemble H3 . 3-containing nucleosomes in vitro ( Lewis et al . , 2010 ) . However , it is important to note that ATRX and DAXX have roles in promoting heterochromatin that do not require histone chaperone activity ( Sadic et al . , 2015 ) . Because ATRX can also act as a H3K9me3 reader ( Eustermann et al . , 2011; Iwase et al . , 2011 ) , restrictive effects of ATRX and DAXX on HSV infection may not require the H3 . 3 chaperone activity of the ATRX/DAXX complex during lytic infection . Individual and combinatorial depletion of other histone loaders and nucleosome formation complexes will be required to identify the cellular factors needed for the initial loading of heterochromatin on the HSV genome . We also cannot rule out the possibility that the observed effects are due to an indirect effect as a consequence of depleting ATRX . Depleting cells of a histone chaperone will likely result in changes to cellular chromatin and altered gene expression . However , changes in viral chromatin density in ATRX-KO cells are not observable until after 4 hr post infection , so we do not believe the observed to be effects to be the consequence of enhanced global histone mobility . Additionally , we observed the association of ATRX with viral gene promoters by 4 hr post infection . This finding is consistent a direct role for ATRX in viral chromatin regulation . Furthermore , as both siRNA and CRISPR depletions of ATRX have very similar effects on viral yield and viral gene expression , we believe we can rule out off-target effects . Our results suggest that DAXX may have ATRX-independent functions during viral infection . Here we report that siRNA depletion of DAXX in ATRX-KO cells significantly decreased the expression of gB transcripts during productive ICP0-null HSV infection . Similar dual anti-viral and pro-viral effects have recently been reported for PML ( Newhart et al . , 2013; Xu et al . , 2016 ) , and the H3 . 3-specific chaperone , HIRA , has been shown to both enhance and restrict viral infection depending on context ( Placek et al . , 2009; Rai et al . , 2017 ) . It is possible that DAXX is restrictive while in complex with ATRX but at a later stage of infection , after dispersion of PML-NBs and degradation of ATRX , DAXX may promote HSV gene expression or replication . Our results contribute to a growing body of literature that is revealing ATRX-independent functions of DAXX that may have important implications for viral restriction , cancer biology , and epigenetics ( He et al . , 2015; Hoelper et al . , 2017 ) . We hypothesize that ATRX restricts viral infection by maintaining viral heterochromatin stability , resulting in DNA that is less accessible to transcription and viral replication factors ( Figure 9 ) . Though de novo formation of HSV-associated heterochromatin was unaffected by the absence of ATRX ( Figure 9B ) , decreased histone retention or histone reloading in the absence of ATRX/DAXX could conceivably explain the observed loss of H3 and H3K9me3 on HSV genomes , elevated levels of viral transcription , and the enhanced accumulation of nascent viral DNA that we observed by 8 hpi in ATRX-KO cells . Decreased histone retention may lead to increased histone dynamics , reduced nucleosome density , and transiently expose transcription factor binding sites , resulting in elevated transcription ( Huang and Zhu , 2014 ) and increased accessibility for viral replication factors . Similarly , if nucleosomes are lost from viral DNA during transcription or DNA synthesis , new nucleosomes will need to be rapidly formed and modified with repressive histone tail modifications to maintain epigenetic silencing . These ideas are consistent with our findings that ATRX promotes the maintenance and continued accumulation of viral heterochromatin , both histones and histone modifications , during times associated with viral transcription and DNA synthesis . When viral DNA synthesis was blocked with PAA , HSV DNA in Control cells continued to accumulate H3 and H3K9me3 from 4 to 8 hpi . In contrast , HSV DNA-associated H3 and H3K9me3 did not accumulate significantly in ATRX-KO cells from 4-8hpi , suggesting that in the absence of ATRX , histone exchange rates had reached an equilibrium by 4 hpi . While there may be more than one way to explain these observations , we think the model that best fits these observations is one in which ATRX promotes heterochromatin accumulation by mediating modified histone retention . ATRX is known to be recruited to heterochromatin through the interaction of its ADD domain with the H3K9me3 histone tail modification ( Iwase et al . , 2011 ) , and is known to be required for both H3 . 3 and H3K9me3 maintenance on regions of cellular DNA ( Lewis et al . , 2010; Sadic et al . , 2015; He et al . , 2015 ) . It can also be recruited by direct interaction with heterochromatin protein 1 ( HP1 ) ( Eustermann et al . , 2011; Bérubé et al . , 2000 ) , a chromatin remodeling factor which has been reported to localize to HSV gene promoters during infection ( Ferenczy and DeLuca , 2009 ) . ATRX may act to recruit DAXX to viral DNA through its interaction with H3K9me3 or HP1 and promote rapid replacement of histones that dissociate from viral DNA ( Figure 9C ) . These histones could then be rapidly modified with H3K9me3 by histone lysine methyltransferases SUV39H1 or SETDB1 that are known to be recruited by HP1 , thus maintaining heterochromatin composition and density on the HSV DNA . Our model is also likely relevant to the maintenance of DNA virus latency . The EBV protein BNRF1 has been shown to bind to DAXX and separate it from the ATRX/DAXX complex ( Tsai et al . , 2011 ) . That study also showed that depletion of DAXX or ATRX led to EBV reactivation in lymphoblastoid cells . Through its interaction with DAXX , BNRF1 was shown to promote increased mobilization and turnover of H3 . 3 during FRAP experiments ( Tsai et al . , 2014 ) . The observed increase in H3 . 3 turnover has been proposed to facilitate expression of viral latent genes and the establishment of latent EBV infection . ATRX and DAXX have also been detected localizing to viral genomes in neurons of mice latently infected with HSV-1 ( 63 ) . These reports are consistent with our model that ATRX/DAXX promotes heterochromatin integrity to enhance epigenetic silencing of viral genes and proposes that this mechanism may be important for both lytic and latent DNA virus infections . Further studies are needed to define how ATRX reduces the loss of chromatin during times of increased viral transcription and replication . This model may also be relevant to cellular heterochromatic loci that are maintained by ATRX . Transcription by RNA Pol II evicts histones which are then quickly replaced ( Schwabish and Struhl , 2004 ) by the FACT complex , which has been implicated in both the disassembly and reassembly of histone components during transcription ( Belotserkovskaya et al . , 2003; Mason and Struhl , 2003; Xin et al . , 2009 ) . The result is a dynamic equilibrium of nucleosome formation , retention , and eviction . Based on our observations of viral chromatin and viral gene expression in the absence of ATRX , we expect that as the level of transcription increases in the absence of ATRX , the equilibrium between eviction and deposition could shift in favor of eviction and result in an apparent reduction in H3 . 3 and H3K9me3 . Supporting this idea , ATRX has been shown to be required to silence transcription of an eGFP-retrotransposon reporter in response to continuous dox-induced transcription ( Sadic et al . , 2015 ) . In that study , ATRX depletion had minor effects on the initial establishment of H3K9me3 at the reporter element , but it was required for the efficient re-establishment of H3K9me3 after transcriptional challenge was ended . Similarly , ATRX protected against expression of minor satellite sequences in mouse embryonic stem cell derived neurons upon neuronal stimulation ( Noh et al . , 2015 ) . Like ATRX , PML was not strictly required for the initial deposition of histone H3 or H3K9me3 on viral DNA but required for maintenance of heterochromatin from 4 to 8 hpi . PML was recently reported to influence ATRX/DAXX complex activity and loss of PML may increase H3 . 3 turnover at PML-associated domains ( PADS ) ( Delbarre et al . , 2017 ) . Because ATRX is required for heterochromatin maintenance at both pericentric heterochromatin and telomeres ( Lewis et al . , 2010; Goldberg et al . , 2010; McDowell et al . , 1999 ) , it is not surprising that pericentric heterochromatin and telomeres have both been shown to exhibit high levels of chromatin stability with very low histone turnover ( Kraushaar et al . , 2013 ) . Consequently , we suggest that a major function of ATRX in heterochromatin maintenance is in stabilizing chromatin density and promoting histone retention . However , increased histone retention is not the only possible mechanism of action . ATRX could also be promoting the re-establishment of heterochromatin in the wake of RNA Pol II mediated transcription , perhaps in a fashion similar to the FACT-complex . Likewise , ATRX could be stabilizing chromatin by inhibiting transcription or replication through a DNA damage pathway . Further mechanistic studies will be required to define the precise mechanism through which ATRX exerts its influence on viral chromatin . Our finding that ATRX restricts HSV gene expression by promoting the maintenance of viral heterochromatin raises the idea that there are at least two stages of epigenetic regulation involved in restriction of viral gene expression: 1 . Initial loading of heterochromatin; 2 . Maintenance of heterochromatin during chromatin stress , such as transcription and replication . Although viral chromatin structure is known to be dynamic ( Gibeault et al . , 2016 ) , before this report , the formation of viral chromatin and the maintenance of viral chromatin were not understood to be separate mechanisms . Importantly , DNA viruses have evolved specific mechanisms for dismantling or disabling the ATRX/DAXX pathway rather than blocking the initial formation of heterochromatin . This argues there may be some role for the early formation of chromatin in the viral life cycle . This is supported by reports that HIRA and ASF1A promote histone occupation on HSV DNA during the first few hours of infection but do not appear to be restrictive during this same period of time ( Oh et al . , 2012; Placek et al . , 2009 ) . Conversely , the heterochromatin maintenance mechanism mediated by ATRX appears to act as a critical restrictive hurdle to productive infection . The multiple stages of epigenetic restriction of viral genomes uncovered here involve cellular chromatin factors . Therefore , their roles in epigenetically silencing viral DNA may reflect their functions in cellular chromatin assembly and maintenance . Cellular chromatin structure must be maintained during transcription , DNA synthesis , and mitosis . Thus , it is likely that multiple mechanisms exist to deal with each of these chromatin stressors . The stages of viral epigenetic regulation by the host cell as defined here will serve to identify host epigenetic factors that may function in similar host cell pathways , and this knowledge could provide insights critical to the development of epigenetic therapies for both viral and host diseases .
HFF , U2OS , and HEK293T cells were obtained from the American Type Culture Collection ( Manassas , VA ) . hTERT immortalized fibroblasts were a kind gift from Robert Kalejta and originally described in Bresnahan and Shenk , 2000 . HFFs expressing a short hairpin against PML ( shPML ) were a kind gift from Thomas Stamminger ( Wagenknecht et al . , 2015 ) . All cells are regularly tested for the presence of mycoplasma contamination . Cells used in this study were mycoplasma free . All fibroblast cells and HEK293T cells were maintained in Dulbecco's Modified Eagle's medium ( DMEM; Corning , Corning NY ) supplemented with 10% ( v/v ) fetal bovine serum ( FBS ) in humidified 5% CO2 incubators at 37°C . Cells were serially passaged by trypsin-EDTA ( 0 . 05%; Corning ) treatment and transfer to fresh media . shPML fibroblasts were maintained under 5 µg/ml puromycin . HSV-1 KOS was used as the wild type virus in this study . HSV-1 KOS 7134 ( Cai and Schaffer , 1992 ) was used for experiments requiring an ICP0-null HSV . HSV-1 7134R is an ICP0+ virus . HSV-1 d109 does not express HSV immediate-early genes ( Samaniego et al . , 1998 ) . Cells were infected with viruses at the indicated MOI in PBS containing 1% ( w/v ) glucose and 1% ( v/v ) bovine calf serum ( BCS ) . Infections were carried out at 37°C with constant agitation with 1 hr adsorption time unless otherwise indicated . After 1 hr , the inoculum was removed and replaced with DMEM supplemented with 1% BCS ( DMEV ) and then incubated at 37°C until times indicated . ICP0-null viral yields were determined by infecting U2OS cells with serial dilutions of viral lysates collected at indicated times . For experiments testing the effects of viral DNA synthesis , PAA ( Sigma-Aldrich; St . Louis , MO ) was added to the media ( 400 µg/ml ) with 10 mM HEPES at the time of infection and maintained until time of harvest . Actinomycin-D ( 5 µg/ml; Sigma ) and flavopiridol ( 1 µM; Selleck Chemicals , Houston , TX ) were added to cells 4 hpi and 1 hr prior to infection , respectively , and maintained until time of harvest . Cells were pre-treated with α-amanitin ( 2 µg/ml; Sigma ) for 16 hr prior to infection and maintained until time of harvest . To block viral attachment and entry to cells , heparin was added to the viral inoculum ( 50 µg/ml ) , and heparin was maintained for the duration of the infection . DNA oligos coding encoding sgRNAs specific for ATRX were designed using the Optimized CRISPR Design tool ( http://crispr . mit . edu/; Zhang Lab , MIT ) and cloned into the lentiCRISPR v2 vector ( Busskamp et al . , 2014 ) . HEK293T cells were grown to 70% confluency in 6-well plates and either lentiCRISPR v2 expressing Cas9 or co-expressing an sgRNA targeting ATRX were co-transfected with the psPAX2 and pVSV-G packaging plasmids using Effectene Transfection Reagent per manufacturer’s instructions ( Qiagen ) . hTERT immortalized fibroblasts seeded in a 6-well plate were transduced with supernatant ( 1:1 ) harvested from HEK293T cells 48 hr post transfection that had been filtered through a 0 . 45 µm sterile syringe filter ( MilliporeSigma; Burlington , MA ) . Transduced cells were maintained for two days then placed under puromycin selection ( 5 µg/ml ) and expanded for two weeks . Oligo sequences for anti-ATRX sgRNA expression: sgATRX 1 F TCTACGCAACCTTGGTCGAA sgATRX 1 R TTCGACCAAGGTTGCGTAGA sgATRX 2 F CGAAACTAACAGCTGAACCC sgATRX 2 R GGGTTCAGCTGTTAGTTTCG HFFs were grown to confluency in T-150 flasks , and then left at confluency for 2–3 days . The confluent HFFs were infected with HSV-1 HSV at a MOI of 10 and incubated at 34°C . At 4 hpi , the medium was replaced with fresh medium containing 0 . 5 µM of the nucleoside analog , EdC . At 34 hpi , the HSV-infected HFFs were washed twice with 5 mL DMEV per wash , and 20 mL of fresh EdC-free media was added to the cells . The infected HFFs were incubated for an additional 2 hr at 34°C . At 36 hpi , cells were harvested by scraping the flasks in 2 . 5 mL DMEV and 2 . 5 mL sterile skim milk . The viral lysate was then sonicated using a Misonix S-4000 probe sonicator on setting 30 for 3 cycles: 30 s on , 30 s off . The HSV-EdC lysate was aliquoted and stored at −80°C . Infection of Vero cells with serially diluted virus was used to determine viral titer via plaque assay . To deplete cells of endogenous proteins , siRNAs were transfected into 1 × 105 cells plated in a 12-well dish using Lipofectamine RNAiMAX Reagent ( Invitrogen ) . At 48 hr post transfection , cells were re-seeded into assay-appropriate dishes and infected 24 hr later . siRNAs used in this study: siNT: Dharmacon ON-TARGETplus Non-targeting pool D-001810–10 siATRX: Dharmacon ON-TARGETplus ATRX pool D-006524–00 siIFI16: Dharmacon ON-TARGETplus IFI16 pool D-20004–00 siDAXX: Dharmacon ON-TARGETplus DAXX pool D-004420–10 The following antibodies were used in this study: IFI16 Abcam ab50004 PML Bethyl A301-167A ( immunoblot only ) PML Abcam ab96051 ( immunofluorescence only ) DAXX Sigma D7810 ( immunoblot only ) GAPDH Abcam ab8245 ( immunoblot only ) ICP8 ( 74 ) ICP4 ( 75 ) ICP0 East Coast Bio H1A207 ATRX Abcam 97508 H3 Abcam ab1791 ( ChIP ) H3K9me3 Abcam ab8898 ( ChIP ) H3K27me3 Active Motif 39156 ( ChIP ) H3 . 3 Millipore 09–838 ( ChIP ) Negative control rabbit IgG Millipore NG1893918 ( ChIP ) Cells were harvested at times indicated in lithium dodecyl sulfate ( LDS ) sample buffer and incubated at 95°C for 10 min . Protein samples were run on NuPAGE 4–12% bis-Tris gels ( Invitrogen ) . Proteins were transferred from the gel to nitrocellulose or PVDF membranes for imaging via LI-COR Odyssey imager ( Lincoln , NE ) or film , respectively . Membranes were blocked in either LI-COR blocking solution or 5% powdered milk-PBS containing 0 . 1% Tween 20 ( PBST ) for 1 hr . After blocking , membranes were incubated with primary antibody overnight at 4°C . Membranes were washed 3 times with PBST , incubated in secondary antibody for 1 hr at room temperature ( IRDye 800CW Goat anti-Mouse IgG and IRDye 800CW Goat anti-Rabbit IgG for LI-COR; α-mouse IgG+HRP and α-Rabbit IgG+HRP , Cell Signaling 7076 s and 7074 s , respectively for film ) . Membranes were washed 3 times in PBST . Nitrocellulose membranes were imaged with a LI-COR Odyssey imager . PVDF membranes were incubated with SuperSignal West Pico chemiluminescent substrate ( Thermo Scientific ) and exposed to film . Transfected or infected cells were fixed with 2% methanol-free formaldehyde ( ThermoFisher ) , permeabilized with 0 . 1% Triton-X 100 . Permeabilized cells were incubated with primary antibodies for 30 min at 37°C and washed twice with 0 . 1% PBST for 5 min each , followed by one wash with PBS . Alexa Fluor 594- and 647-conjugated secondary antibodies ( Invitrogen; A11032 and a21246 , respectively ) and DAPI were incubated with cells for 30 min at 37°C . All antibodies were used at 1:500 . The coverslips were washed as described above and either mounted on slides with ProLong Gold antifade reagent ( Invitrogen ) or incubated with click chemistry reagents as follows . After incubation with secondary antibodies , coverslips were incubated with 10 µM biotin-picolyl azide ( Click Chemistry Tools , Scottsdale , AZ ) , 10 mM sodium ascorbate ( Sigma-Aldritch A4034 ) , and 2 mM CuSO4 ( Fisher ) for 2 hr in the dark . Coverslips were washed twice with 0 . 1% PBST and once with PBS . Coverslips were then incubated with Alexa Fluor 488-conjugated streptavidin ( 1:1000; Invitrogen S32354 ) for 30 min in the dark . Coverslips were washed twice in PBS and mounted as described above . Images were acquired using NIS-Elements AR imaging software ( Nikon ) controlling a Nikon Ti-E inverted microscope system using a Plan Apo 100x/1 . 45 objective with a Zila sCMOS camera ( Andor ) and SPECTRA X light engine ( Lumencor ) or a Nikon Ti-E inverted spinning disk confocal system using laser lines 488 , 514 , 561 and 640 and a Plan Apo 100x/1 . 45 objective . Image J ( FIJI ) was used to minimally adjust contrast and generate 3D projections of exported images . The Image and Data Analysis Core ( IDAC ) at Harvard Medical School developed a custom MATLAB-based software for nuclear foci and co-localization detection in microscopy images based upon their previously described nuclear foci detection software ( Cicconet et al . , 2017; Cicconet , 2017 ) . In brief , unaltered images were analyzed by the software to detect signal intensity over a set threshold to define foci above background . The software then uses DAPI staining to generate a mask that defines nuclear areas in the image . Only foci above the fluorescence threshold and within the nuclear masks were scored . The software determines colocalization based on the distance between the centers of nuclear foci in a reference channel ( vDNA ) to foci in other channels of the image . The centers of compared foci must be ≤ 5 pixels ( ~350 nm at approximately 70 nm per pixel ) in distance to be considered colocalized ( user definable ) . An additional MATLAB script uses nuclear masks and foci center x , y coordinate data from each channel to determine distances of nearest neighbor foci in a second channel from foci in the reference channel and generate distance frequency data ( Cicconet , 2018 ) . The script then generated the same number of random x , y points within the nuclear mask as were detected in the second channel . The distances from the centers of reference channel foci to nearest neighbor random points was then calculated and reported to within a user defined radius ( 60 pixel radius for data in this paper ) of each reference focus . The resulting frequency distributions of reference-to-test channel distances and reference-to-random points were compared by the non-parametric Kolmogorov-Smirnov test . If the test rejects the null hypothesis at 5% significance , the test returns a value of 1 , and 0 if the null hypothesis is not rejected . The test also returns an asymptotic p-value and a list of distances in pixel values which can then be plotted and further analyzed using GraphPad Prism ( as it was here ) or other data analysis packages . Foci detection and colocalization software and nearest neighbor analysis script are available for use: doi:10 . 5061/dryad . 95fs76f . Total RNA was isolated using the Qiagen RNeasy kit per manufacturer’s instructions . Resulting RNA was quantified , treated with DNAse ( DNAfree Kit , Ambion ) , and reverse transcribed ( High Capacity cDNA RT Kit , Applied Biosystems ) to produce cDNA . Quantitative PCR was performed using Fast SYBR Green reagents ( ThermoFisher ) on a StepOnePlus from Applied Biosystems ( ThermoFisher ) . Oligos for qPCR reactions are as follows: ICP27 ( mRNA ) for GCATCCTTCGTGTTTGTCATT ICP27 ( mRNA ) rev GCATCTTCTCTCCGACCCCG ICP8 ( mRNA ) for GTCGTTACCGAGGGCTTCAA ICP8 ( mRNA ) rev GTTACCTTGTCCGAGCCTCC gB ( mRNA ) for TGTGTACATGTCCCCGTTTTACG gB ( mRNA ) rev GCGTAGAAGCCGTCAACCT 18S ( mRNA ) for GCCGCTAGAGGTGAAATTCTTG 18S ( mRNA ) rev CTTTCGCTCTGGTCCGTCTT ChIP experiments were carried out as previously described in detail ( Lee et al . , 2016 ) with minor changes . Briefly , infected cell monolayers were fixed in 1% formaldehyde ( Sigma-Aldrich ) at 37°C for 15 min . Fixation was quenched with glycine at a final concentration of 125 mM for 3 min . Cells were washed 3 times in ice cold PBS supplemented with 1 mM phenylmethanesulfonylfluoride ( PMSF ) , collected in PBS + PMSF , and pelleted at 2000 rpm for 5 min at 4°C . Cell pellets were re-suspended in 1 mL lysis buffer ( 1% SDS , 10 mM EDTA , 50 mM Tris , pH 8 . 1 ) , transferred to 15 mL polystyrene tubes and sonicated at 4°C in a Diagenode Biorupter on high setting for 9 cycles of 5 min each ( 30 s ON , 30 s OFF ) . An aliquot of recovered chromatin was used for each IP reaction in 1 ml IP dilution buffer ( 150 mM NaCl , 10 mM Na2HPO4 , 2 mM EDTA , 1 . 1% Triton , 0 . 1% SDS ) . An aliquot of a no antibody IP reaction was retained as 10% input sample . IPs were performed with 2 . 5 μg ( 6 . 0 μg for ATRX ab ) of ChIP grade antibody ( see Antibodies ) per reaction and incubated overnight at 4°C . 20 μl of MagnaChIP protein A magnetic beads ( Millipore ) were added for 2 . 5–3 hr at 4°C with rotation , washed 3 times with a cold low-salt buffer ( 150 mM NaCl , 20 mM Tris·HCl , pH 8 . 1 , 2 mM EDTA , 1% Triton X-100 , and 0 . 1% SDS , 1 mM PMSF ) and 3 times with cold LiCl wash buffer ( 50 mM HEPES pH 7 . 5 , 250 mM lithium chloride , 1 mM EDTA , 1% NP-40 , 0 . 7% sodium deoxycholate , 1 mM PMSF ) , and washed once with cold Tris-EDTA buffer ( 10 mM Tris-HCl , pH 8 , 1 mM EDTA ) . DNA-protein complexes were eluted in 100 μl of elution buffer ( 1% SDS , 0 . 1 M NaHCO3 ) at 65°C for 20 min . Protein-DNA crosslinks were reversed by adding NaCl ( 0 . 2 M final concentration ) and incubating at 95°C for 30 min followed by 1 hr RNase A ( Ambion ) treatment at 37°C and Proteinase K treatment ( Roche ) at 45°C for 2 hr . DNA was purified using a QIAquick PCR purification kit ( Qiagen ) end eluted twice with 50 μl buffer EB for a final volume of 100 μl . Quantitative-PCR was performed to quantify DNA using Fast SYBR Green reagents ( ThermoFisher ) on a StepOnePlus from Applied Biosystems ( ThermoFisher ) . Oligos for qPCR reactions are as follows: | Cells carefully package their DNA , tightly wrapping the long , stringy molecule around spool-like groups of proteins called histones . However , the genes that are draped around histones are effectively silenced , because they are ‘hidden’ from the molecular actors that read the genetic information to create proteins . A cell can control which of its genes are active by using proteins to move histones on or off specific portions of DNA . For example , a protein known as ATRX associates with a partner to load histones onto precise DNA regions and switch them off . Wrapping DNA around histones can also be a defense mechanism against viruses , which are tiny cellular parasites that hijack the molecular machinery of a cell to create more of themselves . For instance , the herpes simplex virus , which causes cold sores and genital herpes , injects its DNA into a cell where it is used as a template to create new viral particles . By packaging the DNA of the virus around histones , the cell ensures that this foreign genetic information cannot be used to make more invaders . However , the details of this process remain unknown . In particular , it is still unclear what happens immediately after the virus penetrates the nucleus , the compartment that shelters the DNA of the cell . Here , Cabral et al . explored this question by dissecting the role of ATRX in silencing the genetic information of the herpes simplex virus . The viral DNA was labeled while inside the virus itself , and then tracked using microscopy imaging techniques as it made its way into the cell and inside the nucleus . This revealed that , almost immediately after the viral DNA had entered the nucleus , ATRX came in contact with the foreign molecule . One possibility was that ATRX would be responsible for loading certain forms of histones onto the viral DNA . However , after Cabral et al . deleted ATRX from the cell , histones were still present on the genetic information of the virus , but this association was less stable . This indicated that ATRX was only required to keep histones latched onto the viral DNA , but not to load the proteins in the first place . Overall , these results show that using histones to silence viral DNA in done in several steps: first , the foreign genetic material needs to be recognized , then histones have to be attached , and finally molecular actors should be recruited to keep histones onto the DNA . Knowing how cells ward off the herpes simplex virus could help us find ways to ‘boost’ this defense mechanism . Armed with this knowledge , we could also begin to understand why certain people are more likely to be infected by this virus . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | ATRX promotes maintenance of herpes simplex virus heterochromatin during chromatin stress |
Mutations in the synaptic gene SHANK3 lead to a neurodevelopmental disorder known as Phelan-McDermid syndrome ( PMS ) . PMS is a relatively common monogenic and highly penetrant cause of autism spectrum disorder ( ASD ) and intellectual disability ( ID ) , and frequently presents with attention deficits . The underlying neurobiology of PMS is not fully known and pharmacological treatments for core symptoms do not exist . Here , we report the production and characterization of a Shank3-deficient rat model of PMS , with a genetic alteration similar to a human SHANK3 mutation . We show that Shank3-deficient rats exhibit impaired long-term social recognition memory and attention , and reduced synaptic plasticity in the hippocampal-medial prefrontal cortex pathway . These deficits were attenuated with oxytocin treatment . The effect of oxytocin on reversing non-social attention deficits is a particularly novel finding , and the results implicate an oxytocinergic contribution in this genetically defined subtype of ASD and ID , suggesting an individualized therapeutic approach for PMS .
Phelan McDermid syndrome ( PMS ) is a neurodevelopmental disorder characterized by intellectual disability ( ID ) , absent or delayed speech , neonatal hypotonia , attention deficits and autism spectrum disorder ( ASD ) . The neurobehavioral manifestations of PMS are caused by heterozygous mutations/deletions in the SHANK3 gene leading to a reduced expression of the SHANK3 protein . Shank3 is a key structural component of the glutamatergic postsynaptic density ( PSD ) , and interacts with glutamate receptors and cytoskeletal elements to regulate glutamate signaling and synaptic plasticity ( Kreienkamp , 2008 ) . It has been estimated that more than 80% of individuals with PMS meet ASD diagnostic criteria and that ~0 . 5–1% of ASD cases , 1–2% of ID cases , and up to 2% of cases with both ASD and ID harbor a SHANK3 mutation , which makes it one of the more common single locus causes of ASD and ID ( Gong et al . , 2012; Soorya et al . , 2013; Leblond et al . , 2014 ) . Despite its prevalence , PMS is less well studied than other single locus genetic disorders such as Fragile X or Rett syndromes . To date , no pharmaceutical compounds targeting core symptoms of PMS are available . To address this lack of effective therapeutics , several mouse lines with distinct Shank3 gene mutations have been developed to help understand the neurobiology of the syndrome and as a means of ultimately developing and assessing potential therapeutics ( Bozdagi et al . , 2010; Peça et al . , 2011; Wang et al . , 2011; Yang et al . , 2012; Kouser et al . , 2013; Kloth et al . , 2015; Bidinosti et al . , 2016; Mei et al . , 2016; Wang et al . , 2016 ) . The various Shank3-deficient mouse models have displayed PMS and ASD-related behavioral phenotypes including impaired social behavior , increased repetitive behaviors , and motor deficits , as well as altered synaptic transmission and neuronal morphology in the brain ( Bozdagi et al . , 2010; Peça et al . , 2011; Wang et al . , 2011; Yang et al . , 2012; Kouser et al . , 2013; Drapeau et al . , 2014; Wang et al . , 2016 ) . Recently , the translational relevance of these mouse models has been highlighted by our observation that the hormone IGF-1 improves motor and synaptic deficits observed in a Shank3-deficient mouse line ( Bozdagi et al . , 2013 ) , a result which directly led to a safety and preliminary efficacy clinical trial of IGF-1 in children with PMS ( Kolevzon , 2014a ) . Here , we report the generation and characterization of the Shank3-deficient rat , representing a novel genetic model of PMS , which demonstrates clear PMS-related behavioral and electrophysiological phenotypes that can be ameliorated by intracerebroventricular ( ICV ) oxytocin administration .
Founder Shank3-deficient rats were generated using zinc-finger nucleases ( ZFN ) technology , targeting exon 6 of the ankyrin repeat domain . This domain was targeted because five patients had been described with mutations in it ( Figure 1A ) ( Durand et al . , 2007; Moessner et al . , 2007; Hamdan et al . , 2011; Boccuto et al . , 2013; Yuen et al . , 2015 ) . Interestingly , the predicted truncated protein generated upon ZFN targeting of the rat Shank3 gene is quite similar to one of the human mutations that have been described ( Hamdan et al . , 2011 ) ( Figure 1A , middle sequence ) . In rat , this mutation leads to a significant reduction in the number of Shank3 transcripts ( Figure 1—figure supplement 1A ) and reduces expression levels of the full-length Shank3a protein ( Figure 1B and Figure 1—figure supplement 1B ) . We also observed that the levels of the PSD scaffolding protein Homer1 are decreased in the Shank3-deficient rats ( Figure 1—figure supplement 1C ) , consistent with changes in the PSD and with well-replicated findings from Shank3-deficient mice . We noted that , at weaning , there was a modest reduction in the number of homozygous knockout ( KO ) animals from Heterozygous ( Het ) x Het matings , compared to expectation ( 292:555:200 , corresponding to ratios of 0 . 53:1:0 . 36 ) . 10 . 7554/eLife . 18904 . 003Figure 1 . Gene-targeting of the rat Shank3 gene . ( A ) The top schematic shows the Shank3 protein domains [ankyrin repeats domain ( ANK ) , a Src homology 3 ( SH3 ) domain , a PDZ domain , and a sterile α-motif ( SAM ) domain] and indicates the published de novo mutations observed within the ANK domain in PMS patients ( light blue text , top schematic ) ( Durand et al . , 2007; Moessner et al . , 2007; Hamdan et al . , 2011; Boccuto et al . , 2013; Yuen et al . , 2015 ) . The 68 bp deletion that we introduced in exon 6 of the Shank3 rat gene ( middle schematic ) produces a stop codon in exon six that truncates the Shank3 protein . In a PMS patient , the c . 601–1G>A mutation in intron 5 of the SHANK3 gene abolishes the normal acceptor site and leads to utilization of a cryptic acceptor site introducing a premature stop codon in exon 6 ( red lines , bottom schematic ) , which also results in a similar truncated Shank3 protein ( Hamdan et al . , 2011 ) . Lower-case letters , genomic sequence; upper-case letters , amino acids; * , premature stop codons . ( B ) Western blot showing anti-Shank3 staining , using antibodies targeted against the SH3 domain . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 00310 . 7554/eLife . 18904 . 004Figure 1—figure supplement 1 . The introduced mutation in rat targets exon 6 and leads to reduced overall Shank3 transcripts and to decreased levels of the full-length Shank3 and Homer proteins . ( A ) Schematic based on the integrative genome browser ( IGB ) output from RNA sequencing ( RNAseq ) of a wild-type ( WT ) and a knockout ( KO ) sample , showing the mapped reads at exons 4 to 7 of the Shank3 gene and the deletion in exon 6 . ( B ) Left; Western blot analyses using anti-Shank3 antibodies ( targeted toward amino acids 840–857 ) and anti-βIII-tubulin in PFC postsynaptic density ( PSD ) samples in WT and Shank3-heterozygous ( Het ) and Knockout ( KO ) rats . Right; quantification of Shank3 protein levels , based on Western blot analysis of three samples per genotype normalized to βIII-tubulin and then to the WT levels . ( C ) Left; Western blot analyses using anti-Homer-1b/c antibodies and anti-βIII-tubulin in PFC-PSD samples in WT and Shank3-Het and KO rats . Right; quantification of Homer protein levels , based on Western blot analysis of three samples per genotype normalized to βIII-tubulin and then to the WT levels . *** , p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 004 To evaluate the impact of Shank3-deficiency in rats , basic developmental processes as well as early motor and sensory function were assessed . These assessments were carried out as previously described ( Brunner et al . , 2015 ) and included weight , stomach milk content , body temperature , locomotion , grooming , rearing , pup ultrasonic vocalization , geotaxis , and the righting reflex . We found no genotype-related deficits in these basic developmental or functional processes . In addition , when tested on the elevated plus maze , Shank3-Het and KO rats did not exhibit increased anxiety-like behaviors ( See Materials and methods for details and Supplementary file 1 for results ) . These results enabled us to examine more complex PMS and ASD-relevant behaviors . We first measured preference for a social stimulus versus an object ( Figure 2—figure supplement 1A ) , as well as juvenile social play ( Figure 2—figure supplement 1B ) and adult dyadic social interactions ( Supplementary file 1 ) in freely interacting animals . We found no differences between Shank3-deficient and wildtype ( WT ) rats in any of these measures of social behavior or social preference . Next , we used social habituation-dishabituation and social discrimination ( SD ) tests to examine social recognition memory ( SRM ) . In the SD experiments , the subject rat was ultimately tested in its ability to discriminate between familiar and unfamiliar juvenile rats simultaneously introduced to it for 5 min . We used two versions of the SD test , in order to examine both short- and long-term SRM ( with short and long referring to the interval between social memory acquisition and recall ) . To assess short-term SRM , the SD test was performed 30 min after the subject encountered the familiar rat for 5 min ( Figure 2A ) , while , in the more challenging long-term SRM , which requires longer exposure to the social stimulus ( Gur et al . , 2014 ) , the same test was performed 24 hr after a 1 hr encounter with the familiar rats ( Figure 2B ) . We found that Shank3-deficiency does not impair short-term SRM as shown by the comparable findings between WT and Shank3-deficient rats on the social habituation-dishabituation ( Figure 2—figure supplement 1C ) and short-term SD tests ( Figure 2A ) . In contrast , in two independent cohorts , we found that Shank3-Het and KO rats were unable to discriminate between novel and familiar social stimuli in the long-term SD test ( Figure 2B ) as they spent equal time investigating both the familiar and novel social stimuli . Notably , total investigation time ( toward both familiar and unfamiliar rats ) did not differ across genotypes in any of the SD tests ( Figure 2—figure supplement 1D ) , indicating there is not a decreased interest in social exploration . Moreover , the fact that the Shank3-Het and KO rats performed well on the short-term SD test and were able to perceive and remember their conspecifics , even when only given 5 min for the first interaction , also ruled out perceptual deficits . 10 . 7554/eLife . 18904 . 005Figure 2 . Shank3-deficient rats exhibit deficits in social memory . ( A–B ) Above each figure are schematics of the short- and long-term social discrimination ( SD ) paradigms . The examined adult subject is shown in white , and the juveniles are shown in black and grey . Bar plots show behavior in short-term ( WT , n = 12; Het , n = 13; KO , n = 9 ) and long-term ( WT , n = 31; Het , n = 37: KO , n = 25 ) SD tests . For long-term SD , similar results were observed in two independent cohorts , therefore the results were pooled and presented here as a single cohort . Bars ± SEM at left show test subjects average investigation time of a familiar and unfamiliar juvenile rat . The light overlaid gray lines in A and B show the corresponding individual subject data that comprise each bar . Right scatter plots , presented with mean ± SEM , show the ratio of the investigation time ( RDI= ( Unfamiliar-Familiar ) / ( Unfamiliar+Familiar ) for individual subjects . ( C ) Above the figure is a schematic of the long-term object location memory paradigm . The same plotting conventions as bar plots in A and B are used , but here they quantify investigation times of an object ( WT , n = 12; Het , n = 12; KO , n = 12 ) in a novel or familiar location . ( D ) Above , Contextual fear conditioning paradigm schematic . Scatter plots ( mean ± SEM ) represent percent time freezing during retrieval of a 1-day-old conditioned fear memory ( WT , n = 12; Het , n = 12; KO , n = 6 ) . * , p<0 . 05 , ** , p<0 . 01 , *** , p<0 . 001; see Supplementary file 1 for detailed statistical results . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 00510 . 7554/eLife . 18904 . 006Figure 2—figure supplement 1 . General social behaviors are unaffected in Shank3-deficient rats . ( A ) Bars ± SEM show the average investigation times of simultaneously presented juvenile rat ( gray bars ) and object ( black bars ) for WT ( n = 12 ) , Shank3-Het ( n = 12 ) and KO ( n = 11 ) during the social preference test . The light overlaid gray lines show the corresponding individual subject data that contribute to each bar . Scatter plots on the right show the ratio of these investigation times ( RDI = ( Social- object ) / ( Social+ object ) for individual subjects and the group mean ± SEM in thin overlaid black lines . ( B ) Same plot conventions as in A , but here points show individual rat data for the number of social and non-social behaviors during the juvenile social play assessment in WT ( left , n = 11 ) and KO rats ( right , n = 11 ) . ( C ) Performance of WT ( n = 11 ) , Shank3-Het ( n = 11 ) and Shank3-KO ( n = 10 ) rats on the habituation-dishabituation social recognition , see Materials and methods for a detailed description of this task . ( D ) Identical plotting conventions as in the scatter plots in Figure 2A , B but points show the total investigation time ( the sum of investigation time of both the familiar and unfamiliar juvenile rat ) for each individual subject on the short-term SRM ( left ) and long-term SRM ( right ) paradigms . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . See Materials and methods section and Supplementary file 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 006 To determine whether this observed impairment is selective to social memory or if it also involves more general memory processes , we tested the rats on two long-term non-social memory paradigms that , similarly to the SD test , are known to be hippocampal-dependent , specifically , the object location memory test and the contextual fear conditioning memory test . In contrast to the impaired behavior in the long-term SD test , Shank3-Het and KO rats performed similarly to their WT littermates in the object location memory and contextual fear conditioning memory tests ( Figure 2C and D ) . These results indicate that Shank3 deficiency selectively impairs long-term social memory , but leaves intact both short-term social memory and non-social long-term memory . Attention deficits are often associated with PMS . Thus , we assessed performance in the attentionally demanding 5-choice serial reaction time ( 5-CSRT ) task in which rats must respond quickly to briefly presented light cues ( Figure 3 ) ( Mar et al . , 2013 ) . This task requires training the rats in stages where the duration of the light stimulus is slowly decreased from 32 to 1 s by halving the stimulus duration across sessions once performance criteria are met ( i . e . accuracy rates higher than 80% for two consecutive days with omission rates lower than 20% ) . Shank3-Het and KO rats learned the task and were able to reach baseline , similar to WT controls . However , both the Shank3 Het and KO rats had lower accuracy and lower omission rates , when compared to WT rats , even after extensive training . Moreover , after reaching baseline criterion Shank3-deficient rats did not maintain even this level of performance across the 10-day test period , during which they performed significantly fewer correct trials ( Figure 3A ) , made more errors ( Figure 3B ) , and exhibited higher omission rates than WT rats ( Figure 3C ) . Even on trials with a correct response , Shank3-deficient rats responded more slowly and with more variable latencies than WT rats ( Figure 3D ) . 10 . 7554/eLife . 18904 . 007Figure 3 . Shank3-deficient rats exhibit deficits in attention . ( A ) Traces and clouds indicate mean percentage of trials with a correct response ± SEM ( WT , n = 10; Het , n = 13; KO , n = 12 ) across 10 5-CSRT sessions . The right side in all panels is the cross-rat median ( dot ) and middle quartiles ( vertical lines ) . ( B ) Traces represent mean percentage of trials where an incorrect response was made . ( C ) Mean percentage of trials with no cued response . ( D ) Average reaction times on trials with a correct response . Results were observed in two independent cohorts; therefore , the results were pooled . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 00710 . 7554/eLife . 18904 . 008Figure 3—figure supplement 1 . Shank3-deficient rats exhibit normal motivation for reward and task performance is at WT levels when cue durations are long . ( A ) Traces and clouds show the average latency ± SEM from a correct screen touch to entry into a reward receptacle for WT ( black , n = 10 ) , Shank3 Het ( blue , n = 13 ) , or KO rats ( red , n = 12 ) across ten 5-CSRT sessions . ( B ) Scatter plots , presented with mean ± SEM , show percent of correct , incorrect , or omitted trials for each individual subject during this 5-CSRT test training stage . In this 5-CSRT training stage , the light cues remain on for 5 s . *p<0 . 05 , **p<0 . 01; See Supplementary file 1 for detailed statistical results . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 008 Slow , inaccurate and omitted responses to very brief visual stimuli are commonly interpreted as reflecting an attention deficit ( Robbins , 2002 ) . While changes in accuracy might also be attributed to deficits in sensory perception , we excluded this possibility by carrying out studies with less bright visual cues and Shank3-deficient rats performed at WT levels ( not shown ) . It was only when the duration of the light cues was shortened that the deficits were manifested , which indicates impaired vigilance and/or spatial attention . Furthermore , these deficits were not due to decreased motivation for food , because the latency of Shank3-deficient rats to collect reward after a correct response was similar to WT rats ( Figure 3—figure supplement 1A ) , as was task performance when light cues were of a longer duration ( Figure 3—figure supplement 1B ) . In summary , these results indicate that Shank3-deficient rats are impaired in an attentionally demanding task , which , when considered together with the deficits in long-term social memory , indicates this model demonstrates face validity for some features of PMS . Memory deficits , while not considered core symptoms of autism , have been associated with ASD ( Boucher et al . , 2012 ) . Both impaired working ( Barendse et al . , 2013 ) and episodic memory ( Maister et al . , 2013 ) have been observed in human subjects with ASD , which has been attributed to aberrant connectivity of the hippocampus and medial prefrontal cortex ( mPFC ) ( Ben Shalom , 2003 ) . The behavioral deficits we observe in Shank3-deficient rats are consistent with dysfunction in these circuits . The hippocampus and mPFC are both important for SRM ( Watson et al . , 2012; Harvey and Lepage , 2014; Jacobs and Tsien , 2014 ) . Moreover , performance in the 5-CSRT task depends on mPFC function ( Rogers et al . , 2001 ) and attention , working memory , and decision-making require intact hippocampal-prefrontal functional connectivity ( Jones and Wilson , 2005 ) . We therefore evaluated the effect of Shank3 deficiency on synaptic function and plasticity in hippocampal-PFC circuitry . Extracellular field excitatory postsynaptic potential ( fESPs ) recordings at Schaffer collateral-CA1 synapses were similar between genotypes , with no differences in paired-pulse facilitation ( Supplementary file 1 ) or the input-output relationship ( Figure 4—figure supplement 1 ) . These results suggest that basal synaptic transmission is generally intact in Shank3-deficient rats . We found , however , in independent cohorts , that plasticity at these synapses was not intact . Long-term potentiation ( LTP ) induced by high-frequency stimulation ( HFS ) was reduced in both Shank3-Het and KO rats ( Figures 4A and 6A , and Figure 6—figure supplement 1A ) , while mGluR-dependent long-term depression ( LTD ) was reduced only in KO rats ( Figure 4B ) . 10 . 7554/eLife . 18904 . 009Figure 4 . Synaptic plasticity is impaired in Shank3-deficient rats . ( A ) High-frequency stimulation ( HFS , arrow ) -induced long-term potentiation ( LTP ) at hippocampal Schaffer collateral-CA1 synapses ( n = 6 rats/genotype , 1–2 slices per rat ) . ( B ) Long-term depression induced by the mGluR agonist DHPG ( 50 µM , 5 min ) is indicated by the horizontal line ( n = 6 rats/genotype , six slices per rat ) . ( C ) HFS-induced LTP in the prelimbic PFC after stimulation of ipsilateral CA1 in ventral hippocampus of intact anesthetized WT ( n = 5 ) , Shank3 Het ( n = 6 ) and KO ( n = 6 ) rats . Inset shows a schematic of the target location of the stimulating and recording electrodes in vivo . Summary data are presented as mean ± SD . *p<0 . 05 , ***p<0 . 001; See Supplementary file 1 for detailed statistical results . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 00910 . 7554/eLife . 18904 . 010Figure 4—figure supplement 1 . Basal synaptic transmission is intact in Shank3-deficient rats . Input-output relationship at Schaffer collateral-CA1 synapses ( n = 5 rats/genotype , three slices per rat ) . Summary data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 010 Innervation of the mPFC by the hippocampus ( Marquis et al . , 2006 ) is important for attention and working memory , both of which we have shown here are impaired in Shank3-deficient rats . We therefore assessed hippocampal-PFC synaptic transmission in vivo by stimulating hippocampal CA1/subicular regions and recording fEPSPs in the prelimbic area of the PFC in anesthetized rats . We found that LTP was reduced in both Shank3-Het and KO rats ( Figure 4C ) . We also observed that there were no genotype associated differences in the input-output relationship of evoked local field potentials in coronal PFC slices , which demonstrates that these changes were specific to hippocampal-prefrontal circuitry ( Supplementary file 1 ) . In summary , these results suggest that Shank3-deficiency impairs plasticity in both the projections from hippocampus to the PFC and within intrinsic hippocampal circuits . Oxytocin has a central role in SRM formation ( Ferguson et al . , 2000; Gur et al . , 2014 ) and hence we reasoned that it may underly some of the altered behaviors we observed; in addition and more broadly , oxytocin modulates mammalian social behavior and may be dysregulated in ASD ( Harony and Wagner , 2010 ) . To determine whether the behavioral deficits we observed in Shank3-deficient rats could be improved with a pharmacological intervention strategy , we tested Shank3-deficient rats on long-term SD and 5-CSRT tasks following an injection of either oxytocin or saline into the left lateral ventricle . We observed that oxytocin improved both the long-term social memory and attention deficits in Shank3-Het and KO rats ( Figure 5A and B ) . Notably , in WT animals , oxytocin had no effect on long-term SRM ( Figure 5—figure supplement 1A ) or 5-CSRT task performance ( Figure 5—figure supplement 1B ) . 10 . 7554/eLife . 18904 . 011Figure 5 . Oxytocin improves social memory and attentional deficits in Shank3-deficient rats . Above each figure is a schematic depicting the sequence of oxytocin administration and behavioral testing on the long-term social discrimination ( SD ) paradigm ( in A ) or 5-CSRT task ( in B ) . In A , the test subject is shown in white and the juveniles are shown in black and grey . ( A ) Left bars ( ± SEM ) show the test subject’s average investigation time of a familiar and unfamiliar juvenile rat on long-term SD test following treatment with oxytocin or vehicle for Shank3 Het ( n = 15 ) and KO ( n = 8 ) rats . The light overlaid gray lines show the corresponding individual subject data that comprise each bar . Bars on right show the ratio of the investigation time ( RDI= ( Unfamiliar-Familiar ) / ( Unfamiliar+Familiar ) for individual subjects with the light overlaid gray lines representing the corresponding individual subject data . ( B ) Bars ( ± SEM ) represent the percentage of correct ( left ) , omitted ( middle ) , and incorrect ( right ) trials in saline ( solid bars ) and oxytocin ( open bars ) of Shank3 Het ( n = 7 ) and KO ( n = 6 ) rats . Color conventions are identical to those used in Figure 3A–C . The light overlaid gray lines show the corresponding individual subject data that went into each bar . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001; See Supplementary file 1 for detailed statistical results . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 01110 . 7554/eLife . 18904 . 012Figure 5—figure supplement 1 . Oxytocin has no effect on social memory or attention in WT rats . ( A ) Performance on the long-term social discrimination test following treatment with oxytocin or vehicle in WT rats ( n = 7 ) . Plotting conventions are identical to the bar plots in Figure 2A–B . ( B ) Bars ± SEM represent the percentage of correct ( left ) , omitted ( middle ) , and incorrect ( right ) trials in saline ( solid bars ) and oxytocin ( open bars ) of WT rats ( n = 6 ) . Color conventions are identical to those used in Figure 3A–D . The light overlaid gray lines in A and B show corresponding individual subject data . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 012 Since oxytocin reversed the behavioral deficits in Shank3-deficient rats and has been previously shown to enhance the induction of synaptic plasticity in several systems ( Benelli et al . , 1995; Tomizawa et al . , 2003; Fang et al . , 2008; Ninan , 2011; Lin et al . , 2012; Gur et al . , 2014 ) , we also examined the effect of oxytocin on LTP both in vitro and in vivo . As previously reported , in acute hippocampal slices derived from WT rats , oxytocin enhanced LTP induction after a weak stimulation of one train of 100 Hz pulses ( Tomizawa et al . , 2003; Lin et al . , 2012 ) but we did not observe this oxytocin-dependent enhancement in Shank3-deficient rats ( Figure 6—figure supplement 1A ) . In WT slices , oxytocin had no effect on LTP induced by stronger stimulation ( 4 × 100 Hz stimulation trains ) , but it greatly enhanced LTP in slices prepared from Shank3-deficient rats ( in independent cohorts , Figure 6A and Figure 6—figure supplement 1B ) . In vivo , oxytocin also reversed the impaired LTP at hippocampal-prefrontal synapses in Shank3-deficient rats ( Figure 6B ) . These results indicate that oxytocin treatment restores LTP induction at hippocampal and hippocampal-prefrontal synapses in Shank3-deficient rats and is consistent with a mechanism whereby deficits in synaptic plasticity across hippocampal-prefrontal circuitry underly PMS-relevant behavioral deficits we observed in these animals . 10 . 7554/eLife . 18904 . 013Figure 6 . Oxytocin improves synaptic plasticity deficits in Shank3-deficient rats . ( A ) Traces depict Hippocampal HFS-induced LTP ( 4 × 100 Hz ) in WT and Shank3-deficient rats ( n = 4 rats/genotype , six slices per rat ) . Application of 1 µM of oxytocin is indicated by the horizontal line . ( B ) LTP in the hippocampal to prefrontal synaptic pathway recorded in vivo in WT ( n = 3 ) , Shank3 Het ( n = 4 ) and KO ( n = 4 ) rats . Oxytocin ( 2 ng ) or saline was administered in the lateral ventricle 5 min before LTP induction . ***p<0 . 001 . See Supplementary file 1 for detailed statistical results . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 01310 . 7554/eLife . 18904 . 014Figure 6—figure supplement 1 . Oxytocin enhances LTP induced by a single pulse at 100 Hz in WT but not in Shank3-deficient rats . ( A ) Traces show HFS-induced LTP ( 1 × 100 Hz ) recorded from hippocampal slices in WT and Shank3-deficient rats ( n = 4 rats/genotype , six slices per rat ) . Brain slices were treated with oxytocin ( 1 µM , indicated by the horizontal line ) . ( B ) These traces summarize the results from independent replication of the results shown in Figure 6A . The plotting conventions are identical as in Figure 6A , but here Ns in ACSF and oxytocin are; WT , n = 4 rats/5 slices per rat; Het , n = 4 rats/4 slices per rat; KO , n = 5 rats/6 slices per rat . Application of 1 µM of oxytocin is indicated by the horizontal line . See Supplementary file 1 for summary statistics . *p<0 . 05 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 18904 . 014
Deficits in attention and the presence of ASD-related behaviors are common in PMS . In our genetically modified rat model of PMS , we found that Shank3 deficiency impairs social memory and attention . These impairments are accompanied by attenuated synaptic plasticity within the hippocampus and in the monosynaptic pathway connecting the hippocampus to the PFC . Hippocampal synaptic plasticity deficits have also been reported in Shank3-mouse models and were associated with impairments in actin cytoskeleton remodeling and with changes in the levels of glutamatergic receptors and other PSD scaffolding components including Homer , which we also found to be decreased in the Shank3-deficient rat ( Figure 1—figure supplement 1C ) ( Bozdagi et al . , 2010; Wang et al . , 2011; Duffney et al . , 2013; Kouser et al . , 2013; Wang et al . , 2016 ) . The fact that long-term social memory was selectively impaired , in contrast to more general memory processes , may reflect the complexity of the information involved in social memory ( Adolphs , 2010; Wiley , 2013 ) , which may require more elaborate synaptic mechanisms than simpler forms of memory . These results may also relate to previous findings that correlate ASD-associated episodic memory deficits with the complexity of the memoranda ( Lind , 2010 ) , which suggests that sub-optimal thresholds for synaptic plasticity interfere with the ability of complex stimuli , such as social stimuli , to activate memory formation . Deficits in attentionally demanding tasks , like the 5-CSRT task , may also result from blunted synaptic plasticity , where brief events must be rapidly encoded and reliably stored by neural circuits to promote appropriate and timely behavior . It is interesting , and important , to compare findings in different rodent models of PMS and relate them to symptoms observed in patients . In particular , while there is some overlap between the social communicative deficits observed in PMS and those observed in what is considered more classical autism , there are important differences . In contrast to the more typical conception of ASD ( whether accurate or not ) , PMS-associated ASD is not associated with a social aversion or lack of social approach ( Soorya et al . , 2013; Kolevzon , 2014b ) . A recent study of functional neuroimaging showed that differential fMRI changes in response to social versus non-social sounds are preserved in PMS , in contrast to studies of idiopathic autism ( iASD ) ( Wang et al . , 2016 ) . These findings , together with the fact that not all individuals with PMS are diagnosed with ASD , indicate that animal models for PMS should not necessarily present with social behavioral deficits and imply that behavioral phenotypic diversity in PMS patients may reflect additional processes . This may explain some of the discrepancies that have been reported across distinct Shank3-mouse models ( Harony-Nicolas et al . , 2015 ) , and those that we report here in the Shank3-deficient rat . These discrepancies could be attributed to ( 1 ) mutations/deletions in existing rodent models were targeted to different parts of the Shank3 gene , ( 2 ) varying genetic background , ( 3 ) different behavioral paradigms , ( 4 ) varying age and sex , ( 5 ) a focus on heterozygotes or knockouts , or other factors . Beyond these methodological issues , from an evolutionary perspective , mice and rats are separated by millions of years , which likely explain many of the observed discrepancies in behavioral repertoires between the two species . Just as mice and rats differ , so do mice and rats greatly differ from human , and thus one should be careful not to overly anthropomorphize these model systems or look for perfect overlap across species . Further investigation of the mechanisms underlying synaptic plasticity deficits and the molecular pathways affected by Shank3 mutation and the consequent synaptic changes in rat and mouse Shank3 models will likely provide new targets for therapeutic treatments and allow comparison between two different species , thus providing more reliable molecular targets for future drug development studies . Although all the mutations that have been studied , including the one we introduced to the rat model , lead to the deletion of the longest Shank3 isoform , most leave intact other shorter isoforms , whose role is still unclear . Because many patients lack the entire SHANK3 gene , understanding the role of different Shank3 isoforms is of great interest , and contrasting existing rodent models will be useful in clarifying this important question . To date , treatment strategies for PMS are non-specific and do not address key issues of cognition , language , motor development , etc . There are approaches that are applied in PMS that are adopted from the general population . For example , when ASD is present , one might make use of behavioral interventions ( Kolevzon et al . , 2014b ) , but these strategies need to be improved and designed for the specific needs of PMS patients . Recently , the field has begun to translate basic neurobiological findings gleaned from rodent models into promising pharmacological treatments for a host of genetic disorders such as Fragile X syndrome , Tuberous sclerosis , Rett syndrome , and PMS ( Bear et al . , 2004; Ehninger and Silva , 2009; Tropea et al . , 2009; Kolevzon et al . , 2014a ) . The peptide oxytocin is a powerful regulator of mammalian social behavior , which has been shown to improve various aspects of social cognition and social behavior in human and non-human primates , by increasing social memory , enhancing social reward , and modulating social attention ( Chang et al . , 2012; Guastella et al . , 2012; Parr et al . , 2016 ) . Oxytocin has been also studied in clinical trials as a potential therapeutic for iASD-associated social impairments , yet these efforts have led to equivocal results ( Guastella et al . , 2016 ) . Hitherto , the effect of oxytocin has not been investigated in PMS or in animal models of PMS , and its effect on non-social ASD and PMS-associated deficits has not been evaluated . Given the evidence for a key role for oxytocin in SRM , we tested this compound and observed important effects on both behavioral and electrophysiological deficits . Interestingly , an ameliorative effect of oxytocin treatment on autism-like behaviors has recently been reported in other genetically defined mouse models of ASD ( Tyzio et al . , 2014; Peñagarikano et al . , 2015 ) . Our results show , for the first time to our knowledge , a beneficial effect of oxytocin on attention to non-social stimuli . Social and attention deficits are often co-morbid in PMS ( Soorya et al . , 2013; Kolevzon , 2014b ) , which is perhaps reflective of a shared etiology and may explain why oxytocin improved both deficit types in Shank3 deficient rats . The ameliorative effect of oxytocin treatment on the long-term social memory , attention and synaptic plasticity deficits in Shank3-deficient rats , support a mechanism whereby deficits in synaptic plasticity across hippocampal-prefrontal circuitry may underly the behavioral deficits we observed . These results also imply that exogenous oxytocin administration might have therapeutic potential in human PMS patients . In our view , genetically modified rat models are especially valuable for behavioral and functional studies . They have several species-specific advantages over mouse models , including a more complex behavioral repertoire and larger brains that readily facilitate high-density electrophysiological recordings . Moreover , rats remain the primary choice of the pharmaceutical industry for studying the pharmacokinetic properties of novel drugs that may have therapeutic potential in human . Our production of a Shank3-deficient rat model , which demonstrates both construct and face validity , will pave the way for future studies that investigate the mode by which Shank3 mutations alter brain activity during behavior and to further study the effects of oxytocin and other potential therapeutics for PMS . Our findings open new avenues of research to study the effect of Shank3-deficiency on the development and function of oxytocin neurons . Future studies should investigate if Shank3-deficiency affects hypothalamic processes that regulate oxytocin production and secretion ( Brownstein et al . , 1980 ) or whether it perturbs pathways further downstream . Additionally , it would be important to determine if oxytocin treatments during critical developmental windows could ameliorate PMS-associated behavioral deficits , and whether other forms of ASD , defined either etiologically or by the presence of specific biomarkers , may show such broad response to oxytocin . These , and other experiments , would support the use of oxytocinergic agonists for multiple forms of ASD , including additional genetically defined forms of ASD .
Shank3-deficient rats were generated using zinc-finger nucleases ( ZFN ) on the outbred Sprague-Dawley background . The design , cloning , and validation of the ZFN , as well as embryonic microinjection and screening for positive founder rats was performed by SAGE Labs ( Boyertown , PA ) . Briefly , sixteen pairs of ZFN were designed targeting exons 4 , 5 and 6 of the Shank3 gene to disrupt the Shank3 ANK domain . These pairs were assembled by PCR and sub cloned into the pZFN expression plasmid . ZFN were then transfected into the rat C6 cell line and tested for disruption activity using the Surveyor endonuclease ( CEL-1 ) assay ( Kulinski et al . , 2000 ) . The best performing ZFN pair GACGCCCCTGTACCATAGTgccctaGGGGGCGGGGATGCC ( recognizing the GACGCCCCTGTACCATAGT and the GGGGGCGGGGATGCC sequences located between 130163213–130163231 and 130163238–130163252 , respectively , in the Shank3 gene ( NCBI reference sequence NC_005106 . 3 ) , were used for embryo microinjection . Positive Sprague-Dawley founder animals were mated to produce F1 heterozygous breeder pairs . Genomic DNA sequencing using the ZFN Primer F ( GGAGGGACTCAATGCAGAAA ) identified a 68 bp deletion in exon six leading to a premature stop codon , as shown in Figure 1A . All rats were kept under veterinary supervision in a 12 hr light/dark cycle at 22 ± 2°C . Unless otherwise indicated , animals were pair-caged with food and water available ad libitum . All animal procedures were approved by the Institutional Animal Care and Use Committees at the Icahn School of Medicine at Mount Sinai , the University of Haifa , and the Albert Einstein College of Medicine . Age-matched males were used in all tests unless otherwise noted . Minimal group sample sizes were decided in advance of the described experiments based on the standards in the field and using on-line sample size/power calculation tools , specifically the ‘biomath’ online tool; ( http://biomath . info/power/ ) . Any animals excluded from analysis or testing due to poor performance or for any other reason are noted in the appropriate Methods section . Pregnant Shank3-Het rats were shipped to PsychoGenics ( Tarrytown , NY ) where they were checked daily for litters . On the day of birth ( defined as postnatal day 0 [P0] ) , the dam and her litter were left undisturbed . On P2 , pups were tattooed using non-toxic ink applied under the skin of their paw and a tail snip sample was taken for genotyping . All pups were weighed daily from P2 to P21 . Some animals were assessed for milk content score from P2 until their abdomen was covered by fur ( around P10 ) to look for genotypic differences in the ability to breast-feed . The milk score for each pup within a litter is dependent on the overall level of milk in all pups from that litter at the time of observation: 3 , normal milk content , normal milk level is defined in relation to littermates' milk content; 2 , stomach not full , but milk is easily detected; 1 , trace amounts of milk; 0 , absence of milk . Additionally , the age at which the animals’ eyes open was documented . Animals were checked for survival once per day and survival rates were recorded . Neonatal phenotyping assessments , including body temperature , isolation tests ( measuring frequencies of square crossing , pivot , grooming , rearing , ultrasonic vocalization , sniffing , geotaxis , and righting reflex were conducted at either P7 or P15 as previously described ( Brunner et al . , 2015 ) . All tests took place between 10:00 AM and 3:00 PM and were performed by an examiner blind to subject genotype ( Supplementary file 1 ) . At P15 , animals were tested in the geotaxis and righting reflex tests . For the SRM experiments , a guide cannula was implanted in naïve rats at 8 . 5–9 . 5 weeks of age . For the 5-CSRT task , surgery occurred after baseline performance was measured for 10 sessions ( after criterion performance was first met on with a 1 s cue duration ) . For surgery , rats were deeply anesthetized by subcutaneous injection of 10% ketamine ( 0 . 09 ml/100 g ) and Dormitor ( 0 . 05 ml/100 g ) or with isoflurane . Anesthetized animals were fixed in a stereotaxic apparatus with the head flat , and a small hole was drilled in the skull to facilitate implantation of guide cannula . The cannula was inserted 0 . 1 mm ( SRM task ) or 1 . 25 mm ( 5-CSRT task ) above the left lateral ventricle . The coordinates were AP = −1 mm , ML = 1 . 5 mm , DV = −3 . 6 mm ( SRM experiments ) or AP = −0 . 75 mm , ML = 1 . 75 mm , DV = −4 mm ( 5-CSRT task ) both targeting the left lateral ventricle . Cannula were secured in position with acrylic dental cement and stabilized by at least two bone screws . A dummy cannula was placed in the guide cannula to prevent the tube from clogging . Antibiotics ( amoxicilin 15% , 0 . 07 ml/100 g ) and a pain killer ( Calmagin 0 . 03 ml/100 g ) were administered directly before and immediately following surgery and one day post-surgery . Animals were allowed to recover for at least 7 days before experimentation began . Post-surgery , 5-CSRT task rats were retrained to stable performance before being subjected to microinjections . In the oxytocin 5-CSRT experiments , two implanted KOs , 1 Het and WT were not subjected to microinjections because they failed to achive pre-surgery performance levels . In the long-term SRM task , 1 WT and 2 KOs lost their caps , 1 Het and 1 KO died during surgery , and 1 Het was extremely stressed and thus was not subjected to oxytocin injection . All experiments were run and analyzed blind to genotype . The injectate for intracranial microinjections was prepared fresh daily by dissolving oxytocin ( American Peptide Company ) in 0 . 9% saline to achieve a final concentration of 250 nM ( 0 . 25 ng/µl ) ( Benelli et al . , 1995 ) . On test days , the fluid line and the microinjector were filled with mineral oil before drawing up the drug or saline control . Prior to microinjection , animals were gently restrained while the dummy cannula was removed and the microinjectors loaded with injectate were placed into the animal’s guide cannula . The injectate was administered at a rate of 1 µl/min for 4 min ( in total 1 ng of oxytocin was injected ) . Concentration was chosen based on published studies of the effect of ICV oxytocin injection on behavior ( Ferguson et al . , 2000; Gur et al . , 2014 ) . After injection , the microinjector remained in the guide cannula for 1 min to allow for diffusion . In the SRM task , microinjections occurred 10–20 min before the first social encounter ( Figure 5A and Figure 5—figure supplement 1A ) . In the 5-CSRT task , animals were immediately placed in the operant chamber after oxytocin administration and testing started 10 min after ( Figure 5B and Figure 5—figure supplement 1B ) . The order of saline and oxytocin injections was randomized across rats . At the end of each of experiment rats were sacrificed with an overdose of isoflurane , brains were removed and placed in 4% paraformaldehyde overnight , and 200-µm-thick sections were prepared . Correct positioning of the cannula over the left lateral ventricles was confirmed visually in each rat . SPSS ( SPSS 19; IBM ) or the R environment were used for statistical analyses . Parametric approaches such as t-tests or analysis of variance were used if data was found to be normally distributed by the Kolmogorov-Smirnov test . When the condition of normality was not met , non-parametric approaches for statistical comparisons were used such as Kruskal-Wallis or the Wilcoxon signed rank test . A threshold of p<0 . 05 was used to test statistical hypotheses . A detailed description of the results of all analyses , including the test used , sample size , mean or median , SEM or SD , and p values can be found in Supplementary file 1 . | Phelan-McDermid syndrome is a genetic disorder on the autism spectrum that affects how children develop in several ways , with additional symptoms including attention deficits , delays in learning to speak and motor problems . This syndrome is known to be caused by changes in a single gene known as SHANK3 that disrupt communication between brain cells involved in memory and learning . However , we do not know how these changes relate to the symptoms of Phelan-McDermid syndrome . To understand how genetic changes affect the human brain , researchers often carry out experiments in rats or other small rodents because they have brains that are similar to ours . Harony-Nicolas et al . genetically modified rats to carry changes in the SHANK3 gene that reflect those found in people with Phelan-McDermid syndrome . The rats had disabilities related to those seen in Phelan-McDermid syndrome , including limits in long-term social memory and reduced attention span . They also showed changes in the connections between important parts of the brain . Therefore , studying these rats could help us to understand the link between molecular and cellular changes in the brain and how they affect people with Phelan-McDermid syndrome , and associated symptoms . Previous studies have shown that a chemical called oxytocin , which is naturally produced by the brain , helps to form bonds between individuals and can cause positive feelings in relation to certain memories . Harony-Nicolas et al . found treating the rats with oxytocin boosted social memory and led to improvements in other symptoms of Phelan-McDermid syndrome . In particular , oxytocin treatment helped to increase the attention span of the rats . Rats with changes in the SHANK3 gene will be a useful tool for future research into Phelan-McDermid syndrome , particularly in understanding how it affects the connections between brain cells , leading to the symptoms of Phelan-McDermid syndrome . A future challenge will be to find out whether oxytocin has the potential to be developed into a therapy to treat Phelan-McDermid syndrome in humans . Since there is evidence that SHANK3 is involved in other forms of autism , these rats will also be useful in understanding the other ways in which autism can develop . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Oxytocin improves behavioral and electrophysiological deficits in a novel Shank3-deficient rat |
Synaptojanin and endophilin represent a classic pair of endocytic proteins that exhibit coordinated action during rapid synaptic vesicle endocytosis . Current models suggest that synaptojanin activity is tightly associated with endophilin through high-affinity binding between the synaptojanin proline-rich domain ( PRD ) and the endophilin SH3 domain . Surprisingly , we find that truncated synaptojanin lacking the PRD domain sustains normal synaptic transmission , indicating that synaptojanin's core function in vivo resides in the remaining two domains that contain phosphoinositide-phosphatase activities: an N-terminal Sac1 phosphatase domain and a 5-phosphatase domain . We further show that the Sac1 domain plays an unexpected role in targeting synaptojanin to synapses . The requirement for Sac1 is bypassed by tethering the synaptojanin 5-phophatase to the endophilin membrane-bending Bin–Amphiphysin–Rvs ( BAR ) domain . Together , our results uncover an unexpected role for the Sac1 domain in vivo in supporting coincident action between synaptojanin and endophilin at synapses .
Synaptic vesicle ( SV ) endocytosis occurs through rapid and coordinated actions of endocytic proteins ( De Camilli and Takei , 1996; Dittman and Ryan , 2009; Saheki and De Camilli , 2012 ) . A classic example is the functional pair of synaptojanin and endophilin ( Gad et al . , 2000; Schuske et al . , 2003; Song and Zinsmaier , 2003; Verstreken et al . , 2003; Dickman et al . , 2005; Milosevic et al . , 2011; Sullivan , 2011 ) . Synaptojanin is a neuronal phosphoinositide phosphatase that hydrolyzes phosphatidylinositol-4 , 5-bisphosphate ( PI ( 4 , 5 ) P2 ) to facilitate SV recycling at presynaptic terminals ( McPherson et al . , 1996; Cremona et al . , 1999; Harris et al . , 2000; Verstreken et al . , 2003 ) . Deletion of synaptojanin leads to severe synaptic defects , including depletion of SVs , accumulation of endocytic intermediates , and subsequent failure in synaptic transmission ( Cremona et al . , 1999; Harris et al . , 2000; Verstreken et al . , 2003; Van Epps et al . , 2004; Dickman et al . , 2005 ) . Overexpression of synaptojanin causes PI ( 4 , 5 ) P2 deficiency and learning deficits in Down syndrome model mice ( Voronov et al . , 2008 ) . While the importance of synaptojanin is well documented , the precise mechanisms for its role in SV recycling remain elusive . Genetic studies have shown that the function of synaptojanin is tightly linked to the endocytic protein endophilin ( Verstreken et al . , 2002; Schuske et al . , 2003; Dickman et al . , 2005 ) . Mutant animals lacking either synaptojanin or endophilin share identical defects at synapses . These defects are not exacerbated in double mutants , supporting that synaptojanin and endophilin function in the same pathway . Current models suggest that synaptojanin is transiently recruited to endocytic sites via direct binding between the endophilin SH3 domain and the synaptojanin proline-rich domain ( PRD ) ( Schuske et al . , 2003; Verstreken et al . , 2003; Milosevic et al . , 2011 ) . In vitro binding assays provide evidence for a biochemical interaction between PRD and SH3 ( Ringstad et al . , 1997; de Heuvel et al . , 1997 ) , and blocking PRD-SH3 interactions by peptides induces abnormal accumulation of endocytic intermediates at synapses ( Gad et al . , 2000 ) . However , we recently found that truncated endophilin lacking the SH3 domain has synaptic activity in vivo ( Bai et al . , 2010 ) , suggesting that synaptojanin and endophilin interact through PRD-SH3 independent mechanisms . Alternatively , synaptojanin may be recruited through redundant SH3 harboring proteins , such as amphiphysin ( Micheva et al . , 1997 ) and intersectin ( Evergren et al . , 2007; Pechstein et al . , 2010 ) . Synaptojanin harbors two phosphatase domains in addition to the PRD domain ( McPherson et al . , 1996 ) . The N-terminal Sac1 domain removes the phosphate group on the 3- and 4-position from the inositol ( Guo et al . , 1999; Krebs et al . , 2013 ) , and the adjacent 5-phosphatase targets the phosphate on the 5-position ( Cremona et al . , 1999; Chang-Ileto et al . , 2011 ) . This configuration of tandem phosphatases is unique to synaptojanin , as other phosphoinositide phosphatases ( e . g . , OCRL and SHIP1/2 ) have single catalytic domains that are linked to protein- or membrane-binding domains such as PH , SH2 , and C2 domains ( Pirruccello and De Camilli , 2012 ) . Interactions through the non-catalytic domains often enhance phosphatase specificity in membrane recognition through coincident detection of multiple targets ( Carlton and Cullen , 2005 ) . While it is thought that synaptojanin's tandem phosphatase domains act together to degrade multiple types of phosphoinositides at synapses ( Guo et al . , 1999 ) , the precise role of synaptojanin's tandem phosphatase domains is unclear . Here , we show that the functional core of synaptojanin resides in its tandem phosphatase domains rather than the PRD domain . Our results reveal an unexpected mechanism whereby the Sac1 domain displays a non-catalytic function to support coordinated action between synaptojanin and endophilin at synapses .
We investigated the requirement of synaptojanin PRD using both behavioral and electrophysiological phenotypes as in vivo assays . In Caenorhabditis elegans , the unc-26 gene encodes a highly conserved synaptojanin homologue with identical domain structure to the mammalian synaptojanin ( Harris et al . , 2000 ) ( Figure 1A ) . Mutant worms lacking unc-26 synaptojanin have significantly decreased locomotion rates and largely diminished excitatory postsynaptic currents ( EPSCs ) at neuromuscular junctions ( Harris et al . , 2000 ) ( Figure 1B–F and Table 1 ) . Because the density of active zone markers ( e . g . , RIM/UNC-10 ) remains unchanged in unc-26 mutants ( Ch'ng et al . , 2008 ) , reduced EPSC frequency and amplitude cannot be explained by fewer synapses . Instead , these defects are consistent with previous reports showing reduced SV pools and a corresponding decrease in synaptic transmission due to the cumulative effects of impaired endocytosis over time ( Cremona et al . , 1999; Harris et al . , 2000; Verstreken et al . , 2003; Dickman et al . , 2005 ) . 10 . 7554/eLife . 05660 . 003Figure 1 . Synaptojanin UNC-26 lacking the PRD domain fully supports locomotion , endogenous activity , and evoked synaptic currents . ( A ) Domain structure of synaptojanin UNC-26 . Synaptojanin contains three functional modules: a Sac1 phosphatase domain , a 5-phosphatase domain ( 5Pase ) , and a proline-rich domain ( PRD ) . Single-copy transgenes encoding GFP-tagged UNC-26 full-length ( FL; residues 1–1113 ) and ∆PRD ( residues 1–986 ) were introduced into synaptojanin unc-26 ( s1710 ) mutant worms . The pan-neuronal promoter Prab-3 was used to drive transgene expression . ( B ) C . elegans locomotion is restored by neuronal expression of full-length synaptojanin ( UNC-26FL ) or synaptojanin lacking the PRD domain ( UNC-26∆PRD ) . Representative trajectories ( 20 animals ) of 30 s locomotion are shown for each genotype . The starting points for each trajectory are aligned for clarity . ( C–F ) Electrophysiological recordings show that GFP-tagged synaptojanin UNC-26∆PRD is fully functional at synapses . Representative traces and summary data for endogenous EPSC rates ( C–D ) and for evoked EPSC amplitude ( E–F ) are shown for the indicated genotypes . The number of worms analyzed for each genotype is indicated in the bar graphs . *** , p < 0 . 0001 when compared to wild-type ( wt ) controls . ### , p < 0 . 0001 when compared to unc-26 mutants . Error bars represent standard error of the mean ( SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 00310 . 7554/eLife . 05660 . 004Figure 1—figure supplement 1 . Mouse synaptojanin ∆PRD is functional in C . elegans neurons . Truncated version of mouse synaptojanin 1 ( 1–1045 ) that lacks the PRD domain was expressed in C . elegans nervous system under control of the pan-neuronal promoter Psnb-1 . Electrophysiological recordings at NMJs were carried out using wt ( N2 ) , unc-26 ( s1710 ) , and transgenic worms carrying mSYJ∆PRD . Summary data for locomotion rates ( A ) , representative traces and summary data for endogenous EPSC rates ( B–C ) , and representative traces and summary data for evoked EPSC amplitude ( D–E ) are shown for the indicated genotypes . ### , p < 0 . 0001 when compared to unc-26 mutants . Error bars indicate SEM . The number of worms analyzed for each genotype is indicated in the bar graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 00410 . 7554/eLife . 05660 . 005Table 1 . Summary of data from electrophysiological recordings and locomotion analysesDOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 005Evoked EPSC Amp . ( nA ) Endogenous EPSCLocomotion speed ( µm/s ) Frequency ( Hz ) Amp . ( pA ) Wild type ( N2 ) 3 . 2 ± 0 . 2 ( n = 23 ) 49 . 8 ± 2 . 5 ( n = 30 ) 22 . 4 ± 1 . 0140 ± 8 ( n = 65 ) unc-26 ( s1710 ) –0 . 9 ± 0 . 1 ( n = 15 ) †12 . 0 ± 1 . 5 ( n = 16 ) †20 . 7 ± 0 . 628 ± 3 ( n = 65 ) †Si[Prab-3::unc-26::gfp]3 . 1 ± 0 . 2 ( n = 16 ) #50 . 8 ± 5 . 4 ( n = 16 ) #23 . 2 ± 1 . 4139 ± 8 ( n = 60 ) #Si[Prab-3::unc-26∆PRD::gfp]3 . 4 ± 0 . 3 ( n = 16 ) #47 . 4 ± 5 . 0 ( n = 17 ) #24 . 8 ± 1 . 3141 ± 10 ( n = 60 ) #Ex[Psnb-1::mSYJ1∆PRD]3 . 1 ± 0 . 3 ( n = 7 ) #49 . 6 ± 6 . 1 ( n = 11 ) #21 . 5 ± 0 . 6143 ± 10 ( n = 37 ) #Ex[Prab-3::gfp::unc-26 ( C378S , D380N ) ]3 . 5 ± 0 . 2 ( n = 10 ) #53 . 1 ± 6 . 6 ( n = 10 ) 23 . 8 ± 1 . 3135 ± 9 ( n = 60 ) #Ex[Prab-3::gfp::unc-26∆PRD ( C378S , D380N ) ]3 . 2 ± 0 . 3 ( n = 10 ) #50 . 7 ± 5 . 7 ( n = 10 ) 24 . 2 ± 0 . 9130 ± 9 ( n = 60 ) #Ex[Prab-3::gfp::unc-26 ( D716A ) ]0 . 8 ± 0 . 1 ( n = 14 ) 10 . 0 ± 1 . 8 ( N = 15 ) 20 . 9 ± 0 . 828 ± 3 ( n = 60 ) Ex[Prab-3::gfp::unc-26∆Sac1]1 . 2 ± 0 . 2 ( n = 10 ) 7 . 2 ± 1 . 9 ( N = 10 ) 21 . 3 ± 1 . 231 ± 3 ( n = 60 ) Ex[Prab-3::unc-26Sac1 + Prab-3::unc-26∆Sac1]1 . 0 ± 0 . 2 ( n = 10 ) 16 . 1 ± 1 . 8 ( n = 10 ) 20 . 4 ± 0 . 532 ± 3 ( n = 60 ) Ex[Prab-3::unc-26Sac1::IntN + Prab-3::IntC::unc-26∆Sac1]3 . 2 ± 0 . 4 ( n = 10 ) #53 . 1 ± 7 . 1 ( n = 10 ) #24 . 8 ± 1 . 2100 ± 7 ( n = 60 ) #Ex[Prab-3::unc26∆Sac1::rab-3]1 . 0 ± 0 . 2 ( n = 10 ) 13 . 2 ± 2 . 3 ( n = 10 ) 20 . 4 ± 0 . 9Ex[Prab-3::unc-26∆Sac1::snb-1]1 . 4 ± 0 . 2 ( n = 11 ) 17 . 9 ± 2 . 6 ( n = 11 ) 21 . 1 ± 0 . 8Ex[Prab-3::bem1PX::unc-26∆Sac1]0 . 8 ± 0 . 1 ( n = 7 ) 6 . 3 ± 0 . 6 ( n = 7 ) 18 . 8 ± 1 . 2Ex[Prab-3::plc∂PH::unc-26∆Sac1]1 . 0 ± 0 . 2 ( n = 7 ) 8 . 5 ± 1 . 8 ( n = 7 ) 18 . 7 ± 1 . 1Ex[Prab-3::btkPH::unc-26∆Sac1]1 . 2 ± 0 . 2 ( n = 11 ) 11 . 4 ± 1 . 2 ( n = 11 ) 18 . 4 ± 0 . 8Ex[Prab-3::apa-2::unc-26∆Sac1]1 . 1 ± 0 . 1 ( n = 11 ) 12 . 8 ± 1 . 8 ( n = 11 ) 19 . 4 ± 0 . 8Ex[Prab-3::apb-1::unc-26∆Sac1]0 . 9 ± 0 . 1 ( n = 11 ) 13 . 2 ± 2 . 5 ( n = 11 ) 18 . 2 ± 1 . 5Ex[Prab-3::apm-2::unc-26∆Sac1]1 . 1 ± 0 . 1 ( n = 10 ) 15 . 3 ± 2 . 7 ( n = 10 ) 21 . 0 ± 0 . 9Ex[Prab-3::aps-2::unc-26∆Sac1]1 . 3 ± 0 . 2 ( n = 9 ) 12 . 7 ± 1 . 4 ( n = 9 ) 20 . 1 ± 0 . 9Ex[Prab-3::unc-57::unc-26∆Sac1]2 . 5 ± 0 . 3 ( n = 11 ) #28 . 0 ± 4 . 1 ( n = 11 ) §23 . 0 ± 1 . 4Ex[Prab-3::dyn-1::unc-26∆Sac1]1 . 3 ± 0 . 3 ( n = 10 ) 14 . 4 ± 2 . 9 ( n = 10 ) 19 . 4 ± 0 . 9Ex[Prab-3::itsn-1::unc-26∆Sac1]0 . 9 ± 0 . 1 ( n = 9 ) 13 . 9 ± 1 . 4 ( n = 9 ) 21 . 5 ± 1 . 4Ex[Prab-3::unc-57::unc-26∆Sac1∆PRD]3 . 0 ± 0 . 3 ( n = 11 ) #36 . 7 ± 5 . 4 ( n = 11 ) §21 . 7 ± 1 . 3Ex[Prab-3::unc-57::unc-26∆Sac1 ( D716A ) ]0 . 8 ± 0 . 2 ( n = 9 ) 13 . 0 ± 2 . 3 ( n = 9 ) 21 . 0 ± 0 . 7Ex[Prab-3::unc-57BAR::unc-26∆Sac1]2 . 9 ± 0 . 2 ( n = 12 ) #23 . 9 ± 2 . 5 ( n = 12 ) #21 . 3 ± 0 . 8Ex[Prab-3::rEndoBAR::unc-26∆Sac1]3 . 1 ± 0 . 4 ( n = 13 ) #28 . 5 ± 4 . 6 ( n = 13 ) #24 . 1 ± 1 . 4Ex[Prab-3::mAmphBAR::unc-26∆Sac1]1 . 4 ± 0 . 2 ( n = 10 ) 19 . 2 ± 2 . 3 ( n = 10 ) 20 . 8 ± 0 . 8Ex[Prab-3::mNadrin2BAR::unc-26∆Sac1]1 . 7 ± 0 . 2 ( n = 11 ) ‡16 . 1 ± 3 . 1 ( n = 11 ) 23 . 8 ± 1 . 9Ex[Prab-3::rEndoBAR∆N::unc-26∆Sac1]1 . 3 ± 0 . 2 ( n = 12 ) 9 . 1 ± 0 . 9 ( n = 12 ) 18 . 9 ± 0 . 4Ex[Prab-3::rEndoBAR ( K76E , K78E ) ::unc-26∆Sac1]1 . 5 ± 0 . 2 ( n = 10 ) 13 . 5 ± 1 . 3 ( n = 10 ) 19 . 3 ± 0 . 7N2Prab-3::unc-26∆PRD ( D716A ) overexpression1 . 5 ± 0 . 3 ( n = 9 ) *24 . 6 ± 3 . 8 ( n = 10 ) *25 . 7 ± 0 . 9Prab-3::unc-26∆PRD overexpression2 . 9 ± 0 . 3 ( n = 9 ) 53 . 5 ± 4 . 6 ( n = 9 ) 25 . 4 ± 1 . 3Prab-3::unc-26∆Sac1∆PRD ( D716A ) overexpression3 . 5 ± 0 . 3 ( n = 10 ) 49 . 0 ± 7 . 5 ( n = 10 ) 25 . 9 ± 1 . 9unc-57 ( e406 ) ; unc-26 ( s1710 ) –0 . 8 ± 0 . 2 ( n = 9 ) †8 . 6 ± 0 . 8 ( n = 10 ) †21 . 9 ± 1 . 127 ± 3 ( n = 60 ) †Si[Psnb-1::unc-57∆SH3::mCherry]; Si[Prab-3::unc-26∆PRD::gfp]3 . 2 ± 0 . 2 ( n = 9 ) ¶50 . 3 ± 4 . 1 ( n = 9 ) ¶23 . 1 ± 1 . 0142 ± 9 ( n = 62 ) ¶Si[Psnb-1::rEndoBAR::unc-26∆Sac1∆PRD]3 . 0 ± 0 . 3 ( n = 10 ) ¶50 . 9 ± 4 . 1 ( n = 10 ) ¶26 . 5 ± 1 . 1109 ± 4 ( n = 68 ) ¶*p < 0 . 001 when compared with N2 . †p < 0 . 0001 when compared with N2 . ‡p < 0 . 05 when compared with unc-26 mutant . §p < 0 . 001 when compared with unc-26 mutant . #p < 0 . 0001 when compared with unc-26 mutant . ¶p < 0 . 0001 when compared with unc-57; unc-26 double mutants . Si: single-copy transgene ( MosSci insertion ) . Ex: extrachromosomal array . ‘Amp . ’ indicates amplitude . To determine whether the PRD of synaptojanin is required for endocytosis , we expressed a truncated version of C . elegans synaptojanin UNC-26 ( residues 1–986; ∆PRD ) that lacks PRD in unc-26 null mutant worms . In transgenic animals , a single copy of the transgene ( unc-26∆PRD::gfp ) driven by a pan-neuronal promoter ( Prab-3 ) was inserted into chromosome X to avoid confounding issues of overexpression ( Frøkjaer-Jensen et al . , 2012 ) . We reasoned that if the PRD domain is essential , the truncated UNC-26∆PRD should not rescue mutant defects . Surprisingly , similar to full-length UNC-26 , UNC-26∆PRD fully restored locomotion , endogenous EPSCs , and evoked responses in unc-26 mutant worms ( Figure 1 and Table 1 ) . To test the functional conservation between vertebrate and nematode synaptojanin , we expressed a truncated version of mouse synaptojanin 1 ( mSyj1∆PRD , residues 1–1045 ) in unc-26 mutants . We found that truncated mSyj1∆PRD also restored locomotion and synaptic transmission to wild type ( wt ) levels ( Figure 1—figure supplement 1 and Table 1 ) , indicating that synaptojanin from both invertebrate and vertebrate animals remains largely active in the absence of PRD . To assay for membrane recycling , we employed FM4-64 , a fluorescent lipophilic dye that is internalized by endocytosis ( Betz et al . , 1996; Kay et al . , 1999 ) . In wt animals , dye was readily internalized in response to KCl stimulation , evident by the high level of FM4-64 fluorescence ( 3527 ± 412 arbitrary units [a . u . ]; n = 12 ) in the neuron ganglion after washing ( Figure 2A–B ) . Approximately 43% of internalized FM4-64 ( 1411 ± 150 a . u . ; n = 12 ) was released after KCl stimulation , indicating that FM4-64 was internalized into recycling vesicles . By contrast , the dye uptake in unc-26 mutant worms was significantly lower: reduced by ∼40% compared to controls ( 2123 ± 172 a . u . ; n = 11 ) , consistent with defects in membrane recycling . About 32% of internalized dye ( 686 ± 117 a . u . ; n = 11 ) by the unc-26 mutants was released upon KCl challenge ( Figure 2A–B ) . Expression of the single-copy Prab-3::unc-26∆PRD::gfp transgene fully restored FM4-64 uptake ( 3885 ± 505 a . u . ; n = 10 ) and the KCl-dependent dye release ( 1569 ± 243 a . u . ; n = 10 ) ( Figure 2A–B ) , indicating that the recovery of vesicle recycling processes does not require UNC-26PRD . 10 . 7554/eLife . 05660 . 006Figure 2 . Synaptojanin UNC-26∆PRD recovers the recycling vesicle pool and sustains synaptic transmission upon repetitive stimuli . ( A ) A schematic diagram is shown to illustrate the FM4-64 loading and unloading procedure . Experimental details are discussed in the ‘Materials and methods’ section . ( B ) FM4-64 loading and unloading at the head ganglion were compared for wt ( n = 12 ) , unc-26 mutant ( n = 11 ) , and rescued worms with a single-copy transgene encoding GFP::UNC-26∆PRD ( n = 10 ) . The expression of GFP::UNC-26∆PRD significantly rescued both dye uptake and unloading ( p < 0 . 01 and p < 0 . 01 , respectively; compared to unc-26 mutants ) . ( C ) Acetylcholine currents were evoked by 2-Hz light pulses in worms carrying Punc-17::ChR2::mCherry . Representative traces of light-evoked EPSCs during repeated stimulation are shown for the indicated genotypes . ( D ) Mean values of currents normalized relative to the first EPSC were significantly reduced in unc-26 mutant . The expression of GFP::UNC-26∆PRD in unc-26 mutants restores the amplitude of subsequent currents , suggesting that the UNC-26∆PRD is functional to support synaptic transmission upon repeated stimuli . The number of worms analyzed for each genotype is indicated in the graph . * , p < 0 . 05 and ** , p < 0 . 01 when compared to wt controls . # , p < 0 . 05 and ## p < 0 . 01 when compared to unc-26 mutants . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 006 We next asked if synapses rescued by truncated UNC-26∆PRD sustain synaptic transmission upon repetitive stimuli . Cholinergic neurons of transgenic animals carrying Punc-17::ChR2 ( H134R ) ::YFP were activated by 2-Hz photostimulation , and evoked EPSCs were recorded at neuromuscular junctions ( Liewald et al . , 2008; Liu et al . , 2009 ) . For all successive stimuli , the amplitudes of EPSCs in unc-26 mutant worms were significantly reduced compared to those in control worms ( Figure 2C–D ) . These results are consistent with previous findings showing that unc-26 mutant synapses exhibit more depression in synaptic transmission after repeated stimulation , due to impaired endocytosis . Expression of the single-copy Prab-3::unc-26∆PRD::gfp transgene recovered EPSC amplitudes of successive stimuli , supporting the notion that truncated UNC-26 functions sufficiently to supply SVs during sustained activity . Together , these results argue against an essential role of the synaptojanin PRD domain at synapses . To further test the functional importance of the endophilin SH3 , synaptojanin PRD interactions , we studied double-mutant worms that lack both endophilin unc-57 and synaptojanin unc-26 . Consistent with previous findings , synaptic defects in the unc-57; unc-26 double-mutant worms were similar to unc-57 and unc-26 single mutants ( Schuske et al . , 2003 ) ( Figure 3 and Table 1 ) , confirming that these genes function in the same genetic pathway . While the SH3-PRD scaffolding model predicts that SH3 and PRD are essential , we found that co-expression of single copies of mutant UNC-57 lacking SH3 ( UNC-57∆SH3::mCherry ) and mutant UNC-26 lacking PRD ( UNC-26∆PRD::GFP ) restores synaptic activities in unc-57; unc-26 double mutants ( Figure 3C–F and Table 1 ) . Indeed , electron microscopy analyses show that the number of SVs was nearly normal in these animals ( Figure 4A ) . Using quantitative Western blots , we found that mutant UNC-26∆PRD and mutant UNC-57∆SH3 were expressed at ∼32% and ∼75% of endogenous levels of UNC-26 and UNC-57 , respectively , suggesting that the rescue activity of these transgenes was not due to compensatory artifacts of overexpression ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 05660 . 007Figure 3 . The SH3-PRD interaction is dispensable for synaptic activity . ( A ) A schematic drawing showing interactions between synaptojanin ( UNC-26 ) PRD and endophilin ( UNC-57 ) SH3 . Single-copy transgenes encoding UNC-26∆PRD::GFP and UNC-57∆SH3::mCherry were co-expressed in unc-57; unc-26 double mutants . Pan-neuronal promoters Prab-3 and Psnb-1 were used to drive expression of UNC-26∆PRD::GFP and UNC-57∆SH3::mCherry , respectively . Summary data for locomotion rate are shown in ( B ) . Representative traces and summary data for endogenous EPSC rates ( C–D ) and for evoked EPSC amplitude ( E–F ) are shown for the indicated genotypes . The number of worms analyzed for each genotype is indicated in the bar graphs . *** , p < 0 . 0001 when compared to wt controls . ### , p < 0 . 0001 when compared to unc-57; unc-26 double mutants . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 00710 . 7554/eLife . 05660 . 008Figure 3—figure supplement 1 . UNC-57 and UNC-26 are not overexpressed in transgenic animals . Monoclonal antibodies against UNC-57BAR and UNC-26 5-phosphatase were developed as described in the ‘Materials and methods’ . ( A ) Immunoblots for UNC-57 detection . 35 µg of total proteins , extracted from worms of indicated genotypes , were loaded on SDS-PAGE gels . ( B ) Immunoblots for UNC-26 detection . 100 µg of total proteins were analyzed . Tubulin was used as loading controls and was detected by a monoclonal antibody . Immunoreactive bands were visualized using enhanced chemiluminescence and were quantified using a Bio-Rad ChemiDoc MP imaging system . * , proteolytic products or nonspecific bands . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 00810 . 7554/eLife . 05660 . 009Figure 4 . UNC-26∆PRD and UNC-57∆SH3 restore the number of SVs and recover synaptopHluorin retrieval in unc-57; unc-26 double mutants . ( A ) Electron microscopy images of neuromuscular junctions were collected from the ventral nerve cords of adult hermaphrodites . Synaptic profiles of 15 synapses of the wt , 18 synapses of the unc-57; unc-26 double mutants , and 10 synapses of the single-copy transgenic UNC-26∆PRD::GFP; UNC-57∆SH3::mCherry animals were analyzed . Arrowheads indicate dense projections . Synaptic vesicle ( SV ) number was counted in a blind manner . *** , p < 0 . 0001 when compared to wt controls . ### , p < 0 . 0001 and ## , p < 0 . 001 when compared to unc-57; unc-26 double mutants . Scale bar: 100 nm . Error bars indicate SEM . ( B ) Representative images ( left ) and summary data ( right ) for axonal synaptopHluorin ( SpH ) fluorescence in the dorsal nerve cord are shown for the indicated genotypes . Rescue experiments are done using extrachromosomal arrays carrying Psnb-1::unc-26∆PRD and Prab-3::unc-57∆SH3 ( without any fluorescent tags ) . The number of worms analyzed for each genotype is indicated . *** , p < 0 . 0001 compared to wt controls . ### , p < 0 . 0001 when compared to unc-57; unc-26 mutants . Scale bar: 2 µm . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 009 Finally , we utilized synaptopHluorin ( SpH ) to measure changes in surface synaptobrevin ( Dittman and Kaplan , 2006; Bai et al . , 2010 ) . In SVs , SpH fluorescence is quenched by the acidic pH of the vesicle lumen . Following SV exocytosis , SpH fluorescence on the plasma membrane is dequenched ( Dittman and Ryan , 2009 ) . The unc-57; unc-26 double mutants had a 63% increase in SpH axon fluorescence compared to control animals , consistent with a defect in recycling SV proteins from plasma membranes . Co-expression of UNC-57∆SH3 and UNC-26∆PRD fully rescued the SpH defects ( Figure 4B–D ) , demonstrating that UNC-26∆PRD and UNC-57∆SH3 are functional to support SV endocytosis . Overall , these data demonstrate that endophilin and synaptojanin can support synaptic activity even in the absence of the SH3-PRD interaction . Therefore , additional uncharacterized mechanisms must exist to support synaptojanin function at synapses . We next investigated whether synaptojanin's unique configuration of tandem phosphoinositide phosphatase domains ( Figure 1A ) mediates the cooperation between synaptojanin and endophilin at synapses . The N-terminal Sac1 domain degrades phosphoinositides by hydrolyzing the 3- and 4-position phosphates ( Guo et al . , 1999 ) , whereas the central 5-phosphatase domain converts PI ( 4 , 5 ) P2 into PI ( 4 ) P by removing the 5-position phosphate from the inositol ring ( Cremona and De Camilli , 2001; Chang-Ileto et al . , 2011 ) . We found that inactivation of 5-phosphatase ( D716A mutation ) ( Whisstock et al . , 2002 ) completely abolished UNC-26 rescuing ability in restoring EPSC levels and locomotion ( Figure 5 and data not shown ) , indicating that the enzymatic activity of 5-phosphatase is required . By contrast , mutations ( C378S , D380N ) ( Guo et al . , 1999; Hughes et al . , 2000 ) that inactivate Sac1 had little impact on UNC-26 activity , independent of the presence of the PRD domain ( Figure 5 , Figure 5—figure supplement 1 , and Table 1 ) . These data are consistent with previous reports showing that the mouse synaptojanin with inactivated Sac1 supports SV endocytosis in response to persistent activity ( Mani et al . , 2007 ) , and that human patients with synaptojanin Sac1 mutations show no severe symptoms until reaching 20–40 years of age ( Krebs et al . , 2013 ) . Together , these findings indicate that synaptojanin is able to support synaptic transmission , largely independent of its Sac1 phosphatase activity . 10 . 7554/eLife . 05660 . 010Figure 5 . Synaptojanin phosphatase domains have distinct functions . The 5-Pase domain hydrolyzes phosphoinositides , while Sac1 plays an essential but non-catalytic role at synapses . Evoked EPSCs from wt , unc-26 ( s1710 ) mutant , and indicated transgenic strains were compared . Transgenes were GFP-tagged UNC-26 variants , including Sac1 dead ( C378S , D380N; full-length UNC-26 ) , 5-Pase dead ( D716A , full-length UNC-26 ) , and ∆Sac1 ( residues 494–1113 ) . Transgenes were driven by Prab-3 . Representative traces ( A ) and summary data ( B ) for evoked EPSC amplitudes are shown for the indicated genotypes . *** , p < 0 . 0001 when compared to wt controls . ### , p < 0 . 0001 when compared to unc-26 mutants . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01010 . 7554/eLife . 05660 . 011Figure 5—figure supplement 1 . Sac1-inactivated synaptojanin supports synaptic transmission in a PRD independent manner . Two mutations ( C378S and D380N ) that inactivate the Sac1 lipid phosphatase activity were introduced into the Sac1 domain of UNC-26∆PRD ( residues 1–986 ) . The mutant UNC-26∆PRD ( C378S , D380N ) was expressed in neurons using Prab-3 . Representative traces and summary data for endogenous EPSC rates ( A–B ) and for evoked EPSC amplitude ( C–D ) are shown for the indicated genotypes . ### , p < 0 . 0001 when compared to unc-26 mutants . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 011 To ask whether the entire Sac1 domain plays any role in synaptojanin function , we generated a truncated UNC-26 that lacks the Sac1 domain ( UNC-26∆Sac1 , lacking residues 1–493 ) . Surprisingly , we found that removal of the Sac1 domain severely disrupted the rescuing activity of UNC-26 ( Figure 5 ) , suggesting that the physical presence of Sac1 is required . Although isolated Sac1 and 5-phophatase fold correctly ( Tsujishita et al . , 2001; Manford et al . , 2010 ) , it remains possible that Sac1 deletion may perturb the folding of UNC-26 . To address this issue in vivo , we used an intein-mediated protein ligation method to reconnect Sac1 to UNC-26 post-translationally ( Figure 6A ) . We reasoned that if Sac1 truncation causes protein misfolding , UNC-26 would remain inactive after reconnecting with Sac1 . However , if truncated UNC-26 fragments retain correct folding structure , protein ligation should lead to active full-length UNC-26 . 10 . 7554/eLife . 05660 . 012Figure 6 . Sac1 must be physically linked to UNC-26 5-phosphatase to support synaptic transmission . Split-intein mediated ligation was used to post-translationally reconnect Sac1 to the remainder of the UNC-26 synaptojanin protein ( A ) . The Sac1 domain ( 1–493 ) was linked to the N-terminal half of NpuDnaE to generate Sac1::IntN . The C-terminal half of NpuDnaE was fused with the N-terminus of the UNC-26∆Sac1 fragment . Three extra residues ( CFN ) remain in the ligated product . Representative traces and summary data of evoked EPSCs are shown in ( B ) . Co-expression of Sac1::IntN and IntC::UNC-26∆Sac1 significantly rescued the synaptic defects in unc-26 mutant worms . ### , p < 0 . 0001 when compared to unc-26 mutants . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01210 . 7554/eLife . 05660 . 013Figure 6—figure supplement 1 . Transgenic worms that only express either Sac1::IntN or IntC::UNC-26∆Sac1 did not show functional improvements . Representative traces ( A ) and summary data ( B ) for evoked EPSC amplitude are shown for the indicated genotypes . Transgenes encoding Sac1::IntN and IntC::UNC-26∆Sac1 were designed as described in the Figure 4 . ‘n . s . ’ indicates p > 0 . 05 when compared to unc-26 mutants . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 013 We fused UNC-26 fragments with split DnaE intein from Nostoc punctiforme ( NpuDnaE ) ( Figure 6A ) , as this intein system has been shown to be active in C . elegans ( Wong et al . , 2012 ) . Co-expression of Sac1::IntN and IntC::UNC-26∆Sac1 significantly rescued the synaptic defects in unc-26 mutants ( Figure 6B and Table 1 ) , suggesting that UNC-26 became functional upon post-translational ligation of Sac1 . By contrast , transgenic worms that only express either Sac1::IntN or IntC::UNC-26∆Sac1 did not show functional improvements ( Figure 6—figure supplement 1 and Table 1 ) . These data suggest that the two phosphatase domains of synaptojanin have distinct roles: the 5-phosphatase domain hydrolyzes phosphoinositides , while Sac1 plays a non-enzymatic role at synapses . Importantly , we found that Sac1 needs to be physically linked to UNC-26 to support synaptic transmission , as co-expression of UNC-26 fragments without the split inteins did not significantly rescue synaptic defects ( Figure 6B ) . To gain insights into the non-enzymatic function of Sac1 , we investigated the possibility that Sac1 guides 5-phosphatase for synaptic localization . We quantified synaptic abundance of GFP-tagged UNC-26 variants . Full-length UNC-26 is enriched at synapses ( synapse/axon ratio = 3 . 4 ± 0 . 1 fold , Figure 7A–B ) . However , removal of the Sac1 domain significantly reduced synaptic enrichment of UNC-26∆Sac1 ( synapse/axon ratio = 2 . 4 ± 0 . 1 fold; Figure 7A–B ) , indicating that Sac1 has a critical role in retaining UNC-26 at synapses . Consistent with this idea , isolated Sac1 domains ( both wt and the C378S , D380N mutant ) are localized to synapses ( Figure 7—figure supplement 1 ) . It is likely that Sac1 and PRD act together to enhance synaptic distribution of synaptojanin , as deletion of both PRD and Sac1 domain further decreased synaptic enrichment ( GFP::UNC-26∆Sac1∆PRD synapse/axon ratio = 1 . 9 ± 0 . 1 fold; Figure 7A–B ) . 10 . 7554/eLife . 05660 . 014Figure 7 . Sac1 is a synaptic targeting domain . ( A–B ) Removal of the Sac1 domain of synaptojanin perturbs synaptic targeting of synaptojanin . Representative images ( A ) showing various versions of GFP::UNC-26 distribution in the dorsal nerve cord . Synaptic enrichment of GFP::UNC-26 was quantified using ∆F/F = ( Fpeak − Faxon ) /Faxon and was compared for the indicated genotypes ( B ) . Scale bar: 2 µm . ( C–D ) Sac1 is required for dominant negative inhibition . The D716A mutation that blocks 5-phosphatase activity was introduced into UNC-26∆PRD and UNC-26∆Sac1∆PRD . These UNC-26 variants were expressed in nervous system of wt worms . The stimulus-evoked EPSC amplitudes were significantly reduced in worms carrying UNC-26∆PRD ( D716A ) mutant proteins . By contrast , animals expressing either UNC-26∆PRD ( with a functional 5Pase ) or UNC-26∆Sac1∆PRD ( D716A ) mutant proteins showed normal levels of synaptic activity . The number of worms analyzed for each genotype is indicated in the bar graphs . ** , p < 0 . 001 when compared to wt controls . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01410 . 7554/eLife . 05660 . 015Figure 7—figure supplement 1 . GFP-tagged Sac1 domains localize to synapses . ( A ) Representative images showing GFP , GFP::UNC-26Sac1 , and GFP::UNC-26Sac1 ( C378S , D380N ) distribution in the dorsal nerve cord axons . Synaptic enrichment of GFP was quantified using ∆F/F = ( Fpeak − Faxon ) /Faxon . Scale bar: 2 µm . Summary data are shown in ( B ) . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 015 Interestingly , we found that UNC-26∆PRD with inactivated 5-phosphatase ( D716A ) , but not the version with wt phosphatase domains , exhibits significant levels of dominant-negative inhibition , presumably by competing with wt UNC-26 . When D716A UNC-26∆PRD was expressed in wt worms , the evoked EPSC amplitude was reduced by ∼50% ( Figure 7C–D and Table 1 ) . The dominant negative effect was removed by elimination of the Sac1 domain ( Figure 7C–D ) , indicating that the inactivated 5-phosphatase alone does not produce inhibitory activity . Together , these data argue that Sac1 plays a role in localizing synaptojanin to synapses . Because our data suggest that Sac1 acts as a targeting domain rather than an enzyme , we speculated that it might be possible to bypass the Sac1 requirement by directly tethering the UNC-26∆Sac1 mutant with non-enzymatic proteins . We utilized three categories of proteins as candidate targeting tethers: ( 1 ) SV proteins SNB-1 ( Synaptobrevin ) and RAB-3; ( 2 ) lipid-binding domains that recognize specific phosphoinositides ( Lomasney et al . , 1996; Várnai et al . , 1999; Stahelin et al . , 2007 ) ; and ( 3 ) endocytic protein machinery adaptor AP2 subunits and accessory proteins ( Figure 8 and Figure 8—figure supplement 1 ) . Among all proteins tested , UNC-57 endophilin was the only molecular tether that significantly restored UNC-26 activity in supporting endogenous EPSCs and evoked responses , no matter whether the PRD domain is present ( Figure 8A–E and Table 1 ) . The 5-phosphatase activity was still required for proper functioning of this chimeric UNC-57::UNC-26∆Sac1 protein , as the D716A mutation disrupted its ability to rescue synaptic defects ( Figure 8—figure supplement 2 ) . Overall , our findings indicate that the Sac1 domain has a non-enzymatic role in guiding synaptojanin 5-phosphatase , which can be replaced by endophilin . 10 . 7554/eLife . 05660 . 016Figure 8 . Endophilin functionally substitutes for the Sac1 domain . ( A ) A schematic drawing showing the chimeric UNC-57 endophilin::UNC-26∆Sac1 protein . Other endocytic accessory proteins including DYN-1 dynamin and ITSN-1 intersectin were tethered to UNC-26∆Sac1 using an identical strategy . Transgenes were expressed in all neurons using Prab-3 . ( B–C ) Chimeric UNC-57 endophilin::UNC-26∆Sac1 proteins restore evoked EPSCs in unc-26 mutant worms . Other tethers failed to rescue synaptojanin defects . Electrophysiological data in Figure 8B–C and Figure 8—figure supplement 1 were collected blindly . ( D–E ) The PRD domain is not required for the endophilin tether to bypass the Sac1 requirement of synaptojanin . The number of worms analyzed for each genotype is indicated in the bar graphs . ### , p < 0 . 0001 when compared to unc-26 mutants . ‘n . s . ’ indicates p > 0 . 05 when compared to unc-26 mutants . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01610 . 7554/eLife . 05660 . 017Figure 8—figure supplement 1 . Targeting UNC-26∆Sac1 to synaptic vesicles , phosphoinositides , and endocytic adaptor protein AP2 does not recover synaptojanin function . Summary data for evoked EPSC amplitudes are shown for the indicated genotypes in ( A–C ) . Non-enzymatic proteins that are involved in SV cycle and phosphoinositide recognition were tethered to the N-terminus of UNC-26∆Sac1 , except that SNB-1 and RAB-3 were tethered to the C-terminus of UNC-26∆Sac1 . Protein tethers used for these analyses were SV proteins SNB-1/synaptobrevin and RAB-3; phosphatidylinositol lipid-binding domains ( PI4P binding Bem1PX domain , PI ( 4 , 5 ) P2 binding PLCδ PH domain , and PI ( 3 , 4 , 5 ) P3 biding Btk PH domain ) ; and endocytic adaptor AP-2 subunits ( APA-2/AP2α , APB-1/AP2β2 , APM-2/AP2μ2 , and APS-2/AP2δ2 ) . The number of worms analyzed for each genotype is indicated in the bar graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01710 . 7554/eLife . 05660 . 018Figure 8—figure supplement 2 . The endophilin tether does not bypass the requirement for a functional synaptojanin 5-phosphatase domain . The 5-phosphatase activity is required for the UNC-57::UNC-26∆Sac1 chimeric protein . All chimeric proteins were expressed under the control of Prab-3 . Representative traces ( A ) and summary data ( B ) for evoked EPSC amplitude are shown for the indicated genotypes . The number of worms analyzed for each genotype is indicated in the bar graphs . ### , p < 0 . 0001 , when compared to unc-26 mutants . ‘n . s . ’ indicates p > 0 . 05 when compared to unc-26 mutants . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 018 We next investigated the molecular requirements for endophilin to functionally replace the Sac1 domain . Endophilin contains two domains: an N-terminal BAR domain that bends membranes , and a C-terminal SH3 domain that interacts with PRD domains . We found that tethering UNC-26∆Sac1 to the endophilin BAR domain of either worm UNC-57 or rat endophilinA1 significantly restores synaptic transmission in unc-26 mutants ( Figure 9A–B and Table 1 ) . In contrast , UNC-57 SH3 tethered UNC-26∆Sac1 failed to rescue EPSC defects in unc-26 mutants ( data not shown ) , even though the SH3 domain enhanced the synaptic enrichment of UNC-26∆Sac1 ( Figure 9—figure supplement 1 ) . Together , these data indicate that endophilin BAR , rather than the SH3 domain , is the functional core for the Sac1 substitution . The functional difference between the BAR domain and the SH3 domain is likely due to their distinct binding partners and the potential for differential targeting to sub-synaptic regions . 10 . 7554/eLife . 05660 . 019Figure 9 . Endophilin BAR domain and its membrane interactions are required for bypassing Sac1 . ( A–B ) Endophilin BAR is sufficient to bypass the Sac1 requirement . UNC-26∆Sac1 was tethered to worm and rat endophilinA1 BAR ( rEndoBAR ) , respectively . Mutations that disrupt BAR-membrane interactions abolished the rescue activity of the chimeric UNC-26∆Sac1 . *** , p < 0 . 0001 when compared to transgenic unc-26 mutants carrying rEndoBAR wt::UNC-26∆Sac1 . ( C–E ) Expression of a single-copy transgene encoding the rEndoBAR::UNC-26 5Pase chimera significantly restores locomotion and evoked EPSCs in unc-57; unc-26 double mutants . Representative traces ( upper ) and summary data ( lower ) for locomotion ( C ) , and evoked EPSCs ( D–E ) are shown for the indicated genotypes . ### , p < 0 . 0001 when compared to unc-26 mutants . ( F ) A schematic diagram showing that the synaptojanin Sac1 domain and the endophilin BAR domain cooperate to promote SV endocytosis . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 01910 . 7554/eLife . 05660 . 020Figure 9—figure supplement 1 . UNC-57 SH3 domain enhances synaptic enrichment of UNC-26∆Sac1 . ( A ) Distribution of GFP::UNC-57SH3::UNC-26∆Sac1 ( domain structure , upper ) in the dorsal nerve cord axon is shown ( lower ) . Scale bar: 2 µm . ( B ) Summary data indicates that UNC-57 SH3 domain enhances the enrichment of UNC-26∆Sac1 at synapses . The number of worms analyzed for each genotype is indicated in the bar graphs . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 02010 . 7554/eLife . 05660 . 021Figure 9—figure supplement 2 . Specificity in BAR proteins for bypassing the Sac1 requirement . BAR domains of mouse Nadrin2 ( residues 1–248 ) and mouse amphiphysin ( residues 1–250 ) were tethered to the N-terminus of UNC-26∆Sac1 . All chimeric proteins were expressed under the control of Prab-3 in unc-26 mutants . Representative traces ( A ) and summary data ( B ) for evoked EPSC amplitude are shown for the indicated genotypes . # , p < 0 . 05 when compared to unc-26 mutants . Error bars represent SEM . n . s . indicates non-significant ( p = 0 . 07 , compared with unc-26 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 02110 . 7554/eLife . 05660 . 022Figure 9—figure supplement 3 . UNC-26 Sac1 does not bind UNC-57 in solution . GST pull-down assays were performed as described in the ‘Materials and methods’ . Briefly , GST::UNC-57 ( 10 µg ) was immobilized on glutathione beads . Maltose-binding protein ( MBP ) ::UNC-26 fragments were incubated with beads for 2 hr . Twenty percent of samples and 7% of total MBP::UNC-26 fragments were subjected to SDS-PAGE and visualized by staining with Coomassie Brilliant Blue G-250 . The Sac1 domain did not bind to full-length UNC-57 , nor UNC-57 BAR ( left panel ) . The 5-phosphatase domain also did not exhibit UNC-57 binding ( middle panel ) . In contrast , the PRD domain exhibited a significant amount of binding to UNC-57 SH3 ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05660 . 022 Interestingly , we found that expression of a single copy of the rEndoBAR::unc-26 5Pase ( ∆Sac1∆PRD ) transgene significantly restored locomotion , endogenous EPSCs , double mutants ( Figure 9C–E and Table 1 ) , supporting the notion that synaptojanin UNC-26 and endophilin UNC-57 execute coincident action at synapses . We next asked if BAR domains from other proteins could also functionally replace the Sac1 domain . When tethered to UNC-26∆Sac1 , BAR domains from nadrin2 ( Galic et al . , 2012 ) and amphiphysin ( Peter et al . , 2004 ) slightly restored evoked EPSCs ( Figure 9—figure supplement 2 ) , indicating a trend of membrane-bending BAR domains to promote synaptojanin function . While synaptojanin Sac1 and endophilin BAR do not bind each other in solution ( de Heuvel et al . , 1997; Ringstad et al . , 1997 ) ( Figure 9—figure supplement 3 ) , they both bind membranes ( Guo et al . , 1999; Farsad et al . , 2001 ) . Therefore , we next asked whether chimeric UNC-26∆Sac1::BAR requires the BAR-membrane interactions for its function . Indeed , we found that the disruption of BAR-membrane interactions by either deleting the N-terminal amphipathic helix ( ∆N ) or by decreasing the positively charged residues ( K76E , R78E ) ( Gallop et al . , 2006 ) abolished rescue activity to restore synaptic transmission ( Figure 9A–B ) . These data demonstrate that the membrane-binding activity of endophilin BAR is indispensible for Sac1 substitution . Together , these findings suggest that synaptojanin Sac1 and endophilin BAR are functionally coupled through membrane interactions to support SV recycling ( Figure 9F ) .
Synaptojanin and endophilin are a classic example of the coordinated actions of endocytic proteins for rapid SV endocytosis ( Dittman and Ryan , 2009; Saheki and De Camilli , 2012 ) . While synaptojanin and endophilin have distinct biochemical properties , disruption of either protein leads to similar defects in SV recycling , indicating that synaptojanin and endophilin are functional partners in vivo ( Schuske et al . , 2003; Verstreken et al . , 2003; Van Epps et al . , 2004; Dickman et al . , 2005 ) . Currently , the molecular basis for such functional cooperation is attributed solely to the specific and high-affinity binding between the synaptojanin PRD domain and the endophilin SH3 domain ( de Heuvel et al . , 1997; Ringstad et al . , 1997 ) . Post-translational modifications such as phosphorylation may regulate SV endocytosis by controlling PRD-SH3 interactions ( Lee et al . , 2004; Irie et al . , 2005 ) . Although the current model suggests that the PRD-SH3 interactions are required , our data show that synaptojanin and endophilin remain active without the PRD and the SH3 domains , respectively . Similarly , mutant mouse synaptojanin that is defective in PRD-SH3 binding supports rapid endocytosis upon short stimuli ( at 10 Hz ) in mammalian neurons ( Mani et al . , 2007 ) . These results show that the SH3-PRD interactions are not essential for synaptojanin activity in SV endocytosis . Synaptojanin PRD and endophilin SH3 domains are positively selected and maintained during evolution , evident by the sequence conservation among animal species . This observation suggests that the SH3-PRD interactions are important for cellular activities; however , the precise role of these interactions is unclear . It is worth noting that our findings do not rule out a modulatory role of the SH3-PRD interactions at synapses . For example , in cultured hippocampal neurons , the SH3-PRD interactions enhance the fidelity and speed of SV endocytosis after intense stimulation ( Mani et al . , 2007 ) . Thus , it is possible that the PRD-SH3 interaction facilitates co-localization of endophilin and synaptojanin , helping the Sac1-dependent mechanism to sustain membrane recycling during persistent activity . In addition , while the PRD domain alone is not sufficient to promote SV endocytosis , it may become required in some situations , for example , when the Sac1-membrane interactions are reduced . Furthermore , PRD and SH3 domains may function in other important cellular processes , contributing to their conservation . Indeed , photoreceptor neurons in the zebrafish synaptojanin mutant exhibit significant defects in endosomes and the Golgi apparatus ( George et al . , 2014 ) , suggesting that synaptojanin is needed for membrane-trafficking events at other intracellular organelles . The SH3-PRD interactions may be important for targeting membrane organelles in the cell body , where synaptojanin is less abundant than at synapses . Our results indicate that the tandem phosphatase domains , Sac1 and 5-phosphatase , are essential for synaptojanin activity in SV endocytosis . The configuration of linked phosphatase domains is a unique feature of synaptojanin , and is reflected in its name derived from janus , the God of two faces ( Majerus and York , 2009 ) . Interestingly , while both phosphatase domains have catalytic activities in vitro ( Cremona et al . , 1999; Guo et al . , 1999 ) , the enzymatic activity of the 5-phosphatase domain is the only one required for synaptojanin function at synapses . This is consistent with previous findings showing that the major phosphoinositide defect in synaptojanin knockout mice is the abnormal accumulation of PI ( 4 , 5 ) P2 ( Cremona et al . , 1999 ) . However , the 5-phosphatase activity alone is not enough . Expression of the 5-phosphatase , with or without the PRD domain , fails to restore synaptic activity . Unexpectedly , an enzymatic-dead version of Sac1 restores the activity of the 5-phosphatase domain to support synaptic transmission . These data suggest that the Sac1 domain possesses a novel targeting activity , which the PRD domain does not have . Phosphoinositide phosphatases often harbor multiple lipid-binding domains to detect coincident signals for restricted localization on membranes ( Carlton and Cullen , 2005 ) . Synaptojanin is a highly dynamic protein that is transiently recruited to endocytic intermediates . The timing of synaptojanin recruitment is likely to be critical because SV endocytosis is a rapid process that occurs on the time scale of seconds ( Balaji and Ryan , 2007 ) . Our findings suggest that the targeting activity of Sac1 allows synaptojanin to recognize endocytic intermediates . In agreement with this notion , tethering synaptojanin 5-phosphatase to endophilin bypasses the requirement of the Sac1 domain and revives synaptojanin activity , suggesting that synaptojanin 5-phosphatase functions at sites where endophilin resides . Therefore , we propose that the Sac1 domain acts together with the 5-phosphatase as coincident detectors for membranes enriched in PI ( 4 , 5 ) P2 and endophilin . How the synaptojanin Sac1 domain recognizes endophilin-membrane complexes is currently unknown . Biochemical studies have shown that the Sac1 domain does not directly bind endophilin in solution ( Figure 9—figure supplement 3 ) ( de Heuvel et al . , 1997; Ringstad et al . , 1997 ) . Here , we speculate that membranes serve as the molecular connection to couple these proteins , as both the Sac1 domain and the endophilin BAR domain bind membranes ( Guo et al . , 1999; Farsad et al . , 2001 ) . The endophilin BAR domain induces defects in lipid packing ( Gallop et al . , 2006 ) and consequently increases the exposure of lipid head groups . One possibility is that the synaptojanin Sac1 domain recognizes the lipid-packing defects generated by endophilin BAR , which in turn stimulates neighboring 5-phosphatase to degrade exposed PI ( 4 , 5 ) P2 head groups ( Chang-Ileto et al . , 2011 ) . Alternatively , it is also possible that membranes stimulate direct binding between the Sac1 domain and the endophilin BAR domain . Nonetheless , our results show that the Sac1 domain is a crucial targeting domain for synaptojanin function . We propose that the Sac1 domain allows synaptojanin to detect endocytic membranes on which endophilin resides .
Strain maintenance and genetic manipulations were performed as described ( Brenner , 1974 ) . All C . elegans strains were maintained at 20°C on agar nematode growth media ( NGM ) plates seeded with OP50 bacteria . The N2 strain ( Bristol , England ) was used as wt . Mutant unc-26 ( s1710 ) and unc-57 ( e406 ) strains were obtained from the Caenorhabditis Genetics Center and were subsequently outcrossed 10× times to the N2 strain . The following strains were used in this study: BJH188 unc-57 ( e406 ) ; unc-26 ( s1710 ) BJH180 unc-26 ( s1710 ) ; pekSi8 [Prab-3::unc-26::gfp , cb-unc-119 ( + ) ] BJH88 unc-26 ( s1710 ) ; pekSi7 [Prab-3::unc-26∆PRD::gfp , cb-unc-119 ( + ) ] BJH40 unc-26 ( s1710 ) ; pekEx15 [Psnb-1::mSyj1∆PRD] BJH298 unc-57 ( e406 ) ; pekSi19 [Psnb-1::unc-57∆SH3::mCherry , cb-unc-119 ( + ) ]; unc-26 ( s1710 ) ; pekSi7 [Prab-3::unc-26∆PRD::gfp , cb-unc-119 ( + ) ] BJH52 unc-26 ( s1710 ) ; pekEx27 [Prab-3::gfp::unc-26 ( D716A ) ] BJH55 unc-26 ( s1710 ) ; pekEx30 [Prab-3::gfp::unc-26 ( C378S , D380N ) ] BJH312 unc-26 ( s1710 ) ; pekEx66 [Prab-3::gfp::unc-26∆PRD ( C378S , D380N ) ] BJH43 unc-26 ( s1710 ) ; pekEx18 [Prab-3::unc-26Sac1 , Prab-3::unc-26∆Sac1] BJH46 unc-26 ( s1710 ) ; pekEx21 [Prab-3::unc-26∆Sac1] BJH49 unc-26 ( s1710 ) ; pekEx24 [Prab-3::unc-26Sac1::IntN] BJH48 unc-26 ( s1710 ) ; pekEx23 [Prab-3::IntC::unc-26∆Sac1] BJH145 unc-26 ( s1710 ) ; pekEx39 [Prab-3::unc-26Sac1::IntN , Prab-3::IntC::unc-26∆Sac1] KP5105 NuIs269 [Punc-129::gfp::unc-26] BJH360 pekEx80 [Punc-129::gfp::unc-26∆Sac1] BJH53 pekEx28 [Punc-129::gfp::unc-26∆Sac1∆PRD] BJH338 pekEx92 [Prab-3::gfp::unc-26∆PRD ( D716A ) ] BJH344 pekEx98 [Prab-3::gfp::unc-26∆Sac1∆PRD ( D716A ) ] BJH310 unc-26 ( s1710 ) ; pekEx64 [Prab-3::plc∂ PH:unc-26∆Sac1] BJH313 unc-26 ( s1710 ) ; pekEx67 [Prab-3::bem1 PX::unc-26∆Sac1] BJH314 unc-26 ( s1710 ) ; pekEx68 [Prab-3::btk PH::unc-26∆Sac1] BJH317 unc-26 ( s1710 ) ; pekEx71 [Prab-3::aps-2::unc-26∆Sac1] BJH319 unc-26 ( s1710 ) ; pekEx73 [Prab-3::itsn-1::unc-26∆Sac1] BJH320 unc-26 ( s1710 ) ; pekEx74 [Prab-3::dyn-1::unc-26∆Sac1] BJH321 unc-26 ( s1710 ) ; pekEx75 [Prab-3::apm-2::unc-26∆Sac1] BJH322 unc-26 ( s1710 ) ; pekEx76 [Prab-3::apa-2::unc-26∆Sac1] BJH330 unc-26 ( s1710 ) ; pekEx84 [Prab-3::unc-57::unc-26∆Sac1] BJH332 unc-26 ( s1710 ) ; pekEx86 [Prab-3::unc-57::unc-26∆Sac1 ( D716A ) ] BJH333 unc-26 ( s1710 ) ; pekEx87 [Prab-3::apb-1::unc-26∆Sac1] BJH335 unc-26 ( s1710 ) ; pekEx89 [Prab-3::unc-57BAR::unc-26∆Sac1] BJH336 unc-26 ( s1710 ) ; pekEx90 [Prab-3::rEndoBAR::unc-26∆Sac1] BJH337 unc-26 ( s1710 ) ; pekEx91 [Prab-3::mAmphBAR::unc-26∆Sac1] BJH340 unc-26 ( s1710 ) ; pekEx94 [Prab-3::rEndoBAR ( K76E , R78E ) ::unc-26∆Sac1] BJH341 unc-26 ( s1710 ) ; pekEx95 [Prab-3::rEndoBAR∆N::unc-26∆Sac1] BJH343 unc-26 ( s1710 ) ; pekEx97 [Prab-3::unc-26∆Sac1::snb-1] BJH345 unc-26 ( s1710 ) ; pekEx99 [Prab-3::mNadrin2BAR::unc-26∆Sac1] BJH347 unc-26 ( s1710 ) ; pekEx101 [Prab-3::unc-57::unc-26∆Sac1∆PRD] BJH348 unc-26 ( s1710 ) ; pekEx102 [Prab-3::unc-26∆Sac1::rab-3] BJH396 zxIs6 [Punc-17::ChR2 ( H134R ) ::YFP; lin-15+] , acr-16 ( ok789 ) BJH397 zxIs6 , acr-16 ( ok789 ) ; unc-26 ( s1710 ) BJH398 zxIs6 , acr-16 ( ok789 ) ; unc-26 ( s1710 ) ; pekSi7 [Prab-3::unc-26∆PRD::gfp , cb-unc-119 ( + ) ] BJH399 pekEx122 [Prab-3::gfp::unc-26∆PRD] BJH400 pekEx123 [Punc-129::gfp::unc-26Sac1 ( C378S , D380N ) ] BJH401 pekEx124 [Punc-129::gfp::unc-57SH3::unc-26∆Sac1] BJH403 pekEx126 [Punc-129::gfp] BJH405 unc-57 ( e406 ) ; unc-26 ( s1710 ) ; pekSi24 [Psnb-1::rEndoBAR::unc-26∆Sac1∆PRD; cb-unc-119 ( + ) ] BJH406 pekEx127 [Punc-129::gfp::unc-26Sac1] BJH402 nuIs122 [Pacr-2::synaptopHluorin] BJH407 unc-57 ( e406 ) ; unc-26 ( s1710 ) ; nuIs122; pekEx125 [Psnb-1::unc-26∆PRD; Prab-3::unc-57∆SH3] BJH408 unc-57 ( e406 ) ; unc-26 ( s1710 ) ; nuIs122 Psnb-1 and Prab-3 promoters were used for neuronal rescue experiments , and Punc-129 for imaging analyses . cDNAs of unc-26 , unc-57 , snb-1 , rab-3 , dyn-1 , apa-2 , apb-1 , apm-2 , aps-2 , and itsn-1 were amplified from total mRNA extracted from wt worms . cDNAs encoding rat endophilin A1 , mouse Nadrin2 , and mouse amphiphysin were amplified from a cDNA library from Clontech ( Mountain View , CA , USA ) . Transgenic strains for rescue experiments were generated by microinjection of various plasmids ( 2 ng µl−1 ) together with co-injection markers , including Pmyo-2::his11::gfp ( BJP-B36 , 2 ng µl−1 ) , Pvha-6::gfp ( BJP-B197 , 10 ng µl−1 ) , Pmyo2::NLS-MaxFP Green ( KP-JB473 , 2 ng µl−1 ) , and Pttx-3::DsRed ( KP-JB761 , 50 ng µl−1 ) . For dominate-negative inhibition experiments , plasmids ( BJP-M13 , BJP-M82 , and BJP-M185 ) were injected at ∼60 ng µl−1 . For imaging experiments , variants of Punc-129::gfp::unc-26 were injected at 15 ng µl−1 , unless specified . Blank vector pBluescript was used as an injection filler to bring final DNA concentration to 100 ng µl−1 . Integrated transgenes were obtained by UV irradiation of strains carrying extrachromosomal arrays . Transgenic worms were outcrossed at least 10 times . Mos1-mediated single-copy transgene insertion methods were used to generate transgenic animals carrying single-copy transgenes ( Frøkjaer-Jensen et al . , 2012 ) . The Mos1 target sites used in this study are ttTi5605 ( chromosome II , for unc-57 transgenes ) and ttTi14024 ( chromosome X , for unc-26 transgenes ) . The following constructs were used to generate single-copy transgenes: BJP-B178 [Prab-3::unc-26::gfp for ttTi14024 ( X ) ] , BJP-B179 [Prab-3::unc-26∆PRD::gfp for ttTi14024 ( X ) ] , BJP-B384 [Psnb-1::unc-57∆SH3::mcherry for ttTi5605 ( II ) ] , and BJP-M208 [Prab-3::rEndoBAR::unc-26 5Pase for ttTi5605 ( II ) ] . Transgenic worms carrying single copy insertion of transgenes were outcrossed at least 4 times . Worm movement on 10 cm agar plates without bacterial lawn was recorded for 30 s . Young adults ( reared at 20°C ) were transferred to room temperature 1 hr prior to behavior tests . Videos of individual animals were captured on a CCD camera ( MU130 , AmScope , Irvine , CA ) mounted on a stereomicroscope using 0 . 8× magnification . The center of mass was determined for each animal on each video frame using open-source object tracking scripts developed by Jesper S Pedersen ( http://www . phage . dk/plugins/wrmtrck . html ) in ImageJ ( NIH , Bethesda , MD ) . Average speed was determined for each animal . Statistical analysis was performed using Igor Pro 6 ( Wavemetrics , Lake Oswego , OR ) . Average values and standard error of the mean ( SEM ) were reported . p values were generated using one-way ANOVA followed by Dunnett's test . Young adult worms were immobilized on Sylgard-coated coverslips with cyanoacrylate glue ( Histoacryl Blue , Aesculap , Center Valley , PA ) . Animals were dissected in extracellular solution via a dorsolateral incision . Gonads and intestines were removed to reveal the underlying ventral nerve cord and body-wall-muscle quadrants as previously described ( Richmond et al . , 1999; Bai et al . , 2010 ) . The worm prep was mounted onto a fixed stage upright microscope ( BX51WI , Olympus , Japan ) equipped with a 60× water-immersion objective lens . Whole-cell patch clamp recordings were carried out at 20°C . A body wall muscle cell was voltage clamped at −60 mV to record postsynaptic currents . Evoked EPSC responses were induced by applying a 0 . 4 ms , 30 μA pulse , generated by a stimulus isolator ( A365 , WPI , Sarasota , FL ) , through a borosilicate pipette ( ∼2 MΩ ) placed in close apposition to the ventral nerve cord . Series resistance was compensated to 70% for the evoked EPSC recording . The currents were amplified using EPC-10 ( HEKA , Germany ) . The signals were sampled at 10 kHz using Patchmaster ( HEKA ) following low-pass filtering at 2 kHz . Patch pipettes ( 2–5 MΩ ) were pulled using borosilicate glass and were fire polished . Extracellular solution contains ( in mM ) 150 NaCl , 5 KCl , 1 CaCl2 , 5 MgCl2 , 10 glucose , and 10 HEPES and was titrated to pH 7 . 3 with NaOH , 330 mOsm with sucrose . Internal solution contains 135 CH3O3SCs , 5 CsCl , 5 MgCl2 , 5 EGTA , 0 . 25 CaCl2 , 10 HEPES , and 5 Na2ATP and was adjusted to pH 7 . 2 using CsOH . All chemicals were purchased from Sigma ( St . Louis , MO ) . Electrophysiological data were analyzed with open-source scripts developed by Eugene Mosharov ( http://sulzerlab . org/Quanta_Analysis_8_20 . ipf; Mosharov and Sulzer , 2005 ) in Igor Pro 6 ( Wavemetrics ) . Average values and SEM were reported . Statistical analysis was performed using Igor Pro 6 . p values were generated using one-way ANOVA followed by Dunnett's test . A p-value < 0 . 05 was considered to be significant . NGM plates ( 60 mm ) were seeded with 250 µl of OP50 bacteria and 4 µl of 100 mM all-trans retinal ( Sigma , St . Louis , MO ) . Seeded retinal plates were kept in the dark at 4°C and were used within 7 days . Channelrhodopsin-2 transgenic worms ( L4 hermaphrodites ) were transferred from regular plates to retinal plates in the dark at room temperature and then grown for an additional 16 hr before electrophysiological experiments . A TILL Oligochrome light source ( Till Photonics , Germany ) was controlled by TTL signals from a HEKA EPC-10/2 amplifier . Blue light ( 460–500 nm ) through a GFP filter set ( 49012 , Olympus ) was used to excite channelrhodopsin-2 . The light intensity output from the 60× objective ( 1 . 0 NA ) was 12 mW/mm2 , quantified by an XR2100 power meter ( X-Cite , Canada ) . Light pulses ( 8 ms duration , 2 Hz ) were used to evoke post-synaptic currents at neuromuscular junctions . Approximately , 5 adult hermaphrodites were loaded at room temperature into a 100 μm specimen chamber containing space-filling bacteria and M9 buffer . These worms were frozen instantaneously at ∼ −180°C in either a Leica EM PACT2 ( Leica , Germany ) or a BAL-TEC HPM010 ( Bal-Tec , Liechtenstein ) system . The frozen worms were fixed in a Leica EM AFS2 machine using 1% osmium in 0 . 1% UA in acetone fixative and then embedded in Eponate 12 from Ted Pella , Inc ( Redding , CA ) . Serial sections were cut at a thickness of 40 nm , collected on pioloform covered slotted grids ( notchnum 1 × 2 mm oval ) from Ted Pella , Inc . , and counterstained in 6% aqueous uranyl acetate for 1 . 5 hr , followed by Reynolds lead citrate for 7 min . Images were obtained on a JEOL JEM 1400 transmission electron microscope ( JEOL , Japan ) operating at 120 KV . Micrographs were collected using the Gatan Ultrascan 1000XP , 2k × 2k high-resolution camera ( Pleasanton , CA ) . Synapse profiles were used to count the number of synaptic vesicles ( ∼30 nm in diameter ) . Each profile represents a single section that passes through the dense projection . Vesicle counting was performed blindly . p values were generated using one-way ANOVA followed by Dunnett's test . Animals were immobilized with 2 , 3-Butanedione monoxamine ( 30 mg ml−1; Sigma–Aldrich ) and were mounted on 2% agarose pads for imaging . Fluorescence images were collected on an inverted Olympus FV-1000 confocal microscope with an Olympus PlanApo 60× Oil 1 . 4 NA objective at 5× zoom . GFP was excited using a 488 nm argon laser ( 0 . 5% laser power ) . Images of fluorescent slides ( Chroma Technology Group , Rockingham , VT ) were captured during each imaging session to provide a fluorescence standard for comparing fluorescence intensities between animals . Line scans were analyzed with custom-written scripts developed by Jeremy Dittman ( Weill Cornell Medical College; Dittman and Kaplan , 2006 ) in Igor Pro ( Wavemetrics , OR ) . Background signal was subtracted before analysis . ‘Synaptic enrichment’ ( % ∆F/F ) is defined as ( Fpeak − Faxon ) /Faxon . All the values reported in the figures are mean ± SEM . Young adult worms were immobilized in standard medium ( 150 mM NaCl , 5 mM KCl , 2 mM CaCl2 , 4 mM MgCl2 , 10 mM glucose , and 10 mM HEPES [pH 7 . 3] ) on Sylgard-coated coverslips with cyanoacrylate glue ( Histoacryl Blue , Aesculap ) . The head neuron ganglion was exposed by a small incision using a sharp needle . After the ganglion was exposed , a second incision was made at the middle section of worm body to release internal pressure . Dissected worms were gently rinsed with standard medium . To stimulate FM4-64 ( Invitrogen , Carlsbad , CA ) loading , dissected worms were incubated with high-potassium buffer ( 85 mM KCl , 70 mM NaCl , 2 mM CaCl2 , 4 mM MgCl2 , 10 mM glucose , and 10 mM HEPES [pH 7 . 3] ) in the presence of 10 µM FM4-64 dye for 1 min . Stimulated preparations were then incubated with standard medium containing 10 µM FM4-64 dye for 2 min to allow for vesicle recycling to proceed . To remove surface-bound dye , dissected worms were gently washed in a Ca2+-free low-K+ buffer ( 0 . 5 mM EGTA and 1 mM ADVASEP-7 [Sigma] ) for 5 min . Dye unloading from releasable vesicles was carried out by incubation with high-potassium medium without FM4-64 dye for 5 min . Imaging was done on a Zeiss LSM 780 confocal microscope ( Zeiss , Germany ) with a 40×/0 . 8 objective . FM4-64 was excited with a 561-nm laser ( 1% laser power ) , and fluorescence emission was collected between 643 nm and 751 nm . A set of Z-stack images ( 9–12 sections , step size 1 . 71 µm ) was obtained for each worm . Images were imported into ImageJ for data analysis ( NIH ) . For all images in each Z-stack , an area of interest ( AOI ) was defined by drawing a circle of 45 µm in diameter around the neuron ganglion . Total fluorescence of each image was obtained by integrating fluorescence pixel intensity within the AOI . Background fluorescence was subtracted from all images . The fluorescence signal ( arbitrary unit , a . u . ) with the highest value from each stack was used for comparison . Statistical analyses were performed using one-way ANOVA followed by Dunnett's test . All the values reported in the figures are mean ± SEM . All versions of recombinant UNC-57 and UNC-26 proteins were expressed in an Escherichia coli strain BL21 ( DE3 ) as fusion proteins . DNA fragments encoding UNC-57 full-length , UNC-57 BAR ( residues 1–283 ) , and UNC-57 SH3 ( residues 284–379 ) were inserted into pGEX4T-1 using BamHI and NotI sites . Recombinant GST::UNC-57 variants were immobilized onto glutathione beads ( Genscript , Piscataway , NJ ) . UNC-26 Sac1 ( residues 1–493 ) , 5-phosphatase ( residues 494–986 ) , and PRD ( residues 987–1113 ) were fused to the C-terminus of the maltose-binding protein ( MBP ) using overlapping PCR . DNA fragments encoding MBP::UNC-26 variants were subcloned into PET28a using NdeI and XhoI sites to produce C-terminal His6-tagged fusion proteins . Recombinant MBP-UNC-26Sac1-his6 proteins were purified using Ni-NTA Agarose ( Qiagen , Valencia , CA ) and were eluted in HEPES buffer ( 50 mm HEPES , pH 7 . 4 , 150 mM NaCl ) plus 250 mM imidazole . For antibody development , DNA encoding UNC-57 BAR::mCherry was inserted into PET28 using BamHI and NotI sites , and DNA encoding SUMO::UNC-26 ( residues 494–986 ) was inserted into PET28 using NcoI and NotI sites . Purification of GST- and His6-tagged proteins was performed essentially as previously described ( Bai et al . , 2004 ) . GST-pull down assay was performed as previously described with modification ( Bai et al . , 2004 ) . GST-tagged UNC-57 proteins ( 10 µg ) were immobilized on glutathione beads ( Genscript ) . Recombinant MBP::UNC-26::His6 fragments ( 2 µM ) were then incubated with beads in a binding buffer composed of 20 mM HEPES , 150 mM NaCl , 1% Triton X-100 , and 1 mM Dithiothreitol ( DTT ) . After 2 hr , the beads were washed 3 times with binding buffer , and the sample was treated with SDS sample buffer , subjected to SDS-PAGE , and visualized by staining with Coomassie Brilliant Blue G-250 . Monoclonal antibodies to C . elegans UNC-57 BAR and UNC-26 5-phosphatase proteins were generated in the FHCRC Monoclonal Antibody Core Facility . Recombinant proteins ( UNC-26 5-phosphatase [residues 467–986] and UNC-57∆SH3 [residues 1–283] ) were used as antigens . Mice ( e . g . , Swiss Webster , A/J , and C57BL/6 ) were immunized , and immune splenocytes were isolated from mice showing positive antisera titers ( ELISA , enzyme-linked immunosorbent assay ) . Isolated splenocytes were electrofused ( BTX , Harvard Apparatus , Holliston , MA ) to the NS-1/FOX-NY myeloma cell line . Antibody secreting hybridomas were identified using standard ELISA screens , and monoclonal hybridomas were isolated by limiting dilution subcloning . Monoclonal antibodies to UNC-57 and UNC-26 were further screened and characterized by Western blot analysis of recombinant proteins and C . elegans detergent extracts . Worms were reared on 10–15 enriched peptone plates ( 15 cm ) at 20°C . Adult animals were collected using the sucrose flotation method , frozen in liquid nitrogen , and homogenized for 20 s using a mini bead-beater16 ( Biospec Product , Bartlesville , OK ) in M9 buffer with 2 mM Ethylenediaminetetraacetic acid ( EDTA ) , 1 mM DTT , and 1× protease inhibitor cocktail ( Sigma ) . Worm samples were then cooled for 1 min on ice . The beat/cool cycle was repeated 6 times to completely lyse the worms . Triton-X100 ( final 1% ) was next added to extract total protein . After 15 min of incubation , crude protein extracts were centrifuged at 14 , 000 rpm in an Eppendorf centrifuge ( 5417C , Eppendorf AG , Germany ) for 10 min at 4°C to remove the beads , worm debris , and unbroken worms . The protein concentration of worm extracts was determined by the Pierce bicinchoninic acid ( BCA ) assay ( Thermo Scientific , Waltham , MA ) . Primary antibodies against UNC-57BAR and UNC-26 5-phophatase were used for Western blot . Immunoreactive bands were visualized using enhanced chemiluminescence and were quantified using a Bio-Rad ChemiDoc MP imaging system ( Bio-Rad , Hercules , CA ) . | Nerve cells called neurons can rapidly carry information around the body . Each neuron forms connections called synapses with several other cells to build networks for information exchange . At most synapses , electrical activity in one neuron results in the release of chemicals called neurotransmitters from storage compartments called synaptic vesicles . The neurotransmitters leave the cell and cross the gap between the two neurons to activate the next cell . After the neurotransmitters have been released , the synaptic vesicles need to be regenerated via a recycling process called endocytosis . This recycling process is very important for synapses to work properly , but it is not clear exactly how it occurs . Two of the proteins involved are called synaptojanin and endophilin . Synaptojanin is made up of three structural units ( or ‘domains’ ) , including the proline-rich domain and the Sac1 domain . It has been proposed that interactions between endophilin and the proline-rich domain of synaptojanin are essential for vesicle recycling . Here , Dong et al . studied nematode worms that carry mutant forms of synaptojanin . The experiments show that the Sac1 domain , but not the proline-rich domain , is required for the synapses to work properly . However , the Sac1 domain is not required if synaptojanin is artificially linked to endophilin . Dong et al . 's findings suggest that synaptojanin uses its Sac1 domains to work with endophilin . A future challenge will be to understand the details of how this cooperative action occurs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2015 | Synaptojanin cooperates in vivo with endophilin through an unexpected mechanism |
Neutrophils release neutrophil extracellular traps ( NETs ) which ensnare pathogens and have pathogenic functions in diverse diseases . We examined the NETosis pathways induced by five stimuli; PMA , the calcium ionophore A23187 , nigericin , Candida albicans and Group B Streptococcus . We studied NET production in neutrophils from healthy donors with inhibitors of molecules crucial to PMA-induced NETs including protein kinase C , calcium , reactive oxygen species , the enzymes myeloperoxidase ( MPO ) and neutrophil elastase . Additionally , neutrophils from chronic granulomatous disease patients , carrying mutations in the NADPH oxidase complex or a MPO-deficient patient were examined . We show that PMA , C . albicans and GBS use a related pathway for NET induction , whereas ionophores require an alternative pathway but that NETs produced by all stimuli are proteolytically active , kill bacteria and composed mainly of chromosomal DNA . Thus , we demonstrate that NETosis occurs through several signalling mechanisms , suggesting that extrusion of NETs is important in host defence .
Neutrophils are the most abundant white blood cell in the circulation and serve as the first line of host defence against pathogen attack . They are terminally differentiated , short lived cells that emerge from the bone marrow ready to react to the presence of pathogens ( Amulic et al . , 2012; Kolaczkowska and Kubes , 2013 ) . Once a foreign molecule or endogenous threat is identified the neutrophil has a battery of mechanisms it can deploy to insure optimum removal of the hazard . These include the ability to phagocytose , degranulate and produce reactive oxygen species ( ROS ) . The neutrophil can also produce chemokines and cytokines to alert other cells in the vicinity to the danger and thus maximise the host’s immune response ( Scapini and Cassatella , 2014 ) . Another form of defence utilised by the neutrophil is the release of decondensed chromatin decorated with antimicrobial peptides that can capture the pathogen in a process termed neutrophil extracellular trap ( NET ) formation ( Brinkmann et al . , 2004 ) . NETosis has been primarily examined in response to phorbol 12-myristate 13-acetate ( PMA ) , a potent mitogen and a robust NET inducer . Neutrophils also initiate NETosis in response to microbial infections and , similarly to PMA , these activate protein kinase C ( PKC ) , which in turn leads to calcium fluxes within the cell and activation of the NADPH oxidase signalling cascade resulting in the production of reactive oxygen species ( ROS ) ( Hakkim et al . , 2011; Kaplan and Radic , 2012 ) . The hydrogen peroxide ( H2O2 ) produced is in turn consumed by myeloperoxidase ( MPO ) to produce hypochlorous acid as well as other oxidants ( Papayannopoulos et al . , 2010 ) . The production of ROS is responsible for the activation of the azurosome , a protein complex composed of MPO , the serine protease neutrophil elastase ( NE ) and cathepsin G among other granular proteins . The generation of oxidants by MPO liberates NE from the azurosome , allowing it to translocate to the nucleus where it aids in the decondensation of the chromatin by proteolyzing histones ( Metzler et al . , 2014 ) . Finally , the cytoplasmic milieu mixes with the nuclear material as the nuclear and subsequently the plasma membrane break down , resulting in release of the NET . This study describes the different pathways leading to NETs by comparing the induction of NETosis by several stimuli . Primary neutrophils from healthy donors were treated with five stimuli: ( 1 ) PMA , ( 2 ) the calcium ionophore A23187 , ( 3 ) the bacterial toxin nigericin that acts as a potassium ionophore , ( 4 ) the dimorphic fungus Candida albicans and ( 5 ) the gram-positive bacteria Group B Streptococcus ( GBS ) and examined for the production of NETs . We tested a range of inhibitors against proteins involved in NETosis to clarify the essential elements in NET induction . To study the role of ROS in NETosis , we tested neutrophils isolated from chronic granulomatous disease ( CGD ) patients . These patients have mutations in genes coding for subunits of the NADPH oxidase complex and as such their neutrophils cannot make ROS ( Heyworth et al . , 2003 ) . Thus , these patients are highly susceptible to bacterial and fungal infections . We also tested neutrophils from a patient with a mutation in MPO . Citrullination is a post-transcriptional modification resulting in the conversion of arginine to citrulline and is catalysed by a group of calcium-dependent proteins known as peptidylarginine deiminases ( PADs ) ( Fuhrmann et al . , 2015 ) . Recent studies have shown that citrullination occurs during NETosis ( Lewis et al . , 2015 ) . We therefore also investigated if histone H3 is citrullinated during the induction of NETosis in response to the different stimuli . Finally , we showed that the NETs generated by the five stimuli have similar properties and that NETosis is a unique form of cell death , different from classical cell death pathways involving apoptosis and necroptosis .
We selected five representative and well-described NET inducers that are effective over a 2 . 5–4 hr time period: ( 1 ) PMA , ( 2 ) the calcium ionophore A23187 which is produced during the growth of Streptomyces chartreusensis , ( 3 ) the potassium ionophore nigericin which is derived from the bacteria Streptomyces hygroscopicus , ( 4 ) Candida albicans hyphae and ( 5 ) Streptococcus agalactiae or Group B streptococcus ( GBS ) and examined NETosis ( Figure 1A ) . We visualised and quantified NETs in samples that were fixed and stained with antibodies directed against a complex of histone 2A , histone 2B and chromatin ( Losman et al . , 1992 ) and against neutrophil elastase ( NE ) . Finally , the DNA was stained with the DNA-intercalating dye Hoechst 33342 . We used the DNA stain to count the total number of neutrophils and NETs were quantified based on the presence of extracellular chromatin and a size exclusion protocol previously described ( Brinkmann et al . , 2012 ) . Activating neutrophils with each of the stimuli resulted in a similar NET structure containing extracellular DNA co-localised with NE and chromatin ( Figure 1B–G ) . 10 . 7554/eLife . 24437 . 003Figure 1 . NETosis induction by a range of stimuli . Primary human neutrophils were stimulated for the indicated times with 50 nM PMA , 5 µM A23187 , 15 µM nigericin , MOI 5 opsonized C . albicans or MOI 10 opsonized group B streptococcus ( GBS ) , fixed with 2% PFA and incubated with a DNA stain ( Hoechst ) and immunolabeled with antibodies directed against Neutrophil Elastase ( NE ) and chromatin ( A–G ) . ( A ) NETosis rate was quantified by immunofluorescence . Graph shows mean ± SEM from independent experiments with three different donors . ( B–G ) Representative confocal microscopy images of ( B ) non-stimulated neutrophils ( - ) or NETs induced by ( C ) PMA ( D ) A23187 , ( E ) nigericin ( F ) C . albicans or ( G ) GBS and stained with Hoechst ( blue ) and immunolabeled for NE ( green ) and chromatin ( red ) . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00310 . 7554/eLife . 24437 . 004Figure 1—source data 1 . This data is the mean values of three independent NETosis assays in response to the five stimuli of interest and was used to generate the histogram in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00410 . 7554/eLife . 24437 . 005Figure 1—figure supplement 1 . NET induction over time with the five stimuli of interest . Primary human neutrophils were stimulated for the indicated times with 50 nM PMA ( B ) , 5 µM A23187 ( C ) , 15 µM nigericin ( D ) , MOI 5 C . albicans ( E ) or MOI 10 GBS ( F ) , fixed with 2% PFA and incubated with a DNA stain ( Hoechst ) and immunolabeled with antibodies directed against Neutrophil Elastase ( NE ) and chromatin . NETosis rate was quantified by immunofluorescence . Graph shows mean ± SEM from independent experiments with three different donors . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 005 Figure 1—figure supplement 1 shows that PMA ( B ) , A23187 ( C ) and nigericin ( D ) produced NETs with similar kinetics over a 3–4 hr time course . C . albicans ( E ) and GBS ( F ) induced a slower rate of NETosis and non-stimulated cells remained NET free for the duration of the experiment . Videos 1–6 also visualise the induction of NETosis in response to the aforementioned stimuli over a 6 hr time course . All stimuli resulted in the release of extracellular DNA; however , the temporal order of nucleus decondensation and plasma membrane rupture was varied . 10 . 7554/eLife . 24437 . 006Video 1 . No NETosis in non-stimulated primary neutrophilsPrimary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00610 . 7554/eLife . 24437 . 007Video 2 . PMA induced NETosis in primary neutrophilsPrimary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) , stimulated with 50 nM PMA and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . The appearance of the green colour indicated NETosis . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00710 . 7554/eLife . 24437 . 008Video 3 . A23187 induced NETosis in primary neutrophilsPrimary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) , stimulated with 5 µM A23187 and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . The appearance of the green colour indicated NETosis . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00810 . 7554/eLife . 24437 . 009Video 4 . Nigericin induced NETosis in primary neutrophilsPrimary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) , stimulated with 15 µM nigericin and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . The appearance of the green colour indicated NETosis . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 00910 . 7554/eLife . 24437 . 010Video 5 . C . albicans induced NETosis in primary neutrophils . Primary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) , stimulated with MOI 5 C . albicans and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . The appearance of the green colour indicated NETosis . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 01010 . 7554/eLife . 24437 . 011Video 6 . GBS induced NETosis in primary neutrophils . Primary neutrophils were stained with Draq5 ( blue ) and cell impermeable Sytox Green ( green ) , stimulated with MOI 10 GBS and imaged for 6 hr using a Leica SP8 AOBS confocal microscope . The appearance of the green colour indicated NETosis . Video is representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 011 PMA is a direct protein kinase C ( PKC ) activator which , in turn , leads to calcium fluxes within the cell and both of these processes are required for PMA-induced NETosis ( Gupta et al . , 2014; Fuchs et al . , 2007 ) . As anticipated , PMA-induced NET formation was blocked by the PKC inhibitor Gö6976 ( Figure 2A ) ( Gray et al . , 2013 ) . C . albicans and GBS NET induction was also blocked by the PKC inhibitor , albeit to a lesser degree ( Figure 2C ) . The two ionophores , conversely , did not require PKC ( Figure 2B ) . 10 . 7554/eLife . 24437 . 012Figure 2 . Differential requirements for PKC and calcium and a lack of requirement of transcription for NET induction by the stimuli of interest . ( A–C ) NETosis rate in PKC inhibited neutrophils . Primary neutrophils were pre-treated with the PKC inhibitor Gö6976 ( 1 µM ) for 30 min and stimulated with ( A ) PMA , ( B ) A23187 or nigericin , and ( C ) C . albicans or GBS for 2 . 5–4 hr and analysed for NET production by immunofluorescence . ( D–F ) NETosis rate in neutrophils pre-treated with the calcium chelator BAPTA-AM ( 10 µM ) for 30 min and stimulated with ( D ) PMA , ( E ) A23187 or nigericin and ( F ) C . albicans or GBS for 2 . 5–4 hr and analysed for NET production by immunofluorescence . ( G–I ) NETosis rate in neutrophils pre-treated with actinomycin D ( 1 µg/ml ) for 30 min and stimulated with ( G ) PMA , ( H ) A23187 or nigericin and ( I ) C . albicans or GBS for 2 . 5–4 hr and then analysed for NET production by immunofluorescence . Graphs show mean ± SEM from three independent experiments . *p<0 . 05 , NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 01210 . 7554/eLife . 24437 . 013Figure 2—source data 1 . This data is the mean values of three independent NETosis assays in response to the five stimuli of interest in the presence of the PKC inhibitor Gö6976 ( Figure 2A–C ) , the calcium chelator BAPTA-AM ( Figure 2D–F ) and actinomycin D ( Figure 2G–I ) and was used to generate the histograms in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 013 PMA ( Figure 2D ) and nigericin ( Figure 2E ) , induced NETosis was impaired by the calcium chelator BAPTA-AM . This chelator reduced NET formation only slightly in response to A23187 ( Figure 2E ) . Early work on neutrophil signalling revealed that ionomycin can induce a massive influx of calcium into the neutrophil , reaching a concentration greater than 1 µM ( Gennaro et al . , 1984 ) . This abundance of intracellular calcium may have overwhelmed the ability of the BAPTA-AM to chelate the calcium at the concentration used . Pre-treatment of the neutrophils with higher concentrations of the calcium chelator resulted in spontaneous NET formation ( data not shown ) therefore making the A23187 calcium requirements difficult to assess . A previous study demonstrated a role for calcium in NETosis , however , suggesting that A23187 does in fact require the calcium flux it induces to produce NETs ( Gupta et al . , 2014 ) . Notably , calcium chelation did not impair NETosis in response to C . albicans or GBS ( Figure 2F ) . Finally , as previously shown for PMA and C . albicans , A23187 , nigericin and GBS NET induction is independent of transcription ( Figure 2G–I ) ( Sollberger et al . , 2016 ) . Generation of reactive oxygen species ( ROS ) is a hallmark of PMA-induced NETosis . Figure 3A confirms that PMA induced a ROS burst in primary human neutrophils , peaking after 20 min of stimulation . This ROS burst was largely abolished by pre-treating the neutrophils with the ROS scavenger pyrocatechol . A23187 ( Figure 3B ) also induced a ROS burst , although with slower kinetics than PMA , peaking around 80 min post stimulation . Pyrocatechol also prevented this ROS burst . In contrast , nigericin did not induce any ROS production ( Figure 3B ) . Opsonized C . albicans generated ROS ( Figure 3C ) peaked , like PMA 20 min after activation . GBS-induced ROS production to similar levels but with slower kinetics . ROS release by both microbes was abrogated by pyrocatechol . PMA , A23187 , C . albicans and GBS induced ROS returned to basal levels 2 hr post-stimulation . 10 . 7554/eLife . 24437 . 014Figure 3 . Diverse stimuli have different ROS requirements for NETosis . ROS production by neutrophils ( A–C ) . ROS production was measured over a 2-hr time course in the presence or absence of the ROS scavenger pyrocatechol ( pyro , 30 µM ) in response to ( A ) PMA , ( B ) A23187 or nigericin and ( C ) C . albicans or GBS stimulation . Shown is a representative of three independent experiments . ( D–F ) NETosis rate of neutrophils pre-treated for 30 min with pyrocatechol or ( G–I ) NETosis rate of healthy control neutrophils and CGD patients stimulated with ( D and G ) PMA , ( E and H ) A23187 or nigericin and ( F and I ) C . albicans or GBS . ( A–C ) Graphs show mean ± SD from a representative of three independent experiments . ( D–F ) Graphs shows mean ± SEM from three independent experiments . ( G–I ) Graphs show mean ± SEM from five to seven independent experiments using neutrophils from five independent CGD patients ( each represented by a red circle ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 01410 . 7554/eLife . 24437 . 015Figure 3—source data 1 . This data is the mean values of three independent NETosis assays in response to the five stimuli of interest in the presence of the ROS scavenger pyrocatechol and was used to calculate the histograms in Figure 3D–F . This data also shows the means from seven independent experiments with CGD patient neutrophils and was used to generate the histograms in Figure 3G–I . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 01510 . 7554/eLife . 24437 . 016Figure 3—figure supplement 1 . No ROS production in CGD patient neutrophils , S . aureus requires ROS for NET production and C . albicans produces ROS . ( A ) ROS production by CGD neutrophils . Neutrophils from a healthy control donor and a CGD patient were examined for the production of ROS in response to PMA stimulation over a 2-hr time course . Graph shows mean ± SD from a representative of five independent ROS assays carried out with CGD patient neutrophils . ( B ) CGD patient neutrophils are impaired for S . aureus-induced NETosis . Healthy and CGD patient neutrophils were stimulated with S . aureus at a MOI of 20 for 4 hr and NETosis was examined as previously described . Graph shows mean ± SEM from three independent experiments . ( C ) C . albicans produces ROS . C . albicans-induced ROS was measured over a 3-hr time course in the presence or absence of either neutrophils ( PMN ) or pyrocatechol ( pyro ) . Graph shows mean ± SD from a representative of three independent experiments . ( D ) NETosis in response to C . albicans utilises ROS generated from C . albicans . Healthy neutrophils or C . albicans were pre-treated with pyrocatechol for 30 mins . Cells were then stimulated with C . albicans at a MOI of 5 for 3 hr and NETosis was examined as previously described . Graph shows mean ± SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 016 To test whether ROS were required for NETosis , we pre-incubated neutrophils with pyrocatechol before stimulation . As expected , ROS was absolutely required for PMA-induced NETosis ( Figure 3D ) . Conversely , pyrocatechol did not affect the level of NETosis in response to A23187 or nigericin ( Figure 3E ) . Interestingly , C . albicans-induced NETosis was impaired in the presence of the ROS scavenger , but GBS-induced NETosis was not ( Figure 3F ) . To confirm the ROS requirements for NETosis we isolated neutrophils from five patients with chronic granulomatous disease ( CGD , mutations outlined in Table 1 ) . As expected , the neutrophils from these patients were deficient in ROS production ( Figure 3—figure supplement 1A ) . As previously described ( Fuchs et al . , 2007 ) , CGD patient neutrophils did not undergo NETosis in response to PMA ( Figure 3G ) and were also significantly impaired in NET production in the presence of S . aureus ( Figure 3—figure supplement 1B ) . 10 . 7554/eLife . 24437 . 017Table 1 . CGD patient donors . Nomenclature for genotypes is according to den Dunnen and Antonarakis ( den Dunnen and Antonarakis , 2001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 017DonorAgeNucleotide changeMutationAmino acid changeResidual activity 124CYBB c . 742dupAinsertionp . Ile248AsnfsX36No 225CYBB c . 868C > Tnonsensep . Arg290XNo 318CYBB c . 1421T > Gmissensep . Leu474ArgNo 426CYBB c . 868C > Tnonsensep . Arg290XNo 529CYBA c . 371C > Tmissensep . Ala124ValYes Notably , and confirming our data with the ROS scavenger , neither A23187 nor nigericin required ROS to generate NETs ( Figure 3H ) . Intriguingly , and in contrast to data obtained with the ROS scavenger , there was no significant difference in the levels of NETosis comparing healthy vs . CGD patient neutrophils in response to C . albicans . This discrepancy was explained by the fact that C . albicans can induce a ROS burst itself in the absence of neutrophils and this was inhibited in the presence of the ROS scavenger pyrocatechol ( Figure 3—figure supplement 1C ) . Indeed , we confirmed that C . albicans produces sufficient ROS to allow NETosis . We pre-incubated the fungus with pyrocatechol ( Figure 3—figure supplement 1D ) and showed that scavenging C . albicans-derived ROS also inhibited NET production . Moreover , by inhibiting ROS in both the neutrophils and the fungus NETosis was inhibited to a greater extent . Thus , the amount of ROS produced by the C . albicans was sufficient to allow NETs induction in CGD neutrophils . The amount of NETs were , however , decreased in CGD neutrophils infected with GBS ( Figure 3I ) when compared with cells isolated from healthy donors . Overall , these data show that ROS generated by the NADPH oxidase , while absolutely essential for PMA-induced NETosis , are not necessary for NET production in response to both ionophores and only partially required for C . albicans and GBS-induced NETosis . We explored whether myeloperoxidase ( MPO ) is also differentially required by the different stimuli . Aminobenzoic acid hydrazide ( ABAH ) is a potent and irreversible small molecule inhibitor of MPO . Pre-treatment of neutrophils with ABAH did not induce spontaneous NETosis and , as anticipated , significantly decreased PMA-induced NET formation ( Figure 4A ) . This was confirmed with neutrophils isolated from a MPO-deficient patient stimulated with PMA ( Figure 4D ) . Similar to the lack of ROS requirement in NETosis in response to ionophores , the MPO inhibitor did not affect NET production by A23187 or nigericin ( Figure 4B ) and neutrophils from a MPO-deficient individual underwent NETosis in response to both stimuli ( Figure 4E ) . Interestingly , and contrary to the subtle role of ROS in NETosis induction , NETs induced by C . albicans or GBS stimulation required MPO ( Figure 4C and F ) . 10 . 7554/eLife . 24437 . 018Figure 4 . Myeloperoxidase is essential for PMA , C . albicans and GBS-induced NETosis but not for A23187 and nigericin-induced NET formation . ( A–F ) NETosis rate in response to ( A and D ) PMA , ( B and E ) A23187 or nigericin and ( C and F ) C . albicans or GBS . ( A–C ) Primary neutrophils were pre-treated for 30 min with 500 µM ABAH or a DMSO control , stimulated as indicated for 2 . 5–4 hr and analysed for NET production by immunofluorescence . Graphs show mean ± SEM from three independent experiments . ( D–F ) Healthy control neutrophils and neutrophils from a MPO-deficient patient were stimulated as outlined above . Graphs show mean ± SD from a representative of two independent experiments from a single MPO-deficient donor . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 01810 . 7554/eLife . 24437 . 019Figure 4—source data 1 . This data is the mean of three independent NETosis assays in response to the five stimuli of interest in the presence of the MPO inhibitor ABAH and was used to generate the histograms in Figure 4A–C . This data also shows the raw data used to calculate the mean of a representative experiment using MPO-deficient neutrophils used to generate the histograms in Figure 4D–F . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 019 Pre-treatment of healthy neutrophils with a highly specific small molecule NE inhibitor ( Macdonald et al . , 2001 ) did not result in spontaneous NETosis and significantly impaired PMA , C . albicans and GBS induced NETs ( Figure 5A and C ) . NETosis in response to A23187 and nigericin did not require NE ( Figure 5B ) . These data show that ionophores do not require the molecules relevant in other forms of NET induction such as ROS , MPO and NE . 10 . 7554/eLife . 24437 . 020Figure 5 . Neutrophil elastase is required for PMA , C . albicans and GBS-induced NETosis but not for A23187 or nigericin NET production . ( A–C ) NETosis rate of neutrophils during NE inhibition . Primary neutrophils were pre-treated for 30 min with a neutrophil elastase inhibitor ( GW311616A , 20 µM ) or a DMSO control and stimulated for 2 . 5–4 hr with ( A ) PMA , ( B ) A23187 or nigericin and ( C ) C . albicans or GBS and analysed for NET production by immunofluorescence . Graphs show mean ± SEM from three independent experiments . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02010 . 7554/eLife . 24437 . 021Figure 5—source data 1 . This data is the mean of three independent NETosis assays in response to the five stimuli of interest in the presence of a NE inhibitor and was used to generate the histograms in Figure 5A–C . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 021 Stimulation of neutrophils with A23187 , nigericin , C . albicans and GBS , but not PMA , resulted in Histone H3 citrullination ( cit-H3 ) within 90 min as shown by Western blot analysis , ( Figure 6A ) confirming previous publications for both PMA and the calcium ionophore ( Neeli and Radic , 2013 ) . The citrullination data were confirmed by quantifying NETosis and the number of cit-H3-positive cells by immunofluorescence . Very few of the PMA induced NETs ( Figure 6B ) stained positively for cit-H3 . Each of the other stimuli induced both NETosis and citrullination to varying levels ( Figure 6C and D ) , confirming the data seen by Western blot analysis . 10 . 7554/eLife . 24437 . 022Figure 6 . Citrullination of histone H3 occurs during NETosis but is not required for NET induction . ( A–D ) Histone H3 was citrullinated during NETosis in response to all stimuli bar PMA . ( A ) Primary neutrophils were stimulated for 90 min with PMA , A23187 , nigericin , C . albicans or GBS , lysed and assayed for the presence of citrullinated histone H3 and GAPDH by SDS-PAGE electrophoresis and Western immunoblotting . ( B–D ) NETosis rate and percentage of citrullinated cells in response to ( B ) PMA , ( C ) A23187 or nigericin and ( D ) C . albicans or GBS . Graphs show mean ± SD from a representative of two independent experiments . ( E–G ) NETosis rate in neutrophils pre-treated with the PAD inhibitor Cl-amidine at 200 µM , BB-Cl-amidine at 10 µM , TDFA at 200 µM , or DMSO as control and stimulated with ( E ) PMA , ( F ) A23187 or nigericin and ( G ) C . albicans or GBS and analysed for NET production by immunofluorescence . Graphs show mean ± SEM from 10 independent experiments . *p<0 . 05 , NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02210 . 7554/eLife . 24437 . 023Figure 6—source data 1 . This data is the mean of ten independent NETosis assays in response to the five stimuli of interest in the presence of the PAD inhibitors and was used to generate the histograms in Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02310 . 7554/eLife . 24437 . 024Figure 6—figure supplement 1 . PAD inhibitors reduce histone H3 citrullination . Neutrophils were pre-treated with ( A ) 200 µM Cl-amidine , ( B ) 10 µM BB-Cl-amidine or ( C ) 200 µM TDFA , stimulated with A23187 , C . albicans or GBS for 3–4 hr , fixed , stained with an antibody against citrullinated histone H3 and hoechst and analysed for percentage of cells citrullinated on histone H3 by immunofluorescence . Graphs show mean ± SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02410 . 7554/eLife . 24437 . 025Figure 6—figure supplement 2 . PAD inhibitors do not prevent NETosis . ( A–J ) Neutrophils were pre-treated with Cl-amidine at 200 µM ( light grey bars ) , BB-Cl-amidine at 10 µM ( dark grey bars ) , TDFA ( black bars ) at 200 µM , or DMSO as control ( white bars ) , stimulated with PMA , A23187 , nigericin , C . albicans or GBS for 2 . 5–4 hr and analysed for NET production by immunofluorescence . Graphs show mean ± SD from 10 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 025 We next explored if citrullination was required for NET formation . Neutrophils were pre-treated with three inhibitors of PAD proteins: Cl-amidine and BB-Cl-amidine , both of which inhibit PAD2 and PAD4 , and Thr-Asp-F-amidine ( TDFA ) , a potent specific PAD4 inhibitor . Treating neutrophils with Cl-amidine , BB-Cl-amidine or TDFA did not induce NETosis spontaneously ( Figure 6E ) . PMA induced NETosis was not affected by these inhibitors , consistent with the data obtained with GSK199 , another PAD4-selective inhibitor ( Figure 6E ) ( Lewis et al . , 2015 ) . Importantly , in response to A23187 or nigericin stimulation , NETosis remained intact after incubation with the three inhibitors ( Figure 6F ) . Similarly , these inhibitors did not affect C . albicans or GBS-induced NETosis ( Figure 6G ) . These data are the combination of 10 independent experiments with different donors . Each individual experiment is shown in Figure 6—figure supplement 2 . The three inhibitors reduced citrullination in response to A23187 , C . albicans and GBS , confirming that these inhibitors were active ( Figure 6—figure supplement 1A–C ) . These data show that histone H3 citrullination occurs during the course of NETosis in response to A23187 , nigericin , C . albicans and GBS but not PMA-induced NETs . Moreover , the inhibitor assays demonstrate that PAD2 and PAD4 are not required for NETosis . We next examined the properties of the NETs generated by the different stimuli . We began by examining the proteolytic activity of the NETs . As previously shown for PMA ( Papayannopoulos et al . , 2010 ) , the induction of NETosis with all five stimuli resulted in the degradation of histone H3 at both 90 and 180 min ( Figure 7A ) . Stimulation with PMA and nigericin resulted in strong degradation at the 90 min time point and a total loss of the full length histone H3 at 180 min . Conversely , A23187 , C . albicans and GBS stimulation led to less degradation overall . This was further confirmed by assaying the degradation of FITC-labelled casein in the presence of NETs isolated from healthy neutrophils treated with the five stimuli ( Figure 7B ) . NETs from all five stimuli were capable of degrading FITC-labelling casein to a similar level indicating they are proteolytically active . The supernatant of non-stimulated neutrophils was used to determine the background level of degradation and a known concentration of trypsin was added as a positive control . 10 . 7554/eLife . 24437 . 026Figure 7 . NETs are proteolytically active , kill bacteria and are mainly composed of chromosomal DNA . ( A ) NETosis leads to histone H3 degradation . Primary neutrophils were stimulated for 90 and 180 min with PMA , A23187 , nigericin , C . albicans or GBS , lysed and assayed for the presence of histone H3 and GAPDH by SDS-PAGE electrophoresis and Western immunoblotting . Shown is a representative of three independent experiments . ( B ) Isolated NETs are proteolytically active . NETosis was induced for 4 hr , NETs were isolated after treatment with AluI for 20 min , the DNA content was determined and 200 ng/ml DNA was tested for its proteolytic activity using the Pierce Fluorescent Protease Assay Kit according the manufacturer’s instructions . 100 µl of non-stimulated neutrophil supernatant was used to determine the background activity and 125 ng/ml trypsin was added as a positive control . ( C ) NETs can kill E . coli . Neutrophils were stimulated to produce NETs for 4 hr . Phagocytosis was inhibited by the addition of Cytochalasin D and E . coli at a MOI of 1 were added in the presence or absence of 50 U/ml DNase 1 . After 1 hr the cells , NETs and E . coli were collected ( selected samples were sonicated ) , serially diluted , plated on tetracycline-resistant agar plates and incubated for 24 hr at 37°C followed by CFU counts to determine killing . ( D ) NETs are primarily composed of chromosomal DNA . 4 hr post NET induction the NETs were isolated by MNase treatment , followed by proteinase K treatment . NET DNA was isolated by phenol-chloroform extraction and the ratio of S18 to S16 DNA was analysed by real-time PCR . Graphs show mean ± SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 026 Next we tested the ability of the NETs to kill E . coli . Healthy neutrophils were treated for 4 hr with the stimuli to induce NETosis , phagocytosis was blocked with the addition of Cytochalasin D and E . coli were added for 1 hr in the presence or absence of DNase 1 ( to degrade the NETs ) . NETs produced by all five stimuli were capable of limited killing that was blocked in the presence of DNase 1 ( Figure 7C ) . The NETs were sonicated post E . coli incubation to release bacteria potentially trapped in clumps and skewing the killing counts . This did not affect the bacterial counts indicating that the NETs were in fact killing the bacteria . We also examined the NETs for the presence of mitochondrial DNA as this is seen in response to autoimmune stimuli such as ribonucleoprotein immune complexes ( Lood et al . , 2016 ) . NETs produced in response to PMA , nigericin , C . albicans and GBS contained around 10% mitochondrial DNA and A23187 NETs contained around 25% ( Figure 7D ) . This is in line with previous work demonstrating that NETs are mainly generated from chromosomal DNA ( Lood et al . , 2016 ) . Taken together this data revealed that NETs produced in response to all stimuli tested can degrade proteins , kill bacteria and mostly contain nuclear DNA . To conclude , we compared NETosis , apoptosis , necrosis and necroptosis in neutrophils . We treated neutrophils with a caspase-3 inhibitor to block apoptosis or necrostatin to prevent necroptosis and measured NET formation revealing that NETosis is not affected by either of these inhibitors ( Figure 8A–C ) . 10 . 7554/eLife . 24437 . 027Figure 8 . NETosis is a unique form of cell death different from apoptosis , necrosis and necroptosis . ( A–C ) NETosis occurs in the presence of apoptosis and necroptosis inhibitors . Primary human neutrophils were pre-treated for 30 min with 20 µM caspase-3 inhibitor or 30 µM necrostatin or a DMSO control and stimulated with ( A ) PMA , ( B ) A23187 or nigericin and ( C ) C . albicans or GBS for 2 . 5–4 hr and analysed for NET production by immunofluorescence . Graphs show mean ± SEM from three independent experiments . ( D ) NETosis rate in the presence of the apoptosis inducer staurosporine . Primary neutrophils were stimulated for 2–6 hr with staurosporine ( 500 nM ) or PMA and analysed for NET induction by immunofluorescence . Graphs show mean ± SEM from three independent experiments . ( E ) NETosis rate in response to necrosis or necroptosis inducers . Primary neutrophils were stimulated with α-hemolysin ( 25 µg/ml ) to induce necrosis or with TNF-α ( 50 ng/ml ) , Z-VAD-FMK ( 50 µM ) and a SMAC mimetic ( 100 nM ) or cycloheximide ( 25 µg/ml ) to induce necroptosis for 6 hr and analysed for NET production by immunofluorescence . Graphs show mean ± SEM from three independent experiments . NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02710 . 7554/eLife . 24437 . 028Figure 8—source data 1 . This data is the mean of three independent NETosis assays in response to the five stimuli of interest in the presence of necrostatin or caspase 3 inhibitor and was used to generate the histograms in Figure 8A–C . This data also shows the mean of three independent NETosis experiments in response to staurosporine ( Figure 8D ) , hemolysin ( to induce necrosis ) , TNF-α , Z-VAD-FMK and a SMAC mimetic or cycloheximide ( to induce necroptosis ) and was used to generate Figure 8E . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 02810 . 7554/eLife . 24437 . 029Figure 8—figure supplement 1 . Apoptosis , necrosis and necroptosis can be induced in primary neutrophils , NETosis results in LDH release . ( A ) Staurosporine induced caspase-3 cleavage in neutrophils . Primary neutrophils were pre-treated with a caspase-3 inhibitor for 30 mins , stimulated with staurosporine for 3 hr; cell lysates were generated and assayed for the presence of cleaved caspase-3 and β-actin by SDS-PAGE electrophoresis and Western immunoblotting . Data shown is a representative of three independent experiments . ( B ) Staurosporine does not induce LDH release . Neutrophils were stimulated with staurosporine for 21 hr and LDH release was measured as per the manufacturer’s instructions . ( C ) Neutrophils were stimulated for 21 hr with α-hemolysin and LDH release was measured . ( D ) Neutrophils were pre-treated with a necrostatin inhibitor for 30 min and stimulated with TNF-α , Z-VAD-FMK and a SMAC mimetic or cycloheximide ( CHX ) for 21 hr and LDH release was measured . ( C–D ) Graphs show mean ± SD from a representative of two independent experiments . ( E ) LDH is released in NETosis . Neutrophils were treated with the indicated stimuli for 4 hr and LDH release was measured . Graph shows mean ± SEM from three independent experiments . Treatment of neutrophils with triton-X100 was used to normalise the data with triton treatment set to 100% LDH release . DOI: http://dx . doi . org/10 . 7554/eLife . 24437 . 029 Importantly , the apoptosis inducer staurosporine did not induce NET formation even after 6 hr incubation ( Figure 8D ) . As a control , we showed that staurosporine-induced apoptosis in neutrophils as demonstrated by the presence of cleaved caspase-3 . This cleavage was blocked by a caspase-3 inhibitor ( Figure 8—figure supplement 1A ) . These data were confirmed with the pan-caspase inhibitor Z-VAD-FMK ( data not shown ) . As expected , staurosporine did not induce LDH release ( Figure 8—figure supplement 1B ) . Finally , we induced necrosis with α-hemolysin from Staphylococcus aureus or necroptosis with a cocktail of TNF-α , Z-VAD-FMK and a SMAC mimetic or cycloheximide ( CHX ) and measured NETosis . Neither necrosis nor necroptosis activation induced NETosis beyond the level seen in the non-stimulated cells over a 6-hr time course ( Figure 8E ) . LDH assays confirmed that α-hemolysin and the necroptosis cocktail-induced cell death ( Figure 8—figure supplement 1C and D ) . As a control , we verified that necrostatin blocked cell death due to necroptosis ( Figure 8—figure supplement 1D ) . Lastly , we investigated whether LDH release occurs during NETosis . Figure 8—figure supplement 1E showed that PMA , nigericin , C . albicans and GBS stimulation resulted in LDH release greater than non-stimulated cells at 4 hr . A23187 treatment also resulted in LDH release but to a lesser extent . In conclusion , these data show that NETosis is a unique form of cell death that does not utilise components of the pathways associated with apoptosis , necrosis or necroptosis .
Recent work focusing on the various stimuli that lead to NETosis has yielded contradictory results in response to the same stimuli . These discrepancies may arise from technical issues such as differences in neutrophil isolation protocols , the culture of neutrophils in different types of cell culture media and the concentration of stimuli used . With this in mind , we aimed here to confirm that our stimuli of interest generate NETs by use of the benchmark for NETosis analysis: quantifiable analysis along with fixed cell imaging and live cell videos . Using these methods , we demonstrate here that NETs can be robustly induced by a broad range of stimuli , including PMA , the ionophores A23187 and nigericin and the more physiologically relevant stimuli C . albicans and GBS . Using neutrophils isolated from healthy donors and patients as well as small molecule inhibitors , we show that NET formation occurs through different signalling pathways . The NETs generated by all five stimuli were proteolytically active , kill bacteria and composed mainly of chromosomal DNA . We also show that NETosis is distinct from other cell death pathways such as apoptosis , necrosis and necroptosis . The data shown here demonstrates that PMA-induced NETs require PKC , calcium , ROS , MPO and NE . Conversely , Histone H3 is not citrullinated upon stimulation with PMA and transcription plays no role as has been previously shown ( Sollberger et al . , 2016 ) . Clarifying the mechanisms of NET formation in response to C . albicans and GBS proved to be challenging , perhaps due to the need to culture fresh microbes for each experiment . Similar to PMA both microbes require PKC , MPO and NE and do not require transcriptional activation for NET formation . However , both microbes induce the citrullination of histone H3 . Despite this PAD2/PAD4 activity is not required for NETosis in response to the microbes . The role of ROS is less conclusive as there is a discrepancy between comparing healthy neutrophils treated with ROS scavengers and cells isolated from CGD patients . Indeed , C . albicans induces significantly less NETs when ROS were pharmacologically abrogated in neutrophils from healthy donors . In contrast , neutrophils from CGD patients produced similar amounts of NETs as those isolated from healthy controls . This suggests that the ROS used by C . albicans do not originate from the NADPH oxidase complex . We show here that indeed C . albicans itself can produce ROS thus circumventing the ROS requirements of the neutrophil by producing sufficient levels of ROS to allow CGD neutrophils to form NETs . These results are very much in line with the clinical phenotype in which patients with CGD suffer more frequently from invasive infections with A . fumigatus than with C . albicans ( Henriet et al . , 2012 ) . Recently , a study demonstrated that CGD patient neutrophils produce significantly less NETs in response to A . fumigatus than healthy neutrophils further adding to our evidence that neutrophils from CGD patients react diversely to fungal infections ( Gazendam et al . , 2016 ) . The role of ROS in GBS-induced NETosis is confounded by the ability of healthy neutrophils to produce normal levels of NETs in the presence of the ROS scavenger but a significantly reduced level of NETosis in the CGD patient neutrophils . The CGD patient data suggest a requirement for NAPDH oxidase-dependent ROS for GBS induced NETs that is perhaps not revealed by the ROS scavenger . As seen in Figure 3C , there was some residual ROS production in response to GBS in the presence of the ROS scavenger . This level of ROS may be sufficient for the GBS to induce NETosis in a manner similar to C . albicans in which the amount of ROS produced by the fungus allows NETosis to occur in the CGD patient neutrophils . It must also be noted that although the ability of GBS to induce NETosis was significantly reduced in the CGD patients , a high level of NETosis still occurred in these cells . Taken together , these data suggest that while PMA absolutely relies on NADPH oxidase derived ROS for NETosis C . albicans and GBS can circumvent this need to some degree . This could be due to the ability of the microbes to generate ROS themselves that is then hijacked by the neutrophil to generate NETs in the absence of a self-source of ROS . This is in contrast with NETosis induction by S . aureus which was dependent on the ability of the neutrophils to generate ROS as outlined in the original study on NETosis in CGD patient neutrophils ( Fuchs et al . , 2007 ) . This suggests that NETosis induced by physiologically relevant stimuli is also very diverse in the signalling pathways utilised and hence challenging to clarify . Indeed , Leishmania amazonensis can induce NETosis in the absence of ROS production ( Rochael et al . , 2015; DeSouza-Vieira et al . , 2016 ) . The role of ROS production in the generation of NETs is further confounded by recent work demonstrating that the mitochondria can also be a source of ROS in NETosis in response to calcium ionophores ( Douda et al . , 2015 ) or ribonucleoprotein-containing immune complexes ( RNP ICs ) ( Lood et al . , 2016 ) . These data demonstrate that outside of the NADPH-oxidase complex the neutrophil has other ROS sources that are sufficient to induce NETosis . Conversely , activation with the ionophores A23187 and nigericin does not require PKC , ROS , MPO or NE or transcriptional activation and calcium only has a limited role . Histone H3 is citrullinated upon activation by these stimuli; however , pre-treatment with PAD inhibitors does not affect the ability of the neutrophils to make NETs . This suggests that citrullination ( of histone H3 at least ) is a consequence of NETosis , and that PAD4 is not required for NET formation . Ionophore and PMA-induced NETosis appears to be distinct , at least in the few components of the signal transduction cascade already described . A23187 is a calcium ionophore that causes a massive influx of calcium and nigericin stimulates potassium effluxes in cells which also results in the influx of calcium demonstrating the similarity between the ionophores in their mechanism of NET induction ( Yaron et al . , 2015 ) . This flooding of the neutrophil with calcium ions could thus result in perturbation of the membrane potential and cell death that ultimately releases NETs . While the method to initiate NET induction by the ionophores is very different to that seen in response to PMA , the end product appears to be similar . Auto-antibodies directed against citrullinated proteins are commonly found in the serum of rheumatoid arthritis patients ( van Venrooij et al . , 2004 ) and as such elucidating the origin of these modified proteins is of great therapeutic interest . Recent research suggests that PAD enzymes , in particular PAD4 , are associated with the induction of NETosis . Consistent with this , PAD4-deficient mice do not generate NETs in response to a calcium ionophore which is in direct contrast to the data presented here ( Martinod et al . , 2013 ) . However , the readout for NETosis used in this study was the presence of extracellular DNA decorated with citrullinated histone H3 . Since deficiency in PAD4 results in no citrullination of histones this study lacks a readout in the PAD4-deficient cells that would confirm the presence or absence of NETs such as staining with antibodies against NE or MPO on the extracellular DNA . It is also important to note that these experiments were carried out using murine neutrophils which may not behave similarly to human cells ( Bardoel et al . , 2014 ) . Importantly , studies examining the requirements of the PAD enzymes in human NETosis , using the same PAD inhibitors , also demonstrate a very limited inhibition of NETosis in response to a calcium ionophore and S . aureus ( Hosseinzadeh et al . , 2016; Lewis et al . , 2015 ) . Thus , it appears that while PAD enzymes might be important for murine neutrophils to generate NETs , this effect is not seen in human neutrophils . One additional discrepancy between our data and published reports is the observation that PAD inhibitors ( both selective and pan-PAD ) show efficacy in multiple mouse models of SLE and RA ( Knight et al . , 2015; Ghari et al . , 2016; Kawalkowska et al . , 2016 ) . These studies suggest that citrullination is important in disease pathogenesis and as such could affect NET function . We do not understand how NETs function as antimicrobials , immune cell activators or in coagulation . It is possible that these NET functions are altered by the citrullination of NETs components . Indeed , the phenotype observed in the PAD4-deficient mice could potentially be attributed to the effectiveness of NETs and not necessarily NET formation . A recent review highlights the wide range of proteins and pathways required for NETosis in response to a variety of stimuli with emphasis on the questions surrounding the role of citrullination in NETosis ( Konig and Andrade , 2016 ) . It states that ionophores and bacterial pore-forming toxins induce a pathway within neutrophils that results in the citrullination of histones but that is distinct from NETosis . They term this form of neutrophil cell death leukotoxic hypercitrullination ( LTH ) and suggest it is not antimicrobial but a bacterial strategy to kill neutrophils . The data shown here demonstrates that in the presence of the calcium ionophore or nigericin nuclear decondensation occurs and results in the extrusion of DNA , chromatin and peptides from neutrophils , albeit in a different manner to that utilised by PMA , C . albicans and GBS . Whether these extruded DNA and proteins are antimicrobial , however , requires further investigation . The different mechanisms of neutrophil cell death have been studied in detail and as such the data presented here can be included in the body of evidence that NETosis is in fact a distinct cell death mechanism utilised to aid in pathogen killing by neutrophils ( Fuchs et al . , 2007; Remijsen et al . , 2011b , 2011a ) . However , two recent studies on the role of necroptosis in NETs induction present contrasting evidence for and against the requirements of necroptosis ( Amini et al . , 2016; Desai et al . , 2016 ) . Our data strengthens the argument that necroptosis is a separate cell death signalling cascade that is not required by neutrophils to induce NETosis . Our observations show that there are different paths to NETosis in human cells . The elucidation of these pathways is of importance due to the ancient nature of chromatin release as a form of host defence as has been identified in both the animal and plant kingdoms . Therefore , it is unsurprising that NETosis is induced through a wide range of pathways ( Tran et al . , 2016 ) . Consequently , the clarification of these different pathways to NETosis has definite therapeutic relevance . There is a genuine need to identify NET inhibitors to alleviate or prevent many diseases including cystic fibrosis , thrombosis , malaria and sepsis ( Kaplan and Radic , 2012; Brinkmann and Zychlinsky , 2012 ) . NETs are present in the sputum of cystic fibrosis ( CF ) patients and contribute to the viscosity of the sputum ( Manzenreiter et al . , 2012 ) . NETs are evident in the thrombus in deep vein thrombosis ( DVT ) and disease activity is reflective of NET markers in the plasma ( Fuchs et al . , 2012 ) . Many autoimmune diseases such as Systemic lupus erythematosus ( SLE ) and vasculitis also show a very strong NET phenotype with regard the presence of autoantibodies against proteins readily released from neutrophils in the process of NETosis such as anti-dsDNA and anti-neutrophil cytoplasmic autoantibodies ( Hakkim et al . , 2010; Kessenbrock et al . , 2009 ) . This study will aid in the development of tools to help combat the detrimental effects of NETosis while balancing this with the need for the neutrophils to fulfil their purpose in the presence of a pathogen and induce their unique cell death program .
Gö6976 ( PKC , Biozol ) , BAPTA-AM ( Thermo Fisher Scientific ) , Actinomycin D , GW311616A ( NE ) and pyrocatechol ( Sigma-Aldrich ) , 4-Aminobenzoic acid hydrazide ( ABAH , Cayman chemical ) , Cl-amidine ( Causey and Thompson , 2008 ) , BB-Cl-amidine ( Knight et al . , 2015 ) , TDFA ( Jones et al . , 2012 ) , caspase-3 inhibitor and necrostatin ( Merck-Millipore ) . Blood samples were collected according to the Declaration of Helsinki with study participants providing written informed consent . All samples were collected with approval from the ethics committee–Charité –Universitätsmedizin Berlin . Healthy neutrophils were isolated from blood donated anonymously at the Charité Hospital Berlin . Candida albicans clinical isolate SC5314 was cultured overnight at 30°C in YPD media . GBS was grown on a 6% sheep blood agar plate overnight at 37°C , sub-cultured in Todd-Hewitt broth ( Sigma-Aldrich ) for 2–3 hr until the OD600nm reached 0 . 5 . Staphylococcus aureus was prepared as previously described ( Fuchs et al . , 2007 ) . E . coli XL1-Blue ( Stratagene ) was cultured overnight at 37°C in LB plus tetracycline . The C . albicans , GBS and S . aureus were opsonized for 30 min at 37°C with 10% human serum before addition to the neutrophils . This also ensured hyphal growth of the C . albicans . Human neutrophils were isolated by centrifuging heparinized venous blood over Histopaque 1119 ( Sigma-Aldrich ) and subsequently over a discontinuous Percoll ( Amersham Biosciences ) gradient as previously described ( Fuchs et al . , 2007 ) . Experiments were performed in RPMI-1640 ( w/o phenol red ) supplemented with 10 mM HEPES and 0 . 05% human serum albumin . Cells were seeded at 105/well ( 24-well plate ) for NET experiments and stimulated with PMA , staurosporine , cycloheximide ( Sigma-Aldrich ) , A23187 ( Santa Cruz Biotechnology Inc . ) , Nigericin ( InvivoGen ) , Candida albicans SC5314 hyphae , GBS , α-hemolysin ( generated as previously described [Virreira Winter et al . , 2016] ) , Staphylococcus aureus ( prepared as previously described [Fuchs et al . , 2007] ) , TNF-α ( Thermo Fisher Scientific ) , z-VAD-FMK ( Enzo ) or SMAC mimetic ( Birinapant , ChemieTek ) for 2–6 hr . Where applicable , cells were pre-treated inhibitors for 30 min before stimulation . Neutrophils seeded on glass coverslips were stained and quantified as previously described ( Brinkmann et al . , 2012 ) . Briefly , cells were fixed in 2% paraformaldehyde ( PFA ) post-NET induction , permeabilized on 0 . 5% Triton-X100 , blocked for 30 min in blocking buffer . Neutrophils were then stained with the following primary antibodies: anti-neutrophil elastase ( Calbiochem: 481001 , RRID:AB_212213 ) and antibodies directed against the subnucleosomal complex of Histone 2A , Histone 2B , and chromatin ( [Losman et al . , 1992] , generated in house ) . The secondary antibodies donkey anti-mouse Cy3 ( Jackson ImmunoResearch Labs: 715-175-150 , RRID:AB_2340819 ) and donkey anti-rabbit Alexa Fluor488 ( Life Technologies: A11008 , RRID:AB_143165 ) were used . Finally , the samples were stained with Hoechst 33342 ( Immunochemistry: 639 , RRID:AB_2651135 ) and mounted with Mowiol . Image acquisition was c using a Leica DMR upright fluorescence microscope equipped with a Jenoptic B/W digital microscope camera and analysed using ImageJ/FIJI software . Primary human neutrophils ( 106 ) were washed once by centrifugation ( 300 g , 10 min , RT ) in imaging medium ( 20 mM HEPES , 2 . 5 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 0 . 1% Human Serum Albumin , pH 7 . 4 ) ( Sigma Aldrich ) and then resuspended in 4 ml imaging medium containing 2 µM draq5 ( Biostatus ) and 0 . 5 µM Sytox Green ( Thermo Fischer Scientific ) . Each well of an eight-well ibidi treat dish ( ibidi ) was filled with 200 µl of that suspension and cells were allowed to settle down for 30 min at imaging temperature . NETs were induced by PMA , A23187 , nigericin , C . albicans or GBS at the concentrations outlined above . Imaging was performed with a Leica SP8 AOBS confocal microscope equipped with a motorized stage and temperature-controlled chamber at 36°C . Images ( 2048*2048 pixels ) were acquired at 0 . 5% maximal laser intensities every 2 min for each well for a total duration of 6 hr . Neutrophils were seeded at concentration of 1 × 105 cells per well in 200 μl RPMI ( w/o phenol red ) supplemented with 10 mM HEPES , 0 . 05% human serum albumin , 50 μM luminol and 1 . 2 units/ml horseradish peroxidase and pre-treated with pyrocatechol for 30 min at 37°C . The cells were then stimulated for 2 hr with the indicated stimuli and luminescence was measured over time in a VICTOR Light luminescence counter from Perkin Elmer . Neutrophil lysates were generated from 5 × 106 cells 90 , 180 min ( cit-H3 and H3 ) or 3 hr ( caspase-3 ) post stimulation by lysis in RIPA buffer ( 50 mM Tris-HCl Ph 8 . 0 , 150 mM NaCl , 1 mM EDTA , 1% NP-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 10 mM sodium fluoride , 25 mM sodium pyrophosphate ) supplemented with protease inhibitor cocktail ( Sigma-Aldrich ) , 20 µM neutrophil elastase V inhibitor and 20 µM cathepsin G inhibitor ( 219372 , both Calbiochem ) . Protein lysates were quantified by bicinchoninic acid assay ( BCA assay , Pierce ) according to manufacturer’s instructions . Protein lysates were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis followed by analysis via Western immunoblotting using an anti-citrullinated Histone H3 primary antibody ( abcam: ab5103 , RRID:AB_304752 ) , an anti-histone H3 antibody ( Active Motif: 39164 , discontinued ) an anti-cleaved Caspase-3 antibody ( 9661 , RRID:AB_2341188 ) , anti-β-actin ( 5057S , RRID:AB_10694076 ) or anti-GAPDH ( all Cell Signaling Technology: 5014S RRID:AB_10693448 ) and anti-rabbit HRP ( Jackson ImmunoResearch Labs: 111-035-144 , RRID:AB_2307391 ) . NETs were generated as described above using 1 . 5 × 106 cells/point . The NETs were isolated as previously described ( Barrientos et al . , 2013 ) . Briefly , the samples were treated with 4 U/ml AluI , the NETs were collected , their DNA was quantified using the Quant-iT PicoGreen dsDNA Assay Kit ( Thermo Fischer Scientific ) and the protease activity of 200 ng/ml of NET DNA was quantified using the Pierce Fluorescent Protease Assay kit according to the manufacturer’s instructions . 100 µl of non-stimulated neutrophil supernatant was used to calculate background protease activity and 125 ng/ml trypsin was used as a positive control . NETs were generated as described above using 1 × 106 cells/point and stimulated for 4 hr . Bacterial killing was assayed as previously described ( Ermert et al . , 2009 ) . Briefly , once NETosis was induced ( visualised by light microscopy ) , the cells were treated with 10 µg/ml Cytochalasin D ( Sigma-Aldrich ) for 15 min to block phagocytosis . A subset of samples were treated with DNase 1 at 50 U/ml to degrade the NETs prior to killing . A tetracycline-resistant E . coli strain was added to the neutrophils at a MOI of 1 and incubated at 37°C for 1 hr . The cells and E . coli were collected , a subset of samples were sonicated to release any trapped bacteria , serially diluted , plated on tetracycline-treated agar plates and incubated at 37°C for 24 hr . Finally CFUs were counted . NETs were generated as described above using 1 . 5 × 106 cells/point for 4 hr . NETs were released , DNA was isolated and analysed for nuclear ( S18 ) versus mitochondrial ( S16 ) content by Q-PCR as previously described ( Lood et al . , 2016 ) . Neutrophils were seeded at 1 × 105 cells/well in a 96-well plate and treated for 21 hr with the indicated stimuli . LDH release was quantified from the supernatants using Cytotox 96 Non-Radioactive Cytotoxicity Assay ( Promega ) according to the manufacturer’s instructions . Data are presented as mean ± SEM unless otherwise noted and were analysed using a two-sided Student t test . All analyses were performed using GraphPad Prism software ( Version 6 . 04 ) . Results were considered significant at p<0 . 05 ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . | The immune system protects the body against microorganisms that can cause infections and diseases . Neutrophils are a type of immune cell that patrol the blood in search of germs . Once they encounter potentially harmful microbes , neutrophils eradicate them in different ways . One way to catch the germs is by using ‘neutrophil extracellular traps’ , or NETs for short , to confine and kill the invaders . NETs are web-like structures made up of anti-microbial proteins and the neutrophil’s own DNA . The process of making NETs kills the neutrophil , as it eventually explodes to release the NETs . NETs play a key role in disease prevention , but producing too many NETs or producing them at the wrong time or in the wrong place can actually make certain diseases worse . Therefore , it is important to fully understand the signaling pathways and molecules the neutrophils use to make NETs . Kenny et al . exposed neutrophils from healthy people to five different compounds known to cause the cells to make NETs , including some harmful molecules , a fungus and a bacterium . Microscopy was then used to count how many neutrophils made NETs in response to each of the five stimuli . Further experiments showed that neutrophils from patients with an immune system disorder produced fewer NETs when stimulated with some of the compounds , while the other stimuli caused neutrophils to produce the same levels of NETs as healthy individuals . Kenny et al . also revealed that neutrophils use several different ways to produce and release NETs , depending on the stimulus used . Some of the ways required reactive oxygen species , such as hydrogen peroxide and enzymes , while others produced NETs without the need for these molecules . Lastly , Kenny et al . showed that the way the cells die after the NET is released is unique from other pathways that are known to kill cells . Future work will aim to identify a single molecule that can block neutrophils from releasing NETs at the wrong time and place , without affecting the important role NETs play in fighting germs . Such a molecule could be developed into a drug for people with diseases like lupus or rheumatoid arthritis , where the release of NETs makes the disease worse not better . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2017 | Diverse stimuli engage different neutrophil extracellular trap pathways |
Trisomy , the presence of a third copy of one chromosome , is deleterious and results in inviable or defective progeny if passed through the germ line . Random segregation of an extra chromosome is predicted to result in a high frequency of trisomic offspring from a trisomic parent . Caenorhabditis elegans with trisomy of the X chromosome , however , have far fewer trisomic offspring than expected . We found that the extra X chromosome was preferentially eliminated during anaphase I of female meiosis . We utilized a mutant with a specific defect in pairing of the X chromosome as a model to investigate the apparent bias against univalent inheritance . First , univalents lagged during anaphase I and their movement was biased toward the cortex and future polar body . Second , late-lagging univalents were frequently captured by the ingressing polar body contractile ring . The asymmetry of female meiosis can thus partially correct pre-existing trisomy .
During female meiosis , a G2 oocyte containing four genome copies undergoes two asymmetric cell divisions depositing one genome in a single haploid egg , while the other three genomes are segregated into polar bodies . These divisions are mediated by meiotic spindles that are asymmetrically positioned against the oocyte cortex with the pole-to-pole axis of the spindle perpendicular to the cortex . Both the inheritance of only one of the four genome copies and the distinct perpendicular positioning of the meiotic spindle are remarkably conserved among animal phyla suggesting a selective advantage ( Maro and Verlhac , 2002; Fabritius et al . , 2011a; Maddox et al . , 2012 ) . Several advantages of asymmetric meiosis have been suggested previously , yet none are applicable to all animals . Asymmetric meiotic spindle positioning maximizes the volume of a single egg , helps prevent interference with the meiotic spindle by the sperm aster ( McNally et al . , 2012 ) , and preserves predetermined embryonic polarity gradients . Here , we suggest a previously unrecognized advantage of asymmetric meiosis , the ability of meiotic spindles to correct trisomy by preferentially depositing the extra chromosome copy into a polar body . Accurate segregation of homologous chromosomes to opposite spindle poles depends on a physical attachment , or chiasma , between homologous chromosomes . A chiasma consists of a crossover , which holds the two homologous chromosomes together in a bivalent so that kinetochores can be properly oriented toward opposite poles ( Moore and Orr-Weaver , 1998; Miller et al . , 2013 ) . When a chiasma does not form , univalent chromosomes may maintain sister cohesion and move to poles independent of their homologs at anaphase I as can occur in Saccharomyces cerevisiae ( Buonomo et al . , 2000 ) . If a univalent chromosome biorients , loses cohesion , and segregates sister chromatids at anaphase I ( e . g . , Nicklas and Jones , 1977; Lemaire-Adkins and Hunt , 2000; Kouznetsova et al . , 2007 ) , the resulting single chromatid will segregate randomly at anaphase II . Random segregation of homologs at anaphase I or single chromatids at anaphase II should result in equal frequencies of haplo and diplo ova in the case of a trisomy ( Figure 1A ) and equal frequencies of nullo and diplo ova in the case of a crossover failure . 10 . 7554/eLife . 06056 . 003Figure 1 . Trisomy correction during meiosis I . ( A ) Illustration showing expected outcomes of female meiosis in XXX wild-type worms , assuming the extra univalent X ( red ) does not lose cohesion ( yellow ) between sister chromatids during anaphase I and assuming random segregation . ( B ) Illustration of a spindle with chromosomes at the metaphase plate with poles marked ‘P’ ( left ) and a projection of the cross-sectional view down the pole-to-pole axis at the metaphase plate ( right ) . ( C–E ) Z projections of fixed meiotic embryos viewed down the pole-to-pole spindle axis . Meiotic embryos from XXX wild-type mothers were stained with DAPI and anti-tubulin antibody . ( C ) Metaphase I spindles with 7 chromosomes; right images of X-fluorescence in situ hybridization ( FISH ) show two X chromosomes on the spindle . See also Supplementary file 2 . ( D ) Metaphase II spindles with 6 chromosomes; right images of X-FISH show one X on the spindle and 2–3 foci in the polar body . See also Supplementary file 3 . ( E ) Metaphase II spindle with 7 chromosomes . ( F ) Frequency of each spindle class among the progeny of XXX wild-type mothers . Insets show polar bodies , marked by asterisks , which were used to identify metaphase II spindles . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 003 Deviations from random segregation are suggested by observations of X chromosomes in Caenorhabditis elegans . In C . elegans , the single unpaired X chromosome from an XXX mother is inherited with unexpectedly low frequency with twice as many haploX ova produced as diploX ova ( Hodgkin et al . , 1979 ) . HIM-8 is a zinc finger protein that binds to specific DNA sequences that are enriched on the X chromosome . him-8 mutants have a pairing defect that is completely specific for the X chromosome , resulting in two X univalents and five autosomal bivalents in 95% of diakinesis oocytes ( Phillips et al . , 2005 ) . If the two X univalents segregated randomly , him-8 mutants would be expected to produce equal frequencies of nulloX ova and diploX ova . However , Hodgkin et al . ( 1979 ) demonstrated a fivefold preponderance of nulloX ova over diploX ova in him-8 . Using sex-reversed him-8 XX males , these authors showed the opposite effect in spermatogenesis . Rather than producing the 50% haploX , 25% diploX , 25% nulloX sperm expected from random segregation , him-8 XX males produced 86% haploX , 3% diploX , 11% nulloX sperm , indicating symmetric distribution of univalents during male meiosis . Thus , achiasmate maternal X chromosomes are inherited with unexpectedly low frequency in worms . Five mechanisms might reduce the frequency of trisomic offspring from trisomic or him-8 mothers . First , trisomic embryos might die during embryonic development resulting in undercounting of XXX offspring . This is unlikely in C . elegans because both XXX and him-8 mutant mothers produce a very low frequency of dead embryos ( Hodgkin et al . , 1979; Supplementary file 1 ) . A second possibility is that mitotic non-disjunction in the XXX mother results in a mosaic gonad that contains both diploX and triploX oocytes . Selective apoptosis of XXX germ line cells ( Bhalla and Dernburg , 2005 ) would then enrich for XX germ line cells . This does not contribute to the segregation bias in C . elegans , as the most mature diakinesis oocytes in him-8 and wild-type XXX worms have 7 rather than 6 DAPI-staining bodies ( Phillips et al . , 2005; this study ) . A third possibility is that a univalent present during metaphase I or a single chromatid present during metaphase II would be broken or otherwise degraded during anaphase . A fourth possibility is that many XXX progeny look normal because of the stochastic nature of dosage compensation and thus are undercounted . A fifth possibility is that univalent chromosomes present at metaphase I are preferentially placed in the first polar body . Here , we demonstrate that indeed biased deposition of univalent X chromosomes into the first polar body reduces the frequency of trisomic zygotes resulting from oocytes with unpaired X chromosomes .
Elimination of the extra chromosome from an oocyte starting with a trisomy would result in rescue to a euploid state . It has previously been shown that C . elegans XXX wild-type oocytes have a paired bivalent X and an unpaired univalent X chromosome in pachytene ( Goldstein , 1984 ) . We picked wild-type XXX adult hermaphrodites from the progeny of an XXX strain ( AV494 , Mlynarczyk-Evans et al . , 2013 ) based on their characteristic dumpy morphology as described by Hodgkin et al . ( 1979 ) . Meiotic embryos from XXX mothers were fixed and stained for microtubules and DNA . Chromosomes are well separated by bundles of microtubules during C . elegans female meiotic metaphase . This unique morphology facilitates counting of individual chromosomes on the metaphase plate when viewed down the pole-to-pole axis of the spindle ( Figure 1B ) . We found that 100% of metaphase I meiotic embryos from XXX wild-type worms had 7 DAPI-staining bodies on the spindle ( Figure 1C , F ) , consistent with 6 bivalents and a single univalent X . Two chromosomes were labeled with an X-specific fluorescence in situ hybridization ( FISH ) probe in these spindles ( Figure 1C , Supplementary file 2 ) . This result shows that a mosaic gonad resulting from mitotic nondisjunction cannot explain the low frequency of XXX offspring from XXX worms . If the univalent segregated randomly during anaphase I , 50% of metaphase II spindles should have 6 DAPI-staining bodies ( 6 bivalents ) and 50% should have 7 DAPI-staining bodies ( 6 bivalents and 1 univalent ) . Instead , 71% of metaphase II spindles contained only 6 DAPI-staining bodies and only 29% contained 7 DAPI-staining bodies ( Figure 1D–F ) . These frequencies match the 2:1 ratio of X to XX ova previously interpreted from genetic studies ( Hodgkin et al . , 1979 ) and are significantly different than the 50% expected from random segregation ( one-tailed p = 0 . 004 , Pearson's chi-squared test ) . This result eliminates the possibilities that XXX mothers have many XXX offspring that are undercounted due to incomplete penetrance of the XXX dumpy phenotype or that hermaphrodite nulloX sperm contributes significantly to the low frequency of XXX self-progeny . The finding that all assayed metaphase I spindles had 7 chromosomes also indicates that our method of identifying XXX worms as dumpy individuals is accurate and the high frequency of metaphase II spindles with 6 chromosomes is not a result of misidentifying diploid worms as XXX worms . FISH with an X-specific probe revealed that in 6/6 metaphase II embryos with only 6 DAPI-staining bodies , a single hybridization signal was present in the spindle and 2–3 hybridization signals were present in the polar body ( Figure 1D; Supplementary file 3 ) . Because only a single X-hybridization signal was observed in the first polar body in 5/5 spindles from diploids , these results demonstrate that single X univalents are deposited in the first polar body with greater than 50% frequency . To further investigate the mechanism leading to preferential loss of univalents during meiosis I , we utilized him-8 worms as a more tractable model . It has previously been shown that diakinesis stage him-8 oocytes have 5 autosomal bivalents and two X univalents ( Phillips et al . , 2005 ) . If segregation of the two X univalents was random , these worms should produce equal numbers of nulloX and diploX ova . Instead , him-8 mutants produce a fivefold higher frequency of nulloX ova over diploX ova ( Hodgkin et al . , 1979 ) , indicating that both maternal X univalents are lost at some time after diakinesis in a large fraction of embryos . To determine when maternal X univalents are preferentially lost , we imaged both live embryos within him-8 worms expressing GFP::tubulin and mCherry::histone ( Figure 2A–E ) and also fixed embryos stained with DAPI and anti-tubulin antibodies ( Figure 2F–J ) . We assayed the number of chromosomes ( defined here as DAPI-staining or mCherry:histone-positive bodies that would include univalents and bivalents ) present at metaphase of meiosis I and II ( Figure 2K ) . At meiosis I metaphase , 7 chromosomes were present in 96% of him-8 embryos ( Figure 2B , K ) , with the remainder having 6 chromosomes . If the two univalents segregated randomly without losing cohesion , 25% of metaphase II spindles would be expected to have 5 autosomes and no X , 50% would have 5 autosomes and 1 X , and 25% would have 5 autosomes and 2 X chromosomes . Instead , 40% of him-8 metaphase II embryos had 5 chromosomes , 55% had 6 chromosomes , and only 5% had 7 chromosomes ( Figure 2K ) . These frequencies differ significantly from those expected from unbiased segregation ( chi-square test , two-tailed p < 0 . 0001 ) , closely match the ratio of nulloX to diploX ova inferred by Hodgkin et al . ( 1979 ) and support the hypothesis that the majority of X univalents are eliminated between metaphase I and metaphase II . These maternal chromosome counts are also unaffected by nulloX or diploX sperm that might contribute to phenotype-based progeny counts . 10 . 7554/eLife . 06056 . 004Figure 2 . X univalents are preferentially lost between metaphase I and metaphase II in him-8 mutants . Z projections of living ( A–E ) and fixed ( F–J ) C . elegans meiotic embryos viewed down the pole-to-pole spindle axis at metaphase I ( A , B , F , G ) or metaphase II ( C , D , E , H , I , J ) . mCherry::Histone H2B and GFP::tubulin label the chromosomes and spindle , respectively , in live embryos . Fixed embryos were stained with DAPI , anti-tubulin antibody , and a LacO FISH probe that recognizes a LacO array integrated on the X chromosome . Asterisks indicate polar bodies . Insets show polar bodies that did not fit in the image frame . In ( E ) , ‘s’ denotes a sperm outside of the embryo . Percentages are shown for each outcome ( K , L ) . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 00410 . 7554/eLife . 06056 . 005Figure 2—figure supplement 1 . zim-2 embryos also deposit unpaired chromosome V univalents into the first polar body . Z projections of fixed meiotic embryos viewed down the pole-to-pole spindle axis . Embryos were stained with DAPI , anti-tubulin antibody , and a LacO FISH probe that recognizes a LacO array integrated on chromosome V . This array is larger than the array inserted on X; thus , the foci are larger than those shown in Figure 2 . ( A ) Metaphase I embryo with 7 DAPI chromosomes and two LacO ( V ) univalents . ( B ) Metaphase II embryo with 5 DAPI chromosomes and no LacO ( V ) chromosome on the spindle and two in the first polar body . ( C ) Metaphase II embryo with 6 DAPI chromosomes and one LacO ( V ) chromosome on the spindle and one in the first polar body . ( D ) Early anaphase II spindle with 2 LacO ( V ) chromosomes in the spindle and none in the first polar body . In ( B ) and ( C ) , polar bodies are marked by asterisks . In ( D ) , the polar body is shown as an inset because it was in a distant focal plane . ( E ) Quantification of the frequencies of each class . ( F ) Diakinesis chromosome counts from the zim-2 and him-8 strains bearing Lac operator arrays . Left panel shows representative zim-2 diakinesis nucleus with 6 chromosomes , right two panels show two examples of zim-2 diakinesis nuclei with 7 chromosomes . Bar = 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 005 To confirm that the two chromosomes lost between these stages were indeed the X univalents , we used FISH with a lac operator probe to detect a multi-copy lac operator array integrated on the X chromosomes in a him-8 background ( Figure 2F–J ) . FISH revealed two X univalents and 7 total DAPI-staining bodies on 93% of all the him-8 metaphase I plates ( Figure 2G , L ) . At metaphase II , FISH revealed at least two X hybridization foci in the first polar body and none on the metaphase plate when the spindle had 5 DAPI-staining bodies ( Figure 2H ) . When 6 DAPI-staining bodies were present on the metaphase II plate , we always observed one X hybridization focus each on the metaphase plate and in the first polar body ( Figure 2I ) . Finally , metaphase II embryos containing 7 DAPI-staining bodies had two X hybridization foci on the metaphase plate and none in the first polar body ( Figure 2J ) . Together , these results demonstrate that both achiasmate X univalents are deposited into the first polar body in 40% of him-8 embryos as compared with the 25% expected from random segregation . To test whether achiasmate autosomes are also placed in the first polar body with higher than random frequency , we analyzed a strain with a lac operator array integrated on chromosome V and bearing a loss of function mutation in the him-8 homolog zim-2 , which contributes to chromosome V pairing ( Phillips and Dernburg , 2006 ) . Unlike the situation with him-8 and pairing of the X , redundancy between ZIM proteins may contribute to chromosome V pairing . Phillips and Dernburg reported only 72% of diakinesis oocytes with 7 rather than 6 DAPI-staining bodies in a zim-2 mutant , and our zim-2 strain with lacO ( V ) had only 62% of diakinesis oocytes with 7 DAPI-staining bodies ( Figure 2—figure supplement 1F ) . FISH revealed two distinct chromosome V hybridization foci and 7 DAPI-staining bodies on 41% of metaphase I spindles ( Figure 2—figure supplement 1A , E ) . Starting with 41% achiasmate V's , random segregation should yield 10% of metaphase II embryos with both V's in the first polar body ( 25% of 41% ) . Instead , FISH revealed 27% of metaphase II embryos had five DAPI-staining bodies on the metaphase plate and chromosome V hybridization foci only in the first polar body ( Figure 2—figure supplement 1B , E ) . Likewise , random segregation of achiasmate V's should yield 10% metaphase II spindles with 7 DAPI-staining bodies on the metaphase II spindle , two distinct chromosome V hybridization foci on the spindle , and none in the first polar body . Only 5% of this embryo class was observed ( Figure 2—figure supplement 1D , E ) . These frequencies are significantly different than those expected from random segregation ( chi-square test , two-tailed p < 0 . 0002 ) . The discrepancy in the fraction of zim-2 oocytes with 7 DAPI-staining bodies at diakinesis vs metaphase I raises the possibility that chromosomes might be systematically undercounted in zim-2 metaphase I spindles ( but not in wild-type , him-8 , or XXX metaphase I spindles ) . If this is the case , the two V univalents must be positioned close together on the spindle because 0/57 zim-2 metaphase I plates with 6 DAPI-staining bodies had two widely spaced lacO ( V ) FISH foci and the deviation between expected and observed nulloV metaphase II spindles would be even greater . Two results strongly indicate that the same mechanisms acting on univalent X's in him-8 mutants also act on V univalents in the zim-2 mutant . First , the fivefold preponderance of metaphase II spindles with 5 DAPI-staining bodies over those with 7 DAPI-staining bodies is similar to him-8 . Second , the presence of lacO FISH signal only in the first polar body of metaphase II embryos with 5 DAPI-staining bodies on the spindle is the same in him-8 and zim-2 . Thus , achiasmate autosomes , like achiasmate X chromosomes , are preferentially deposited into the first polar body . To understand the mechanism by which univalent X chromosomes are preferentially deposited in polar bodies , we examined their orientation and position in the spindle . Antibodies specific for the cohesin subunit , REC-8 , label a cruciform on metaphase I bivalents ( Figure 3A , B ) and a single band on metaphase II chromosomes ( Figure 3A , C ) . The single REC-8 bands on wild-type metaphase II chromosomes and on him-8 metaphase I univalents were both oriented perpendicular to the pole-to-pole axis of the spindle ( Figure 3D ) , indicating that him-8 X univalents biorient at metaphase I . him-8 worms expressing GFP::KNL-2 , which labels the C . elegans cup-shaped meiotic kinetochores ( Dumont et al . , 2010 ) , were also analyzed for biorientation and yielded the same conclusion as analysis by REC-8 antibody ( Figure 3—figure supplement 1 ) . We also examined the localization of GFP:AIR-2 , the aurora B kinase that is essential for the loss of cohesion at anaphase I and which is loaded between homologs of wild-type bivalents in a chiasma-dependent fashion ( Rogers et al . , 2002 ) . The fluorescence intensity of GFP::AIR-2 was 2 . 3 times higher on bivalents than univalents at metaphase I of him-8 embryos ( Figure 3—figure supplement 2B , C ) . AIR-2 is normally re-loaded between sister chromatids at metaphase II . GFP::AIR-2 on metaphase II chromosomes was 1 . 7 times higher than on him-8 metaphase I univalents ( Figure 3—figure supplement 2C ) , indicating that the reduced amount of AIR-2 on metaphase I univalents was not simply a consequence of the smaller size of a univalent relative to a bivalent . AIR-2 is required for the crossover dependent , prometaphase , partial removal of REC-8 from between homologs in a wild-type bivalent , an event proposed to be essential for loss of cohesion at anaphase I ( Severson and Meyer , 2014 ) . Consistent with the low levels of AIR-2 , him-8 univalents had 1 . 7 ± 0 . 3 times the intensity of REC-8 staining as the inter-homolog region of bivalents in the same spindle ( Figure 3B; n = 8 embryos , two-tailed p = 0 . 04 , chi square relative to expected 1 . 0 ) . 10 . 7554/eLife . 06056 . 006Figure 3 . X univalents biorient at metaphase I in him-8 embryos . ( A ) Cartoon diagram of REC-8 staining on bivalents and univalents . ( B and C ) Anti-REC-8 staining of metaphase I and metaphase II embryos with bivalents ( yellow arrow head ) and univalents ( white arrow head ) . In him-8 embryos , univalents at metaphase I have a single band of REC-8 with the same orientation seen on normal chromosomes at metaphase II . ( D ) Quantification of the orientation of univalents by offset angle from the metaphase plate , 0° corresponds to perfect biorientation and 90° corresponds to perfect mono-orientation . Cortical pole is on the left in all images . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 00610 . 7554/eLife . 06056 . 007Figure 3—figure supplement 1 . Imaging of GFP::KNL-2 demonstrates that him-8 univalent chromosomes biorient at metaphase of meiosis I . ( A–B ) Z-projections through fixed GFP::KNL-2 embryos stained with DAPI and anti-tubulin antibody . ( A ) Metaphase I wild-type and him-8 embryos showing the distinct KNL-2 cups around bivalents ( yellow arrow head ) and univalents ( white arrow head ) . ( B ) Metaphase II wild-type and him-8 embryos showing the characteristic KNL-2 cups around bioriented chromosomes at the metaphase plate . ( C ) Quantification of the orientation of chromosomes by offset angle from the metaphase plate , 0° corresponds to perfect biorientation and 90° corresponds to perfect mono-orientation . Cortical pole is on the left in all images . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 00710 . 7554/eLife . 06056 . 008Figure 3—figure supplement 2 . Reduced levels of AIR-2 are loaded on him-8 X univalents at meiosis I . ( A and B ) Z projections of fixed metaphase I , GFP:AIR-2 embryos stained with DAPI and anti-tubulin antibody . AIR-2 is loaded between homologs of both wild-type ( A ) and him-8 bivalents ( B ) , whereas less AIR-2 was observed on him-8 univalents ( arrow heads in B ) . ( C ) Relative pixel intensity ratios show that him-8 metaphase II chromosomes load 1 . 7 times as much AIR-2 as X univalents at metaphase I , and metaphase I bivalents load 2 . 3 times as much AIR-2 as X univalents at metaphase I . n for both refers to total number of embryos counted , where each metaphase II embryo bore 5–7 chromosomes and each metaphase I embryo bore 2 univalents and 5 bivalents . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 008 Because X univalents biorient at metaphase I but load half as much AIR-2 , which is required for loss of cohesion at anaphase I in C . elegans ( Kaitna et al . , 2002; Rogers et al . , 2002 ) , and retain twice as much REC-8 , we hypothesized that bioriented univalents might be pulled toward both spindle poles and lag on the anaphase spindle as they fail to lose cohesion . To test this possibility , we did time-lapse imaging of him-8 embryos expressing GFP::tubulin and mCherry::histone , focusing specifically on the events of anaphase I . 90% of him-8 embryos at anaphase I had one or two lagging chromosomes ( n = 119 ) compared to 2% of wild-type embryos ( n = 52 ) ( Figure 4A , B ) . In 51% of living him-8 embryos with lagging chromosomes at anaphase I , two discrete lagging chromosomes could be resolved . Each lagging chromosome eventually moved as a single unit either toward the cortex or into the embryo in 98% of embryos ( n = 179 ) ( Figures 4B , 5A ) , indicating that cohesion between sister chromatids is maintained and that univalents are not broken or destroyed during anaphase . At anaphase II , only 10% of him-8 embryos exhibited lagging chromosomes ( n = 60 ) and 0/22 wild-type embryos had lagging chromosomes , suggesting that lagging chromosomes are caused by the presence of univalents at meiosis I . 10 . 7554/eLife . 06056 . 009Figure 4 . X univalents lag at anaphase I . ( A ) Time-lapse images of a living wild-type embryo undergoing anaphase I show chromosomes separating as two distinct masses . ( B ) Time-lapse images of a living him-8 embryo show a lagging chromosome at anaphase I . ( C–E ) Z projections of fixed anaphase I embryos . ( C ) LacO FISH labeling of a wild-type strain with a LacO array integrated on the X chromosome shows normal segregation of two X homologs from one X bivalent . ( D ) LacO FISH shows that a lagging chromosome in him-8 is the X . ( E ) LacO FISH labeling of a zim-2 strain with a LacO array integrated on chromosome V showing a univalent V lagging at anaphase I . Cortical pole is to the left in all images . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 00910 . 7554/eLife . 06056 . 010Figure 5 . The contractile ring moves inward past the lagging chromosomes of him-8 embryos . ( A ) Time-lapse sequence of anaphase I in a him-8 strain with GFP::PH , GFP::Tubulin , and mCherry::Histone H2B . The plasma membrane ingresses past the lagging chromosomes to engulf them in the polar body . ( B ) Fraction of him-8 anaphase I embryos in which a lagging chromosome eventually resolved toward the cortex or eventually resolved into the embryo ( interior ) . Lagging univalents resolved more frequently toward the cortex during both early and late anaphase . Depletion of NMY-2 , the myosin required for polar body formation , eliminated only the late anaphase bias . Pairwise two-tailed p values by Fisher's exact test: him-8 late vs him-8 nmy-2 ( RNAi ) late = 0 . 02 , him-8 early vs him-8 nmy-2 ( RNAi ) early = 0 . 80 , him-8 early vs him-8 late = 1 . 0 , him-8 nmy-2 ( RNAi ) early vs him-8 nmy-2 ( RNAi ) late = 0 . 26 . p values from Pearson's chi-squared test: him-8 late vs 50% = 0 . 003 , him-8 nmy-2 ( RNAi ) late vs 50% = 0 . 32 , him-8 early vs 50% = 0 . 02 , him-8 nmy-2 ( RNAi ) early vs 50% = 0 . 38 . ( C ) Top , diagram illustrating how the position of scission by the contractile ring along the pole-to-pole spindle axis was scored . Bottom , representative images from time-lapse sequences showing scission at different positions along the length of the spindle . ( D ) Average position of contractile ring scission along the pole-to-pole spindle axis in wild-type embryos and in him-8 embryos with no lagging chromosomes , lagging chromosomes that end up at the cortex ( cortical ) , or lagging chromosomes that end up in the embryo ( interior ) . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 010 To confirm that the lagging chromosomes are bioriented X univalents , we used LacO ( X ) FISH . We found that the X-specific FISH probe labeled one or two lagging chromosomes at anaphase I of him-8 embryos , indicating that lagging chromosomes are X univalents ( 13/14 ) ( Figure 4D ) . 36% of fixed him-8 anaphase I embryos with lagging chromosomes had two distinct FISH-positive chromosomes lagging . Another 21% had a single FISH-positive lagging body , but no other FISH-positive chromosomes on the spindle indicating that the two X univalents were likely too close to resolve in these embryos . The remaining 36% had one FISH-positive lagging chromosome and one FISH-positive chromosome in one of the main chromosome masses ( one embryo had a lagging chromosome that was a bivalent ) . These results suggest that one or both X univalents lag in up to 90% of him-8 embryos . Similar results were obtained for chromosome V in zim-2 mutants , where 40% of metaphase I embryos have univalent V's ( Figure 2—figure supplement 1 ) . 27% ( 4/15 or over half of anaphase I spindles expected to have V univalents ) had a lagging chromosome . 100% ( 4/4 ) of these lagging chromosomes were chromosome V as assayed by LacO ( V ) FISH ( Figure 4E ) . These results indicate that achiasmate autosomes lag at anaphase I , just like achiasmate X chromosomes . After establishing that lagging chromosomes were univalents , we next asked if univalents that lagged were subject to biased segregation at anaphase I . To analyze this , we conducted time-lapse imaging of embryos from him-8 worms expressing GFP::tubulin and mCherry::histone , as well as embryos from him-8 worms expressing these along with GFP::PH ( plextrin homology domain ) to label the plasma membrane ( Figure 5A ) . Our time-lapse analysis revealed that 65% of lagging chromosomes eventually moved toward the cortex and the forming polar body of him-8 embryos during anaphase I ( n = 181 ) ( Figure 5B ) , indicating that preferential expulsion of lagging univalents into the first polar body could contribute to the higher than random frequency of metaphase II spindles with 5 autosomes and no X . Because the polar body contractile ring ingresses inward toward the midpoint of the late anaphase spindle where it normally scissions , we hypothesized that the preferential resolution of lagging chromosomes toward the cortex might result from inhibition of contractile ring scission until the ring ingresses past lagging chromosomes ( Figure 5C ) . In wild type , ingression of the polar body contractile ring initiates when homologs have separated by 2 . 3 μm and polar body scission completes when homologs have separated by 5 . 6 μm ( Fabritius et al . , 2011b ) . The bias of univalents that moved toward the cortex before initiation of contractile ring ingression could not be caused by engulfment by the polar body . Therefore , we separated lagging univalents into two categories , early-resolving univalents that moved to one pole while the main chromosome masses were separated by less than 4 . 0 µm and late-resolving univalents that moved to one pole only after the main chromosome masses were separated by greater than 4 . 0 µm . If late-resolving univalents were engulfed during polar body formation , elimination of contractile ring activity would reduce the fraction of lagging chromosomes resolving toward the cortex . Indeed , RNAi depletion of the non-muscle myosin , NMY-2 , which causes complete loss of cortical furrowing and polar body formation ( Fabritius et al . , 2011b ) , in him-8 embryos resulted in a significant ( p = 0 . 02 ) reduction in the percentage of late-lagging univalents resolving toward the cortex from 64% to 43% ( Figure 5B ) . As a complementary approach , we asked if more rapid polar body ring ingression would have the opposite effect of NMY-2 depletion . We previously showed that the depletion of the myosin phosphatase , MEL-11 , doubled the rate of polar body ring ingression ( Fabritius et al . , 2011b ) . Therefore , we hypothesized that the inactivation of MEL-11 might enhance the preferential engulfment of lagging univalents by the first polar body . Unlike NMY-2 depletion , which generates 100% dead embryos , the lethality of mel-11 mutants is rescued by wild-type sperm so the chromosome constitution of progeny from a mel-11 mother can be scored by phenotype . Mating otherwise wild-type males bearing the recessive X-linked marker , lon-2 , to him-8 hermaphrodites allows measurement of the frequency of nulloX ova ( which give rise to lon male progeny ) and diploX ova ( which give rise to XXX dumpy progeny ) ( Hodgkin et al . , 1979 ) . Random segregation of univalents should generate a 1:1 ratio of nulloX:diploX ova . We found that mel-11 increased the segregation bias of him-8 by sevenfold from 3:1 to 23:1 ( Table 1 ) . This result indicates that more rapid furrow ingression captures more lagging univalents in the first polar body resulting in more nulloX ova . 10 . 7554/eLife . 06056 . 011Table 1 . Enhancement of the segregation bias in him-8 mutants by mutations in the myosin phosphatase , mel-11DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 011Self-progeny countsGenotypeTemperature ( °C ) % XO male% XX hermaphrodite% XXX DpyTotal progenymel-11 ( sb55 ) unc-4200 . 299 . 8NC1763mel-11 ( sb55 ) unc-4; him-82049*51NC374unc-4; him-8203466NC1442mel-11 ( it126 ) unc-4150 . 699NC790mel-11 ( it126 ) unc-4; him-81558*38 . 63 . 4873Ratio of nulloX ova/diploX ova calculated from progeny of cross with lon-2 malesMaternal genotypeTemperature ( °C ) # NulloX ( ion male progeny ) # DiploX ( dpy progeny ) Nullo/diploTotal progenymel-11 ( it26 ) unc-42510NA785mel-11 ( it26 ) unc-4; him-825160722 . 9595unc-4; him-82598313 . 2677mel-11 increases the frequency of male progeny from him-8 mothers . mel-11 ( sb55 ) and mel-11 ( it26 ) worms produce high frequencies of dead embryos , which cannot be scored for sex at 25°C ( Wissmann et al . , 1999 ) . Percent male ( XO ) , hermaphrodite ( XX ) , and dumpy ( XXX ) progeny from self-fertilizing mel-11 , him-8 , or him-8 mel-11 double mutant worms were therefore scored at 15°C and 20°C . Only progeny that developed to the L4 or adult stage were counted . *Two-tailed p < 0 . 0001 by binomial test compared with him-8 alone . 100% of mel-11 ( it26 ) self progeny die as embryos at 25°C , but this lethality is rescued by mel-11 ( + ) sperm ( Kemphues et al . , 1988 ) . The progeny of mel-11 ( it26 ) hermaphrodites crossed with lon-2 males could therefore be scored at 25°C . When lon-2 ( + ) hermaphrodites are crossed with lon-2 males ( lon-2 is a recessive X-linked marker ) , 50% of the ova will be fertilized by sperm with a single lon-2 X chromosome . Fertilization of a nulloX ova by a lon-2 X sperm will result in a lon-2 male . Fertilization of a diploX ova by a lon-2 X sperm will result in a XXX dumpy worm . Random segregation of the unpaired X chromosomes in him-8 would result in a ratio of nulloX/diplo X ova of 1 . 0 . The mel-11; him-8 double mutant showed a sevenfold increase in the ratio of nullo/diploX ova relative to him-8 alone , indicating an increased efficiency of eliminating maternal unpaired X chromosomes . To test our hypothesis that a lagging chromosome inhibits contractile ring scission to allow univalent capture , we asked whether the presence of late-lagging univalents might cause misplacement of the contractile ring from the 50% spindle length scission point observed in wild-type embryos ( Fabritius et al . , 2011b ) by time-lapse imaging of the plasma membrane marker GFP::PH ( Figure 5A ) . Spindle length was measured between the outside edges of the main chromosome masses and only in frames in which both chromosome masses were in focus ( Figure 5C ) . In him-8 embryos , when there were no lagging univalents , the contractile ring ingressed normally to 50% spindle length as measured from the outside edge of the main chromatin mass in the interior of the embryo . When a lagging univalent was seen segregating into the polar body , the contractile ring was seen ingressing deeper into the embryo to 40% spindle length ( Figure 5C , D ) . Alternatively , when a lagging univalent was seen segregating into the embryo , the contractile ring ingressed inward to a shallower depth at 59% spindle length ( Figure 5C , D ) . To further test the idea that a late-lagging univalent might influence the choice of the scission point , we imaged formation of the first polar body in wild-type or him-8 worms expressing GFP:UNC-59 ( septin ) and mCherry: histone . Septins are polymerizing GTPases that assemble in the contractile ring with myosin II , F-actin , and anillins ( Green et al . , 2013 ) . In wild type or him-8 with early-resolving univalents , GFP:UNC-59 labeled a flat washer–shaped contractile ring that moved down to the midpoint of the elongating anaphase spindle as reported previously for GFP:NMY-2 ( Fabritius et al . , 2011b ) . The septin ring transformed into a tube ( Figure 6A , 255 s ) as previously described for myosin and ANI-1 ( Dorn et al . , 2010 ) . When cortical furrowing relaxed at the end of telophase I , the septin tube moved outward to the embryo surface , then flopped over , and remained as a separate entity next to the chromosomes in the first polar body ( Figure 6A , 420 s ) . In 5/10 him-8 embryos in which the septin-labeled ring reached the lagging univalent , the univalent was trapped in the septin tube and moved with the septin tube outward during cortical relaxation ( Figure 6B ) . In these cases , the univalent remained trapped in the septin tube adjacent to the polar body as shown in Figure 6B , C . In 3/10 cases , the septin ring passed the univalent before the tube was formed , and in these cases , the univalent joined the main mass of chromatin in the polar body . In the 2/10 cases where the univalent did not end up in the polar body , the univalent slipped out of the septin tube into the embryo before the septin tube moved toward the embryo surface . These results are consistent with a model where the septin tube traps late-lagging univalents until scission occurs on the embryo side of the septin tube . 10 . 7554/eLife . 06056 . 012Figure 6 . Lagging chromosomes are captured by the septin tube and expelled with polar bodies . Time-lapse imaging of embryos expressing GFP::septin and mCherry::histone . ( A ) Time-lapse images of a living wild-type embryo undergoing anaphase I show the conversion of a flat washer–shaped contractile ring into a tube during formation of the first polar body . ( B ) Time-lapse images of a living him-8 embryo show two lagging chromosomes at anaphase I ( arrows ) as one moves into the polar body early on , while the second is trapped in the septin tube and is extruded with the first polar body . ( C ) 2 time points of a him-8 embryo during telophase I showing the lagging chromosome trapped in the septin tube . Bar = 4 μm . Times are from the onset of homolog separation . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 012 To further test the idea that late-lagging univalents are physically trapped in the septin tube , we tried to influence the integrity of the septin tube without blocking polar body scission . Septins act together with anillins ( Green et al . , 2013 ) . C . elegans has three anillins: ANI-1 that is required for polar body scission , ANI-2 that is required for gonad development , and ANI-3 that has no known function ( Maddox et al . , 2005 ) . We hypothesized that ANI-3 might play a non-essential structural role in the polar body septin tube and that ani-3 ( RNAi ) might therefore allow late-lagging univalents to slip out of the tube back into the embryo . Indeed , RNAi of ANI-3 initiated on L4 him-8 hermaphrodites ( which have already completed spermatogenesis ) significantly ( p < 0 . 001 binomial test ) reduced the fraction of XO male progeny from 37% ( n = 9 mothers , 1960 progeny ) to 27% ( n = 11 mothers , 2123 progeny ) , whereas ani-3 ( RNAi ) had no significant effect on wild-type worms ( wt: 0 . 04% XO , 1% dead , n = 11 mothers; ani-3 [RNAi]: 0 . 05% XO , 1% dead , n = 17 mothers ) . This result suggests that compromising the integrity of the septin tube may reduce the efficiency of trapping lagging univalents in the septin tube . ANI-3 depletion did not significantly increase the frequency of XXX dumpy progeny from him-8 mothers ( him-8: 3% XXX , 5% dead; him-8 ani-3[RNAi]: 4% XXX , 6% dead ) . This apparent inconsistency might be explained if additional ANI-3–independent mechanisms act to reduce the number of XXX progeny ( see below ) . Lagging chromosomes were resolved prior to contractile ring ingression in 45% of embryos with lagging chromosomes at anaphase I . These were resolved toward the cortex 64% of the time ( n = 72 ) , and NMY-2 depletion had no significant effect on this class of embryos ( p = 0 . 8 ) ( Figure 5B ) . These results suggest that an additional mechanism biasing univalent movement toward the cortex might be at work earlier in the cell cycle . During wild-type meiosis , bivalents congress to the metaphase plate on an 8-µm long spindle that is oriented parallel to the cortex . Upon anaphase promoting complex activation , the meiosis I spindle shortens to 4 . 8 μm in length ( Yang et al . , 2003 ) , then one spindle pole moves to the cortex in a dynein-dependent manner , and homolog separation initiates ( Ellefson and McNally , 2009 , 2011 ) . We found that univalents were misaligned toward the spindle poles in fixed him-8 embryos at late metaphase I , when the meiotic spindle is shortened but not yet rotated ( Figure 7A–B ) . In 46% of these embryos , both univalents were misaligned toward the same pole ( Figure 7B ) , close to the 50% expected from random positioning . In fixed him-8 embryos at the onset of anaphase , when spindles are shortened and rotated but chromosomes are not yet separated , 57% had one or both univalents closer to the cortical pole ( 38 + 19%; Figure 7F , G ) . No embryos had both univalents closer to the interior spindle pole . We hypothesized that one of two mechanisms might link spindle rotation with the early anaphase preference for univalent movement toward the cortex . Univalents might stochastically align closer to one spindle pole before rotation and bias the movement of that pole to the cortex . Alternatively , the cortex-proximal pole might acquire distinct biochemical properties after rotation due to cortical contact and subsequently generate more pulling force on the lagging univalents and pull them preferentially toward the cortex . 10 . 7554/eLife . 06056 . 013Figure 7 . Early bias of univalent X chromosomes might occur at the metaphase to anaphase I transition . Representative cartoon diagrams and Z projections from fixed embryos stained with DAPI , anti-tubulin antibody , and LacO ( X ) FISH probe . Cortex is at the top . ( A–C ) Both X univalents on metaphase I spindles that were shortened ( 5 . 3–7 . 2 μm spindle length ) but still parallel to the embryo cortex were frequently ( 46% ) aligned closer to the same spindle pole . ( D–G ) One or both univalents on MI spindles that had rotated but homologs had not yet separated were closer to the cortex and future polar body in 38 + 19% of embryos . Both univalents were never observed closer to the interior spindle pole . Yellow dashed lines indicate the metaphase plate . ( H and I ) Time-lapse images of two univalents ( arrows in H ) or one univalent ( arrowhead in I ) offset from the metaphase plate just before rotation of the univalent-proximal pole to the cortex . Time zero is initiation of spindle rotation . Bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06056 . 013 To test whether spindle rotation is involved with the him-8 segregation bias , we utilized mei-2 ( ct98 ) , a partial loss of function katanin mutant , which causes a failure of meiotic spindle rotation but still allows polar body formation and production of viable progeny ( McNally et al . , 2006 ) . If offset univalents bias spindle rotation or if the cortex-proximal pole exerts greater pulling on univalents after rotation , then a mei-2 ( ct98 ) him-8 double mutant should have a reduced frequency of male progeny relative to him-8 alone . At 20°C , the mei-2 ( ct98 ) him-8 double mutant produced only 21% male progeny ( n = 1440 progeny from 14 parents ) , which is significantly less than the 36% male progeny produced by the him-8 single mutant ( n = 964 progeny from 8 parents; p < 0 . 0001 by one-tailed binomial test ) and is significantly different than mei-2 ( ct98 ) alone ( 0% males; n = 925 progeny from 8 parents ) . The reduction in male progeny is unlikely to be due to effects on spermatogenesis , as sperm is unaffected by katanin-null mutants ( Mains et al . , 1990 ) . This result is consistent with either spindle rotation-based models for the early anaphase segregation bias . The role of spindle rotation is not conclusive; however , since mei-2 ( ct98 ) meiotic spindles have other phenotypes besides spindle rotation failure ( McNally et al . , 2006 ) . Absolute distinction between the two possible rotation models would require unambiguous tracking of both univalents before , during , and after spindle rotation . This was not possible in any of 201 time-lapse sequences . In 8 particularly clear time-lapse sequences , one or both univalents could be identified unambiguously 10–100 s before initiation of spindle rotation . In 6/8 of these cases , the univalent-proximal pole rotated to the cortex ( Figure 7I , H ) . In 1/8 cases , the univalent-proximal pole rotated away from the cortex . In 1/8 cases , the two univalents were offset to opposite poles both before and during rotation . If the cortical environment conferred a stronger pulling force on the cortical pole after rotation , then lagging univalents crossing the midpoint of the anaphase spindle should be common . Time-lapse imaging of spindles after rotation revealed that among 30 embryos in which one or two lagging chromosomes were already positioned closer to one pole at anaphase I onset and the lagging chromosome resolved early , the lagging chromosome resolved toward the pole that it was already close to in 80% of these embryos ( data not shown ) . Among the 20% of embryos in which the lagging chromosome moved to the opposite pole after spindle rotation , the chromosome moved toward the cortex 3 times , toward the embryo 3 times , and in one instance , the two lagging chromosomes resolved to opposite poles . These observations are not consistent with a cortical pole that generates a stronger pulling force after rotation but instead favor the idea that the offset position of univalents before rotation increases the probability that the univalent-proximal pole will move to the cortex . Two results suggested that additional factors might contribute to the overall inheritance of univalent X chromosomes . Both ani-3 ( RNAi ) and mei-2 ( ct98 ) reduced the frequency of male self-progeny from him-8 worms without increasing the frequency of triploX self-progeny . We therefore tested whether aneuploid sperm might influence phenotypic outcomes by LacO ( X ) FISH on pronuclear stage embryos from self-fertilized him-8 mothers ( not shown ) . Before pronuclear meeting , male pronuclei are distinguished from female pronuclei by the presence of sperm asters . We observed 90% haploX , 8% nulloX , and 2% diploX male pronuclei ( n = 52 ) . These values are significantly different than the 50% , 25% , 25% expected from random segregation ( two-tailed p < 0 . 0001 by chi square ) and are similar to the frequencies obtained by Hodgkin et al . ( 1979 ) using genetic tests with sex-reversed him-8 XX males . 10% nulloX sperm thus make a small contribution to reducing the frequency of XXX self-progeny .
Hodgkin et al . ( 1979 ) showed that C . elegans that are trisomic for the X chromosome or that fail to form a chiasma between the normal two X homologs have fewer trisomic offspring than expected from random segregation . Our results explain this phenomenon by demonstrating that two cellular pathways preferentially segregate X univalents into the first polar body . Mechanisms reducing the frequency of trisomic offspring have not been investigated in other model organisms because in both mouse and Drosophila , animals with trisomy X are not fertile ( Schupbach et al . , 1978; Tada et al . , 1993 ) , and there are no mutants , like him-8 , that specifically block crossover formation on one specific chromosome in females . However , women with trisomy 21 or trisomy X are often fertile and have been reported to have more than 50% euploid offspring ( Bovicelli et al . , 1982; Neri , 1984; Ratcliffe et al . , 1991; Robinson et al . , 1991; Stewart et al . , 1991 ) . Triploid oysters provide a stronger example of apparent female-specific correction to a diploid state . Eggs produced by triploid females and fertilized with sperm from diploid males result in 57% diploid , 31% triploid , and 12% aneuploid offspring , whereas eggs produced by diploids and fertilized by sperm from triploids result in 15% diploid and 85% aneuploid offspring ( Gong et al . , 2004 ) . Gaging the likelihood that the phenomenon described here for C . elegans might be relevant to trisomic humans or triploid oysters is complicated by two issues . In contrast with trisomic C . elegans , triploid oysters ( Guo and Allen , 1994 ) and trisomic human oocytes sometimes form trivalent structures rather than a separate bivalent and univalent . Only 42–16% of diplotene oocytes from fetuses with trisomy 21 , trisomy 13 , or trisomy 18 exhibited a separate bivalent and univalent ( Roig et al . , 2005; Robles et al . , 2007 ) . It is difficult to predict the behavior of trivalents on the spindle . In addition , it is not clear whether a univalent present during anaphase I of a human or oyster oocyte would be more likely to move to one pole intact as in C . elegans or to separate equationally . We speculate that single chromatids resulting from equational separation of univalents at anaphase I could be subjected to asymmetric segregation at anaphase II . Our results suggest that any chromosome that lags during late anaphase might be prefentially expelled simply due to the conserved asymmetric nature of polar body formation . There is one example where a univalent chromosome exhibits the opposite of the segregation bias reported here in C . elegans . In the 44–78% of oocytes from XO mice in which the univalent segregates intact at anaphase I , the univalent is preferentially retained in the embryo ( Lemaire-Adkins and Hunt , 2000 ) . This appears to be a difference between worms and mice rather than a difference between a trisomy and a monosomy since sex-reversed XO C . elegans produce an excess of nulloX ova ( Hodgkin , 1980 ) . Discerning the overall significance of preferentially placing univalents into the first polar body is a complex problem . In the case of an XXX mother or a mother with a mosaic ovary containing trisomic and diploid oocytes , these pathways would increase the frequency of normal haploid eggs relative to that expected from random distribution of a single univalent ( Figure 1 ) . In the case of diploid oocytes with two univalent autosomes , however , 100% efficient expulsion of univalents into the first polar body would increase the frequency of lethal monosomy . Data shown in Figure 2 , however , show no significant decrease in haploid eggs ( interpreted from the frequency of MII spindles with 6 chromosomes ) from him-8 or zim-2 mothers relative to the 50% that would occur by random distribution . Thus , the efficiency of placing univalents in the first polar body has evolved to a point that corrects trisomy without reducing the frequency of haploid eggs produced from oocytes that failed to form a chiasma between one pair of homologs . The conservation of these mechanisms in other species will have to be elucidated by studies focused specifically on the concept of chromosomal errors that are corrected , rather than caused , by female meiotic spindles .
The genotypes of C . elegans strains used in this work are listed in Supplementary file 1 . For LacO ( X ) FISH , EG7477 , which has lac operator arrays integrated on chromosome II and X , was outcrossed to him-8 males or to wild-type males to eliminate the extra LacO array on chromosome II , generating strains FM299 ( wild-type LacO ( X ) ) and FM300 ( him-8 lacO ( X ) ) . The loss of the chromosome II LacO array and homozygosity for the X chromosome array were confirmed by PCR . RNAi was carried out by feeding bacteria ( HT115 ) induced to express double-stranded RNA ( Timmons et al . , 2001 ) . The clones used were nmy-2 l-3L24 , ani-3 V-12J23 ( Kamath et al . , 2001 ) . Adult hermaphrodites were anesthetized with tricaine and tetramisole and immobilized between a coverslip and agarose pad on a slide . The time-lapse images shown in Figure 2A–E and Figure 4A–B were captured on an Olympus ( Center Valley , PA ) IX71 microscope equipped with a 60× PlanApo NA 1 . 42 oil objective and an ORCA R2 CCD camera ( Hamamatsu Photonics , Hamamatsu City , Japan ) . Hg arc excitation light was shuttered by a Sutter Lambda 10-3 shutter controller ( Sutter Instruments , Novato , CA ) . Images shown in Figure 5A were captured with an Intelligent Imaging Innovations ( Denver , CO ) Marianas Spinning Disk Confocal equipped with a Photometrics ( Tucson , AZ ) Cascade QuantEM 512SC EMCCD , and Zeiss 63× 1 . 4 objective . Image sequences in Figure 6 were captured with a Perkin Elmer-Cetus ( Waltham , MA ) Ultraview Spinning Disk Confocal equipped with an Orca R2 CCD and an Olympus 60× 1 . 4 objective . Meiotic embryos were extruded from hermaphrodites by gentle squishing between coverslip and slide , flash frozen in liquid N2 , permeabilized by removing the coverslip , and then fixed in cold methanol before staining with antibodies and DAPI . Antibodies used in this work were mouse monoclonal anti-tubulin ( DM1alpha , 1:200; Sigma ) , mouse monoclonal DM1alpha:FITC conjugated ( 1:30; Sigma ) , rabbit anti-REC-8 ( from 1:500; Josef Loidl ) , Alexa 594 anti-rabbit , and Alexa 594 anti-mouse ( both from Molecular Probes and used at 1:200 ) . Images in Figure 3 were captured with an Applied Precision Deltavision Deconvolution system equipped with an Olympus PlanApo 60× 1 . 40 objective and a CoolSnap HQ CCD camera ( Photometrics ) . Deltavision z-stacks were captured at 130-nm intervals . Images in Figures 1 , 2F–J , 4C–E , 7 and Figure 2—figure supplement 1 were captured with the Olympus IX71 described above but using an Olympus DSU ( disc scanning unit ) . Z stacks were acquired by taking images every 200 nm ( unless otherwise noted ) from the top to the bottom of the spindle tubulin signal . Deconvolution was performed on most images shown . Deconvolution of time-lapse movies from the IX71 was performed using Huygens Professional X11 ( Scientific Volume Imaging , Hilversum , Netherlands ) , with point spread functions determined from bead images . Deltavision z-stacks were deconvolved using Softworx native deconvolution software , with PSFs calculated from bead images taken on that system . A lac operator oligonucleotide CCACATGTGGAATTGTG AGCGGATAACAATTTGTGG and an oligonucleotide corresponding to an X-specific repeat , XC ( Phillips et al . , 2005 ) TTTCGCTTAGAGCGATTCCTTACCCTTAAATGGGCGCCGG , were each synthesized with 3′ and 5′ Texas Red and used in hybridization to LacO arrays integrated on X or V or to endogenous X sequences . FISH with immunofluorescence was performed as described by Phillips et al . ( 2009 ) with some modifications . Worms were washed in 0 . 8% egg buffer and then placed on slides pre-coated with poly-L-lysine ( Sigma ) . Worms were then gently crushed between the slide and a 25-mm sq . #1 coverslip to extrude meiotic embryos and immediately submerged in liquid nitrogen for 10–15 min . Coverslips were then flicked off to freeze-crack eggshells , and slides were submerged in −20°C methanol for 20–30 min . Slides were then washed in 1× phosphate buffered saline ( PBS ) twice for 10 min and then in 1× PBST ( PBS with 0 . 2% Tween-20 ) for 10 min . Slides were then blocked in 1× PBST with 4% bovine serum albumen ( BSA ) for 30–45 min at room temperature in a moist chamber . Blocking solution was wicked away being careful not to dry out the samples , and FITC-conjugated DM1a was applied 1:30 in 1× PBST with 4% BSA using 20 μl coverwells ( Grace BioLabs ) . Slides were incubated in this antibody for 4 hr at room temperature or left overnight at 4°C . Slides were then washed sequentially in 1× PBST , 1× PBS , and 2× SSCT ( saline sodium citrate buffer with 0 . 4% Tween-20 ) for 10 min each . Following the last wash , slides underwent secondary fixation in 7% formaldehyde in 1× egg buffer for 5 min and were immediately dipped in 2× SSCT to wash off fixative . Slides were then washed in 2× SSCT twice for 5 min each and then pre-hybridized . Pre-hybridization was performed by adding 200 μl of 50% formamide in 2× SSCT with a 200 μl coverwell ( Grace BioLabs ) overnight at 37°C in a moist chamber . After 24 hr , slides were taken out of 37°C incubation and placed at room temperature while the FISH probe was prepared . The FISH probe was prepared by adding 0 . 6 μl of the stock ( 900 ng/μl ) to 30 μl of hybridization buffer ( hybridization buffer was made as described in Phillips et al . , 2009 ) with 300 μl/ml salmon sperm DNA and 0 . 1% Tween-20 per slide . Slides were then incubated in 30 μl of this solution under a hybridization slip ( Grace BioLabs ) at 95°C for 3 min on an OmniSlide ( Thermo Scientific ) and then at 37°C in a moist chamber for 48–72 hr . Following this incubation , slides were washed in 50% formamide in 2× SSCT as before but for two 1-hr incubations . Finally , slides were stained with DAPI by submerging in a Coplin jar filled with 2× SSCT 6 μg/ml DAPI for 10 min and were then washed for 30 min in fresh 2× SSCT . Slides were then wicked dry with a Kimwipe taking care not to dry out the sample and were mounted with 8 μl of DABCO Mowiol and sealed with nail polish . Following 2–3 days for curing , slides were imaged . Chromosome counts were carried out on live embryos in utero or on fixed embryos extruded from the worm by locating metaphase spindles whose chromosomes were all aligned at the metaphase plate . Z stacks were captured at 200-nm intervals . Spindles that were oriented sideways , with their metaphase plates perpendicular to the imaging plane , were reconstructed using ImageJ 3D stack reconstruction , and chromosomes were counted only if individual masses could be discerned . Metaphase II spindles were distinguished from metaphase I spindles by the presence of polar bodies . Time-lapse images of lagging chromosomes in FM125 , FM126 , and FM232 were acquired at 10-s intervals beginning at late metaphase I when the spindle is shortening and rotating and continuing through polar body extrusion and the formation of the metaphase II spindle . The direction of resolution was determined from the last frame where the lagging chromosome was still discernable from the segregating chromosome masses . At this frame , spindle length was determined by measuring the distance between the outside edges of the main masses of segregating chromosomes . Lagging chromosomes that decided which way to go when the spindle was more than 4 μm long were classified as late resolving because earlier work indicated that myosin-dependent polar body scission occurs when spindles are longer than 4 μm ( Fabritius et al . , 2011b ) . We confirmed this assumption by finding that 5/5 FM232 ( GFP:PH ) spindles longer than 4 μm exhibited deep cortical furrows . For nmy-2 ( RNAi ) time-lapse sequences , only embryos in which polar body extrusion completely failed were analyzed . The fate of lagging chromosomes was scored based on whether they ended up at the cortex or in the interior prior to the formation of the metaphase II spindle . Often , chromosomes at the cortex were picked up by the metaphase II spindle . These were still scored as cortex-fated lagging chromosomes . | Inside cells , DNA is found packaged into structures called chromosomes . Most human and animal cells contain two sets of chromosomes , one inherited from each parent . Chromosomes from one set pair up with the equivalent chromosome from the other set . However , egg and sperm cells only contain one copy of each chromosome , so that when the egg is fertilized , the resulting cell again has two sets of chromosomes . If there are either more or fewer than two copies of a chromosome in the fertilized cell , this can cause birth defects and conditions such as Down syndrome . An egg cell develops from a cell called an oocyte via a process called meiosis . The oocyte first duplicates its DNA so that it contains four copies of each chromosome . The oocyte then divides , and the resulting cells divide again , to produce four cells that each contains one copy of each chromosome . Only one of these cells is an egg cell: the other three are called polar bodies , and these normally self-destruct . The tiny roundworm C . elegans is a model organism used to study meiosis . Worms can be hermaphrodites or males; the hermaphrodites normally have a pair of ‘X’ sex chromosomes . However , sometimes problems with meiosis can produce hermaphrodite worms with three X chromosomes in each of their cells . In these cells , two of the X chromosomes pair with each other as normal , and one X chromosome remains unpaired . Cortes et al . examined meiosis in mutant worms that had an extra copy of the X chromosome by marking all the chromosomes with a fluorescent tag . This allowed the movement of the chromosomes to be tracked through images taken using a microscope . This revealed that an unpaired X chromosome moves more slowly than a normal paired set . Furthermore , the unpaired chromosomes tend to move toward the region of the oocyte that will develop into a polar body . Thus , when the oocyte divides , the unpaired chromosomes are placed in the polar body and eliminated . This mechanism improves the chance that the correct number of chromosomes will end up in the egg cell . Women with three X chromosomes are often fertile and in most cases produce normal offspring . Further work is needed to see whether human oocytes remove extra chromosomes by a mechanism similar to that seen in the roundworms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"cell",
"biology"
] | 2015 | The asymmetry of female meiosis reduces the frequency of inheritance of unpaired chromosomes |
Pharmacological studies in mammals suggest that norepinephrine ( NE ) plays an important role in promoting arousal . However , the role of endogenous NE is unclear , with contradicting reports concerning the sleep phenotypes of mice lacking NE due to mutation of dopamine β-hydroxylase ( dbh ) . To investigate NE function in an alternative vertebrate model , we generated dbh mutant zebrafish . In contrast to mice , these animals exhibit dramatically increased sleep . Surprisingly , despite an increase in sleep , dbh mutant zebrafish have a reduced arousal threshold . These phenotypes are also observed in zebrafish treated with small molecules that inhibit NE signaling , suggesting that they are caused by the lack of NE . Using genetic overexpression of hypocretin ( Hcrt ) and optogenetic activation of hcrt-expressing neurons , we also find that NE is important for Hcrt-induced arousal . These results establish a role for endogenous NE in promoting arousal and indicate that NE is a critical downstream effector of Hcrt neurons .
Sleep remains among the most persistent and perplexing mysteries in modern biology . Several studies have shown that neuronal centers that regulate sleep and wakefulness lie predominantly in the hypothalamus and brainstem ( Pace-Schott and Hobson , 2002; Saper et al . , 2005 ) , and many of the neurotransmitters and neuropeptides employed by these centers are known . However , it remains unclear how these centers interact with each other , and what specific functions are fulfilled by the neurotransmitters and neuropeptides they employ . Norepinephrine ( NE ) is one of the most abundant neurotransmitters in the central and peripheral nervous systems , and has been implicated in many aspects of physiology and behavior , including cognition , attention , reward , locomotion and arousal ( Berridge and Waterhouse , 2003; Weinshenker and Schroeder , 2007; Sara , 2009 ) . Small molecule activators of NE signaling have been shown to increase wakefulness , whereas inhibitors promote sleep ( Berridge et al . , 2012 ) . These results suggest that NE plays a significant role in promoting arousal . However , the role of endogenous NE in regulating the sleep/wake cycles of vertebrates remains unclear , with contradicting reports concerning the sleep phenotype of mice that do not synthesize NE due to mutation of dopamine beta hydroxylase ( dbh ) ( Hunsley and Palmiter , 2003; Ouyang et al . , 2004 ) . The locus coeruleus ( LC ) in the brainstem is a major arousal promoting center and a major source of NE in the brain ( reviewed in Berridge and Waterhouse , 2003 ) . Optogenetic studies in mice have shown that activation of the LC promotes sleep-to-wake transitions , while inhibition of the LC reduces time spent awake ( Carter et al . , 2010 , 2012 ) . These studies also showed that activation of the LC plays an important role in mediating the arousing effects of hypocretin ( hcrt ) -expressing neurons ( Carter et al . , 2010 , 2012 ) , which constitute an important arousal center in the hypothalamus ( Sutcliffe and de Lecea , 2002 ) . These observations suggest that NE synthesized in the LC could be involved in mediating Hcrt-induced arousal . However , in addition to NE , the LC also produces other neurotransmitters and neuropeptides that affect sleep , including dopamine ( Dzirasa et al . , 2006 ) , neuropeptide Y ( Dyzma et al . , 2010 ) , neurotensin ( Erwin and Radcliffe , 1993 ) and vasopressin ( Born et al . , 1992 ) . Manipulations of the LC likely affect the synaptic levels of these other neurotransmitters and neuropeptides . Therefore , it is unclear whether NE is required to mediate the arousing effects of LC neurons and Hcrt signaling , or if this is accomplished by another factor in LC neurons . To address the role of NE in regulating vertebrate sleep/wake states , we generated dbh mutant zebrafish that do not produce NE . Contrary to dbh null mice ( Thomas et al . , 1995 ) , zebrafish dbh mutants develop normally and are viable . Importantly , they also exhibit a dramatic increase in sleep . Interestingly , despite increased sleep , these animals display a reduced arousal threshold . Using this mutant , we show that NE is important for arousal that is induced by either genetic overexpression of the Hcrt neuropeptide or optogenetic activation of Hcrt neurons . These results clarify the role of NE in regulating vertebrate sleep and establish a role for NE in mediating Hcrt-induced wakefulness .
Previous pharmacological studies in zebrafish ( Rihel et al . , 2010 ) suggested that the noradrenergic system is an important regulator of sleep and wakefulness in zebrafish , similar to mammals ( Berridge et al . , 2012 ) . Therefore , we reasoned that the zebrafish model system could provide a new platform for exploring the role of endogenous NE in sleep . Three classes of receptors mediate NE signaling: the activating alpha1-and beta-adrenergic receptors and the inhibitory alpha2 adrenergic receptors . For each class there exist at least 5 paralogs in the zebrafish genome , making pharmacological manipulations more practical than genetic manipulations . We first tested the effects of prazosin , a well-established alpha1-adrenergic receptor inhibitor , on zebrafish behavior . We found that , compared to larvae exposed to dimethyl sulfoxide ( DMSO ) vehicle alone , larvae exposed to 100 µM prazosin showed lower overall activity ( Figure 1A , D ) and activity when awake ( Figure 1B , E ) during both day and night , as well as an increase in sleep during both day ( +140% ) and night ( +60% ) ( Figure 1C , F; Figure 1—figure supplement 1 ) . This increase in sleep was primarily due to an increase in the number of sleep bouts ( +110% during the day and +70% during the night , Figure 1G ) , as well as a smaller increase in sleep bout duration ( +20% during the day ) ( Figure 1H ) . We also observed a 23% reduction in sleep latency at night in prazosin-treated animals ( Figure 1I ) . 10 . 7554/eLife . 07000 . 003Figure 1 . Prazosin treated larvae are less active and sleep more than vehicle treated controls . Representative activity ( A ) , waking activity ( amount of locomotor activity while awake ) ( B ) and sleep ( C ) traces of vehicle ( blue ) and prazosin ( red ) treated zebrafish larvae . Bar graphs of activity ( D ) , waking activity ( E ) , sleep ( F ) , sleep bout number ( G ) , sleep bout length ( H ) and sleep latency ( I ) from three combined experiments ( n > 180 for each condition ) . Bars represent mean ± s . e . m . * , p < 0 . 05 and *** , p < 0 . 0001 by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 00310 . 7554/eLife . 07000 . 004Figure 1—figure supplement 1 . Prazosin night sleep dose–response curve . Larvae were treated with prazosin over range of concentrations ( n = 12 for each concentration ) on the morning of day 5 of development . Sleep was measured during night 5 . Mean ± s . e . m is shown for each concentration . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 00410 . 7554/eLife . 07000 . 005Figure 1—figure supplement 2 . Clonidine treated larvae are less active and sleep more than vehicle treated controls during the day . Representative activity ( A ) , waking activity ( B ) and sleep ( C ) traces of vehicle ( blue ) and clonidine ( red ) treated zebrafish larvae . Bar graphs of activity ( D ) , waking activity ( E ) , sleep ( F ) , sleep bout number ( G ) , sleep bout length ( H ) and sleep latency ( I ) from two combined experiments ( n > 140 for each condition ) . Bars represent mean ± s . e . m . * , p < 0 . 05; **; p < 0 . 001; *** , p < 0 . 0001 by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 00510 . 7554/eLife . 07000 . 006Figure 1—figure supplement 3 . Bopindolol treated larvae sleep more than vehicle treated controls . Representative activity ( A ) , waking activity ( B ) and sleep ( C ) traces of vehicle ( blue ) and bopindolol ( red ) treated zebrafish larvae . Bar graphs of activity ( D ) , waking activity ( E ) , sleep ( F ) , sleep bout number ( G ) , sleep bout length ( H ) and sleep latency ( I ) from two combined experiments ( n > 140 for each condition ) . Bars represent mean ± s . e . m . n indicates number of larvae . * , p < 0 . 05; ** , p < 0 . 001; *** , p < 0 . 0001 by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 006 We further investigated whether treatment with the alpha2-adrenergic receptor agonist clonindine or the beta1-adrenergic receptor antagonist bopindolol affects the sleep of larval zebrafish . We found that the main effect of clonidine treatment was a reduction in day activity ( −48% ) and increase in day sleep ( +90% ) , with no effect in night activity or sleep ( Figure 1—figure supplement 2 ) , suggesting a day-specific role for alpha2 receptors in sleep regulation . Bopindolol treatment resulted in reduction of night activity ( −40% ) and increase in both day ( +82% ) and night ( +60% ) sleep ( Figure 1—figure supplement 3 ) . Encouraged by our pharmacological studies , we decided to investigate the role of endogenous NE in sleep regulation using genetics . To this end we used the zinc-finger nuclease approach ( Foley et al . , 2009 ) , a well established technique in zebrafish that induces targeted mutations with few off-target lesions ( Doyon et al . , 2008; Meng et al . , 2008; Gupta et al . , 2011 ) , to generate zebrafish containing a mutation in the single zebrafish dopamine beta-hydroxylase ( dbh ) gene ( Chen et al . , 2013a ) , which encodes the enzyme that converts dopamine to NE . The dbh mutant allele that we generated contains a four nucleotide insertion in the third exon , which gives rise to a premature stop codon , resulting in a predicted 178 amino acid peptide lacking the DBH active site , instead of the full length 614 amino acid protein ( Figure 2—figure supplement 1A ) . Although the murine dbh knockout ( Thomas et al . , 1995 ) displays no overt developmental defects , only 10% of dbh−/− pups produced by dbh+/− mothers survive embryonic development and no dbh−/− pups are born to dbh−/− mothers . In contrast to the mouse knockout , female dbh−/− zebrafish produced a normal number of dbh−/− larvae when mated to dbh+/− males ( out of 287 larvae genotyped from 3 independent crosses , 49% were dbh+/− and 51% were dbh−/− ) . We verified that dbh−/− larvae completely lack NE using an ELISA assay ( Figure 2—figure supplement 1B ) . Using in situ hybridization ( ISH ) we also found that the level of dbh mRNA in dbh−/− larvae is much lower than in their dbh+/− siblings , presumably due to nonsense mediated decay of the mutant transcript ( Isken and Maquat , 2007 ) ( Figure 2—figure supplement 1C ) . While only 40% of dbh mutant mice that survive embryogenesis reach adulthood ( Thomas et al . , 1995 ) , dbh−/− zebrafish embryos developed to adulthood normally ( 24/96 ( 25% ) of embryos produced by a male dbh+/− crossed to female dbh+/− , and raised to adulthood , were identified as dbh−/− ) . We used the dbh mutant to determine whether genetic ablation of NE affects the sleep/wake patterns of zebrafish larvae . We observed that dbh mutants display lower overall activity ( Figure 2A , D ) and activity when awake ( Figure 2B , E ) during both day and night . Importantly , we found that dbh−/− larvae sleep much more during both day ( +185% ) and night ( +57% ) than sibling controls ( Figure 2C , F ) . During the day , the increase in sleep was due solely to more frequent sleep bouts ( +220% ) , while during the night both longer ( +17% ) and more frequent ( +25% ) sleep bouts occurred ( Figure 2G , H ) . Furthermore , sleep latency at night in the mutants was reduced by 40% ( Figure 2I ) . It is interesting to note that , as expected , treatment of dbh mutants with 100 μM prazosin did not affect sleep/wake behavior compared to DMSO controls ( Figure 2—figure supplement 2 ) , suggesting that prazosin-induced behavioral effects are specific to NE signaling . 10 . 7554/eLife . 07000 . 007Figure 2 . dbh mutant larvae are less active and sleep more than sibling controls . Representative activity ( A ) , waking activity ( B ) and sleep ( C ) traces of dbh+/+ ( blue ) dbh+/− ( cyan ) and dbh−/− ( red ) zebrafish larvae . Bar graphs show mean ± s . e . m . activity ( D ) , waking activity ( E ) , sleep ( F ) , sleep bout number ( G ) , sleep bout length ( H ) and sleep latency ( I ) from seven combined experiments ( n > 250 for each genotype ) . n indicates number of larvae . *** , p < 0 . 0001 and ns , not significant by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 00710 . 7554/eLife . 07000 . 008Figure 2—figure supplement 1 . Verification of zebrafish dbh mutant . ( A ) Alignment of human , mouse , WT zebrafish and mutant zebrafish DBH amino acid sequences . The coding sequence of the mutant allele terminates before the conserved active site residue ( asterisk ) . ( B ) ELISA quantification of NE levels in WT and dbh−/− zebrafish larvae . Mean ± s . e . m . is shown . *** , p < 0 . 0001 by one-way ANOVA . ( C ) RNA in situ hybridization ( ISH ) for dbh in dbh+/− ( C′ ) and dbh−/− ( C′′ ) sibling larvae in the LC , indicated by white arrows . Larvae were obtained from a dbh+/− to dbh−/− mating and were genotyped after imaging . Schematic diagram indicates the brain region ( box ) containing the LC ( green ) that is shown in the ISH images . a = anterior , p = posterior . Scale bar = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 00810 . 7554/eLife . 07000 . 009Figure 2—figure supplement 2 . Treatment of dbh−/− fish with prazosin does not alter sleep/wake behavior . Representative activity ( A ) , waking activity ( B ) and sleep ( C ) traces of vehicle ( blue ) and prazosin ( red ) treated zebrafish dbh−/− larvae . Bar graphs of activity ( D ) , waking activity ( E ) and sleep ( F ) from two combined experiments ( n > 45 for each condition ) . Bars represent mean ± s . e . m . n indicates number of larvae . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 009 Since dbh mutants exhibit increased sleep , we decided to investigate whether their arousal threshold is also altered . To this end we delivered mechano-acoustic tapping stimuli of variable intensities to larvae while monitoring their behavior using a videotracking system . We delivered a stimulus once a minute during the night , and determined the fraction of larvae that responded to each stimulus . This dataset allowed us to construct dose–response curves ( Figure 3A ) and calculate the tapping intensity at which 50% of larvae responded to the stimulus ( effective tap power 50 , ETP50 ) . Surprisingly , the ETP50 for dbh−/− larvae was 2 . 1 compared to 3 . 1 for their dbh+/+ siblings ( 32% decrease ) ( p < 0 . 0001 by extra sum-of-squares F test ) , suggesting that although the dbh mutants sleep more , they have a reduced arousal threshold . In addition to a lower ETP50 , dbh−/− larvae also exhibited an increased maximal response fraction , from 0 . 43 for dbh−/− compared to 0 . 34 for dbh+/+ ( p < 0 . 0001 by extra sum-of-squares F test ) . Interestingly , treatment with the alpha1-adrenergic antagonist prazosin or the alpha2-adrenergic agonist clonidine did not affect the arousal threshold of zebrafish larvae ( data not shown ) . However , larvae treated with the alpha2-adrenergic antagonist bopindolol showed a 54% decrease in their ETP50 from 5 . 7 to 2 . 6 ( Figure 3B , p < 0 . 0001 by extra sum-of-squares F test ) . The maximal response fraction was increased from 0 . 42 to 0 . 53 but this change did not achieve statistical significance ( p = 0 . 08 ) . 10 . 7554/eLife . 07000 . 010Figure 3 . dbh mutants and bopindolol-treated WT larave exhibit lower arousal threshold . Stimulus-response curves generated by a tapping assay for dbh−/− and sibling control larvae ( A ) , and bopindolol and DMSO treated WT larvae ( B ) . Thirty trials were performed at each stimulus intensity , with a 1 min inter-trial interval . Each data point indicates mean ± s . e . m . Stimulus-response curves were constructed using the non-linear variable slope module in Prism and fitted using ordinary least squares . Dashed lines mark the ETP50 values for different genotypes and drug treatments . dbh−/− have an ETP50 value of 2 . 1 vs 3 . 1 for sibling controls ( 32% decrease , p < 0 . 0001 by extra sum-of-squares F test ) ( A ) , and bopindolol treated animals have an ETP50 value of 2 . 6 vs 5 . 7 for sibling controls ( 54% decrease , p < 0 . 0001 by extra sum-of-squares F test ) ( B ) . n indicates number of larvae . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 010 The Hcrt neuropeptide has been shown to promote wakefulness and inhibit sleep in zebrafish larvae ( Prober et al . , 2006 ) and mammals ( reviewed in Alexandre et al . , 2013; Sakurai , 2013 ) , and stimulation of Hcrt neurons has been shown to promote sleep to wake transitions in mice ( Adamantidis et al . , 2007 ) . Hcrt has been proposed to promote wakefulness in part by stimulating the noradrenergic LC based on several lines of evidence . First , the LC is densely innervated by Hcrt neurons in both zebrafish ( Prober et al . , 2006 ) ( Figure 4—figure supplement 1A , B ) and mammals ( Peyron et al . , 1998; Chemelli et al . , 1999; Date et al . , 1999; Horvath et al . , 1999 ) and LC neurons express the Hcrt receptor ( Horvath et al . , 1999; Bourgin et al . , 2000; Prober et al . , 2006 ) . It is worth noting , however , that the medulla oblongata is another source of NE in the brain , and it also receives Hcrt projections in zebrafish ( Figure 4—figure supplement 1C ) and mammals ( Ciriello et al . , 2003; Zhang et al . , 2004 ) . Second , application of Hcrt peptide depolarizes LC neurons in brain slices ( Hagan et al . , 1999 ) and in vivo ( Bourgin et al . , 2000 ) . Third , acute inhibition of LC neurons using halorhodopsin inhibits the arousing effects of stimulating Hcrt neurons ( Carter et al . , 2012 ) . However , despite the evidence that the LC mediates Hcrt-induced arousal , it is unknown whether this process requires NE . The mammalian genome contains two hcrt receptor ( hcrtr ) paralogs , both of which are required for the increase in locomotor activity and decrease in sleep observed after injection of Hcrt peptide ( reviewed in Alexandre et al . , 2013; Sakurai , 2013 ) . To determine whether the single zebrafish hcrtr ortholog is required for the Hcrt overexpression phenotype in zebrafish larvae , we mated Tg ( hsp:Hcrt ) zebrafish to a previously described hcrtr mutant ( Yokogawa et al . , 2007 ) . We subjected the larval progeny to heat shock ( HS ) during the sixth day post fertilization ( dpf ) to induce Hcrt overexpression ( Prober et al . , 2006 ) . We then compared the amount of sleep on night 6 ( post-HS ) to night 5 ( pre-HS ) . We previously showed that Hcrt protein levels remain elevated for over 48 hr after HS , with no detectable decrease during the first 24 hr after HS ( Prober et al . , 2006 ) , so overexpressed Hcrt protein should remain at high levels throughout the night following HS . hcrtr+/− and hcrtr−/− larvae reacted similarly to the HS , with no significant difference in the amount of sleep between the two genotypes ( Figure 4—figure supplement 2A , B ) . Hcrt overexpression decreased the sleep of Tg ( hsp:Hcrt ) ;hcrtr+/− larvae by 50% ( Figure 4—figure supplement 2C , D ) , as previously described for Tg ( hsp:Hcrt ) ;hcrtr+/+ larvae ( Prober et al . , 2006 ) . However , no decrease in sleep was observed in Tg ( hsp:Hcrt ) ;hcrtr−/− larvae following HS ( Figure 4—figure supplement 2C , D ) . These results indicate that the decrease in sleep following overexpression of Hcrt requires a functional hcrtr . To test the hypothesis that NE is required for Hcrt-induced wakefulness , we mated Tg ( hsp:Hcrt ) +/−;dbh+/− fish to dbh−/− fish and heat-shocked the larval progeny during day 6 . We first asked whether the presence or absence of NE affects the response of larvae to HS . Larvae lacking the hsp:Hcrt transgene were similarly affected by the HS independently of whether they produced NE or not ( no significant difference observed between dbh+/− and dbh−/− larvae , Figure 4A , B ) . We therefore conclude that lack of NE does not affect the response of larvae to HS . We next asked whether NE is required for Hcrt overexpression-induced arousal ( Figure 4C , D ) . As expected , the amount of sleep exhibited by Tg ( hsp:Hcrt ) larvae was reduced following Hcrt overexpression ( Figure 4D ) . Specifically , the sleep of Tg ( hsp:Hcrt ) +/−;dbh+/− larvae was reduced to 40% of the pre-HS value ( Figure 4D ) . However , the sleep of Tg ( hsp:Hcrt ) +/−;dbh−/− larvae was reduced to 75% of the pre-HS value ( Figure 4D ) . Thus , the majority of Hcrt-induced sleep loss is blocked in the absence of NE . Importantly , treatment of Tg ( hsp:Hcrt ) +/− animals with prazosin similarly inhibits Hcrt overexpression-induced arousal ( Figure 4—figure supplement 3 ) suggesting that the dbh mutant phenotype is indeed due to the blocking of noradrenergic signaling . We also confirmed that the suppressed effect of Hcrt overexpression on sleep in dbh and hcrtr mutants is not due to an effect of the mutations on the efficacy of the HS overexpression system ( Figure 4—figure supplement 4 ) . These results indicate that NE plays an important role in Hcrt-mediated arousal , but suggest that Hcrt also inhibits sleep via NE-independent mechanisms . 10 . 7554/eLife . 07000 . 011Figure 4 . Reduced sleep at night due to Hcrt overexpression is suppressed in dbh mutants . ( A , C ) Sleep traces from a single representative experiment; yellow region indicates time of heat shock ( HS ) during day 6 . ( B , D ) Bar graphs show mean ± s . e . m . percentage change in sleep during night 6 compared to night 5 from four combined experiments . ( B ) In the absence of the hsp:Hcrt transgene , HS during day 6 has no significant effect on dbh−/− larvae compared to dbh+/− larvae . ( D ) The reduced sleep induced by Hcrt overexpression is significantly diminished in dbh−/− larvae compared to dbh+/− sibling controls . n indicates number of larvae . *** , p < 0 . 0001 and ns , not significant by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01110 . 7554/eLife . 07000 . 012Figure 4—figure supplement 1 . Hcrt neurons project to dbh expressing cells . Hcrt neurons ( red ) project to dbh expressing cells ( green ) in the LC . ( A ) Dorsal-ventral maximum intensity projection , ( B ) lateral maximal intensity projection ( dorsal to the right ) . Boxed regions are shown at higher magnification in ( A′ , A′′ , B′ ) . ( C ) Projections from Hcrt neurons to dbh-expressing neurons in the medulla oblongata ( MO ) . A dorsal-ventral maximum intensity projection is shown . Single optical sections of the boxed regions are shown at higher magnification in ( C′ , C′′ ) . Schematic diagrams indicate brain regions shown in the confocal images . a = anterior , p = posterior , d = dorsal . Scale bars = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01210 . 7554/eLife . 07000 . 013Figure 4—figure supplement 2 . Reduced sleep at night following Hcrt overexpression requires the hcrtr . ( A , C ) Sleep traces from a single representative experiment; yellow region indicates time of HS during day 6 . ( B , D ) Bar graphs show mean ± s . e . m . percentage change in sleep during night 6 compared to night 5 from four combined experiments . ( B ) In the absence of the hsp:Hcrt transgene , HS during day 6 has no significant effect on hcrtr−/− larvae compared to hcrtr+/− larvae . ( D ) The reduced sleep induced by Hcrt overexpression is significantly diminished in hcrtr−/− larvae compared to hcrtr+/− sibling controls . n indicates number of larvae . *** , p < 0 . 0001 and ns , not significant by one-way ANOVA . We note that zebrafish hcrtr−/− larvae have been shown to exhibit normal baseline locomotor activity and sleep levels ( Appelbaum et al . , 2009 ) , similar to mouse hcrt and hcrt receptor 2 mutants ( Willie et al . , 2003; Mochizuki et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01310 . 7554/eLife . 07000 . 014Figure 4—figure supplement 3 . Reduced sleep at night due to Hcrt overexpression is suppressed by prazosin . ( A , C ) Sleep traces from a single representative experiment; yellow region indicates time of HS during day 6 . ( B , D ) Bar graphs show mean ± s . e . m . percentage change in sleep during night 6 compared to night 5 from two combined experiments . ( B ) In the absence of the hsp:Hcrt transgene , HS during day 6 has no significant effect on the Sleep Night 6/Night 5 ratio , whether the animals are treated with prazosin or vehicle . ( D ) The reduced sleep induced by Hcrt overexpression is significantly diminished by prazosin . n indicates number of larvae . ** , p < 0 . 01 and ns , not significant by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01410 . 7554/eLife . 07000 . 015Figure 4—figure supplement 4 . The heat shock promoter is not suppressed in hcrtr or dbh mutants . Representative images of brains from larvae that do not ( A ) or do ( B ) contain the hsp:Hcrt transgene , fixed 30 min after HS , and processed for ISH using a hcrt-specific probe . hcrt overexpression levels are similar for hcrtr+/− ( B ) , hcrtr−/− ( B′ ) , dbh+/− ( B′′ ) and dbh−/− ( B′′′ ) animals . Larvae were generated by crossing homozygous to heterozygous mutant animals , and brains were genotyped after imaging . Dark spots in ( A–A′′′ ) are pigment cells . Endogenous hcrt expression is not observed in ( A–A′′′ ) because ISH development was stopped before endogenous hcrt is detectable . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 015 As an alternative approach to study the effects of activated Hcrt signaling on larval zebrafish behavior , we generated transgenic zebrafish in which the hcrt promoter ( Faraco et al . , 2006 ) drives expression of channelrhodopsin-2 fused to EYFP ( hcrt:ChR2-EYFP ) . Using immunofluorescence , we verified that ChR2-EYFP is exclusively expressed in Hcrt neurons , and that all Hcrt neurons express ChR2-EYFP ( Figure 5—figure supplement 1A ) . To test whether the hcrt:ChR2-EYFP transgene can activate Hcrt neurons , we exposed larvae to blue light for 30 min , and then performed double fluorescent ISH using probes specific for hcrt and c-fos . We observed that 40% of Hcrt neurons in Tg ( hcrt:ChR2-EYFP ) larvae expressed c-fos , compared to only 5% of Hcrt neurons in control Tg ( hcrt:EGFP ) larvae ( Figure 5A , B ) . This result confirms that the light paradigm employed activates Hcrt neurons . 10 . 7554/eLife . 07000 . 016Figure 5 . Optogenetic activation of Hcrt neurons increases locomotor activity . ( A ) Representative images of Hcrt neurons co-labeled with EYFP or EGFP ( green ) and c-fos ( red ) . Scale bar = 10 μm . ( B ) Mean ± s . e . m . percentage of EYFP- or EGFP-expressing neurons that also express c-fos in Tg ( hcrt:EGFP ) and Tg ( hcrt:ChR2-EYFP ) larvae . ( C ) Representative locomotor activity trace during red light exposure . ( D ) Average locomotor activity relative to WT siblings during red light exposure . Data is pooled from 3 experiments and is represented as mean ± s . e . m . ( E ) Representative locomotor activity trace during blue light exposure . ( F ) Average locomotor activity relative to WT siblings during blue light exposure . Data is pooled from 4 experiments and is represented as mean ± s . e . m . Red and blue boxes in ( C ) and ( E ) indicate periods of red and blue light illumination . ‘A’ is the maximum activity reached for larvae of a particular genotype and ‘TA’ is the time taken to reach maximum activity A . n indicates number of larvae . *** , p < 0 . 0001 by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01610 . 7554/eLife . 07000 . 017Figure 5—figure supplement 1 . Specific expression of ChR2-EYFP in Hcrt neurons and temporal dynamics of hcrt:ChR2-EYFP induced locomotor activity . ( A ) Representative image from a Tg ( hcrt:ChR2-EYFP ) larval brain showing Hcrt ( red ) and EYFP ( green ) immunoreactivity , and merged images ( yellow ) . Arrow indicates a neuron that is shown at higher magnification in the inset images . Schematic diagram indicates the brain region shown in the images . a = anterior , p = posterior . Scale bar = 20 μm . ( B ) . Activation of Hcrt neurons using ChR2 changes the dynamics of locomotor activity . Tg ( hcrt:ChR2-EYFP ) larvae reach a higher maximum level of locomotor activity ( B ) , and they do so faster ( C ) , than their non-transgenic siblings . ‘A’ is the maximum activity reached for larvae of a particular genotype and ‘TA’ is the time taken to reach maximum activity , as diagrammed in Figure 5 . Data represent mean ± s . e . m . n indicates number of larvae . * , p < 0 . 05 by one-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 017 To test the behavioral effect of activating Hcrt neurons using ChR2 , we developed a large-scale and non-invasive optogenetic behavioral assay . We modified the locomotor activity assay by adding an array of blue and red LEDs that uniformly illuminates the 96-well plate , thus allowing simultaneous light stimulation of 96 freely behaving larvae . Larvae at 5 dpf were transferred to a 96-well plate and kept in the dark for 8 hr in the behavioral chamber , after which they were exposed to either red or blue light for 30 min . The behavior of Tg ( hcrt:ChR2-EYFP ) larvae was compared to their non-transgenic siblings throughout the night with an inter-trial interval of 3 hr . A typical response to blue or red light included a burst of activity at light onset lasting approximately 30 s , followed by a return to near baseline activity levels , a gradual increase in activity that reached a plateau for the remainder of the illumination phase , and a burst of activity when the lights were turned off . The bursts of activity observed at light onset and offset were similar for transgenic and non-transgenic larvae and were excluded from behavioral analysis . Tg ( hcrt:ChR2-EYFP ) larvae and their non-transgenic siblings exhibited similar levels of locomotor activity when illuminated with red light ( Figure 5C , D ) , a treatment that does not activate ChR2 . In contrast , when illuminated with blue light the total locomotor activity of Tg ( hcrt:ChR2-EYFP ) larvae increased by 46% compared to non-transgenic siblings ( Figure 5E , F ) . Activation of Hcrt neurons by ChR2 also altered the dynamics of locomotor activity . Specifically Tg ( hcrt:ChR2-EYFP ) larvae displayed a 25% increase in the maximum activity level reached during stimulation ( Figure 5—figure supplement 1B ) and reached this maximum activity level 20% faster ( Figure 5—figure supplement 1C ) than sibling controls . We next asked whether the increase in locomotor activity observed upon activation of zebrafish Hcrt neurons using ChR2 requires the hcrtr . We first confirmed that hcrtr−/− larvae in the absence of the hcrt:ChR2-EYFP transgene respond to blue light in a manner similar to their hcrtr+/− and hcrtr+/+ siblings . Indeed , we observed no significant difference in the locomotor activity of larvae of the three genotypes during exposure to blue light ( Figure 6A , B ) . However , Tg ( hcrt:ChR2-EYFP ) ;hcrtr−/− larvae were 25% less active than their Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/− and Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ siblings in response to blue light . This result indicates that a functional hcrtr is important for Hcrt neuron-induced arousal ( Figure 6C , D ) . 10 . 7554/eLife . 07000 . 018Figure 6 . The hcrtr is required for Hcrt neuron-induced increased locomotor activity . ( A , C ) Representative locomotor activity traces for hcrtr−/− and sibling controls without ( A ) and with ( C ) the hcrt:ChR2-EYFP transgene during blue light exposure . ( B , D ) Average locomotor activity relative to sibling controls . Data is pooled from 3 experiments in both cases . Data is represented as mean ± s . e . m . and is plotted as percentage of hcrtr+/+ ( B ) or hcrt:ChR2-EYFP; hcrtr+/+ ( D ) . n indicates number of larvae . *** , p < 0 . 0001 and ns , not significant by one-way ANOVA followed by Tukey's test to correct for multiple comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 01810 . 7554/eLife . 07000 . 019Figure 6—figure supplement 1 . Hcrt signaling is required for Hcrt neuron-induced increased locomotor activity . In this experiment , larvae of all six genotypes were tested in the same behavioral plates . Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ larvae ( blue stripes ) showed a significant increase in locomotor activity compared to hcrtr+/+ larvae ( blue ) . A similar increase was observed for Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/− larvae ( cyan stripes ) compared to hcrtr+/− larvae ( cyan ) , although the difference was smaller than for hcrtr+/+ larvae , suggesting a dosage effect . Tg ( hcrt:ChR2-EYFP ) ;hcrtr−/− larvae ( red stripes ) were also more active than hcrtr−/− larvae ( red ) , although the difference was much smaller than for Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ larvae ( blue stripes ) compared to hcrtr+/+ larvae ( blue ) . Tg ( hcrt:ChR2-EFYP ) ;hcrtr+/+ larvae ( blue stripes ) were also significantly more active than Tg ( hcrt:ChR2-EYFP ) ;hcrtr−/− larvae ( red stripes ) . These results suggest that most , but not all , of the effect of stimulating Hcrt neurons on locomotor activity is due to Hcrt signaling . Data is pooled from 3 experiments . Data is represented as mean ± s . e . m . and is plotted as percentage of hcrtr+/+ . n indicates number of larvae . * , p < 0 . 05; **p < 0 . 01; *** , p < 0 . 0001 by Steel–Dwass test and correction for multiple comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 019 Mammalian and zebrafish Hcrt neurons coexpress other neurotransmitters and neuropeptides , such as glutamate and dynorphin ( Chou et al . , 2001; Rosin et al . , 2003; Appelbaum et al . , 2009; Liu et al . , 2015 ) , which could play a role in locomotor activity that is induced by stimulation of Hcrt neurons . Determining whether this is the case requires directly comparing locomotor activity in hcrtr−/− larvae with and without the hcrt:ChR2-EYFP transgene . Since these larvae were previously tested in separate experiments , we repeated the experiment but included all six genotypes in the same behavioral plates . As expected , blue light stimulation significantly increased locomotor activity for Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ compared to hcrtr+/+ larvae , and for Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/− compared to hcrtr+/− larvae ( Figure 6—figure supplement 1 ) . Tg ( hcrt:ChR2-EYFP ) ;hcrtr−/− were also more active than hcrtr−/− larvae , although the difference was much smaller than for Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ compared to hcrtr+/+ larvae . Furthermore , Tg ( hcrt:ChR2-EYFP ) ;hcrtr+/+ larvae were significantly more active than Tg ( hcrt:ChR2-EYFP ) ;hcrtr−/− larvae . These results suggest that most , but not all , of the effect of stimulating Hcrt neurons on locomotor activity is due to Hcrt signaling , similar to results obtained in mammals ( Adamantidis et al . , 2007 ) . Thus , similar to Hcrt overexpression ( Prober et al . , 2006 ) , optogenetic activation of only ∼8 larval zebrafish Hcrt neurons ( an average of 40% of the 20 Hcrt neurons are c-fos positive upon stimulation with ChR2; Figure 5B ) increases locomotor activity , consistent with a role for Hcrt neurons in promoting arousal ( Lee et al . , 2005; Mileykovskiy et al . , 2005; Adamantidis et al . , 2007; Carter et al . , 2012 ) . This effect appears to be significantly stronger than the phenotype described in mice ( Adamantidis et al . , 2007; Carter et al . , 2012 ) , suggesting that zebrafish may be more sensitive to the arousing effect of Hcrt . These experiments also demonstrate for the first time a large-scale and non-invasive application of optogenetics to manipulate the activity of a small population of neurons deep in the brain of freely behaving animals . This approach may be generally useful for studies of neuronal circuit function in regulating vertebrate behaviors . Based on our observation that larval zebrafish Hcrt neurons innervate the LC ( Figure 4—figure supplement 1 ) and previous studies in mammals and zebrafish ( Peyron et al . , 1998; Chemelli et al . , 1999; Date et al . , 1999; Horvath et al . , 1999; Kaslin et al . , 2004; Prober et al . , 2006 ) , we sought to explore the functional interaction between the two neuronal populations using a combination of optogenetics and calcium imaging . Using a transient injection approach ( see ‘Materials and methods’ ) , we generated Tg ( hcrt:ChR2-EYFP ) and Tg ( hcrt:EGFP ) larvae in which single LC neurons expressed the genetically encoded calcium indicator GCaMP6s . After paralyzing and mounting these larvae in low melt agarose , we illuminated with a 488 nm laser a region of the brain containing the Hcrt neuron soma with 10 short pulses ( 0 . 3 s each over 3 . 2 s total ) , then imaged GCaMP6s fluorescence in the LC for 30 s ( Figure 7A ) . We reasoned that even though we did not stimulate Hcrt neurons and image the LC simultaneously , the short time interval between the final stimulation pulse and the initiation of imaging ( <0 . 1 s ) , and the relatively slow kinetics of GCaMP6s fluorescence changes ( Chen et al . , 2013b ) , would allow us to detect Hcrt neuron-induced effects on LC neuron activity . Indeed , over multiple trials in several animals we observed a significant increase in GCaMP6s fluorescence in Tg ( hcrt:ChR2-EYFP ) larvae , but not in Tg ( hcrt:EGFP ) larvae ( Figure 7B–E , Videos 1 and 2 ) , indicating that stimulation of Hcrt neurons results in activation of the LC in zebrafish larvae . 10 . 7554/eLife . 07000 . 020Figure 7 . Optogenetic stimulation of Hcrt neurons activates LC neurons . ( A ) Schematic representation of areas stimulated ( Hcrt , yellow ) and imaged ( LC , green ) . a = anterior , p = posterior . ( B , C ) Representative images of a LC cell expressing GCaMP6s in a Tg ( hcrt:EGFP ) ( B ) or Tg ( hcrt:ChR2-EYFP ) ( C ) larva before ( B , C ) and immediately after ( B′ , C′ ) stimulation of Hcrt neuron region . ( D ) GCaMP6s fluorescence intensity for a representative LC neuron in a Tg ( hcrt:EGFP ) ( top ) and a Tg ( hcrt:ChR2-EYFP ) ( bottom ) larva . Single-trial ( gray ) and average ( black ) responses are shown . ( E ) Mean ± s . e . m . ∆F/F ( % ) values averaged for all trials . n indicates number of trials for 5 Tg ( hcrt:EGFP ) and 4 Tg ( hcrt:ChR2-EYFP ) larvae . *** , p < 0 . 001 by one-way ANOVA . See Videos 1 and 2 for examples of GCaMP6s imaging . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 02010 . 7554/eLife . 07000 . 021Video 1 . Pulsed illumination of hcrt:EGFP neurons does not affect GCaMP6s fluorescence in LC neurons . Two trials for a representative LC neuron are shown . Frames labeled ‘Stimulation’ indicate periods during which the soma of Hcrt neurons were illuminated by ten 0 . 3 s pulses of 488 nm light , during which time GCaMP6s was not imaged in the LC . See Figure 7 for quantification . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 02110 . 7554/eLife . 07000 . 022Video 2 . Pulsed illumination of hcrt:ChR2-EYFP neurons increases GCaMP6s fluorescence in LC neurons . Two trials for a representative LC neuron are shown . Frames labeled ‘Stimulation’ indicate periods during which the soma of Hcrt neurons were illuminated by ten 0 . 3 s pulses of 488 nm light , during which time GCaMP6s was not imaged in the LC . See Figure 7 for quantification . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 022 Optogenetic studies in rodents have shown that acute inhibition of LC neurons blocks the wake-promoting effects of acute activation of Hcrt neurons ( Carter et al . , 2012 ) . However , the contribution of NE to this phenotype is unclear . To test the hypothesis that NE is required for the arousing effect of Hcrt neuron stimulation , we assayed the behavioral effects of optogenetically activated Hcrt neurons in dbh−/− larvae ( Figure 8 ) . As we previously observed ( Figure 2D , E ) , dbh−/− larvae were significantly less active during the 30 min of baseline recording than sibling controls ( dbh+/+ 31 . 0 ± 4 . 0 , dbh+/− 39 . 7 ± 4 . 1 and dbh−/− 15 . 7 ± 2 . 6 s of activity , p < 0 . 05 by one-way ANOVA ) . dbh−/− larvae were also 25% less responsive to blue light than both dbh+/+ and dbh+/− larvae ( Figure 8A , B ) . However , when Hcrt neurons were activated optogenetically the requirement for NE became more pronounced , as the ChR2-induced increase in locomotor activity was reduced by 76% in Tg ( hcrt:ChR2-EYFP ) ;dbh−/− larvae compared to both Tg ( hcrt:ChR2-EYFP ) ;dbh+/+ and Tg ( hcrt:ChR2-EYFP ) ;dbh+/− siblings ( Figure 8C , D ) . Thus , while it is possible that supranormal optogenetic activation of Hcrt neurons could make NE more important for Hcrt-induced locomotor activity than it is under normal conditions , these results suggest that NE is an important effector of Hcrt-neuron induced arousal . 10 . 7554/eLife . 07000 . 023Figure 8 . dbh is required for Hcrt neuron-induced increased locomotor activity . ( A , C ) Representative locomotor activity traces for dbh−/− and sibling control larvae without ( A ) and with ( C ) the hcrt:ChR2-EYFP transgene during blue light exposure . ( B , D ) Average locomotor activity relative to sibling controls . Data is pooled from 2 ( B ) or 3 ( D ) experiments . Data is represented as mean ± s . e . m . and is plotted as percentage of dbh+/+ ( B ) or hcrt:ChR2-EYFP; dbh+/+ ( D ) . n indicates number of larvae . * , p < 0 . 05 and *** , p < 0 . 0001 by one-way ANOVA followed by Tukey's test to correct for multiple comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 07000 . 023
The first catecholamine to be identified as a neurotransmitter by Ulf von Euler in 1946 , NE has been implicated in many aspects of physiology and behavior ( Berridge and Waterhouse , 2003; Weinshenker and Schroeder , 2007; Sara , 2009 ) . Several lines of evidence indicate that exogenous NE is a potent arousal-promoting agent ( reviewed in Berridge et al . , 2012 ) . In Drosophila , lack of endogenous octopamine , which is considered the invertebrate equivalent of NE , results in increased sleep ( Crocker and Sehgal , 2008 ) . However , the role of endogenous NE in regulating vertebrate sleep remains unclear . The murine dbh knockout ( Thomas et al . , 1995 ) displays no overt developmental defects . However , only 10% of dbh−/− pups produced by dbh+/− mothers survive embryonic development and no dbh−/− pups are born to dbh−/− mothers; all such embryos die by embryonic day 13 . 5 ( Thomas et al . , 1995 ) . Of the dbh−/− pups that survive embryogenesis , only 40% reach adulthood . Although the basis of the embryonic lethality is unclear , it was hypothesized to be due to abnormal heart development or function , as the hearts of mutant embryos display greater heterogeneity in cell size and orientation ( Thomas et al . , 1995 ) . Supplementation of drinking water during gestation with dihydroxyphenylserine ( DOPS ) , which is converted to NE by L-aromatic-amino-acid decarboxylase ( AADC ) , rescues the embryonic lethal phenotype . Following birth , DOPS supplementation is not required for continued survival of dbh−/− pups . Surprisingly , dbh−/− mice that reach adulthood were initially reported as having normal sleep/wake cycles ( Hunsley and Palmiter , 2003 ) , although a later study reported a 20% increase in sleep ( Ouyang et al . , 2004 ) . These contradictory results could arise from differences in methodology , a relatively subtle and thus poorly reproducible sleep phenotype , or the complication that mice lacking dbh also exhibit higher levels of dopamine , the substrate of DBH and a major neurotransmitter in the brain . In this study we sought to clarify the role of endogenous NE in regulating vertebrate sleep using pharmacology and genetics in zebrafish . Sleep can be distinguished from inactivity using electrophysiology or three behavioral criteria ( Campbell and Tobler , 1984; Borbely and Tobler , 1996; Allada and Siegel , 2008 ) . First , sleep primarily occurs during specific periods of the circadian cycle . Second , animals exhibit an increased arousal threshold during sleep , although they can still be aroused by strong stimuli , thus distinguishing sleep from paralysis or coma . Third , sleep is controlled by a homeostatic system , which can be demonstrated as an increased need for sleep following sleep deprivation . Based on these criteria , it has been shown that rest in a variety of organisms , including zebrafish ( Zhdanova et al . , 2001; Prober et al . , 2006; Yokogawa et al . , 2007 ) , meets the behavioral definition of sleep . Several groups have demonstrated behavioral , anatomical , genetic and pharmacological conservation of sleep between zebrafish and mammals , establishing zebrafish as a simple vertebrate model for sleep research ( Zhdanova et al . , 2001; Kaslin et al . , 2004; Faraco et al . , 2006; Prober et al . , 2006; Renier et al . , 2007; Yokogawa et al . , 2007; Rihel et al . , 2010; Gandhi et al . , 2015 ) . To study the function of NE in regulating zebrafish sleep , we generated a predicted null mutation in the single zebrafish dbh ortholog ( Figure 2—figure supplement 1 ) . Unlike mice , dbh−/− zebrafish larvae develop normally yet show reduced activity and strikingly increased sleep ( +185% during the day and +57% during the night ) compared to sibling controls ( Figure 2 ) . Importantly , these phenotypes were also observed in larvae treated with the alpha-adrenergic inhibitor prazosin ( Figure 1 ) , which blocks NE signaling without affecting dopamine levels . This observation suggests that the zebrafish dbh−/− phenotypes are due to lack of NE and not altered dopamine levels . It is important to note that our pharmacological manipulations ( blocking of the activating alpha1-adrenergic receptors with the antagonist prazosin; activation of the inhibitory alpha2-adrenergic receptors with the agonist clonidine; blocking of the activating alpha2-adrenergic receptors with the antagonist bopindolol ) did not fully recapitulate the activity and sleep phenotypes observed in the dbh mutant . This is not surprising considering the inherent limitations of global exposure to a drug and the fact that each receptor type has at least five paralogs in zebrafish that may have different sensitivities to these drugs . Among these drugs , prazosin most faithfully phenocopied the dbh sleep/wake phenotype ( reduced day and night activity and increased day and night sleep ) with differences observed in sleep bout number . Clonidine reduced activity and increased sleep during the day , but had almost no effect at night . Bopindolol increased sleep during the day and night , but only inhibited activity at night . These observations suggest that alpha1-adrenergic receptors could be the main facilitators of noradrenergic modulation of larval zebrafish sleep . Why do dbh mutant mice fail to exhibit the robust sleep phenotypes observed in NE deficient zebrafish ? One possible explanation is that the chronic loss of NE in dbh mutant mice may induce compensatory mechanisms during development that do not occur in response to chronic loss of NE in zebrafish . Another possibility is that DOPS treatment during gestation affects developmental processes that ultimately affect behavior . An intriguing third potential explanation involves the role of NE in the mammalian sympathetic nervous system in promoting brown adipose tissue ( BAT ) thermogenesis ( Cannon and Nedergaard , 2004 ) . BAT thermogenesis has been shown to promote slow wave sleep ( Dewasmes et al . , 2003 ) and to be required for sleep rebound after sleep deprivation ( Szentirmai and Kapas , 2014 ) . Therefore , genetic ablation of dbh could induce two competing effects . In the periphery , lack of NE reduces BAT thermogenesis and thus inhibits sleep . In the central nervous system , lack of NE interferes with the normal arousal function of the LC and thus inhibits wakefulness . The combination of these two opposing forces could result in a minor sleep increase in dbh−/− mice , and the magnitude of this phenotype might be particularly sensitive to experimental conditions . However , zebrafish larvae and flies do not regulate their body temperature and are thus subject only to the sleep promoting effects of NE ablation , resulting in a dramatic increase in sleep . It would be interesting to test whether mice in which dbh is deleted only centrally , and thus have normal BAT thermogenesis in the periphery , display a similarly dramatic sleep increase . To further characterize the dbh phenotype , we used an automated arousal threshold assay using a mechano-acoustic stimulus . We were surprised to find that although dbh−/− larvae sleep more , they display a lower arousal threshold ( i . e . they respond to stimuli of lower intensity ) , as well as a higher maximal response than sibling controls ( Figure 3A ) . Genetic ablation of dbh is predicted to increase dopamine levels in neurons that normally express dbh , since this enzyme converts dopamine to NE . Indeed , dopamine levels are higher in dbh mutant mice compared to sibling controls ( Thomas et al . , 1995 ) . However , pharmacological inhibition of adrenergic receptors should not interfere with the conversion of dopamine to NE . Interestingly , we were able to pharmacologically replicate the reduced arousal phenotype seen in dbh−/− animals by treatment with bopindolol , a beta-adrenergic inhibitor , but not with the alpha1-adrenergic inhibitor prazosin , or the alpha2-adrenergic agonist clonidine . The ability of bopindolol to recapitulate the reduction in arousal threshold seen in the dbh mutant suggests that this aspect of the mutant phenotype is a consequence of the silencing of beta-adrenergic receptor signaling . Furthermore , the inability of prazosin ( which increases sleep during day and night ) or clonidine ( which increases sleep during the day ) to recapitulate the phenotype suggests that the observed reduction in arousal threshold is not simply a consequence of lighter sleep due to ‘over-sleeping’ in the dbh mutant . These observations suggest that lack of beta-adrenergic receptor signaling potentiates responses to sensory stimuli , and are particularly interesting considering that NE reuptake inhibitors such as atomoxetine , which increase NE signaling , are used to treat patients suffering from Attention Deficit/Hyperactivity Disorder ( ADHD ) ( Garnock-Jones and Keating , 2009 ) . It is important to note that these pharmacological results do not preclude a role of excess dopamine release in the dbh mutant by the previously noradrenergic neurons of the LC and/or medulla oblongata in reducing the arousal threshold . This hypothesis is supported by zebrafish studies implicating dopamine in arousal modulation ( Burgess and Granato , 2007; Mu et al . , 2012 ) as well as work demonstrating that dopaminergic stimulation during anesthesia produces a robust arousal response ( Taylor et al . , 2013 ) . Furthermore , dopamine agonists have been shown to increase locomotion in zebrafish larvae ( Rihel et al . , 2010; Irons et al . , 2013 ) and dopamine promotes locomotor development in zebrafish larvae ( Lambert et al . , 2012 ) . The Hcrt neurons of the hypothalamus are a major arousal promoting center . Several lines of evidence suggest that Hcrt promotes arousal , at least in part , via the LC . First , Hcrt neurons send dense projections to the LC in rodents and zebrafish ( Peyron et al . , 1998; Chemelli et al . , 1999; Date et al . , 1999; Horvath et al . , 1999; Kaslin et al . , 2004; Prober et al . , 2006 ) ( Figure 4—figure supplement 1A , B ) . Second , the LC expresses the Hcrt receptor in rodents ( Horvath et al . , 1999; Bourgin et al . , 2000 ) and zebrafish ( Prober et al . , 2006 ) . Third , application of Hcrt peptide depolarizes LC neurons in brain slices ( Hagan et al . , 1999 ) and in vivo ( Bourgin et al . , 2000 ) . Fourth , optogenetic activation of Hcrt neurons induces c-Fos expression in the LC of mice ( Carter et al . , 2012 ) and GCaMP6 activation in the LC of zebrafish larvae ( Figure 7 ) . Fifth , acute inhibition of LC neurons blocks the effect of acute Hcrt neuron activation on sleep to wake transitions ( Carter et al . , 2012 ) . While these studies suggest that the LC plays an important role in mediating Hcrt-induced arousal , it was unknown whether NE , a neurotransmitter in the LC , is required for this process . This question is important because the LC produces other neurotransmitters and neuropeptides that have been shown to affect sleep , including dopamine ( Dzirasa et al . , 2006 ) , neuropeptide Y ( Dyzma et al . , 2010 ) , neurotensin ( Erwin and Radcliffe , 1993 ) and vasopressin ( Born et al . , 1992 ) . To address this question , we used two independent strategies to drive Hcrt-induced arousal . First , we overexpressed Hcrt using a heat-shock inducible promoter , which was previously shown to decrease sleep in zebrafish larvae ( Prober et al . , 2006 ) . We observed that in the absence of NE , sleep reduction was attenuated ( Figure 4 ) . Second , we developed a large-scale and non-invasive optogenetic assay and used it to activate Hcrt neurons in freely behaving larvae ( Figure 5 ) . Similar to Hcrt overexpression , activation of Hcrt neurons increased locomotor activity , consistent with the increased sleep-to-wake transitions previously described in mammals ( Adamantidis et al . , 2007 ) . This observation provides evidence of a causal relationship between the activity of Hcrt neurons and locomotor activity in zebrafish , and together with data from the mouse ( Adamantidis et al . , 2007 ) supports an evolutionarily conserved role of Hcrt neurons in promoting arousal . We found that the arousing effects of Hcrt overexpression ( Figure 4 ) and Hcrt neuron activation ( Figure 8 ) were dramatically reduced in dbh mutant larvae , indicating that NE is an important downstream effector of Hcrt-mediated arousal . Interestingly , lack of NE did not completely abolish arousal induced by Hcrt overexpression ( Figure 4 ) , suggesting that Hcrt also promotes arousal via NE-independent pathways . Whether these pathways employ other molecules produced by the LC or are LC-independent remains to be tested . Zebrafish Hcrt neurons also send projections to noradrenergic cells in the medulla oblongata , similar to mammals ( Ciriello et al . , 2003; Zhang et al . , 2004 ) , suggesting that these cells might also mediate Hcrt-induced arousal . Furthermore , in humans , Hcrt neurons are reciprocally connected to other important arousal centers such as the cholinergic basal forebrain , the histaminergic tuberomammillary nucleus , the dopaminergic ventral tegmental area and the serotonergic dorsal raphe ( Alexandre et al . , 2013 ) . Many of these connections are also present in zebrafish , with Hcrt neurons projecting to dopaminergic , histaminergic and serotonergic populations ( Panula et al . , 2010 ) . Although functional interactions between Hcrt neurons and these populations have yet to be explored in zebrafish , the anatomical and molecular similarities of zebrafish and mammalian brains suggest that at least a portion of Hcrt-induced arousal is likely to be mediated by these centers . In summary , we have demonstrated that endogenous NE is required to maintain normal levels of wakefulness in a diurnal vertebrate . This study , together with previous reports in Drosophila ( Crocker and Sehgal , 2008 ) and mice ( Ouyang et al . , 2004 ) , support a phylogenetically conserved role for endogenous NE in promoting wakefulness . Furthermore , we showed that NE also plays an important role in mediating arousal induced by either overexpression of the Hcrt peptide or activation of Hcrt neurons . Finally , we established and characterized the zebrafish dbh mutant , a useful tool for studying other processes that are thought to involve NE , including the fight-or-flight response ( Colwill and Creton , 2011 ) , congestive heart failure ( Thomas and Marks , 1978 ) , cognitive disorders such as Alzheimer's ( Chalermpalanupap et al . , 2013 ) and Parkinson's diseases ( Rommelfanger et al . , 2007 ) , and neuropsychiatric disorders including depression and ADHD ( Chamberlain and Robbins , 2013 ) .
All experiments followed standard protocols ( Westerfield , 2000 ) in accordance with the California Institute of Technology Institutional Animal Care and Use Committee guidelines ( animal protocol 1580 ) . Sleep/wake analysis was performed as previously described ( Prober et al . , 2006 ) . Larvae were raised on a 14/10 hr light/dark ( LD ) cycle at 28 . 5°C with lights on at 9 am and off at 11 pm . Dim white light was used to raise larvae for optogenetic experiments to prevent stimulation of ChR2 by ambient light . Individual larvae were placed into each well of a 96-well plate ( 7701-1651 , Whatman , Pittsburgh , PA , United States ) containing 650 μl of E3 embryo medium ( 5 mM NaCl , 0 . 17 mM KCl , 0 . 33 mM CaCl2 , 0 . 33 mM MgSO4 , pH 7 . 4 ) . Locomotor activity was monitored using a videotracking system ( Viewpoint Life Sciences , Lyon , France ) with a Dinion one-third inch Monochrome camera ( Dragonfly 2 , Point Grey , Richmond , Canada ) fitted with a variable-focus megapixel lens ( M5018-MP , Computar , Cary , NC , United States ) and infrared filter . The movement of each larva was recorded using the quantization mode . The 96-well plate and camera were housed inside a custom-modified Zebrabox ( Viewpoint Life Sciences ) that was continuously illuminated with infrared lights . The 96-well plate was housed in a chamber filled with recirculating water to maintain a constant temperature of 28 . 5°C . The parameters used for movement detection were: detection threshold , 15; burst , 29; freeze , 3; bin size , 60 s . Data were analyzed using custom Perl and Matlab ( Mathworks , Natick , MA , United States ) scripts ( Source code 1 ) , which conform to the open source definition . For Hcrt overexpression experiments , videotracker analysis was initiated at 4 dpf . During the afternoon of 6 dpf , the 96-well plate was transferred to a 37°C water bath for 1 hr to induce Hcrt overexpression . For each larva , total sleep during night 6 was divided by the average total sleep on night 5 for all larvae of the same genotype and converted to a percentage to compare among different genotypes . Prazosin hydrochloride ( P7791 , Sigma–Aldrich ) , clonidine hydrochloride ( C7897 , Sigma–Aldrich ) and bopindolol malonate ( SC-200144 , Santa Cruz Biotechnology , Dallas , TX , United States ) were dissolved in dimethyl sulfoxide ( DMSO , 4948-02 , Macron Chemicals , Center Valley , PA , United States ) and added to E3 medium for a final concentration of 0 . 1% DMSO and 100 μM prazosin , 5 μM clonidine or 20 μM bopindolol . At these concentrations , we observed robust behavioral phenotypes without apparent toxicity or abnormal responses to a gentle stimulus . Controls were exposed to 0 . 1% DMSO alone . For sleep/wake experiments drugs were added during the evening of the fourth day of development and recording was performed from the beginning of day 5 until the end of night 6 . For arousal experiments , drugs were added in the afternoon of day 5 , and experiments were performed during night 5 ( 12:30 am to 7:30 am ) . dbh ISH was performed using standard protocols ( Thisse and Thisse , 2008 ) and developed using nitro-blue tetrazolium and 5-bromo-4-chloro-3′-indolyphosphate ( 10760978103 , Roche , Mannheim , Germany ) . A fragment of the dbh gene was used as a probe ( Guo et al . , 1999 ) . NE levels were measured using an ELISA assay ( 17-NORHU-E01-RES , ALPCO , Salem , NH , United States ) according to the manufacturer's instructions . Five larvae were analyzed per sample , with four samples analyzed in triplicate for each genotype . The videotracking system was modified by adding an Arduino ( http://www . arduino . cc/ ) based automated driver to control two solenoids ( 28P-I-12 , Guardian Electric , Woodstock , IL , United States ) that delivered a tap to a 96-well plate containing larvae . This setup allowed us to drive the solenoids with voltage ranging from 0 V to 20 V over a range of 4095 settings ( from 0 . 01 to 40 . 95 ) . In our experiments we used taps ranging from a power setting of 1–36 . 31 . Taps of 14 different intensities were applied in a random order from 12:30 am to 7:30 am during the fifth night of development with an inter-trial-interval of 1 min . Previous studies showed that a 15 s interval between repetitive stimuli is sufficient to prevent behavioral habituation ( Burgess and Granato , 2007; Woods et al . , 2014 ) . The background probability of movement was calculated by identifying for each genotype the fraction of larvae that moved 5 s prior to all stimuli delivered during an experiment ( 14 different tap powers x 30 trials per experiment = 420 data points per larva; average background movement ) . This value was subtracted from the average response fraction value for each tap event ( corrected response = average response – average background movement ) . The response of larvae to the stimuli was monitored using the videotracking software and was analyzed using Matlab ( Mathworks ) and Excel ( Microsoft , Redmond , WA , United States ) . Statistical analysis was performed using the Variable Slope log ( dose ) response curve fitting module of Prism ( Graphpad , La Jolla , CA , United States ) . Samples were fixed in 4% paraformaldehyde ( PFA ) in PBS overnight at 4°C and then washed with 0 . 25% Triton X-100/PBS ( PBTx ) . Brains were manually dissected and blocked for at least 1 hr in 2% goat serum/2% DMSO/PBTx at room temperature or overnight at 4°C . Antibody incubations were performed in blocking solution overnight at 4°C . Primary antibodies were rabbit anti-orexin A ( AB3704 , 1:500; Millipore , Temecula , CA , United States ) , rabbit anti-RFP ( 632496 , 1:100 , Clontech , Mountain View , CA , United States ) and chicken anti-GFP ( GFP-1020 , 1:400 , AvesLabs , Tigard , OR , United States ) . Alexa Fluor secondary antibodies were used ( 1:500 for anti-orexin and anti-RFP and 1:600 for anti-GFP , Life Technologies , Carlsbad , CA , United States ) . Samples were mounted in 50% glycerol/PBS and imaged using a Zeiss LSM 780 confocal microscope ( Zeiss , Oberkochen , Germany ) . To exclude the possibility that the HS promoter response is suppressed in hcrtr and dbh mutants , we performed ISH to compare the level of overexpressed hcrt mRNA in hcrtr and dbh homozygous mutants to their heterozygous mutant siblings . Fish heterozygous for the hsp:Hcrt transgene and for the hcrtr or dbh mutation were mated to hcrtr or dbh homozygous mutants , respectively . Larvae from these crosses were heat shocked during the afternoon at 6 dpf to induce hcrt overexpression , and were fixed 30 min after HS in 4% PFA overnight at room temperature . ISH was performed using digoxygenin ( DIG ) labeled antisense riboprobes ( Thisse and Thisse , 2008 ) . Images were acquired using a Zeiss AxioImagerM1 microscope and samples were then genotyped by PCR . Tg ( hcrt:ChR2-EYFP ) larvae were placed in a 96 well plate in the videotracking system as described above , left in the dark for 8 hr , and then exposed to blue light for 30 min starting at 1 am . Larvae were then fixed in 4% PFA overnight at 4°C . Double-fluorescent ISH was performed using DIG- and 2 , 4-dinitrophenol ( DNP ) -labeled riboprobes and the TSA Plus DNP System ( NEL747A001 KT , PerkinElmer , Waltham , MA , United States ) . Probes specific for c-fos and eyfp were used to quantify the number of eyfp-expressing Hcrt neurons that expressed c-fos . Tg ( hcrt:EGFP ) larvae ( Prober et al . , 2006 ) were used as negative controls . Samples were mounted in 50% glycerol/PBS and imaged using a Zeiss 780 LSM confocal microscope . The videotracking system was modified to include a custom array containing three sets of red and blue LEDs ( 627 nm , MR-D0040-10S and 470 nm , MR-B0040-10S , respectively , Luxeon V-star , Brantford , Canada ) mounted 15 cm above and 7 cm away from the center of the 96-well plate to ensure uniform illumination . The LEDs were controlled using a custom built driver and software written in BASIC stamp editor . A power meter ( 1098293 , Laser-check , Santa Clara , CA , United States ) was used before each experiment to verify uniform light intensity ( ∼400 µW at the surface of the 96-well plate ) and similar red and blue light intensities were used . During the afternoon of the fifth day of development , single larvae were placed into each well of a 96-well plate as described above and placed in the videotracker in the dark for 8 hr . Larvae were then exposed to either red or blue light for 30 min , starting at 1 am . Three trials were performed during the night , with an inter-trial interval of 3 hr . Total activity for each larva was monitored for 30 min before and after light onset , with data collected in 10 s bins . Light onset caused a short burst of locomotor activity lasting for ∼30 s for all genotypes , so data obtained during the minute before and after light onset was excluded from analysis . A large burst of locomotor activity was also observed for all genotypes when the lights were turned off after the 30 min illumination period . This data was excluded from analysis and is not shown in the figures . The total amount of locomotor activity of each larva during the 30 min of light exposure , excluding the minute after light onset , was divided by the average baseline locomotor activity for all larvae of the same genotype . The baseline period was defined as 30 min before light onset , excluding the minute before light onset . Data from multiple experiments were pooled and converted to percentage of wild type larvae . A 1 . 1 kb genomic fragment upstream of the dbh gene was amplified using the primers 5′-ACTTGAACCAGCGACCTTCT-3′ and 5′-GGTTTGAAGGCCTTTCTAAGTTTTT-3′ ( Liu et al . , 2015 ) and cloned 5′ to a transgene encoding GCaMP6s ( Chen et al . , 2013b ) in a plasmid containing Tol2 transposase recognition sequences . Either Tg ( hcrt:ChR2-EYFP ) or Tg ( hcrt:EGFP ) embryos were injected at the 1–2 cell stage with a solution containing 50 ng/µl plasmid , 0 . 04% phenol red and 50 ng/µl Tol2 transposase mRNA . This injection procedure resulted in GCaMP6s expression in no more than 1 LC neuron per animal . At 5 dpf , larvae were paralyzed by immersion in 1 mg/ml α-bungarotoxin ( 2133 , Tocris , Bristol , UK ) dissolved in E3 , embedded in 1% low melting agarose ( EC-202 , National Diagnostics , Atlanta , GA , United States ) and imaged using a 20× water immersion objective on a Zeiss LSM 780 confocal microscope . To stimulate Hcrt neurons , a region of interest ( ROI ) that encompassed the Hcrt neuron soma was illuminated using a 488 nm laser at 100% power . Ten pulses lasting 0 . 3 s each were applied over 3 . 2 s using the bleaching function , and then a ROI encompassing a GCaMP6s-expressing LC soma was imaged at 4 Hz for 30 s . The time between the final stimulation pulse and initiation of imaging was less than 0 . 1 s . This cycle of stimulation and imaging was repeated 8 times for each GCaMP6s-expressing soma . A baseline of 60 s was recorded before the first stimulation . Movies were processed using ImageJ ( Schneider et al . , 2012 ) . The mean fluorescence of each LC neuron was measured by drawing a ROI around the soma . Baseline fluorescence ( Fo ) for each trial was defined as 10 frames immediately preceding the stimulation , and 10 frames post stimulation ( F ) were used to measure the total change in fluorescence ( ∆F/Fo = ( F−Fo ) / Fo ) . | Although the neural circuits that regulate sleep and wakefulness have yet to be fully identified , the importance of at least two brain regions is well established . These are the hypothalamus , a structure deep within the brain that controls a number of basic activities including hunger , thirst and sleep; and the brainstem , which connects the brain with the spinal cord . Specific neurons within the hypothalamus and brainstem regulate the sleep–wake cycle by signaling to one another using chemicals called neurotransmitters and neuropeptides . Throughout the day , some hypothalamic neurons release a neuropeptide called hypocretin , which helps maintain wakefulness . Hypocretin acts on neurons within the brainstem and causes them to release other neurotransmitters that promote wakefulness . However , the identity of these molecules is unclear . One candidate is norepinephrine . Drugs that enhance the effects of norepinephrine increase wakefulness , whereas those that block norepinephrine signaling promote sleep . Despite this , mice that have been genetically modified to lack the enzyme that produces norepinephrine exhibit relatively normal sleep . This may be because in mammals , norepinephrine also has important roles outside the brain , thus complicating the effects of this genetic modification on behavior . Alternatively , while zebrafish that lack norepinephrine are healthy , mice containing this modification die early in development . Treating these mice with a specific drug allows them to survive , but might affect their behavior . To clarify the role of norepinephrine and its interaction with hypocretin , Singh , Oikonomou and Prober created a new animal model by genetically modifying zebrafish . In contrast to mice , zebrafish that were unable to make norepinephrine slept more than normal fish , although they were also lighter sleepers and were more prone to being startled . A genetic modification that increases hypocretin signaling induces insomnia; Singh , Oikonomou and Prober found that this occurs only in animals with normal levels of norepinephrine . Thus , these experiments indicate that hypocretin does indeed promote wakefulness though norepinephrine . The work of Singh , Oikonomou and Prober has clarified the role of norepinephrine in regulating the sleep–wake cycle . These findings could help in the development of drugs that target the neurons that make hypocretin , which may improve treatments for sleep disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Norepinephrine is required to promote wakefulness and for hypocretin-induced arousal in zebrafish |
Ubiquitin-dependent proteolysis can initiate at ribosomes for myriad reasons including misfolding of a nascent chain or stalling of the ribosome during translation of mRNA . Clearance of a stalled complex is required to recycle the ribosome for future use . Here we show that the ubiquitin ( Ub ) pathway segregase Cdc48/p97 and its adaptors Ufd1-Npl4 participate in ribosome-associated degradation ( RAD ) by mediating the clearance of ubiquitinated , tRNA-linked nascent peptides from ribosomes . Through characterization of both endogenously-generated and heterologous model substrates for the RAD pathway , we conclude that budding yeast Cdc48 functions downstream of the Ub ligases Ltn1 and Ubr1 to release nascent proteins from the ribosome so that they can be degraded by the proteasome . Defective RAD could contribute to the pathophysiology of human diseases caused by mutations in p97 .
Maintaining protein homeostasis is key for the healthy lifespan of the organism ( Koga et al . , 2011 ) and cells have evolved sophisticated mechanisms to control the quality of newly synthesized proteins ( Bukau et al . , 2006; Smith et al . , 2011 ) . Dysfunction of these quality control ( QC ) pathways can tip the balance towards the accumulation of toxic proteins and manifestation of disease , most notably neurodegeneration ( Powers et al . , 2009; Hartl et al . , 2011; Dennissen et al . , 2012 ) . There are multiple pools of QC substrates in cells . A relatively well-characterized substrate pool arises from misfolding of nascent chains that have been properly translated . Depending upon the relative rate at which malfolded sequences that comprise degrons form and bind their Ub ligase receptors vs the rate at which translation is completed , ubiquitination and degradation of misfolding proteins may initiate on the ribosome . It has been shown , for instance , that the acetylated N-terminus recognized by the Ub ligase Doa10 ( Hwang et al . , 2010 ) is generated co-translationally ( Gautschi et al . , 2003; Polevoda et al . , 2008 ) , and that the Ub ligase Ubr1 can initiate degradation of an N-end rule substrate prior to completion of its translation ( Turner and Varshavsky , 2000 ) . A second pool of QC substrates arises from failure to complete translation successfully . This can occur because the mRNA is defective or the ribosome stalls due to either irresolvable structures in the mRNA or interactions between the nascent peptide and the ribosome ( Lu and Deutsch , 2008; Kramer et al . , 2009; Ingolia et al . , 2011 ) . Defective mRNAs are recognized by a cytosolic mRNA surveillance machinery that mediates their destruction . There are at least three pathways: no-go decay ( NGD ) targets mRNAs with strong stalls in translation elongation leading to endonucleolytic cleavage of the mRNA , nonsense-mediated decay ( NMD ) targets messages containing premature termination codons ( PTC ) , and non-stop decay ( NSD ) targets mRNAs that lack a stop codon . Within this broad scheme , there are many variations by which aberrant mRNAs are generated ( Parker , 2012 ) . Interestingly , recent data suggest an intimate coupling between QC of defective mRNAs and the nascent peptides they produce ( Shoemaker and Green , 2012 ) . For example , Upf1 is a positive effector of NMD that is also needed for proteasome-mediated degradation of the prematurely terminated products ( Takahashi et al . , 2008; Kuroha et al . , 2009 ) . Quality-control of mRNAs that lack a stop codon and the proteins produced from them has been the focus of considerable recent attention . Translation of these ‘non-stop’ mRNAs yields a ribosome-tethered polypeptide that is targeted for ubiquitination and degradation ( Ito-Harashima et al . , 2007; Dimitrova et al . , 2009; Bengtson and Joazeiro , 2010 ) . Notably , the mRNA may be processed by multiple pathways ( e . g . , NSD and NGD ) because absence of a stop codon can cause translation to stall either because the end of the message is encountered or polylysine encoded by the poly ( A ) tail interacts with the negatively charged exit tunnel of the ribosome . Non-stop mRNAs can arise naturally through premature termination or polyadenylation within ORFs , which are estimated to occur for up to 10% of all transcriptional events ( Frischmeyer et al . , 2002 ) . Proteins encoded by these messages could potentially impose a considerable burden on QC pathways . Thus , disposal of proteins produced from non-stop messages is likely to be of general significance to understanding human diseases that result from defects in protein homeostasis . Stalled ribosomes that trigger NGD/NSD are recognized by the Dom34–Hbs1 complex which is structurally related to the termination factors eRF1 and eRF3 but lacks the residues necessary for both stop codon recognition and hydrolysis of peptidyl-tRNA and instead binds to the empty A site in the stalled ribosome ( Doma and Parker , 2006; Becker et al . , 2011; Tsuboi et al . , 2012 ) . Dom34–Hbs1 works in concert with a member of the ABC family of ATPases , known as ABCE1 in mammalian cells ( Pisareva et al . , 2011 ) or Rli1 in budding yeast ( Shoemaker et al . , 2010; Shoemaker and Green , 2011 ) , to disassemble stalled ribosomes for subsequent reuse . Additionally , the translational GTPase Ski7 , which is closely related to Hbs1 ( and thus to eRF3 ) , has been proposed to recognize ribosomes stalled at the 3′ end of non-stop messages ( van Hoof et al . , 2002 ) . However , no Dom34/eRF1-like molecule has been identified to date that can interact with Ski7 specifically so it remains unknown if it can promote recycling of ribosome subunits . Following dissociation of subunits , the mechanism by which the associated nascent peptide is extricated remains poorly understood . What has been shown so far is that the Ub ligase Ltn1 can bind the ribosome and ubiquitinate NSD substrates that contain a polylysine tract ( Bengtson and Joazeiro , 2010 ) . Ltn1 belongs to the RING domain family of E3 Ub ligases that ubiquitinate substrates targeted for degradation by the 26S proteasome ( Deshaies and Joazeiro , 2009; Finley , 2009 ) . However , the mechanism by which the nascent non-stop peptide is released from the ribosome and degraded remains unknown . Nevertheless , the underlying biochemistry is likely to be important for human health , because point mutations in mouse Ltn1 result in neurodegeneration and a null allele is lethal ( Chu et al . , 2009 ) . Budding yeast Cdc48 ( the human protein is known as p97 or valosin-containing protein [VCP] ) is a member of the AAA ( ATPases associated with diverse cellular activities ) family of proteins . Cdc48 has intrinsic Ub-binding affinity that is greatly enhanced by the binding of UBX domain proteins or the Ufd1–Npl4 heterodimer to its N-terminus and Ufd2 to its C-terminus ( Ye , 2006; Meyer et al . , 2012 ) . Cdc48/p97 is widely believed to function as a ‘segregase’ that uses the free energy released upon ATP hydrolysis to remodel ubiquitinated substrate complexes , usually , but not always , as a prelude to their degradation by the 26S proteasome ( Jentsch and Rumpf , 2007 ) . In the work reported here , we show that Cdc48–Ufd1–Npl4 promotes the release of tRNA-linked ubiquitinated nascent chains as well as aberrant non-stop and prematurely terminated polypeptides from the ribosome . Our findings establish a new function for Cdc48/p97 in RAD .
In an effort to systematically elucidate cellular functions of Cdc48 and its adaptors , mutants were subjected to chemical profiling by determining growth in the presence of various drugs , including translation inhibitors . All mutants were insensitive to anisomycin , an inhibitor that blocks the peptidyl transferase reaction on ribosomes ( not shown ) . However , low concentrations of hygromycin B and paromomycin inhibited growth of multiple cdc48 mutants and the adaptor mutants npl4-1 , ufd1 , ubx1Δ , and ubx2Δ ( Figure 1—figure supplement 1A , B and data not shown ) . Hygromycin B and paromomycin belong to the aminoglycoside class of antibiotics that bind close to the decoding center and affect translational fidelity ( Carter et al . , 2000; Kim and Craig , 2005 ) . Deletion of LTN1 leads to hygromycin-sensitivity and since Ltn1 is involved in QC of NSD pathway substrates ( Bengtson and Joazeiro , 2010 ) , we explored the possibility that Cdc48–Ufd1–Npl4 and one or more Ubx proteins also function in this pathway . To determine if Cdc48 contributes to protein degradation at the ribosome in unperturbed , exponentially growing cells , we purified ribosomes using a rapid , one-step affinity method designed to isolate 60S subunits , 80S monosomes and polysomes ( Oeffinger et al . , 2007; Halbeisen and Gerber , 2009; Halbeisen et al . , 2009 ) . Coomassie blue staining of purified ribosomes confirmed that Cdc48 activity was not required for their assembly ( Figure 1 , Figure 1—figure supplement 2A , lane 3 ) , a conclusion that was further validated by sucrose gradient fractionation of ribosomes from wildtype , cdc48-3 , and ufd1-2 whole cell lysates ( Figure 1—figure supplement 2B ) . However , the mutants do appear to have a slightly lower ratio of polysomes:80S monosomes . Immunoblotting for Ub revealed strong accumulation of Ub conjugates on ribosomes affinity-purified from cdc48-3 cells ( Figure 1A ) . To ascertain if the epitope tag on the ribosome subunit was exacerbating the effect of the cdc48-3 mutation , we isolated ribosomes from untagged cells by pelleting through a sucrose cushion ( Halbeisen and Gerber , 2009 ) . Again , Ub conjugates were observed to accumulate on ribosomes in cdc48-3 and npl4-1 mutants ( Figure 1B ) . Because the proteolytic defects of npl4-1 are manifested at the semi-permissive temperature of 30°C ( Bays et al . , 2001 ) , we carried out the experiment in Figure 1B at 30°C . The data indicate that cdc48-3 does not have to be shifted to the fully non-permissive temperature to observe Ub conjugate accumulation at the ribosome . Additionally , conjugate accumulation on cdc48-3 and npl4-1 ribosomes exceeded that observed for ribosomes isolated from cells treated with the proteasome inhibitor MG132 , a result that will be further discussed below . 10 . 7554/eLife . 00308 . 003Figure 1 . Ub conjugates accumulate on ribosomes isolated from cdc48 and npl4-1 mutants . ( A ) Ribosome assembly is unimpaired in cdc48-3 mutant cells and Ub conjugates accumulate on cdc48-3 ribosomes . Ribosomes were immunoprecipitated ( IP ) from cdc48-3 ( Untagged , UT ) , RPL18BTAP ( WT ) , and cdc48-3 RPL18BTAP cells shifted to the non-permissive temperature ( 37°C ) for 90 min . Purified ribosomes were evaluated by Coomassie blue staining ( Figure 1—figure supplement 2A ) and immunoblotting ( IB ) with Ub and anti-Rpl32 antibodies . Anti-Rpl32 also detects Rps2 . ( B ) WT and mutant cells grown at 24°C were shifted to 30°C for two generations and pdr5Δ cells were either mock-treated with DMSO , or 30 μM MG132 for 30 min . Ribosomes were isolated ( input cell lysates , right panel ) by sedimentation through sucrose cushions . Ribosome pellets were evaluated by IB with Ub and Rpl32 antibodies ( left panels ) . ( C ) The ATPase activity of Cdc48 promotes clearance of Ub conjugates from ribosomes . Mutant cdc48-3 cells containing plasmid-borne WT GAL-CDC48His6 or the Q2 mutant were grown in galactose for 2 hr to induce expression of ectopic Cdc48 . Induced cells were shifted to 35°C for 1 hr to inactivate cdc48-3 before being harvested . Ribosomes were isolated by sedimentation through sucrose cushions and input cell lysates and ribosome pellets were evaluated as in ( A ) . ( D ) Cdc48 binds to purified ribosomes . Ribosomes were affinity-purified from untagged or RPL18BTAP-tagged cells and the levels of bound Cdc48 and Ufd1 were evaluated by IB with the respective antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 00310 . 7554/eLife . 00308 . 004Figure 1—figure supplement 1 . Cdc48 pathway mutants are sensitive to hygromycin B ( A ) and paromomycin ( B ) . Serial fivefold dilutions of wildtype and mutant cells were spotted on YPD plates either lacking ( untreated ) or containing the indicated amount of drug and incubated at the indicated temperature for 3 days . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 00410 . 7554/eLife . 00308 . 005Figure 1—figure supplement 2 . Ribosome assembly is unimpaired in Cdc48 pathway mutants . ( A ) SDS-PAGE and Coomassie blue analysis of affinity-purified ribosomes from untagged ( UT , lane 1 ) , WT ( lane 2 ) and cdc48-3 ( lane 3 ) strains used in Figure 1A . ( B ) Absorbance profile ( 254 nm ) of wildtype and mutant lysates fractionated on 10–50% sucrose gradients . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 005 To determine whether the ATPase activity of Cdc48 was required to prevent accumulation of ribosome-associated Ub conjugates , ribosomes were pelleted from cdc48-3 mutant cells expressing plasmid-borne wildtype CDC48 , or a mutant ( Q2 ) deficient in ATPase activity ( Ye et al . , 2003 ) . Whereas CDC48 complemented the cdc48-3 phenotype , high levels of Ub conjugates were detected on ribosomes isolated from cells expressing the cdc48-Q2 mutant ( Figure 1C ) . Consistent with the idea that Cdc48 ATPase functions at the ribosome to mediate release of ubiquitinated nascent peptides , both Cdc48 and Ufd1 were found to cofractionate with affinity-purified ribosomes isolated under stringent conditions ( Figure 1D ) . We next wished to ascertain if the Ub conjugates were derived from nascent polypeptides that were linked to tRNA . Magnetic beads containing bound ribosomes from wildtype or cdc48-3 cells were treated with TEV protease , which cleaves between Rpl18B and the TAP tag . The TEV cleavage was carried out in the presence or absence of puromycin . Puromycin causes premature termination of translation by binding in the A-site of the ribosome and reacting with peptidyl-tRNA in the P-site , yielding a peptide-puromycin conjugate ( Nathans , 1964 ) . Thus , we expected that puromycin would induce the release of peptidyl-tRNA from the ribosome as a peptide-puromycin conjugate . The supernatants were then bound to UBA Sepharose resin to enrich for Ub conjugates . The rationale behind this step was that ribosome-associated Ub conjugates that remained bound to the ribosome would bind poorly to UBA beads because ribosomes do not efficiently penetrate the pores of a Sepharose resin , whereas Ub conjugates released from the ribosome by puromycin would be able to bind more efficiently to UBA beads . Puromycin treatment of ribosomes from cdc48-3 but not wildtype cells released ubiquitinated proteins that were enriched on the UBA resin ( Figure 2A , top panel ) . A similar result was obtained with ribosomes isolated from npl4-1 and ufd1-1 mutants ( Figure 2—figure supplement 1A , B ) . The conjugates observed in Figure 2A were indeed linked to puromycin because they cross-reacted with an antibody specific for puromycin ( Figure 2A , bottom panel ) . The same was true for Ub conjugates released from npl4-1 and ufd1-2 ribosomes ( Figure 2—figure supplement 1C ) . The fact that no Ub immunoreactivity was recovered from mutant ribosomes in the absence of puromycin treatment validated our assumption that the UBA resin largely excluded mega-complexes . This experiment also suggests that at least some of the Ub conjugates that accumulated on cdc48 and npl4 ribosomes ( Figure 1A , B ) must be ubiquitinated nascent proteins linked to tRNA . 10 . 7554/eLife . 00308 . 006Figure 2 . Ubiquitinated nascent peptides linked to tRNA accumulate on ribosomes in cdc48-3 and ufd1-2 mutants . ( A ) Puromycin-dependent binding to UBA resin of Ub conjugates from cdc48-3 ribosomes . Ribosomes were affinity-purified from untagged or wildtype and mutant RPL18BTAP cells grown at 37°C for 90 min and eluted with TEV protease in the presence or absence of puromycin ( puro ) , followed by incubation with UBA resin . Bound fractions were resolved by SDS-PAGE and immunoblotted with antibodies to Ub ( top panel ) or puromycin ( lower panel ) . ( B ) Transfer of Ub conjugates to puromycin was RNAse A-sensitive . Ribosomes were affinity-purified from cdc48-3 ( UT ) or cdc48-3RPL18BTAP cells as described above in the absence ( lanes 1 , 2 and 3 ) or presence of 200 μg/ml RNAse A . Following elution with TEV protease , samples from the tagged cells ( lane 3 ) were treated with 200 μg/ml RNAse A at 30°C for 10 min before incubation of all samples with puromycin and binding to UBA resin . The bound fractions were evaluated as in ( A ) . ( C ) CTAB-precipitable Ub conjugates accumulate on cdc48-3 ribosomes . Ribosomes affinity-purified from the same strains used in panel ( A ) were treated with RNAse A ( or not ) and subjected to precipitation ( pptn ) with CTAB . Precipitates were resolved by SDS-PAGE and immunoblotted with anti-Ub . 48 Refers to cdc48-3 . ( D ) Ubiquitinated newly-synthesized proteins accumulate on ribosomes isolated from cdc48-3 and ufd1-2 mutants . Cells were pulse-labeled for 90 s with 35S methionine and chased with cold methionine and cycloheximide for the indicated times . Ribosomes were affinity-purified , eluted with TEV protease in the presence of puromycin , and loaded onto UBA resin . The bound fraction was evaluated by SDS-PAGE followed by autoradiography . Densitometry indicated that <5% of the high MW material was released from ufd1-2 ribosomes between the 0- and 10-min time points . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 00610 . 7554/eLife . 00308 . 007Figure 2—figure supplement 1 . Cdc48-Ufd1-Npl4 complex is required to clear ubiquitinated nascent peptides from ribosomes . ( A ) SDS-PAGE/Coomassie blue profiles of ribosomes affinity-purified from the indicated strains carrying the RPL18BTAP allele . UT: wildtype cells lacking the RPL18BTAP allele . ( B ) Ribosomes from the indicated strains grown at 30°C for two generations were pelleted through sucrose cushions and analyzed by immunoblotting ( IB ) for anti-Rpl32 ( bottom panel ) . Resuspended ribosomes were treated with puromycin in high salt and then incubated with UBA resin . The bound fraction was evaluated by SDS-PAGE and IB with anti-ub ( top panel ) . ( C ) Ribosomes were affinity-purified from wildtype and mutant strains carrying the RPL18BTAP allele and eluted from the magnetic beads with TEV protease in the presence of puromycin . The eluate was either evaluated directly by IB with anti-TAP ( bottom panel ) or was incubated with UBA beads . The bound material was fractionated by SDS-PAGE and analyzed by IB with anti-puromycin antibody ( top panel ) . ( D ) Left panel: Total cell lysates for [35S]-labeled cells were fractionated by SDS-PAGE and subjected to autoradiography to show incorporation of label during the 90 sec pulse . Right panel: Ribosomes were affinity-purified from lysates depicted on left . Ribosomes eluted from the magnetic beads with TEV protease were analyzed by SDS-PAGE and autoradiography to reveal newly-synthesized proteins contained in the preparations . UT: wildtype cells lacking the RPL18BTAP allele . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 007 To further confirm that the Ub conjugates that accumulated on ribosomes in cdc48 cells were ubiquitinated nascent peptides linked to tRNA we did three additional experiments . First , ribosomes affinity-purified from cdc48-3 were treated with RNAse A or not before incubation with puromycin followed by binding to the UBA resin . Immunoblotting for Ub revealed that both the control and RNAse-treated ribosomes contained Ub conjugates that bound the UBA beads ( Figure 2B , top panel , lanes 2 and 3 ) . However , ribosomes pre-treated with RNAse no longer yielded puromycin-reactive conjugates ( Figure 2B , bottom panel ) . By contrast , pre-treating lysate with RNAse before affinity purification of ribosomes eliminated recovery of both Ub- , and puromycin immunoreactivity ( Figure 2B , lane 4 , top and bottom panels ) . The second experiment that we did to confirm tRNA linkage was to treat affinity-purified ribosomes with CTAB . CTAB is an ionic detergent that disrupts protein structure and precipitates peptidyl-tRNA molecules selectively ( Hobden and Cundliffe , 1978; Mariappan et al . , 2010 ) . In parallel with results obtained previously using puromycin , Ub conjugates were precipitated from cdc48-3 ribosomes in an RNAse-sensitive manner ( Figure 2C ) . As a final test of the idea that the Ub conjugates that accumulated on cdc48-3 ribosomes were attached to nascent chains , wildtype and mutant cells were pulsed with radioactive methionine for 90 s and then chased with cold methionine . Ribosomes were affinity-purified from cells sampled at different times of chase and treated with puromycin to release tRNA-linked nascent chains . This material was then fractionated on UBA resin to enrich for discharged Ub conjugates . Strikingly , ribosomes from both cdc48-3 and ufd1-2 cells contained far higher levels of radiolabeled high MW Ub conjugates than wildtype ribosomes , and a substantial portion of this signal persisted on ribosomes for 10–20 min ( Figure 2D ) , which is considerably longer than the time required to synthesize a protein . Importantly , the low signal on wildtype ribosomes was not due to inefficient incorporation , because the WT inputs for the UBA fractionation step actually contained slightly more total pulse-labeled species ( Figure 2—figure supplement 1D ) . Taken together , the experiments described thus far indicate that Cdc48 and its cofactors Ufd1 and Npl4 prevent accumulation of ubiquitinated , tRNA-linked nascent chains on the ribosome . For simplicity , we hereafter refer to these species as tUNPs ( tRNA-linked , ubiquitinated nascent polypeptides ) . Notably , other Cdc48 cofactors including Ubx2 , Ubx4 , Ubx6 , and Ufd2 did not appear to contribute to clearance of tUNPs from the ribosome ( Figure 2—figure supplement 1C ) . These cofactors were tested because they had scored positively in the hygromycin-sensitivity assay whereas ubx3Δ , ubx5Δ , and ubx7Δ were hygromycin-insensitive , and therefore not pursued further . Hygromycin-sensitive ubx1Δ/shp1Δ was also not pursued further , because this mutant is extremely sick . The findings summarized above raised the question: what is the composition of the tUNPs ? One possibility is a class of aberrant proteins arising from translation of mRNAs that lack stop codons ( ‘non-stop mRNA’ ) . Translation through the poly ( A ) tail of these mRNAs can result in a ribosome that stalls either because it reaches the end of the message , or the polylysine encoded by the poly ( A ) tail interacts with the negatively charged exit tunnel of the ribosome ( Dimitrova et al . , 2009 ) . Prior work has demonstrated that the Ub ligase Ltn1 targets proteins translated from mRNAs that lack a stop codon or that contain polylysine stretches ( Bengtson and Joazeiro , 2010 ) . We therefore employed the puromycin release/UBA bead capture protocol to test whether deletion of LTN1 diminished the accumulation of tUNPs on ribosomes in cdc48-3 cells . We also tested the impact of ubr1Δ , because of prior reports that Ubr1 promotes co-translational ubiquitination ( Turner and Varshavsky , 2000 ) , and because ubr1Δ mutants , like ltn1Δ , are hygromycin-sensitive ( not shown , but see below ) . The data in Figure 3A demonstrate that additional loss of LTN1 or UBR1 resulted in a marked decrease in the level of tUNPs on cdc48-3 ribosomes . This result encouraged us to evaluate the steady state levels of the Gfp-FLAG-His3Non-Stop ( GFHNS ) fusion protein shown earlier to be an Ltn1 substrate at the ribosome that is degraded by the proteasome ( Bengtson and Joazeiro , 2010 ) . Amongst the Cdc48 pathway mutants examined , GFHNS accumulated maximally in ufd1 mutants ( similar to levels observed for ltn1Δ ) , and also exhibited significant accumulation in npl4-1 and cdc48-3 ( Figure 3B ) . Cycloheximide chase analysis in ufd1-2 cells indicated that accumulation resulted from stabilization of the non-stop reporter ( Figure 3C ) . As observed in earlier studies ( Bengtson and Joazeiro , 2010 ) , the same reporter containing a stop codon was not degraded and accumulated at a much higher level than the NS reporter ( Figure 3C; amounts loaded were adjusted to yield signals of equal intensity at 0 min ) . 10 . 7554/eLife . 00308 . 008Figure 3 . Aberrant proteins derived from non-stop mRNA accumulate in cells deficient in Cdc48–Ufd1–Npl4 function . ( A ) Ltn1 and Ubr1 contribute to accumulation of Ub conjugates on cdc48-3 ribosomes . Ribosomes from the indicated strains grown at 30°C were isolated from input cell lysates ( right panels ) by pelleting through sucrose cushions , treated with puromycin , and incubated with UBA resin . The bound fraction ( left panels ) and inputs were evaluated by SDS-PAGE and immunoblotting with Ub , tubulin ( Tub ) , and Rpl32 antibodies as indicated . The -3 allele of cdc48 was used . ( B ) The non-stop reporter GFHNS accumulates in Cdc48 pathway mutants . Glass bead/SDS extracts from exponential cultures ( grown at 30°C ) of the indicated mutants harboring a plasmid that expresses GFHNS were analyzed by SDS-PAGE and IB with anti-Gfp . Tubulin served as the loading control . ( C ) Cycloheximide chase analysis of cells expressing either the non-stop ( GFHNS ) or stop codon-containing ( GFHStop ) reporters . Glass bead/SDS extracts prepared from aliquots harvested at the indicated times from wildtype and mutant cultures were analyzed by SDS-PAGE and IB with anti-Gfp . Samples from WT expressing GFHNS were also evaluated by IB with anti-tubulin to confirm equal loading . Note that extracts prepared from cells expressing GFHStop were loaded at one-fifth the amount of GFHNS . ( D ) Non-stop protein accumulates in the ubiquitinated state in ufd1 mutants . Glass bead/SDS extracts of wildtype and mutant cells expressing plasmid-borne GFHNS and grown at 37°C for 90 min were evaluated directly ( inputs ) or immunoprecipitated with anti-Flag antibodies after 10-fold dilution with buffer containing Triton X-100 . Bound proteins and inputs were evaluated by SDS-PAGE and IB with anti-Ub and Flag antibodies . UT corresponds to WT cells not expressing GFHNS . ( E ) Cdc48 and Ufd1 interact selectively with non-stop protein . Lysates prepared from cells expressing Flag-tagged GFHNS ( NS ) or GFHStop ( Stop ) reporters as well as cells lacking a plasmid ( UT ) were immunoprecipitated with anti-Flag antibodies . Total cell extracts ( inputs; right panels ) and bound proteins ( left panels ) were evaluated by SDS-PAGE and IB with anti-Cdc48 , Ufd1 , Flag and Ub antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 00810 . 7554/eLife . 00308 . 009Figure 3—figure supplement 1 . Ribosomes from wildtype and mutant cells grown at 37°C for 90 min were sedimented through sucrose cushions . Pellets were resuspended and either evaluated by SDS-PAGE and immunoblotting ( IB ) with anti-Rpl32 ( bottom panel ) or incubated with the indicated amount of RNAse for 10 min at 30°C . Aliquots of the RNAse reactions were evaluated by separation on NuPAGE gels and IB with anti-Gfp ( top panel ) . Species B and C are RNAse-insensitive species whose physical nature is unknown . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 009 If the Cdc48–Ufd1–Npl4 complex functions downstream of Ltn1 to mediate clearance of tUNPs from the ribosome , as suggested by the data in Figure 3A , we anticipated that ubiquitinated species of the non-stop reporter should accumulate upon inactivation of Cdc48 complex . To address this question , we prepared SDS lysates of wildtype and ufd1 mutant cells expressing GFHNS , diluted the lysates with Triton X-100 to sequester the SDS in micelles , and immunoprecipitated with anti-FLAG . Immunoblotting for Ub demonstrated that GFHNS accumulated in an ubiquitinated form in ufd1-1 and ufd1-2 but not in ltn1Δ mutants ( Figure 3D ) . The above data indicate that the non-stop , but not the stop codon-containing reporter was targeted for degradation by the combined actions of Ltn1 and Cdc48–Ufd1–Npl4 . The role of the Cdc48 complex in reporter degradation appeared to be direct , because both Cdc48 and Ufd1 were co-immunoprecipitated with GFHNS , but not GFHStop , even though GFHStop was more abundant ( Figure 3E ) . GFHNS accumulates on 80S ribosomes in ltn1Δ mutants ( Bengtson and Joazeiro , 2010 ) . To evaluate if any of these molecules are linked to tRNA , we sedimented ribosomes rapidly through sucrose cushions and treated the pellet with RNAse A ( or not ) . Strikingly , immunoblotting revealed an RNAse-sensitive species on ribosomes from ltn1Δ , as well as those from cdc48-3 and ufd1-2 mutants , but not from wildtype cells ( Figure 3—figure supplement 1 ) . The fraction of GFHNS linked to tRNA was low . It is not clear whether this was due to hydrolysis of the tRNA linkage in vivo or in vitro . For our subsequent experiments , we decided to focus on Protein A-NS ( PrANS ) , a non-stop reporter utilized in earlier studies ( Wilson et al . , 2007; Bengtson and Joazeiro , 2010 ) . Unlike GFHNS , the RNAse-sensitive , tRNA-linked form of PrANS accumulated more efficiently and could be detected even in total cell extracts . We first examined the steady-state levels of PrANS in several mutants . Interestingly , levels of tRNA-linked PrANS in cdc48-3 and ltn1Δ cell lysates were equivalent to those detected in dom34Δ and ski7Δ , bona fide components of the NSD pathway ( Figure 4A; equivalent loading confirmed by staining with Ponceau S and immunoblotting for tubulin , Figure 4—figure supplement 1A ) . 10 . 7554/eLife . 00308 . 010Figure 4 . Non-stop reporter PrANS accumulates on 60S and 80S ribosomes in a tRNA-linked form in cdc48-3 cells . ( A ) NuPAGE gel analysis reveals accumulation of tRNA-linked PrANS in cdc48-3 and ltn1Δ cells . Wildtype and mutant cells containing reporter plasmid were grown at 30°C for one generation and expression of reporter was induced ( or not ) by the addition of 2% galactose ( ±Gal ) for 2 hr . Glass bead/SDS extracts were fractionated on a NuPAGE gel to preserve tRNA-linked species , and analyzed by IB with PAP to detect protein A . Equal loading was confirmed by Ponceau S staining and IB for tubulin ( Figure 4—figure supplement 1A ) . The right panel shows collapse of the tRNA-linked species in cdc48-3 extract following RNAse treatment . ( B ) tRNA-linked PrANS accumulates on 60S and 80S ribosomes in ltn1Δ and cdc48-3 mutants . Sucrose gradient ( 10–30% ) fractions from cells expressing plasmid-borne PrANS and grown at 30°C for two generations were concentrated by TCA precipitation and evaluated by fractionation on a NuPAGE gel and IB with anti-PAP or anti-Rpl32 ( Figure 4—figure supplement 1B ) . Fractions 15 and 16 ( enriched for 60S ) and 17–19 ( enriched for 80S ribosomes ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 01010 . 7554/eLife . 00308 . 011Figure 4—figure supplement 1 . Fractionation of ribosome subunits on sucrose gradients . ( A ) Ponceau S staining ( top panel ) and anti-tubulin IB ( bottom panel ) of nitrocellulose filter used for Figure 4A . The lanes in Figure 4A with no signal contain lysate from uninduced cells . ( B ) Lysates from wildtype cells containing the PrANS reporter were fractionated on 10–30% sucrose gradients in parallel with lysates from ltn1Δ and cdc48-3 grown at 30°C for two generations ( Figure 4B ) . 0 . 2 ml fractions were collected and aliquots were resolved by SDS-PAGE followed by IB with anti-Rpl32 . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 011 To more precisely determine which subset of ribosomes contained tRNA-linked PrANS , we performed sucrose gradient fractionation . Shallow gradients ( 10–30% ) were set up , enabling good resolution of the subunits , monosomes , and polysomes ( Figure 4—figure supplement 1B ) . Immunoblotting for PrANS revealed that the RNAse-sensitive species peaked in the 80S fractions isolated from ltn1Δ mutants but was found in both the 60S and 80S-containing fractions from cdc48-3 cells ( Figure 4B ) . Additionally , modified forms of slower mobility than the tRNA-linked species were detected in 60S ribosomes from cdc48-3 but not ltn1Δ , suggesting that these high MW species are Ltn1-dependent Ub conjugates . Only a small fraction of the PrANS detected on ribosomes migrated as Ub-conjugated species . This is consistent with what is typically observed for most Ub-dependent substrates upon inhibition of their degradation , and presumably reflects robust cytosolic deubiquitination activity . In addition , overproduction of PrANS from the GAL promoter may overwhelm Ltn1 , which is inabundant ( Bengtson and Joazeiro , 2010 ) . Besides the modified forms of PrANS , unmodified protein was observed throughout the gradient and peaked in the ribosomal fractions . We do not know whether this latter form arose from hydrolysis of the peptidyl-tRNA linkage in cells or in vitro . Our data thus far indicate that in cdc48 , npl4 , and ufd1 mutants , nascent proteins were targeted for degradation by Ltn1- and Ubr1-dependent ubiquitination but became stuck on 60S and 80S ribosomes as peptidyl-tRNAs , implying that Cdc48 activity promotes release of stalled , ubiquitinated peptides , which are then degraded by the proteasome . This hypothesis predicts that ribosomes isolated from cells in which proteasome activity is compromised should accumulate less tUNPs than cdc48 mutants . To test this prediction , we used the sucrose cushion method to isolate ribosomes from cdc48-3 and npl4-1 mutants , as well as pdr5Δ cells treated with or without the proteasome inhibitor MG132 ( pdr5Δ enables cellular accumulation of this compound ) . The ribosome pellet was then treated with puromycin and incubated with UBA resin to enrich for tUNPs . Ribosomes isolated from cells with inhibited proteasomes contained far less tUNPs than ribosomes from cdc48-3 and npl4-1 cells ( Figure 5A; the input lysates and ribosome pellets for this experiment are shown in Figure 1B ) . This result is in striking contrast to what we previously reported for proteasomes isolated from cdc48-3 and MG132-treated cells—both contained very high levels of accumulated Ub conjugates ( Verma et al . , 2011 ) . A similar result was obtained using the cim3-1 mutant instead of MG132 treatment ( Figure 5—figure supplement 1A; cim3-1 is a ts allele of the gene that encodes the proteasome subunit Rpt6 ) . 10 . 7554/eLife . 00308 . 012Figure 5 . Non-stop reporter PrANS accumulates on ribosomes in cdc48-3 cells but is released from ribosomes in cells deficient in proteasome activity . ( A ) Ub conjugates accumulate to a greater extent on ribosomes isolated from cdc48-3 and npl4-1 mutants than on ribosomes isolated from cells treated with proteasome inhibitor MG132 . Puromycin-treated ribosomes from Figure 1B were incubated with UBA resin . The bound fraction was evaluated by SDS-PAGE and immunoblotting with anti-Ub . ( B ) PrANS preferentially accumulates on the ribosome in cdc48-3 cells and in the post-ribosome supernatant in a proteasome mutant . Ribosomes were isolated from cells grown at 37°C for 90 min by pelleting through sucrose cushions . Total cell lysate inputs ( left panel ) , ribosome pellets ( middle panel ) and post-ribosome supernatants ( right panel ) were evaluated by separation on a NuPAGE gel and IB with PAP . Equivalent recovery of ribosomes in the pellet fractions was confirmed by IB with anti-Rpl32 ( bottom panel ) . The same cell cultures were also evaluated by preparing extracts in SDS and resolving aliquots on a Tris-Glycine ( Figure 5—figure supplement 1B , left panel ) or NuPAGE ( Figure 5—figure supplement 1B , right panel ) gel . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 01210 . 7554/eLife . 00308 . 013Figure 5—figure supplement 1 . Maximal Ub conjugate accumulation on ribosomes isolated from cdc48-3 mutant cells . ( A ) Ribosomes from wildtype and mutant cells grown at 37°C for 90 min were sedimented through sucrose cushions . Pellets were resuspended and either evaluated by SDS-PAGE and IB with anti-Rpl32 ( lower left panel ) or incubated with UBA resin following treatment with puromycin . The bound fraction ( upper left panel ) and the input cell lysates ( right panel ) were evaluated by SDS-PAGE and IB with anti-ub ( top panel ) . ( B ) Extracts prepared by the glass bead/SDS method from the indicated strains grown at 37°C for 90 min were analyzed on a Tris-Glycine ( TG; left panel ) or NuPAGE ( right panel ) gel . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 013 We next sought to evaluate the effect of proteasome deficiency on accumulation of a specific peptidyl-tRNA on the ribosome . To evaluate the subcellular distribution of the PrANS species that accumulate upon inhibition of Cdc48 or proteasome function , we sedimented extracts through sucrose cushions to generate ribosome pellets and post-ribosomal supernatants from wildtype , cdc48-3 and cim3-1 cells ( Figure 5B ) . Notably , whereas most of the PrANS fractionated in the ribosome pellet in cdc48-3 cells ( a substantial portion of which remained linked to tRNA ) , the opposite was true in cim3-1 cells , where the bulk of the PrANS was recovered in the supernatant fraction ( Figure 5B ) . Comparison of extracts made by lysing cells directly in SDS ( Figure 5—figure supplement 1B ) or under native conditions ( Figure 5B ) revealed that the proportion of PrANS that accumulated as full length or truncated species in vivo and its susceptibility to proteolytic clipping in native cell extract differed for cdc48-3 and cim3-1 mutants . We do not know the basis for these differences . Our observation that Ubr1 contributed to ubiquitination of tUNPs ( Figure 3A ) led us to wonder about the diversity of the QC mechanisms that underlie their formation . To determine if the role of Cdc48 is limited to the non-stop pathway or whether it plays a more global role in degradation of proteins encoded by aberrant messages , we also assessed the stability of endogenous phosphogluconate dehydrogenase truncated by a premature termination codon ( Gnd1PTC ) in the middle of a folded domain at residue 368 . Over-produced Gnd1PTC was shown to be a substrate of the Ubr1 cytosolic quality control pathway in a prior study ( Heck et al . , 2010 ) . The data in Figure 6A confirmed stabilization of Gnd1PTC in ubr1Δ . Notably , cdc48-3 and ufd1-2 mutations also stabilized Gnd1PTC , but ltn1Δ was without effect ( equivalent loading confirmed in Figure 6—figure supplement 1 ) . Gnd1PTC was also stabilized in upf1Δ . Upf1 is a RING domain Ub ligase ( Takahashi et al . , 2008 ) that plays a crucial role in NMD and stimulates the degradation of aberrant PTC translation products ( Kuroha et al . , 2009 ) . The dependence of Gnd1PTC degradation on Cdc48 and Upf1 suggested that tRNA-linked Gnd1PTC might become stuck on ribosomes , as observed for the non-stop reporters . To test this possibility we used the sucrose cushion method to isolate ribosomes from various mutants , and evaluated the ribosomal pellets for their content of Gnd1PTC . Analysis of total cell lysates revealed strong accumulation of Gnd1PTC in cdc48-3 , ubr1Δ , and upf1Δ cells ( Figure 6B ) . Fractionation of lysates prepared in an identical manner revealed an RNAse-sensitive species on ribosomes but not in post-ribosomal supernatants isolated from cdc48-3 mutants ( Figure 6C ) as well as ufd1-2 , ubr1Δ and upf1Δ mutants ( data not shown ) . In contrast , little tRNA-linked substrate was found in dom34Δ mutants ( Figure 6C ) . Strikingly , the tRNA-linked species was not detected on ribosomes from ltn1Δ ( data not shown ) . 10 . 7554/eLife . 00308 . 014Figure 6 . The Cdc48 pathway is required for degradation of prematurely terminated Gnd1PTC . ( A ) A cycloheximide chase was performed with the indicated mutants grown at 30°C for two generations . Glass bead/SDS extracts were prepared at the indicated times after adding cycloheximide and analyzed by immunoblotting with anti-HA antibody to detect HA-tagged Gnd1PTC . ( B ) A tRNA-linked form of Gnd1PTC accumulates on ribosomes from cdc48-3 mutants . Total cell extracts prepared identically to those used for the ribosome analysis in panel C were immunoblotted with anti-HA to detect Gnd1PTC and anti-tubulin to evaluate loading . cdc48-3 mutant cells were incubated at a semi-permissive ( 30°C ) temperature for 90 min prior to lysis whereas all other strains were grown at 30°C for two generations . ( C ) Ribosomes ( top panels ) and post-ribosomal supernatants ( bottom panel ) were isolated from the same strains in ( B ) by sedimentation through sucrose cushions and analyzed by immunoblotting with anti-HA to detect Gnd1PTC . The ribosome pellets were also immunoblotted with anti-Rpl32 ( middle panel ) to confirm equivalent recovery . The right panel shows collapse of the tRNA-linked species following RNAse treatment of the ribosomal pellet from cdc48-3 . cdc48-3 cells were incubated at either the semi-permissive ( 30°C ) or non-permissive ( 37°C ) temperature prior to lysis , as described in ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 01410 . 7554/eLife . 00308 . 015Figure 6—figure supplement 1 . Ponceau S staining of filters for Figure 6A . Ponceau S staining of filters for Figure 6A . A tubulin loading control is shown for the top filter . DOI: http://dx . doi . org/10 . 7554/eLife . 00308 . 015
Whereas QC of defective mRNAs and recycling of ribosomes stalled on these templates has received considerable attention in prior studies , much less is known about what happens to the nascent polypeptides , some of which accumulate as peptidyl-tRNA . Some proteins encoded by non-stop mRNAs , including the reporter proteins used here ( GFHNS and PrANS ) , are degraded by the ubiquitin-proteasome system ( UPS ) ( Wilson et al . , 2007; Bengtson and Joazeiro , 2010 ) . The Ub ligase Ltn1 binds 60S ribosomes and promotes ubiquitination of stalled GFHNS ( Bengtson and Joazeiro , 2010 ) . Our data suggest that Cdc48–Ufd1–Npl4 acts downstream of Ltn1 to promote release of the non-stop peptide from the ribosome , so that it can be degraded by the proteasome . How Ltn1 and the Cdc48 complex does this remains unknown . Indeed , this remains unknown for any NSD or NGD pathway substrate; the Dom34-Hbs1 complex lacks the residues necessary for hydrolysis of the peptidyl-tRNA bond , and the two budding yeast peptidyl-tRNA hydrolases , Pth1 and Pth2 , are non-essential mitochondrial proteins whose transmembrane configurations are currently unknown ( Rosas-Sandoval et al . , 2002 ) . An interesting question for the future will be to determine whether Ltn1 and Cdc48–Ufd1–Npl4 act together with , or in a redundant pathway parallel to Ski7 or Dom34–Hbs1-Rli1 . We observed direct physical association of Cdc48 and Ufd1 with both affinity-purified ribosomes and the NSD pathway substrate GFHNS . This interaction was highly specific , because Cdc48 and Ufd1 did not bind the same substrate when it was translated from a message that includes a stop codon . While this manuscript was in revision , it was reported that Cdc48 , Ufd1 , and Npl4 are components of the newly identified Ribosome Quality Control ( RQC ) complex , which also contains Ltn1 , Rqc1 and Tae2 and associates with the 60S ribosomal subunit . The latter two proteins , like Ltn1 and Cdc48–Ufd1–Npl4 , are highly conserved and have mammalian homologs ( Brandman et al . , 2012 ) . Both Ltn1 and Rqc1 help link Cdc48–Ufd1–Npl4 to RQC . Rqc1 associates with and is required for the degradation of an Ltn1 substrate that contains an internal polyarginine sequence , but a functional role for Cdc48–Ufd1–Npl4 in RQC was not established by Brandman et al . ( 2012 ) . A key question that emerges from this work is , what is the signal that differentiates ribosomes engaged with NSD or other aberrant substrates from normally translating ribosomes ? Ltn1 ( Bengtson and Joazeiro , 2010 ) and RQC ( Brandman et al . , 2012 ) are reported to be predominantly bound to 60S ribosomal subunits , suggesting that they may act downstream of the initial recognition and splitting of a stalled ribosome . However , in ltn1Δ cells , GFHNS accumulates on 80S monosomes ( Bengtson and Joazeiro , 2010 ) . Likewise , we report here that tRNA-linked PrANS appeared to be concentrated in monosomes in ltn1Δ . By contrast , tRNA-linked PrANS was spread more broadly across 60S and 80S fractions in extracts from cdc48-3 cells . The different fractionation behavior of tRNA-linked PrANS in ltn1Δ and cdc48 mutants may reflect leakiness of the cdc48 mutation ( the experiment was done at the semi-permissive temperature of 30°C ) or Cdc48-independent ribosome splitting . Detailed resolution of the exact sequence of events will require reconstitution of the overall process with defined components . In addition to observing accumulation of engineered NSD pathway reporter proteins in Cdc48 pathway mutants , we also observed extensive accumulation of endogenous , tRNA-linked ubiquitinated nascent polypeptides ( tUNPs ) on ribosomes in these cells . These tUNPs could potentially arise from multiple sources including misfolded domains ( Turner and Varshavsky , 2000 ) , mRNAs whose transcription was prematurely terminated , mRNAs or nascent peptides containing sequences that cause ribosome pausing ( Ingolia et al . , 2011 ) or mRNAs that encode premature termination codons ( e . g . , Figure 6C ) . Thus it is perhaps not surprising that at least two Ub ligases ( Ltn1 and Ubr1 ) contributed to their accumulation . In any event , we believe it is unlikely that there is a single pathway or mechanism for disposing of tUNPs . Even in the seemingly simple case of a ribosome that translates into the poly ( A ) tail of an mRNA that is prematurely polyadenylated within the coding sequence , the 3′-most ribosome may be stalled because it reached the end of the message , or because polylysine encoded by the poly ( A ) tail bound the exit channel of the ribosome ( Ito-Harashima et al . , 2007 ) . Because stalls in translation elongation lead to endonucleolytic cleavage of the mRNA upstream of the 3′-most ribosome ( Tsuboi et al . , 2012 ) , each trailing ribosome will present a unique substrate to the UPS , and may be processed by a different mechanism . The intimacy with which protein biosynthesis is coupled to protein degradation is highlighted by the identification of the translasome , a supercomplex defined by mass spectrometric analysis of the translation initiation factor eIF3 and found to contain both ribosome and proteasome subunits as well as chaperones involved in protein QC ( Sha et al . , 2009 ) . A tight coupling of these processes may be particularly important for neurons , which are post-mitotic and need to remain robust for the life of the organism ( Segref and Hoppe , 2009 ) . Indeed , homozygous lister mice deficient in Listerin ( Ltn1 ) function exhibit severe neurological dysfunction ( Chu et al . , 2009 ) , whereas mutations in human Ubr1 result in the Johanson–Blizzard syndrome characterized by , amongst other phenotypes , frequent mental retardation ( Hwang et al . , 2011 ) . Mutations in p97 contribute to 1–2% of familial ALS ( Johnson et al . , 2010 ) and are the root cause of a syndrome , IBMPFD , that includes frontotemporal dementia ( Watts et al . , 2004 ) . Defects in metabolism of protein aggregates ( Manno et al . , 2010 ) , endosomal trafficking ( Ritz et al . , 2011 ) and autophagy ( Ju et al . , 2009; Tresse et al . , 2010 ) have been previously suggested to contribute to the pathophysiology that results from mutation of p97 . As the molecular mechanisms underlying these diseases are unraveled , it will be interesting to see if ribosome-associated degradation contributes to their etiology .
All strains used in this study are listed in Supplementary file 1 . They were derived from the W303 or S288C backgrounds . Deletion strains were acquired from Open Biosystems ( Waltham , MA ) . All crosses were confirmed by auxotrophic marker selection and genomic PCR . Temperature-sensitive strains were confirmed by incubating the plate at the restrictive temperature for 4–5 days . Cultures were grown in YPD , or if cells contained plasmids , in synthetic selection medium . For expression of PrANS from the GAL1 , 10 promoter , induction was typically for 2 hr . Temperature-sensitive mutants and their congenic wildtypes were grown at 25°C and temperature shifts were for 90 min at 37°C , or two cell division cycles at 30°C . Any exceptions to these regimens are noted in the respective figure legends . Cells were grown as described in the respective figure legends to a final O . D . 600 ( optical density at a wavelength of 600 nm ) between 1 . 0 and 2 . 0 , harvested by centrifugation , and drop-frozen in liquid nitrogen . Frozen cell pellets were thawed and washed with ice-cold buffer containing 50 mM Tris , pH 7 . 5 , 10 mM sodium azide , 10 mM EDTA , 10 mM EGTA , 1× protease inhibitor tablet ( Roche , Basel , Switzerland ) , 10 mM NEM , 50 mM NaF , 60 mM β-glycerophosphate , 10 mM sodium pyrophoshate . An equal volume of glass beads ( Sigma , St Louis , MO; 425–600 μm , acid washed ) was added and the cell pellets were immersed in boiling water for 3 min . They were then suspended in 1× SDS buffer ( 37 . 5 µl/O . D . unit ) and cells were lysed by vortexing in Fast Prep-24 ( MP ) for 45 s at a setting of 6 . 5 , and boiled again for 4 min . Boiled lysates were centrifuged at 16 , 000×g for 1 min . Aliquots were resolved by SDS-PAGE . In some instances , particularly when tRNA-linked substrate was being detected , washed cell pellets were brought up in 1× SDS buffer containing 1× Secure ( Ambion , Carlsbad , CA ) , heated at 65°C for 5 min , lysed by glass beads as described above , and loaded on NuPAGE gels ( Invitrogen , Carlsbad , CA ) . For immunoblot detection , gels were transferred to nitrocellulose and filters were stained with Ponceau S to determine equivalent loading of protein extracts . If loading was not equivalent , the experiment was repeated . The nitrocellulose filters were immunoblotted with desired antibody and developed by ECL or Super Signal ( Pierce , Rockford , IL ) . Anti-Gfp was from Clontech ( Mountain View , CA ) , Peroxidase Anti-Peroxidase Soluble Complex ( for all Protein A detections , PAP ) , anti-Tubulin and anti-Flag were from Sigma , anti-TAP was from Thermo ( Waltham , MA ) , anti-HA ( 12CA5 ) was from Roche , and anti-Ub antibodies were from Chemicon ( Millipore , Billerica , MA ) , Stressgen ( Enzo Life Sciences , Farmingdale , NY ) , and Abcam ( Cambridge , MA ) . Puromycin antibody was a gift from Peter Walter , and anti-Rpl32 a gift from Jonathan Warner . Ribosomes were affinity-purified using TAP-tagged RPL18B by following a protocol established in prior studies with some modifications as follows ( Halbeisen and Gerber , 2009; Halbeisen et al . , 2009 ) . Exponential cultures were grown at either 30°C or 37°C for 1 . 5 hr before being harvested by filtration using 0 . 2 µm Nalgene filters . In some experiments 0 . 1 mg/ml cycloheximide was added during filtration , and was omitted when no difference was observed in the experimental outcome . Cells were washed with a buffer containing 50 mM Tris , pH 7 . 5 , 10 mM NEM , 10 mM sodium pyrophosphate , 0 . 1 mg/ml cycloheximide and frozen in liquid nitrogen . Pellets were ground in liquid nitrogen . Cell powder was weighed and resuspended in twice the volume of lysis buffer ( buffer A ) containing 20 mM Tris–HCl ( pH 8 . 0 ) , 140 mM KCl , 5 mM MgCl2 , 1% Triton-100 , 10% glycerol , 0 . 2 mg/ml heparin , 0 . 1 mg/ml cycloheximide , 5 mM NEM , 0 . 5 mM AEBSF , protease inhibitor cocktail ( EDTA-free; Roche ) , 20 U/ml DNAse I , 0 . 05 U/µl of RNaseOut ( Invitrogen ) , 0 . 05 U/µl of SUPERase-In ( Ambion ) , 0 . 5 mM DTT . Lysate was centrifuged at 5000 rpm for 5 min and the supernatant reclarified at 14 , 000 rpm for 10 min in a refrigerated microfuge ( Eppendorf , Hamburg , Germany; 5417R ) . Lysate was bound to rabbit IgG conjugated to superparamagnetic Dynabeads ( Invitrogen ) as described by ( Oeffinger et al . , 2007 ) . Using beads of this diameter enables the isolation of large complexes such as polysomes . Magnetic beads were washed twice with wash buffer ( buffer B ) containing 20 mM Tris–HCl , pH 8 . 0 , 140 mM KCl , 5 mM MgCl2 , 10% glycerol , 0 . 025 U/µl RNaseOut , 0 . 025 U/µl of SUPERase-In , 1 mM DTT , 0 . 1% NP40 , supplemented with 0 . 1 mg/ml cycloheximide and 5 mM NEM , and then twice with buffer B lacking supplements . Elution was in buffer B containing 0 . 3 U/µl of Tobacco Etch Virus ( TEV ) protease at 16°C for 2 hr with intermittent shaking in the presence ( or absence ) of 2 mM puromycin , following which samples were adjusted to 0 . 6 M KCl and 3 mM puromycin and incubated at 37°C for 7 min . Eluates were removed after immobilization of beads with a magnet and characterized by Coomassie blue staining , immunoblotting , and RNA analysis using RNeasy mini columns ( Qiagen , Valencia , CA ) . A total of 25 A260 units of cultures lysed in buffer A were loaded on top of a 10–50% sucrose gradient prepared in buffer A lacking Triton X-100 , glycerol and NEM . The samples were centrifuged in an SW-41 rotor for 180 min at 35 , 000 rpm at 4°C and fractionated while continuously recording the absorbance at 254 nm with a flow cell UV detector ( ISCO , Lincoln , NE ) . Fractions of 0 . 5 or 0 . 75 ml were collected and analyzed by Coomassie blue staining and immunoblotting with various antibodies , including Rpl32 , to identify 40S , 60S , 80S and polysomal fractions . An aliquot ( 100 µl ) of each fraction was also analyzed for RNA using RNeasy mini columns ( Qiagen ) . Figure 1—figure supplement 2B is representative of this protocol . Ribosomes were also analyzed on 10–30% gradients prepared in 50 mM Hepes , pH 7 . 4 , 2 . 5 mM MgCl2 and 100 mM KOAc using the SW55Ti rotor at 50 , 000 rpm for 100 min with slow acceleration /deceleration . 0 . 2 ml fractions were collected from the top . Aliquots were resolved by SDS-PAGE and immunoblotted with anti-Rpl32 and the Ab for the specific NS substrate . To aid in tRNA detection , samples were also precipitated with 10% TCA following the addition of 0 . 02% deoxycholate . Pellets were washed with ice-cold acetone , resuspended in 2× Laemmli buffer containing RNA secure and analyzed by NuPAGE . Representative data are shown in Figure 4B . For analysis of tRNA-linked substrates in Figures 5B and 6B and Figure 3—figure supplement 1 , we employed the method described by ( Tsuboi et al . , 2012 ) . Ribosomes were sedimented at 49 , 000 rpm for 45 min ( RP100AT4 rotor in Sorvall RC M120EX ultracentrifuge ) through 0 . 5 M sucrose cushions in buffer C containing 20 mM Hepes , pH 7 . 4 , 5 mM MgOAc , 100 mM KOAc , 0 . 5 mM DTT , 100 µg/ml cycloheximide , 200 µg/ml heparin , and the protease and RNAse inhibitor cocktails described above . Ribosome pellets were resuspended in the same buffer in the presence ( or absence ) of 0 . 5 M KCl , 3 mM puromycin , 2 µM Ub aldehyde and incubated for 15 min on ice and 30 min at 25°C . RNAse inhibitors were omitted when ribosomes were treated with 10 or 100 µg/ml RNAse A at 30°C for 10 min . Wildtype and mutant cells growing in synthetic medium were temperature shifted to 36°C for 1 hr and starved for methionine for 30 min . Cells were then harvested by centrifugation , resuspended in one-tenth the original volume , and pulse-labeled with 150 µCi/ml 35S-methionine ( >600 Ci/mmol , cell-labeling grade; Perkin Elmer , Waltham , MA ) for 90 s . Aliquots were immediately withdrawn into collection tubes containing 2× stop solution ( 0 . 2 mg/ml cycloheximide and 500 mM sodium azide ) , and the remainder of the cultures chased for the indicated times with 0 . 2 mg/ml cycloheximide and 10 mM unlabeled methionine . Ribosomes were affinity-purified from cell pellets drop-frozen in liquid nitrogen by resuspending in buffer A and lysing cells with glass beads using a Fast-Prep . Elution with TEV protease was done in the presence of 2 mM puromycin , following which eluates were bound to recombinant Gst-Dsk2 UBA resin . Washed beads were eluted in 2× SDS-PAGE buffer , and aliquots were resolved by SDS-PAGE . Gels were dried and exposed to film ( Kodak , Rochester , NY ) . Cell pellets were resuspended in lysis buffer A described above except that the buffering reagent was 20 mM Hepes , pH 7 . 4 , and glycerol was omitted . Lysate was incubated on ice for ten minutes and then spun at 20 , 000×g for 10 min . Ribosomes were sedimented at 80 , 000 rpm for 65 min ( RP100AT4 rotor in Sorvall RC M120EX ultracentrifuge ) through 1 . 0 M sucrose cushions . Ribosome pellets were resuspended in SDS sample buffer and evaluated by Coomassie blue staining , or immunoboltting with anti-Ub antibodies . | Ribosomes are complex molecular machines that translate the sequence of bases in a messenger RNA ( mRNA ) transcript into a polypeptide that subsequently folds to form a protein . Each ribosome is composed of two major subunits: the small subunit reads the mRNA transcript , and the large subunit joins amino acids together to form the polypeptide . This process stops when the ribosome encounters a stop codon and releases the completed polypeptide . It is critical that cells perform some form of quality control on the polypeptides as they are translated to prevent a build up of incomplete , incorrect or toxic proteins in cells . Problems can occur if a ribosome stalls while translating the mRNA transcript , or if the mRNA transcript is defective . For example , most mRNA transcripts contain a stop codon , but some do not , and these non-stop mRNA transcripts result in a non-stop polypeptide that remains tethered to the ribosome . It is important that the cell identifies and removes these faulty polypeptides so as to leave the ribosome free to translate other ( non-faulty ) mRNA transcripts . A regulatory protein called ubiquitin is responsible for marking and sending proteins that are faulty , or are no longer needed by the cell , to a molecular machine called the proteasome , where they are degraded by a process called proteolysis . In 2010 researchers identified Ltn1 as the enzyme that attaches ubiquitin to non-stop proteins in yeast . Now , building on this work , Verma et al . identify additional proteins involved in this process . In particular , an ATPase enzyme called Cdc48 ( known as p97 or VCP in human cells ) and two co-factors—Ufd1 and Npl4—promote release of the ubiquitinated non-stop polypeptides from the ribosomes , thus committing the marked polypeptide to destruction by the proteasome . Verma et al . also show that the Cdc48-Ufd1-Npl4 complex is involved in other aspects of quality control of newly synthesized proteins within cells . Collectively these processes are known as ribosome-associated degradation . Mutations of the gene that codes for human p97 can cause a number of diseases , including Paget's disease of the bone and frontotemporal dementia , so an improved understanding of ribosome-associated degradation could provide new insights into these diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2013 | Cdc48/p97 promotes degradation of aberrant nascent polypeptides bound to the ribosome |
Bergmann's rule is a widely-accepted biogeographic rule stating that individuals within a species are smaller in warmer environments . While there are many single-species studies and integrative reviews documenting this pattern , a data-intensive approach has not been used yet to determine the generality of this pattern . We assessed the strength and direction of the intraspecific relationship between temperature and individual mass for 952 bird and mammal species . For eighty-seven percent of species , temperature explained less than 10% of variation in mass , and for 79% of species the correlation was not statistically significant . These results suggest that Bergmann's rule is not general and temperature is not a dominant driver of biogeographic variation in mass . Further understanding of size variation will require integrating multiple processes that influence size . The lack of dominant temperature forcing weakens the justification for the hypothesis that global warming could result in widespread decreases in body size .
Bergmann's rule describes a negative relationship between body mass and temperature across space that is believed to be common in endothermic species ( Bergmann , 1847; Brown and Lee , 1969; Kendeigh , 1969; Freckleton et al . , 2003; Carotenuto et al . , 2015 ) . Many hypotheses have been proposed to explain this pattern ( Blackburn et al . , 1999; Ashton , 2002; Watt et al . , 2010 ) including the heat loss hypothesis , which argues that the higher surface area to volume ratio of smaller individuals results in improved heat dissipation in hot environments ( Bergmann , 1847 ) . Though originally described for closely-related species ( i . e . , interspecific; Blackburn et al . , 1999 ) , the majority of studies have focused on the intraspecific form of Bergmann's rule ( Rensch , 1938; Meiri , 2011 ) by assessing trends in individual size within a species ( Langvatn and Albon , 1986; Yom-Tov and Geffen , 2006; Gardner et al . , 2009 ) . Bergmann's rule has been questioned both empirically and mechanistically ( McNab , 1971; Geist , 1987; Huston and Wolverton , 2011; Teplitsky and Millien , 2014 ) but the common consensus from recent reviews is that the pattern is general ( Ashton et al . , 2000; Ashton , 2002; Meiri and Dayan , 2003; Watt and Salewski , 2011 ) . It has recently been suggested that this negative relationship between mass and temperature could result in decreasing individual size across species in response to climate change ( Sheridan and Bickford , 2011 ) and that this may be a ‘third universal response to warming’ ( Gardner et al . , 2011 ) . The resulting shifts in size distributions could significantly alter ecological communities ( Brose et al . , 2012 ) , especially if the rate of size decrease varies among species ( Sheridan and Bickford , 2011 ) . While there is limited empirical research on body size responses to changes in temperature through time ( but see Smith et al . , 1995; Caruso et al . , 2014; Teplitsky and Millien , 2014 ) , the apparent generality of Bergmann's rule across space indicates the likelihood of a similar relationship in response to temperature change across time . The generality of Bergmann's rule is based on many individual studies that analyze empirical data on body size across an environmental gradient ( e . g . , Langvatn and Albon , 1986; Barnett , 1977; Fuentes and Jaksic , 1979; Dayan et al . , 1989; Sand et al . , 1995 ) and reviews that compile and evaluate the results from these studies ( Ashton , 2002; Meiri and Dayan , 2003; Watt et al . , 2010 ) . Most individual studies of Bergmann's rule are limited by: ( 1 ) analyzing only one or a few species ( e . g . , Langvatn and Albon , 1986 ) ; ( 2 ) using small numbers of observations ( e . g . , Fuentes and Jaksic , 1979 ) ; ( 3 ) only including data at the small scales typical of ecological studies ( e . g . , Sand et al . , 1995 ) ; ( 4 ) using latitude instead of directly assessing temperature ( e . g . , Barnett , 1977 ) ; and ( 5 ) focusing on statistical significance instead of the strength of the relationship ( e . g . , Dayan et al . , 1989 ) . The reviews tabulate the results of these individual studies and assess patterns in the direction and significance of relationships across species . Such aggregation of published results allows for a more general understanding of the pattern but , in addition to limitations of the underlying studies , the conclusions may be influenced by publication bias and selective reporting due to studies or individual analyses that do not support Bergmann's rule being published less frequently ( Koricheva et al . , 2013 ) . Previous analyses of publication bias in the context of Bergmann’s rule have found no evidence for selective publication , which supports the idea that it is a general rule ( Ashton , 2002; Meiri et al . , 2004 ) . However , two of the most extensive studies of Bergmann’s rule , which both used museum records to assess dozens of intraspecific Bergmann’s rule relationships simultaneously , found that the majority of species did not exhibit significant positive relationships between latitude and size ( McNab , 1971; Meiri et al . , 2004 ) . As a result , understanding the generality of this ecophysiological rule and its potential implications for global change requires more extensive analysis . A data-intensive approach to analyzing Bergmann's rule , evaluating the pattern using large amounts of broad scale data , has the potential to overcome existing limitations in the literature and provide a new perspective on the generality of the intraspecific form of Bergmann's rule . Understanding the generality of the temperature-mass relationship has important implications for how size will respond to climate change . We use data from VertNet ( Constable et al . , 2010 ) , a large compilation of digitized museum records that contains over 700 , 000 globally distributed individual-level size measures , to evaluate the intraspecific relationship between temperature and mass for 952 mammal and bird species . The usable data consist of 273 , 901 individuals with an average of 288 individuals per species , with individuals of each species spanning an average of 75 years and 34 latitudinal degrees . This approach reduces or removes many of the limitations to previous approaches and the results suggest that Bergmann's rule is not a strong or general pattern .
Most of the species in this study showed weak non-significant relationships between temperature and mass ( Figures 1 and 2 ) . The distribution of correlation coefficients was centered near zero with a mean correlation coefficient of −0 . 05 across species ( Figure 2A ) . Relationships for 79% of species were not significantly different from zero based on false discovery rate-controlled p values and associated z scores , while 14% of species' relationships were significant and negative and 7% were significant and positive ( Figure 2A , Figure 2—figure supplement 1 ) . Temperature explained less than 10% of variation in mass ( i . e . , −0 . 316 < r < 0 . 316 ) for 87% of species , indicating that temperature explained very little of the observed variation in mass for these species ( Figure 2A ) . The weak , non-directional intraspecific relationships indicated by the distribution of correlation coefficients are consistent across taxonomic groups and temporal lags . Mean correlation coefficients for both endothermic classes are −0 . 006 and −0 . 065 , for mammals and birds respectively ( Figure 2B ) . Similarly , correlation coefficient distributions were approximately centered on zero for all of the 30 orders analyzed ( −0 . 2 < r¯ < 0 . 003 for orders with more than 10 species; Figure 3 and Figure 3—figure supplement 1 ) , and for migrant and nonmigrant bird species ( Figure 2—figure supplement 2 ) . Correlation coefficient distributions for temperature-mass relationships using lagged temperatures were centered around zero like those using temperature from the collection year ( Figure 4 and Figure 4—figure supplement 1 ) , indicating that there was not a temporal lag effect on the response of species' masses to temperature . Correlation coefficients did not vary systematically by sample size ( Figure 5A ) , extent of variation in temperature or mass ( Figure 5B , C ) , species' average mass ( Figure 5D ) , or species' average latitude ( Figure 5E ) . While temperature is considered the actual driver , some studies use latitude as a proxy when evaluating variation in size ( Bergmann , 1847; Stillwell , 2010 ) . Using latitude , the mean correlation coefficient was −0 . 05 with no statistically significant latitude-mass relationship for 71% of species ( Figure 2—figure supplement 3 ) , while the respective values for temperature were −0 . 05 and 79% ( Figure 2A ) . Results were robust to a variety of decisions and stringencies about how to filter the size ( Figure 2—figure supplements 4 and 5 ) and species data ( Figure 2—figure supplements 6 and 7 ) .
In contrast to conventional wisdom and several recent review papers , our analysis of 952 species shows little to no support for a negative intraspecific temperature-mass relationship that is sufficiently strong or common to be considered a biogeographic rule . Three quarters of bird and mammal species show no significant change in mass across a temperature gradient and temperature explained less than 10% of intraspecific variation in mass for 87% of species ( Figure 2A ) . This was true regardless of taxonomic group ( Figures 2 and 3 ) , temporal lag in temperature ( Figure 4 ) , species' size , location , or sampling intensity or extent ( Figure 5 ) . These results are consistent with two previous studies that examined museum specimen size measurements across latitude . The first study showed that 22 out of 47 North American mammal species studied had no relationship between latitude and length , and 10 of the 25 significant relationships were opposite the expected direction ( McNab , 1971 ) . The second found a similar proportion of non-significant results ( 42/87 ) , but a lower proportion of significant relationships that opposed the rule ( 9/45 ) for carnivorous mammals ( Meiri et al . , 2004 ) . While more species had significant negative relationships than positive in both our study and these two museum-based studies , in all cases less than half of species had significant negative correlations ( 14–41% ) . In combination with these two smaller studies , our results suggest that there is little evidence for a strong or general Bergmann's rule when analyzing raw data instead of summarizing published results . Our results are inconsistent with recent reviews , which have reported that the majority of species conform to Bergmann's rule ( Ashton , 2002; Meiri and Dayan , 2003; Watt et al . , 2010 ) . While these reviews included results that were either non-significant or opposite of Bergmann's rule , the proportion of significant results in support of Bergmann's rule was higher and therefore resulted in conclusions that supported the generality of the temperature-mass relationship . Generalizing from results in the published literature involves the common challenges of publication bias and selective reporting ( Koricheva et al . , 2013 ) . In addition , because the underlying Bergmann's rule studies typically report minimal statistical information , often providing only relationship significance or direction instead of p-values or correlation coefficients ( Meiri and Dayan , 2003 ) , proper meta-analyses and associated assessments of biological significance are not possible . While several reviews found no evidence for publication bias using limited analyses ( Ashton , 2002; Meiri et al . , 2004 ) , the notable differences between the conclusions of our data-intensive approach and those from reviews suggest that publication bias in literature examining Bergmann's rule warrants further investigation . These differences also demonstrate the value of data-intensive approaches in ecology for overcoming potential weaknesses and biases in the published literature . Directly analyzing large quantities of data from hundreds of species allows us to assess the generality of patterns originally reported in smaller studies while avoiding the risk of publication bias . This approach additionally makes it easier to integrate other factors that potentially influence size into future analyses . The new insight gained from this data-intensive approach demonstrates the value of investing in large compilations of ecologically-relevant data ( Hampton et al . , 2013 ) and the proper training required to work with these datasets ( Hampton et al . , 2017 ) . Our analyses and conclusions are limited to the intraspecific form of Bergmann’s rule . This is the most commonly studied and well-defined form of the relationship , and the one most amenable to analyses using large compilations of museum data . Difficulty in interpreting Bergmann’s original formulation has resulted in an array of different ideas and implementations of interspecific analyses ( Blackburn et al . , 1999; Meiri and Thomas , 2007; Watt et al . , 2010; Meiri , 2011 ) . The most common forms of these interspecific analyses involve correlations between various species-level size metrics and environmental measures and are conducted at various taxonomic levels from genus to class ( e . g . , Blackburn and Gaston , 1996; Diniz-Filho et al . , 2007; Boyer et al . , 2009; Clauss et al . , 2013 ) . Efforts to apply data-intensive approaches to the interspecific form of this relationship will need to address the fact that occurrence records are not evenly distributed across the geographic range of species , and determine how the many interpretations of interspecific Bergmann’s rule are related to one another and the biological expectations for interspecific responses to temperature . The original formulation of Bergmann's rule , and the scope of our conclusions , apply only to endotherms . However , negative temperature-mass relationships have also been documented in ectotherms , with the pattern referred to as the size-temperature rule ( Ray , 1960; Angilletta and Dunham , 2003 ) . In contrast to the hypotheses for Bergmann's rule , which are based primarily on homeostasis ( Gardner et al . , 2011 ) , the size-temperature rule in ectotherms is thought to result from differences between growth and development rates ( Forster et al . , 2011 ) . The current version of VertNet contained ectotherm size data for only seven species , which is not sufficient to complete a comprehensive analysis of the ectotherm size-temperature rule . Future work exploring the ectotherm size-temperature rule in natural systems using data-intensive approaches is necessary for understanding the generality of this rule in ectotherms , and data may be sought for this effort in the literature or via a coordinated effort by museums to continue digitizing size measurements for specimens . A number of mechanisms have been suggested to explain why higher temperatures should result in lower body sizes , including heat loss , starvation , resource availability , migratory ability , and phylogenetic constraints ( Blackburn et al . , 1999 ) . Most of the proposed hypotheses have not been tested sufficiently to allow for strong conclusions to be drawn about their potential to produce Bergmann's rule ( Blackburn et al . , 1999; Watt et al . , 2010; Teplitsky and Millien , 2014 ) and the widely studied heat loss hypothesis has been questioned for a variety of reasons ( James , 1970; McNab , 1971; Blackburn et al . , 1999; Watt et al . , 2010; McNamara et al . , 2016 ) . While no existing hypotheses have been confirmed , it is possible that some processes are producing negative relationships between size and temperature . The lack of a strong relationship does not preclude processes that result in a negative temperature-mass relationship , but it does suggest that these processes have less influence relative to other factors that affect intraspecific size . The relative importance of the many factors besides temperature that can influence size within a species is as yet unknown . Size is affected by abiotic factors such as humidity and resource availability ( Teplitsky and Millien , 2014 ) , characteristics of individuals like clutch size ( Boyer et al . , 2009 ) , and community context , including possible gaps in size-related niches ( Smith et al . , 2010 ) and the trophic effects of primary productivity on consumer size ( Sheridan and Bickford , 2011 ) . Temperature itself can have indirect effects on size , such as via habitat changes in water flow or food availability , that result in size responses opposite of Bergmann's rule ( Gardner et al . , 2011 ) . Anthropogenic influences have been shown to influence the effect of temperature on size ( Faurby and Araújo , 2016 ) , and similar impacts of dispersal , extinctions , and the varying scales of climate change have been proposed ( Clauss et al . , 2013 ) . Because our data primarily came from North America , further analyses focused on species native to other continents could reveal differing temperature-mass relationships due to varying temperature regimes . While our work shows that more species have negative significant relationships between temperature and mass than positive , only 21% of species have statistically significant relationships and it consequently appears that some combination of other factors more strongly drives intraspecific size variation for most endothermic taxa . The lack of evidence for temperature as a primary determinant of size variation in endothermic species calls into question the hypothesis that decreases in organism size may represent a third universal response to global warming . The potentially general decline in size with warming was addressed by assessments that evaluated dynamic body size responses to temperature using similar approaches to the Bergmann's rule reviews discussed above ( Sheridan and Bickford , 2011; Gardner et al . , 2011; Teplitsky and Millien , 2014 ) . The results of these temporal reviews were similar to those for spatial relationships , but the conclusions of these studies clearly noted the variability in body size responses and the need for future data-intensive work ( Sheridan and Bickford , 2011; Gardner et al . , 2011 ) using broader temperature ranges ( Teplitsky and Millien , 2014 ) to fully assess the temperature-size relationship . Our results in combination with those from other studies suggest that much of the observed variation in size is not explained simply by temperature . While there is still potential for the size of endotherms , and other aspects of organismal physiology and morphology , to respond to both geographic gradients in temperature and climate change , these responses may not be as easily explained solely by temperature as has been suggested ( Sheridan and Bickford , 2011; Gardner et al . , 2011 ) . Future attempts to explain variation in the size of individuals across space or time should use integrative approaches to include the influence of multiple factors , and their potential interactions , on organism size . This will be facilitated by analyzing spatiotemporal data similar to that used in this study , which has broad ranges of time , space , and environmental conditions for large numbers of species and individuals . This data-intensive approach provides a unique perspective on the general responses of bird and mammal species to temperature , and has potential to assist in further investigation of the complex combinations of factors that determine biogeographic patterns of endotherm size and how species respond to changes in climate .
Organismal data were obtained from VertNet , a publicly available data platform for digitized specimen records from museum collections primarily in North America , but also includes global data ( Constable et al . , 2010 ) . Body mass is routinely measured when organisms are collected , with relatively high precision and consistent methods , by most field biologists , whose intent is to use those organisms for research and preservation in natural history collections ( Winker , 2000; Hoffmann et al . , 2010 ) . These measurements are included on written labels and ledgers associated with specimens , which are digitized and provided in standard formats , e . g . , Darwin Core ( Wieczorek et al . , 2012 ) . In addition to other trait information , mass has recently been extracted and converted to a more usable form from Darwin Core formatted records published in VertNet ( Guralnick et al . , 2016 ) . This crucial step reduces variation in how these measurements are reported by standardizing the naming conventions and harmonizing all measurement values to the same units ( Guralnick et al . , 2016 ) . We downloaded the entire datasets for Mammalia , Aves , Amphibia , and Reptilia available in September 2016 ( Bloom et al . , 2016a , Bloom et al . , 2016b , Bloom et al . , 2016c , Bloom et al . , 2016d ) using the Data Retriever ( Kironde et al . , 2017; Morris and White , 2013 ) and filtered for those records that had mass measurements available . Fossil specimen records with mass measurements were removed . We only analyzed species with at least 30 georeferenced individuals whose collection dates spanned at least 20 years and collection locations at least five degrees latitude , in order to ensure sufficient sample size and spatiotemporal extent to accurately represent each species' temperature-mass relationship . We conducted sensitivity analyses to determine if these thresholds were appropriate ( Figure 2—figure supplements 6 and 7 ) . We selected individual records with geographic coordinates for collection location , collection dates between 1900 and 2010 , and species-level taxonomic identification , which were evaluated to ensure no issues with synonymy or clear taxon concept issues . To minimize inclusion of records of non-adult specimens , we identified the smallest mass associated with an identified adult life stage category for each species and removed all records with mass values below this minimum adult size . Results were not qualitatively different due to either additional filtering based on specimen lifestage ( Figure 2—figure supplement 4 ) or removal of outliers ( Figure 2—figure supplement 5 ) . Temperatures were obtained from the Udel_AirT_Precip global terrestrial raster provided by NOAA from their website at http://www . esrl . noaa . gov/psd/ , a 0 . 5 by 0 . 5 decimal degree grid of monthly mean temperatures from 1900 to 2010 ( Willmott and Matsuura , 2001 ) . For each specimen , the mean annual temperature at its collection location was extracted for the year of collection . This resulted in a final dataset containing records for 273 , 901 individuals from 952 bird and mammal species ( MSB Mammal Collection ( Arctos ) , 2015; Ornithology Collection Passeriformes - Royal Ontario Museum , 2015; MVZ Mammal Collection ( Arctos ) , 2015; MVZ Bird Collection ( Arctos ) , 2015; KUBI Mammalogy Collection , 2016; CAS Ornithology ( ORN ) , 2015; DMNS Bird Collection ( Arctos ) , 2015; UCLA Donald R , 2015; DMNS Mammal Collection ( Arctos ) , 2015; UAM Mammal Collection ( Arctos ) , 2015; UWBM Mammalogy Collection , 2015; UAM Bird Collection ( Arctos ) , 2015; UMMZ Birds Collection , 2015; CUMV Bird Collection ( Arctos ) , 2015; CUMV Mammal Collection ( Arctos ) , 2015; MLZ Bird Collection ( Arctos ) , 2015; LACM Vertebrate Collection , 2015; CHAS Mammalogy Collection ( Arctos ) , 2016; Ornithology Collection Non Passeriformes - Royal Ontario Museum , 2015; KUBI Ornithology Collection , 2014; MSB Bird Collection ( Arctos ) , 2015; Biodiversity Research and Teaching Collections - TCWC Vertebrates , 2015; TTU Mammals Collection , 2015; CAS Mammalogy ( MAM ) , 2015; Vertebrate Zoology Division - Ornithology , Yale Peabody Museum , 2015; University of Alberta Mammalogy Collection ( UAMZ ) , 2015; UAZ Mammal Collection , 2016; Charles and Conner Museum , 2015; SBMNH Vertebrate Zoology , 2015; Cowan Tetrapod Collection - Birds , 2015; Cowan Tetrapod Collection - Mammals , 2015; NMMNH Mammal , 2015; Schmidt Museum of Natural History_Mammals , 2015; USAC Mammals Collection , 2013; MLZ Mammal Collection ( Arctos ) , 2015; Ohio State University Tetrapod Division - Bird Collection ( OSUM ) , 2015; Collections , 2015; DMNH Birds , 2015; CM Birds Collection , 2015; WNMU Mammal Collection ( Arctos ) , 2015; UCM Mammals Collection , 2015; UWYMV Bird Collection ( Arctos ) , 2015; NCSM Mammals Collection , 2015; Vertebrate Zoology Division - Mammalogy , Yale Peabody Museum , 2015; HSU Wildlife Mammals , 2016; WNMU Bird Collection ( Arctos ) , 2015; UWBM Ornithology Collection , 2015; UCM Birds , 2015; University of Alberta Ornithology Collection ( UAMZ ) , 2015; SDNHM Birds Collection , 2015 ) . The average number of individuals per species was 288 , ranging from 30 to 15 , 415 individuals . The species in the dataset were diverse , including volant , non-volant , placental , and marsupial mammals , and both migratory and non-migratory birds . There were species from all continents except Antarctica , though the majority of the data were concentrated in North America ( Figure 1A ) . The distribution of the species' mean masses was strongly right-skewed , as expected for broad scale size distributions ( Brown and Nicoletto , 1991 ) , with 74% of species having average masses less than 100 g . Size ranged from very small ( 3 . 7 g desert shrew Notiosorex crawfordi and 2 . 6 g calliope hummingbird Stellula calliope ) to very large ( 63 kg harbor seal Phoca vitulina and 5 . 8 kg wild turkey Meleagris gallopavo ) . We fit the intraspecific relationship between mean annual temperature and mass for each species with ordinary least squares linear regression ( e . g . , Figure 1B , C , D and Figure 1—figure supplements 1–12 ) using the statsmodels . formula . api module in Python ( Seabold and Perktold , 2010 ) . The strength of each species’ relationship was characterized by the correlation coefficient , its significance at alpha of 0 . 05 , and the associated z score . When assessing statistical significance with large numbers of correlations it is important to consider the expected distribution of these correlations under the null model that no correlation exists for any species . We addressed this issue by using false discovery rate control ( Benajmini and Hochberg , 1995 ) implemented with the stats package in R ( R Core Team , 2016 ) . This method determines the expected distribution of values for p ( or Z ) in the case where no relationship exists for individual correlation and adjusts observed values to control for excessive false positives . Specifically , it maintains the Type I error rate ( proportion of false positives ) across all tests at the chosen value of alpha and therefore gives an accurate estimate of the number of significant relationships ( Benajmini and Hochberg , 1995 ) . This allows us to estimate the number of species with true positive and negative correlations ( i . e . , those that have values that exceed those expected from the null distribution ) . We then compared the number of species with positive and negative correlation coefficients , and the proportion of those with statistically significant adjusted p-values . We investigated various potential correlates of the strength of Bergmann's rule . Because it has been argued that Bergmann's rule is exhibited more strongly by some groups than others ( McNab , 1971 ) , we examined correlation coefficient distributions within each class and order . Additionally , distributions for migrant and nonmigrant bird species were compared due to conflicting evidence about the impact of migration on temperature-mass relationships ( Ashton , 2002 ) . As a temporal lag in size response to temperature is likely due to individuals of a species responding to temperatures prior to their collection year ( e . g . , Stacey and Fellowes , 2002 ) , we assessed species' temperature-mass relationships using temperatures from 1 to 110 years prior to collection year . We also examined the relationship between species' correlation coefficients and five variables to understand potential statistical and biological influences on the results . We did so with the number of individuals , temperature range , and mass range to determine if the relationship was stronger when more data points or more widely varying values were available . Because it has been argued that Bergmann's rule is stronger in larger species ( Steudel et al . , 1994 ) and at higher latitudes ( Freckleton et al . , 2003; Faurby and Araújo , 2016 ) , we examined variability with both mean mass and mean latitude for each species . We also conducted all analyses using latitude instead of mean annual temperature . The reproducible code for these analyses is available ( https://github . com/KristinaRiemer/MassResponseToTemp; Riemer and White , 2017 ) and archived ( https://zenodo . org/badge/latestdoi/17957630 ) . | Scientists have found that individual animals of the same species tend to be smaller in hotter environments and larger in cooler ones . They named this pattern “Bergmann’s Rule” to describe how temperature can influence the size of an animal . However , most studies of Bergmann’s Rule have only looked at one or a few species at a time . Knowing how many species follow this rule is important because globally rising temperatures could cause lots of species to become smaller . Since the size of organisms affects how much food and space they need , this could disrupt natural systems around the world . To test if Bergmann’s rule can be extended to many species , Riemer , Guralnick , and White assessed the relationship between temperature and body mass for 952 bird and mammal species . Contrary to Bergmann’s Rule , the results showed that most of the species had similar sizes regardless of the temperature of their environment . Only about 140 species were smaller in hotter environments , and about 70 species were larger in hotter environments . This suggest that Bergmann’s Rule does not apply to most species as expected . While most birds and mammals may not get bigger or smaller due to warming global temperatures , the few species that do change in size – and the species that interact with them – may be more likely to become endangered or extinct . If we can determine which animals are at risk , we can prioritize their conservation and design better plans for doing so . Losing even a single species disrupts our ecosystems , on which we rely for critical resources like food , building materials , and clean air . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2018 | No general relationship between mass and temperature in endothermic species |
mTOR inhibition is beneficial in neurodegenerative disease models and its effects are often attributable to the modulation of autophagy and anti-apoptosis . Here , we report a neglected but important bioenergetic effect of mTOR inhibition in neurons . mTOR inhibition by rapamycin significantly preserves neuronal ATP levels , particularly when oxidative phosphorylation is impaired , such as in neurons treated with mitochondrial inhibitors , or in neurons derived from maternally inherited Leigh syndrome ( MILS ) patient iPS cells with ATP synthase deficiency . Rapamycin treatment significantly improves the resistance of MILS neurons to glutamate toxicity . Surprisingly , in mitochondrially defective neurons , but not neuroprogenitor cells , ribosomal S6 and S6 kinase phosphorylation increased over time , despite activation of AMPK , which is often linked to mTOR inhibition . A rapamycin-induced decrease in protein synthesis , a major energy-consuming process , may account for its ATP-saving effect . We propose that a mild reduction in protein synthesis may have the potential to treat mitochondria-related neurodegeneration .
The mTOR complexes coordinate nutrient availability with cell growth and proliferation , promoting protein synthesis and inhibiting autophagy ( Laplante and Sabatini , 2012 ) . Protein homeostasis is often distorted in neurodegenerative diseases , such as Parkinson’s and Alzheimer’s disease , as well as PolyQ and other proteinopathies , making mTOR an attractive therapeutic target ( Bové et al . , 2011 ) . Studies from animal models support mTOR inhibition as a promising therapeutic approach for neurodegenerative diseases , although several distinct beneficial mechanisms have been proposed . Rapamycin , an mTORC1 inhibitor , reduces neuronal cell death in a mouse model of Parkinson's disease , and decreased synthesis of DDIT4 ( DNA-damage-inducible transcript 4 ) was proposed to provide the protective effect by maintaining AKT pro-survival phosphorylation ( Malagelada et al . , 2010 ) . Rapamycin strongly suppresses degeneration of dopaminergic neurons in Drosophila with loss of function mutations of PINK1 and PARKIN , genes in which mutations cause human early onset Parkinsonism; importantly , overexpression of 4E-BP , a protein synthesis inhibitor downstream of the mTORC1 pathway , also rescues neuronal degeneration in these fly mutants . Increased production of GSTS1 , a detoxifying enzyme , was suggested to be the beneficial factor ( Tain et al , 2009 ) . In a mouse model of Alzheimer’s disease , deleting one mTOR allele decreases amyloid-β deposits and ameliorates memory deficits possibly through enhanced autophagy ( Caccamo et al . , 2014 ) . Mitochondrial dysfunctions are frequently observed in neurodegenerative diseases ( Lin and Beal , 2006 ) . Proteins causing neuronal degeneration often have either direct or indirect deleterious effects on mitochondrial functions . For example , α-synuclein inhibits mitochondrial fusion and causes mitochondrial fragmentation followed by a decrease in mitochondrial respiration and neuronal death ( Kamp et al . , 2010; Nakamura et al . , 2011 ) . Pink1 and Parkin are critical in mitochondrial quality maintenance ( Pickrell and Youle , 2015 ) . Amyloid precursor proteins accumulate in mitochondrial import channels , resulting in mitochondrial dysfunction as a hallmark of Alzheimer’s disease pathology ( Devi et al . , 2006 ) . In Huntington's disease , mutant huntingtin also has detrimental effects on mitochondrial function ( Zuccato et al . , 2010 ) . Thus , it appears that mitochondrial dysfunction and bioenergetic collapse could be a critical step towards neuronal death . Mitochondrial dysfunction results in decreased ATP levels in neurons . The delicate influence of ATP level on neuronal survival is best exemplified by maternally inherited Leigh syndrome ( MILS ) , a mitochondrial DNA ( mtDNA ) disease , characterized by severe early childhood neurodegeneration . T8993G in MT-ATP6 , encoding an ATP synthase subunit , is the most common mutation in MILS ( Finsterer , 2008 ) . A unique feature of mtDNA disease is that disease severity is correlated with the mutation load , i . e . , the percentage of mutated mitochondrial DNA copies ( Taylor and Turnbull , 2005 ) . Higher than 90% ATP6 T8993G causes MILS , whereas , 70~90% causes a less severe disease called NARP syndrome with symptoms , such as neuropathy , ataxia , and retinitis pigmentosa , that gradually develop with age . In a cybrid study where patient platelets containing the T8993G mtDNA mutation were fused to human osteosarcoma cells devoid of mtDNA , ATP synthesis was found to be negatively correlated with the mutation load ( Mattiazzi et al . , 2004 ) , indicating that a moderate difference in ATP level can dictate disease severity and the extent of neuronal death . mTOR inhibition by rapamycin greatly attenuates neurodegeneration caused by mitochondrial complex I defects ( Johnson et al . , 2013b ) . This study showed a dramatic therapeutic effect of rapamycin on a mouse model of Leigh syndrome , deficient in Ndufs4 , a nuclearly-encoded component of complex I . The life span is significantly extended , and neuronal degeneration is greatly attenuated . The exact rescue mechanism is unclear , but autophagy or mitochondrial biogenesis was excluded . It is not known if mTORC1 inhibition by rapamycin would have similar beneficial effects on the mutations affecting other respiratory complexes ( Vafai and Mootha , 2013 ) . So far , rapamycin’s effects on neuronal bioenergetics have not yet been explored . Here , we show that rapamycin significantly preserves neuronal ATP levels , particularly when mitochondrial oxidative phosphorylation is impaired by mitochondrial inhibitors . To test the therapeutic potential of rapamycin on neurodegeneration due to energy deficiency , we developed an iPSC-based disease model of maternally inherited Leigh syndrome ( MILS ) , due to a T8993G mtDNA mutation in the ATP6 gene . The MILS neurons exhibited energy defects and degenerative phenotypes consistent with patient clinical observations . Rapamycin treatment significantly alleviated ATP deficiency , reduced aberrant AMPK activation in MILS neurons and improved their resistance to glutamate toxicity . Mechanistically , MILS neurons and neurons treated with mitochondrial inhibitors all exhibited enhanced mTORC1 activity , signified by elevated ribosomal S6 and S6 kinase phosphorylation , indicating a causal link between mitochondrial dysfunction and mTOR signaling in neurons , and providing a rationale for treatment with rapamycin , which reduces protein synthesis , a major energy-consuming process .
The effect of rapamycin on cellular ATP level was examined in neurons derived from human embryonic stem cells , an approach that has been successfully used to model a variety of neurological diseases ( Qiang et al . , 2013 ) . Three mitochondrial drugs were used to mimic mitochondrial oxidative defects: oligomycin , blocking the ATP synthase; rotenone and antimycin-A , inhibiting complexes I and III , respectively , and CCCP , a mitochondrial uncoupler . We first tested whether rapamycin would affect neuronal ATP level . After a 6 hr rapamycin treatment of cultured wild type neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 human ESCs , the ATP level was increased by ~13% compared to neurons treated with DMSO as control . FK-506 ( tacrolimus ) that binds FKBP12 , which is also a rapamycin target protein , but inhibits calcineurin signaling rather than the mTOR pathway ( Taylor et al . , 2005 ) , did not change the ATP level ( Figure 1A ) . Oligomycin treatment alone decreased neuronal ATP level to ~ 64% of that in neurons treated with DMSO , but strikingly , cotreatment with oligomycin plus rapamycin maintained the ATP level at ~86% ( Figure 1A ) . Consistent with the higher ATP level , neurons cotreated with rapamycin showed lower AMPK T172 phosphorylation , an indicator of cellular ATP deficiency , compared to treatment with oligomycin alone ( Figure 1B ) . Similar effects of rapamycin were observed in neurons treated with rotenone and antimycin-A; but , interestingly , rapamycin was not able to preserve ATP when neurons were treated with CCCP ( Figure 1A ) . It should be noted that both oligomycin and rotenone/antimycin-A treatment reduce ATP production by directly inhibiting oxidative phosphorylation; in contrast , CCCP does so by uncoupling electron transport from ATP production , which not only reduces ATP production , but also stimulates oxidative phosphorylation and induces mitochondrial substrate burning and heat production . We suspect that this difference may account for the different effects of co-treatment with rapamycin . These data indicate that rapamycin can increase neuronal ATP levels and preserve cellular energy when oxidative phosphorylation is impaired . 10 . 7554/eLife . 13378 . 003Figure 1 . Rapamycin treatment increased neuronal ATP levels . ( A ) The effect of rapamycin ( RAPA ) on cellular ATP level was examined in 5-week neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 ESCs . Rapamycin was used at 20 nM ( final concentration ) . Mitochondrial dysfunction was mimicked by chemicals disrupting mitochondrial oxidative function: oligomycin ( 2 µM ) , blocking complex V ( ATP synthase ) ; rotenone and antimycin A ( R&A; 1 µM each ) , complex I and III inhibitors; CCCP ( 20 µM ) , a mitochondrial uncoupler . All were prepared in DMSO as vehicle . N-acetylcysteine ( NAC ) was used at 750 µM ( final concentration ) . The treatment was done for 6 hr with neurons grown in duplicate wells from the same batch of differentiation . The relative ATP level for each treatment was calculated as percentage after normalization to DMSO-treated neurons . Bars are mean ± SD , n=3 . *p<0 . 05 . **p<0 . 01 , calculated by two-tailed t-test . ( B ) Immunoblot analysis of cell lysates prepared from neurons treated with oligomycin , rapamycin or both for 6 hr . The intensity of phosphorylated protein was quantified after normalization to non-phosphorylated signal , and was presented as fold change compared to control group treated with DMSO . ( C ) Immunoblot analysis of cell lysates prepared from neurons treated with oligomycin or AICAR for 6 hr . ( D ) Oxygen consumption rate ( OCR ) measurement by Seahorse extracellular flux analyzer . The basal OCRs of neurons treated with rapamycin for 6 hr were compared to neurons treated with DMSO as control . Bars are mean ± SD , n=3 . ( E ) Immunoblot analysis of cell lysates prepared from neurons treated with CCCP for 6 hr . ( F ) Immunoblot analysis of cell lysates prepared from neurons treated with rotenone and antimycin-A for 6 hr . ( G ) Immunoblot analysis of cell lysates prepared from neurons treated with oligomycin or rotenone & antimycin-A for 20 min . ( H ) Immunoblot analysis of cell lysates prepared from NPCs treated with rapamycin , oligomycin , and rotenone/antimycin-A for 6 hr . ( I ) Immunoblot analysis of cell lysates prepared from NPCs treated with oligomycin or rotenone & antimycin-A for 20 min . ( J ) The effect of protein synthesis inhibition on cellular ATP level was examined in 5-week neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 ESCs . Cycloheximide ( CHX ) was used at 20 µg/ml , and 4E1RCat was used at 50 µM . The treatment was done for 2 hr with CHX and 4E1RCat alone , and for 6 hr when combined with mitochondrial inhibitors with neurons grown in duplicate wells from the same batch of differentiation . ( K ) Five-week neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 ESCs were treated with vehicle ( DMSO ) or rotenone & antimycin-A ( R&A ) . Protein synthesis was measured by pulsing for 2 hr with 35S-Cys/Met every 2 hr from 0 to 6 hr , and 35S incorporation into protein and neuronal ATP levels were quantified and normalized to the DMSO-treated controls . Data are mean ± SD , n=3 . ( L ) Five-week neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 ESCs were treated for vehicle ( DMSO ) , rotenone & antimycin-A , rapamycin or both ( R&A Rapa ) for 4 hr . Protein synthesis was measured by labeling for 2 hr with 35S-Cys/Met from 2 to 4 hr . **p<0 . 01 , calculated by two-tailed t-test . ( M ) Protein synthesis in NPCs derived from H9 ESC treated with rotenone & antimycin-A for 6 hr . Data are mean ± SD , n=3 . All the experiments were repeated at least three times . ( see associated Figure 1—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00310 . 7554/eLife . 13378 . 004Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00410 . 7554/eLife . 13378 . 005Figure 1—figure supplement 1 . Five-week neurons differentiated from human neuroprogenitor cells ( NPCs ) derived from H9 ESCs were treated for vehicle ( DMSO ) , cycloheximide and 4E1RCat . Protein synthesis was measured by pulsing for 2 hr with 35S-Cys/Met , and 35S incorporation into protein was quantified and normalized to DMSO-treated control . Data are mean ± SD , n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00510 . 7554/eLife . 13378 . 006Figure 1—figure supplement 2 . The glucose concentration in the medium growing 3-week neurons derived from H9 ESCs treated with DMSO and rapamycin for 8 hr were quantified by YSI 2950 metabolite analyzer . Before drug treatment , 1 ml fresh neuronal growth medium was added to neurons grown in duplicated wells of 12-well plate . Glucose used was calculated by minus from glucose in fresh medium . Bars represent mean ± SD . n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 006 Phosphorylation of ribosomal protein S6 , a target of mTOR complex 1 ( mTORC1 ) signaling , is increased in the brain lysate of Ndufs4 -/- mice , although it is unknown in what type of brain cells , i . e . neurons or glial cells , this occurs ( Johnson et al . , 2013b ) . We found an ~2-fold increase in ribosomal S6 and S6 kinase phosphorylation in neurons treated for 6 hr with oligomycin or rotenone/antimycin-A , but not CCCP ( Figure 1B , E and 1F ) . Rapamycin only partially decreased mTOR S2481 phosphorylation as previously reported ( Hsu et al . , 2011 ) , but almost completely abolished the ribosomal S6 and S6K phosphorylation observed in oligomycin-treated neurons , indicating its dependence on mTORC1 ( Figure 1B ) . We did not observe a consistent change of mTOR phosphorylation at S2481 or S2448 ( not shown ) upon oligomycin or rotenone/antimycin-A treatment . The increased S6 and S6 kinase phosphorylation was not due to AMPK activation , as AICAR , an AMPK agonist , did not alter their phosphorylation ( Figure 1C ) . In fact , AMPK activation is generally associated with decreased mTORC1 activity and S6K phosphorylation as a result of direct phosphorylation of Tsc2 and Raptor by AMPK ( Inoki et al . , 2003; Gwinn et al , 2008 ) . Consistent with this , NPCs treated with oligomycin or rotenone/antimycin-A for 20 min or for 6 hr showed decreased S6 and S6K phosphorylation , increased phosphorylation of AMPK and its substrate , acetyl-CoA carboxylase ( ACC ) ( Figure 1H and 1I ) . In contrast to the 6 hr treatment , 20 min of oligomycin or rotenone/antimycin-A treatment did not significantly alter S6 and S6K phosphorylation in neurons ( Figure 1G ) . Taken together , these results suggest that the increased S6 and S6K phosphorylation in neurons treated for 6 hr with these mitochondria inhibitors is a neuron-selective response to some cumulative effect caused by mitochondrial dysfunction . As rapamycin was able to preserve ATP in neurons treated with rotenone/antimycin-A , which largely abolishes oxidative phosphorylation , it is unlikely that the effect of rapamycin is through increasing ATP production from mitochondrial oxidative phosphorylation . Nevertheless , we measured the basal oxygen consumption rate in neurons treated with rapamycin , and the rate was similar to that in DMSO-treated controls , supporting the conclusion that rapamycin’s effect does not come from increasing oxidative phosphorylation activity ( Figure 1D ) . Rapamycin treatment did not increase neuronal glucose consumption indicating that glycolysis was not increased ( Figure 1—figure supplement 2 ) . Mitochondrial dysfunction leads to increased production of reactive oxygen species causing cellular damage; however , N-acetylcysteine ( NAC ) , a commonly used anti-oxidant , was unable to maintain the ATP level in this assay ( Figure 1A ) . With long-term treatment , both oligomycin and rotenone/antimycin-A treated neurons showed increased ribosomal S6 and S6K phosphorylation , which are indicators of mTORC1 signaling and , indirectly , the rate of protein synthesis , a major cellular energy-consuming process ( Rolfe and Brown , 1997 ) , and the increase in ribosomal S6 phosphorylation in oligomycin-treated neurons was almost completely abolished by rapamycin treatment , which inhibits mTORC1 ( Figure 1B ) . One consequence of rapamycin inhibition of mTORC1 is a decrease in protein synthesis , and to explore the potential ATP-saving effect of protein synthesis inhibition , neurons were treated for 2 hr with cycloheximide ( CHX ) , a widely-used protein synthesis inhibitor , and 4E1RCat , a protein synthesis initiation inhibitor that blocks eIF4E:eIF4G and eIF4E:4E-BP1 interactions ( Cencic et al . , 2011 ) , causing a drop in protein synthesis to ~ 9% and ~55% of that of neurons treated with DMSO as control , respectively ( Figure 1—figure supplement 1 ) . As a result of this inhibition , the neuronal ATP level increased by ~26% and ~14% , respectively ( Figure 1J ) . Similar to rapamycin , cycloheximide and 4E1RCat could also significantly preserve ATP levels in neurons treated with mitochondrial inhibitors ( Figure 1J ) . These results imply that reduction of protein synthesis is an important factor for rapamycin to preserve ATP in neurons . The increased S6 and S6K phosphorylation in neurons treated with mitochondrial inhibitors at later times implies that an increase in protein synthesis should occur . To monitor neuronal protein synthesis during the treatment with mitochondrial inhibitors , we labeled neuronal cultures with 35S-cysteine/methionine at different time points after rotenone/antimycin-A treatment . During the first 2 hr , protein synthesis dropped to ~36% of that of neurons treated with DMSO as control , but , interestingly , protein synthesis recovered and reached ~80% by 4–6 hr despite the sustained decrease in neuronal ATP levels ( Figure 1K ) . Such increased protein synthesis would consume more ATP and presumably aggravate the energy crisis . When rapamycin was added together with rotenone/antimycin-A , after 4 hr treatment , protein synthesis was ~38% of the control , while in neurons treated with rotenone and antimycin-A , the protein synthesis rate was ~62% , and rapamycin treatment alone decreased protein synthesis to ~60% ( Figure 1L ) . Therefore , rapamycin may help to preserve ATP through reducing protein synthesis . Protein synthesis was not restored in NPCs treated with rotenone/antimycin-A for 6 hr ( Figure 1M ) , consistent with the phosphorylation status of S6K and S6 ( Figure 1H ) . Our results so far indicated that rapamycin can preserve the ATP level in neurons treated with mitochondrial oxidative phosphorylation inhibitors , which cause acute mitochondrial dysfunction . To test the therapeutic potential of rapamycin for neurodegeneration resulting from energy deficiency , we developed an induced pluripotent stem ( iPS ) cell model of maternally-inherited Leigh syndrome ( MILS ) , an infantile neurodegenerative disease due to mitochondrial DNA mutation . We obtained a clone of primary fibroblasts ( GM13411 ) derived from a male MILS syndrome patient , who died at 8 months; the patient’s symptoms , disease development and brain pathology were typical of MILS syndrome as described in a clinical report ( Pastores et al . , 1994 ) . The GM13411 MILS patient fibroblast line has a T8993G mutation resulting in a change from a conserved leucine to arginine at amino acid position 156 in ATP6 , a subunit of Complex V/ATP synthase . Three iPS cell lines ( iPSCs ) were established from GM13411 fibroblasts using a standard cocktail of reprogramming retroviruses , expressing OCT4 ( POU5F1 ) , SOX2 , KLF4 , and MYC ( Takahashi et al . , 2007 ) . Pluripotency markers were assessed by RT-PCR ( Figure 2—figure supplement 1 ) and immunostaining ( Figure 2A ) . Healthy control iPS cell lines were derived from BJ male human fibroblasts . Both the T8993G and BJ iPSCs had normal karyotypes ( Figure 2—figure supplement 2 ) . The mitochondrial genomes from these T8993G and BJ iPS cells were sequenced , and no major pathogenic mutations , apart from T8993G , were found compared to mitochondrial genome variation databases ( Figure 2—figure supplement 3 ) . Subsequently , three neural progenitor cell lines ( NPC ) were derived from the respective patient iPS cell lines using the embryoid-body based protocol outlined in Figure 2—figure supplement 4 . Expected neural progenitor markers were present by immunostaining ( Figure 2A ) , and all the T8993G NPC lines retained the T8993G mutation ( Figure 2B and Figure 2—figure supplement 5 ) . Like the GM13411 fibroblasts , the derived iPSCs , NPCs and neurons all had an extremely high T8993G mtDNA mutation load ( Figure 2—figure supplement 6 ) . The staining of neuronal differentiation markers and electrophysiological analysis of patient and BJ neurons is described in supplementary data; representative results for T8993G neurons were shown in Figure 2—figure supplement 7 , 8 . 10 . 7554/eLife . 13378 . 007Figure 2 . Established iPSCs and neuroprogenitor cells ( NPC ) from GM13411 , a MILS fibroblast line . ( A ) T8993G iPSC expressed pluripotency markers that included Tra-1–60 , Lin28 , Tra-1–81 and Nanog . NPCs derived from T8993G iPSCs were stained with anti-Sox2 and Nestin . ( B ) T8993G mutation generates a Sma I restriction enzyme site . T8993G iPSCs and NPC cells still retained the mutation as confirmed by PCR and Sma I digestion . DNA products were separated on agarose gel by electrophoresis . ( C ) Mitochondrial membrane potential analyzed by fluorescence-activated cell sorting ( FACS ) using TMRE staining . Two lines of iPSCs , NPCs and neurons derived from BJ fibroblasts and one from H9 hESCs were used as controls ( WT ) . The relative mitochondrial membrane potential was presented as percentage compared to the mean of control . Bars are mean ± SD , n=3 . The experiment was repeated three times . ( D ) Cellular reactive oxygen species ( ROS ) analyzed by FACS using CM-H2DCFDA staining . Two lines of iPSCs , NPCs and neurons derived from BJ fibroblasts and one from H9 hESCs were used as control ( WT ) . The relative ROS level was presented as percentage compared to the mean of control . Bars are mean ± SD , n=3 . The experiment was repeated three times . ( E ) T8993G NPCs and neurons had higher expression of oxidative stress response genes including SOD1 , GPX1 and GSS . Two lines of iPSCs , NPCs and neurons derived from BJ fibroblasts and one from H9 hESCs were used as control ( WT ) . The gene expression levels were quantified by real-time PCR after normalization to β-actin . The relative expression level was presented as percentage compared to the mean of control . Bars are mean ± SD , n=3 . The experiment was repeated three times . ( F , H , J ) Oxygen consumption rate ( OCR ) measured by Seahorse extracellular flux analyzer . FCCP ( F ) is a mitochondrial uncoupler; rotenone and antimycin A ( R&A ) are complex I and III inhibitors . Error bars represent SD , n=6 . Non-mitochondrial oxygen consumption has been subtracted . The relative percentage of basal and maximum OCR of T8993G iPSC , NPC and neurons at 3 weeks of differentiation were calculated by comparing to the mean of BJ and H9 cells ( WT ) . The original data was in Figure 2—figure supplement 9 . ( G , I , K ) Measurement of lactate secreted by iPSCs , NPCs and neurons at 3 weeks of differentiation . The relative percentage of secreted lactate from T8993G iPSC , NPC and neurons was calculated by comparing to the mean of BJ and H9 cells . Bars represent mean ± SD . n=3 . *p<0 . 05 . Calculated by two-tailed t-test . The experiments were repeated three times . ( see associated Figure 2—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00710 . 7554/eLife . 13378 . 008Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00810 . 7554/eLife . 13378 . 009Figure 2—figure supplement 1 . RT-PCR analysis of pluripotency genes , OCT4 , NANOG , KLF4 and SOX2 in T8993G and BJ iPSCs . The expression of ectopic OCT4 and SOX2 in the retrovirus vectors was also examined . Primary fibroblasts , H9 ESCs and day3-transduced fibroblasts were used as negative and positive controls respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 00910 . 7554/eLife . 13378 . 010Figure 2—figure supplement 2 . Karyotypes of three T8993G iPSC clones , 46 , XY; and one BJ iPSC , 46 , XY . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01010 . 7554/eLife . 13378 . 011Figure 2—figure supplement 3 . Sequencing of mitochondrial DNA extracted from T8993G ( GM13411 ) and BJ iPSCs . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01110 . 7554/eLife . 13378 . 012Figure 2—figure supplement 4 . The outline of the protocol used to differentiate neurons from iPSCs; representative pictures of fibroblasts , iPSCs , embryoid bodies ( EB ) and neural rosettes . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01210 . 7554/eLife . 13378 . 013Figure 2—figure supplement 5 . Sanger sequencing confirmed the T8993G mutation ( upper panel , representative result ) . The sensitivity of Sanger sequencing to detect T8993G mutation copy load was tested by mixing wild-type and T8993G PCR product at the ratios indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01310 . 7554/eLife . 13378 . 014Figure 2—figure supplement 6 . T8993G iPSCs , NPC cells and neurons all had an extremely high T8993G mtDNA mutation load as GM13411 fibroblast . The mutations were confirmed by PCR and Sma I digestion . DNA products were separated on agarose gel by electrophoresis . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01410 . 7554/eLife . 13378 . 015Figure 2—figure supplement 7 . Neuronal marker staining . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01510 . 7554/eLife . 13378 . 016Figure 2—figure supplement 8 . Electrophysiological study of T8993G and BJ 5-week neurons . Representative results of T8993G neurons were shown . a . Evoked voltage dependent sodium and potassium currents recorded in voltage-clamp ( -70mV ) . b . Evoked action potential . c . Spontaneous burst of action potentials . . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01610 . 7554/eLife . 13378 . 017Figure 2—figure supplement 9 . Oxygen consumption rate ( OCR ) analysis by Seahorse extracellular flux analyzer on BJ , H9 and ATP6 T8993G iPSCs , NPCs and 3-week neurons . The data were analyzed by WAVE , software from Seahorse Bioscience . A , C and E present the OCR measurement normalized to protein content , while B , D , F show the OCR percentage change after normalization to measurement point 5 . FCCP ( F ) is a chemical uncoupler of electron transport; rotenone and antimycin A ( D ) are complex I and III inhibitors . Error bars represent SD , n=6 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 017 T8993G iPSCs , NPCs and neurons all exhibited increased mitochondrial membrane potential and cellular ROS level compared to the control BJ- and H9-derived lines ( Figure 2C and 2D ) . ROS responsive genes , such as SOD1 , GPX1 and GSS , were also up-regulated in T8993G NPCs and neurons ( Figure 2E ) . These data are consistent with the previous studies based on a cybrid model , in which the ATP6 T8993G mutation was found to impair the proton passage through ATP synthase , which leads to higher mitochondrial membrane potential and ROS production ( Cortés-Hernández et al . , 2007; Manfredi et al . , 1999; Mattiazzi et al . , 2004; Trounce et al . , 1994 ) . Extracellular flux analysis of oxygen consumption rate ( OCR ) revealed that iPSCs , NPCs and neurons containing T8993G mtDNA had a lower basal OCR but similar maximum OCR to controls ( Figure 2F , H and 2J ) , mechanistically consistent with ATP synthase deficiency . Significantly more lactate was secreted by T8993G iPSCs and NPCs than controls ( Figure 2G and 2I ) , whereas T8993G neurons exhibited only a small increase in secreted lactate ( Figure 2K ) . This result indicates enhanced aerobic glycolysis in proliferating T8993G cells . Recently , Ma et al . ( 2015 ) reported the establishment of iPSCs from the same patient fibroblast line , which have metabolic phenotypes similar to our T8993G iPSC lines . Although they did not differentiate patient neurons from iPSCs , by using somatic cell nuclear transfer ( SCNT ) technology , they replaced the mutant mtDNA and generated corrected pluripotent stem cells , which had a normal metabolic profile , proving that the metabolic phenotypes observed in the patient iPSCs are due to mtDNA mutation . Interestingly , only T8993G neurons showed a significant ATP shortage; in contrast , T8993G iPSCs and NPCs had slightly lower but comparable ATP levels to controls ( Figure 3A ) . The ATP level in MILS neurons dropped to ~73% of healthy control neurons , and , consistently , phosphorylation of AMPK T172 , an indicator of ATP shortage , and its substrate ACC was significantly increased in T8993G neurons but not in T8993G NPCs , iPSCs and fibroblasts ( Figure 3A ) . Similarly , oligomycin , an ATP synthase inhibitor , dramatically reduced ATP levels in wild-type neurons but less significantly in NPCs; and lactate levels increased significantly in NPCs but not neurons ( Figure 3B ) . These results are consistent with the fact that neurons mainly rely on mitochondria for energy production . Immunoblot analysis of representative enzymes in the glycolysis and TCA pathways , and mitochondrial respiratory complexes showed no major differences between T8993G and BJ control NPCs or neurons ( Figure 3C ) . Strikingly , however , neurons did not express detectable hexokinase ( HK2 ) or lactate dehydrogenase ( LDHA ) proteins , the two key enzymes supporting aerobic glycolysis ( DeBerardinis and Thompson , 2012; Dang , 2012 ) . Consistently , the production of lactate in wild type neurons was ~10 fold less than in NPCs ( Figure 3D ) . Presumably , in iPSCs and NPCs , the increased production of lactate allows more NADH to be recycled to NAD+ , which is required for the conversion of glyceraldehyde 3-phosphate into 1 , 3-bisphosphoglycerate , and results in production of more glycolytic ATP . Without LDHA and HK2 , neurons appear unable to compensate for the mitochondrial ATP deficiency through aerobic glycolysis . Moreover , the data also argue that even when short of energy , neurons cannot turn on aerobic glycolysis , at least in MILS neurons . In spite of ATP deficiency in T8993G neurons , we did not detect enhanced autophagy ( data not shown ) , probably due to mTOR activation , which suppresses autophagy . Moreover , there was no marked increase in mitochondrial mass in T8993G neurons ( Figure 3C ) , indicating that mitochondrial biogenesis was not deployed to compensate for cellular energy deficiency . 10 . 7554/eLife . 13378 . 018Figure 3 . Shutoff of aerobic glycolysis during neuronal differentiation exposes mitochondrial ATP synthesis deficiency in T8993G MILS neurons . ( A ) Relative ATP level of T8993G compared to healthy control ( BJ , H9 hESC ) in iPSCs , NPCs and neurons . The relative percentage of ATP levels in T8993G was calculated by comparing to the mean of control cells respectively . Bars are mean ± SD , n=3 . *p<0 . 05 . Calculated by two-tailed t-test . Immunoblot analysis of AMPK Thr172 and ACC Ser79 phosphorylation in cell lysates prepared from primary fibroblasts , iPSCs , NPCs and neurons . ( B ) Cellular ATP level and secreted lactate from H9 NPCs and neurons treated with DMSO and oligomycin for 6 hr . The relative percentage of ATP levels was calculated by comparing to the mean of DMSO-treated cells respectively . Bars are mean ± SD , n=3 . ( C ) Immunoblot analysis of representative enzymes in glycolysis , TCA and mitochondrial respiratory complexes in BJ and T8993G NPCs and neurons . 20 µg protein lysate from each sample were loaded for SDS-PAGE . ( D ) Measurement of lactate secreted by NPCs and neurons derived from human BJ iPSCs at 3 weeks . NPC and differentiated neurons at 3 weeks were incubated in fresh medium for 12 hr , and lactate in the medium is quantified . Bars represent mean ± SD of the absolute concentration of lactate after normalized to protein content . n=3 . All the experiments were repeated at least three times . ( see associated Figure 3—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 01810 . 7554/eLife . 13378 . 019Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 019 T8993G neurons showed finer neuronal fibers than control neurons , and bead-like structures along the T8993G axons were more common ( Figure 4A and 4B ) . Neurons at 3 and 8-weeks of differentiation retained the homoplasmic T8993G mutation of the parental GM13411 fibroblasts ( Figure 4C ) . The bead-like structures , known as neuritic beading , are focal swellings in the axons and dendrites of neurons , which occur upon intracellular ATP decrease ( Takeuchi et al . , 2005 ) . Mitochondrial dysfunction increases neuronal vulnerability to form neuritic beading ( Greenwood et al . , 2007 ) . To challenge the MILS neurons with energy demanding stress , we used a glutamate toxicity assay . Glutamate overdosage is toxic to neurons through two mechanisms - excessive stimulation of neuronal activity and non-receptor oxidative toxicity ( Rothman , 1985; Murphy et al . , 1989 ) . ATP deficiency is the primary trigger for neuronal toxicity caused by glutamate overdosage ( Nicholls et al . , 2007 ) . T8993G neurons were hypersensitive to increased levels of extracellular glutamate ( Figure 4D and 4E ) . In neurons expressing GFP driven by the neuron-specific DCX promoter to highlight neuronal structures , a 6-h treatment with 100 µM glutamate abolished neuronal processes in T8993G neurons , whereas the neuronal fibers from control neurons , differentiated from BJ iPSCs or H9 hESCs , remained intact , and could tolerate up to 300 µM glutamate ( Figure 4D and 4H ) . After a 3-h treatment , neuronal fibers , containing multiple neuritic beadings , were already apparent in T8993G neurons ( Figure 4F ) . Consistently , neuronal ATP levels dropped during glutamate treatment ( Figure 4G ) . To confirm this observation , we used oligomycin , an ATP synthase inhibitor to mimic the defect caused by the ATP6 T8993G mutation . By measuring the basal OCRs with different concentration of oligomycin , 40 nM was found to partially inhibit ATP synthase ( Figure 4—figure supplement 1 ) . Wild type neurons treated with 40 nM oligomycin became sensitive to 100 µM glutamate similar to MILS neurons ( Figure 4H and 4I ) , and neuronal ATP levels fell consistent with their sensitivities to glutamate ( Figure 4J ) . 10 . 7554/eLife . 13378 . 020Figure 4 . Degenerative phenotype of T8993G MILS neurons . ( A ) Phase contrast photo of T8993G and BJ neurons at 8 weeks of differentiation . Scale bar , 20µm . ( B ) The percentage of neuronal processes containing neuritic beads was quantified by counting 20 neuronal processes for T8993G , BJ and H9 neurons as control . Bars are mean ± SD , n=3 . *p<0 . 05 calculated by two-tailed t-test . ( C ) T8993G neurons differentiated at 3 and 8 weeks still retained the original high T8893G mutation load as confirmed by PCR and Sma I digestion . DNA products were separated on an agarose gel by electrophoresis . ( D ) Glutamate-induced toxicity test . Eight-week T8993G and BJ neurons containing DCX promoter-driven GFP were treated with 100 µM glutamate in neuron growth medium . ( E ) To quantify the extent of neuronal process collapse , the number of discernible neuronal processes in a fixed photo area ( 350x350 pixel ) were counted at time points of 0 , 3 , and 6 hr . Bars are mean ± SD , n=3 . *p<0 . 05 calculated by two-tailed t-test . ( F ) Neuritic beads , indicated by white arrows , formed along the axons . ( G ) Cellular ATP of 8-week T8993G and BJ neurons treated with 100 µM glutamate . The relative percentage of ATP level was calculated by comparing to the mean of 0 hr cells respectively . Bars are mean ± SD , n=3 . ( H ) BJ neurons containing DCX promoter-driven GFP were treated with oligomycin and glutamate in neuron growth medium for 6 hr . ( I ) Quantification of the extent of neuronal process collapse . ( J ) Measurement of ATP level . The relative percentage of ATP level was calculated by comparing to the mean of untreated BJ neurons respectively . Bars are mean ± SD , n=3 . All the experiments were repeated at least three times . ( see associated Figure 4—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02010 . 7554/eLife . 13378 . 021Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02110 . 7554/eLife . 13378 . 022Figure 4—figure supplement 1 . Oxygen consumption rate ( OCR ) analysis by Seahorse extracellular flux analyzer on neurons treated with Oligomycin . By measuring the basal OCRs with different concentration of oligomycin , 40 nM was found to partially inhibit ATP synthaseDOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02210 . 7554/eLife . 13378 . 023Figure 4—figure supplement 2 . Effect of AICAR on neuron differentiation . AICAR ( 2 mM ) treatment started from day 2 of BJ neuron differentiation , and DMSO as control . The photo shows neurons after 6 days of treatment . ( B ) AICAR ( 2 mM ) was added to 6-week BJ neurons . Two days of AICAR treatment led to extensive cell death . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02310 . 7554/eLife . 13378 . 024Figure 4—figure supplement 3 . Effect of AICAR on 6-week differentiated neurons . AICAR ( 2 mM ) was added to 6-week BJ neurons . Two days of AICAR treatment led to extensive cell death . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 024 Besides low ATP itself , AMPK activation may also contribute to the deleterious effect of the T8993G mutation . We found that AICAR treatment to activate AMPK from the start of neuronal differentiation led to collapse of neuronal extensions ( Figure 4—figure supplement 2 ) , while treatment of already differentiated neurons with AICAR led to significant cell death ( Figure 4—figure supplement 3 ) . This is consistent with the finding that activation of AMPK suppresses axon initiation and neuronal polarization by phosphorylation of KIF5 , the motor protein for the kinesin light chain ( Amato et al . , 2011 ) . However , it should be noted that the intensity of AMPK T172 phosphorylation triggered by AICAR was stronger in BJ and H9 neurons than in T8993G neurons ( not shown ) . The ATP6 T8993G mutation , similar to oligomycin treatment , hampers proton intake through ATP synthase . Like oligomycin-treated healthy control neurons , neurons differentiated from independent T8993G NPC lines all showed significantly increased ribosomal S6 and S6K phosphorylation compared to BJ and H9 , while T8993G NPCs did not show increased S6 and S6K phosphorylation ( Figure 5A ) . Six-h rapamycin treatment of T8993G neurons increased the ATP level by ~23% compared to those treated with DMSO as control; and consistently , AMPK T172 phosphorylation also decreased ( Figure 5B ) . In spite of significantly lower cellular ATP levels and activated AMPK , the rate of protein synthesis in T8993G neurons was still about 92% of BJ control neurons ( Figure 5C ) . Rapamycin treatment also helped T8993G neurons cope with the stress of glutamate treatment; the neural fibers of T8993G neurons treated with rapamycin sustained a 6 hr 100 µM glutamate treatment ( Figure 5D ) . 10 . 7554/eLife . 13378 . 025Figure 5 . Rapamycin treatment alleviates ATP deficiency and aberrant AMPK activation in T8993G MILS neurons . ( A ) Immunoblot analysis of phosphorylation of ribosomal S6 , S6K and mTOR in 3-week neurons and NPCs . ( B ) Effect of rapamycin on ATP level was examined in T8993G neurons . Five-week T8993G neurons were treated with rapamycin ( 20 nM ) and DMSO for 6 hr . The relative ATP levels of rapamycin-treated neurons were calculated as a percentage compared to the mean of DMSO-treated T8993G neurons . Bars are mean ± SD , n=3 . *p<0 . 05 calculated by two-tailed t-test . AMPK Thr172 phosphorylation was examined by immunoblot analysis and quantified . ( C ) Five-week neurons differentiated from BJ and T8993G NPCs were used to measure protein synthesis rate . Protein synthesis are pulsed for 2 hr with 35S-Cys/Met . 35S incorporation into protein were quantified and normalized to the total protein . Data are mean ± SD , n=3 . ( D ) The effect of rapamycin on glutamate-induced toxicity test . Eight-week T8993G neurons containing DCX promoter-driven GFP were treated with 100 µM glutamate in neuron growth medium with rapamycin ( 20 nM ) or DMSO . To quantify the extent of neuronal process collapse , the number of discernible neuronal processes in a fixed photo area ( 350x350 pixel ) were counted at time points of 0 hr and 6 hr . Bar are mean ± SD , n=3 . *p<0 . 05 calculated by two-tailed t-test . ( E ) Immunoblot analysis of phosphorylation of ribosomal AMPK , S6 , S6K and mTOR in 3-week neurons and NPCs depleting of ATP5A1 . Control ( Ctl ) lysate were from cells infected with scramble shRNA . ( F ) Three-weeks ATP5A1-depleting BJ neurons containing DCX promoter-driven GFP . Neurons were infected with lenti-shRNA ATP5A1 and scramble shRNA from day 2 of differentiation . Relative ATP levels were quantified after normalized to protein content . Data are mean ± SD , n=3 . ( G ) Three-weeks ATP5A1-depleted BJ neurons containing DCX promoter-driven GFP treated with DMSO , rapamycin and cycloheximide ( 200 ng/ml ) for 12 hr . ( H ) Quantification of beaded neuronal process . ( I ) Measurement of ATP level . All the experiments were repeated at least three times . ( see associated Figure 5—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02510 . 7554/eLife . 13378 . 026Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 026 To further confirm the therapeutic effect of rapamycin , we established another model of ATP synthase deficiency using shRNA-mediated knockdown of ATP5A1 , a nuclearly-encoded key component of ATP synthase . ATP5A1 deficiency is also known to cause fatal neonatal mitochondrial encephalopathy ( Jonckheere et al . , 2013 ) . Consistent with our previous observations , ATP5A1-depleted neurons but not NPCs showed increased S6 and S6K phosphorylation similar to T8993G cells ( Figure 5E ) . ATP5A1-depeleted neurons at 3-weeks of differentiation already showed a significant number of beaded neuronal processes compared to neurons infected with scramble control shRNA , and the ATP level of ATP5A-depeleted neurons was only ~60% of the neurons transduced with scramble control shRNA ( Figure 5F ) . Twelve-hr treatment of rapamycin increased ATP level by ~26% and greatly decreased the number of beaded neuronal processes ( Figure 5G , 5H and 5I ) . Cycloheximide , a protein synthesis inhibitor , also had effects comparable to rapamycin in preventing beading ( Figure 5G ) . These data suggest that rapamycin treatment has the potential to benefit Leigh or NARP syndrome patients by counteracting energy deficiency . To understand the mechanism through which long term treatment with oligomycin or MILS ATPase deficiency results in elevated mTORC1 activity , we defined the metabolic profiles of MILS neurons , by measuring representative glycolytic and TCA metabolites and amino acids using gas chromatography mass spectrometry ( GC-MS ) . The levels of 19 amino acids were quantified; we were unable to measure the level of arginine due to instability . As shown in Figure 6A , the overall levels of amino acids in MILS neurons were increased compared to healthy control neurons: 14 amino acids showed significant increase; alanine , asparagine , histidine , isoleucine , leucine , lysine , methionine , phenylalanine , proline , serine and valine increased by ~40 to 80%; cysteine and threonine increased by ~100%; and , glycine increased by 340% . Pyruvate and lactate levels in MILS neurons were two-fold higher , and the TCA intermediates in MILS neurons were also increased; citrate was ~200% of that in control neurons , α-ketoglutarate was ~180% , and succinate was ~140% , whereas fumarate and malate were comparable to the control ( Figure 6B ) . These data are consistent with the ATP6 T8993G mutation defect , which decreases mitochondrial electron transport chain activity , resulting in decreased usage of TCA and glycolytic metabolites . The increased intracellular amino acids levels in MILS neurons may also be attributable to the accumulated glycolytic intermediates and a clogged TCA cycle , because the synthesis and catabolism of amino acids is linked to glycolysis and the TCA cycle; under catabolic conditions the carbon atoms of amino acids are oxidized by the TCA cycle for ATP production ( Stryer et al . , 2002 ) . Since the mTORC1 complex is a sensor of cellular nutrition , and can be activated by amino acids ( Bar-Peled and Sabatini , 2014 ) , we suspect that the observed enhancement of neuronal mTORC1 activity after prolonged mitochondrial OXPHOS inhibition is due to the accumulated amino acids or other nutrients that activate mTORC1 . 10 . 7554/eLife . 13378 . 027Figure 6 . Metabolite profiling of amino acids , TCA and glycolysis intermediates . ( A ) Metabolites measured by gas chromatography mass spectrometry ( GC-MS ) . The metabolites were extracted from 3-week T8993G and control including two BJ and one H9 neurons . Relative cellular amino acids were shown . Bar are mean ± SD , n=3 . ( B ) Metabolites of glycolysis and TCA reactions . The relative amount of metabolites in T8993G neurons was presented as percentage compared to the mean of control . Bar are mean ± SD , n=3 . ( C ) A simplified metabolic flow diagram of glycolysis and the TCA cycle . ( D , E ) 3-week BJ neurons were treated with oligomycin ( 40 nM ) and rotenone and antimycin A ( 1 µM each ) for 6 hr . Bar are mean ± SD , n=3 . *p<0 . 05 . **p<0 . 01 , calculated by two-tailed t-test . ( see associated Figure 6—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02710 . 7554/eLife . 13378 . 028Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 02810 . 7554/eLife . 13378 . 029Figure 6—figure supplement 1 . 10 µg protein lysate prepared form NPCs and 3-week neurons were separated on SDS-PAGE and blotted with respective antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 13378 . 029 To further examine the change of these metabolites during mitochondrial inhibition , we measured metabolite levels in the neurons treated with oligomycin or rotenone and antimycin A ( R&A ) for 6 hr . Similar to MILS neurons , neurons treated with these mitochondrial drugs showed higher levels of amino acids , pyruvate and lactate ( Figure 6D and 6E ) ; although there are some notable differences between mitochondrial inhibitor-treated neurons and MILS neurons . For example , the extent of glycine increase in inhibitor-treated neurons is not as pronounced as in MILS neurons . These differences may be due to the difference in mechanism or extent of mitochondrial inhibition between mitochondrial inhibitors and ATP6 T8993G mutation . The large increase of glycine in MILS neurons may be an anti-oxidant response to chronic mitochondrial oxidative stress . In the mitochondrial matrix , serine is converted to glycine catalyzed by serine hydroxymethyltransferase 2 ( SHMT2 ) , a reaction coupled with covalent linkage of tetrahydrofolate to a methylene group to form 5 , 10-methylene-tetrahydrofolate , which is subsequently used to generate 5 , 10-methenyl-tetrahydrofolate and NADPH by tetrahydrofolate dehydrogenase 2 ( MTHFD2 ) in mitochondria . The NADPH produced through this serine catabolism pathway maintains mitochondrial redox by regenerating the reduced forms of glutathione and thioredoxin ( Fan et al , 2014; Lewis et al , 2014; Martínez-Reyes I and Chandel , 2014 ) , and is critical for tumor survival ( Ye et al , 2014; DeNicola et al , 2015; Kim et al , 2015 ) . In MILS neurons , the expression of SHMT2 , MTHFD2 and other enzymes in the de novo serine synthesis pathway were significantly higher than in control healthy neurons , supporting our hypothesis ( unpublished data ) . We are currently investigating the role of this pathway in neuronal survival under mitochondrial stress . On the other hand , this accumulated glycine can be used to synthesize glutathione itself .
Our observation that in mitochondrially-defective neurons mTORC1 activity increased , rather than decreased concurrent with AMPK activation is contrary to previous reports that pharmacological disruption of mitochondrial function leads to mTORC1 inhibition due to AMPK activation by reduced energy levels ( Zoncu et al . , 2011 ) . In proliferating NPCs , we did find that mitochondrial inhibitors activated AMPK and , consistently , decreased S6 and S6K phosphorylation . Activation of mTORC1 in neurons with mitochondrial dysfunction takes several hours , and appears to be a slow response to some cumulative effect caused by mitochondria dysfunction . This paradoxical observation could be due to unknown differences in the mTORC1 complex between NPCs and neurons . We found that AKT Ser473 phosphorylation was dramatically decreased in wild type neurons compared to NPCs , as was phosphorylation of its substrate PRAS40 at Thr246 ( Figure 6—figure supplement 1 ) . PRAS40 is an inhibitor of mTORC1 signaling , and its overexpression has been shown to inhibit mTORC1 hyperactivation in Tsc2 -/- mutant cells , with AKT-mediated phosphorylation of PRAS40 preventing its inhibition of mTORC1 ( Sancak et al . , 2007; Vander Haar et al . , 2007 ) . An alternative explanation for mTORC1 activation emerged from our metabolite analysis , which showed that MILS neurons and neurons treated with mitochondrial drugs had significantly higher levels of amino acids , which can activate mTORC1 . Indeed , it has been shown that the basal activity of S6 kinase rises progressively with increased concentration of medium amino acids in a nearly linear fashion; and at a 2-fold increased concentration , S6 kinase activity is close to maximal and no longer shows further activation by growth factors ( Hara et al , 1998 ) . Therefore , the increased amino acid levels in MILS neurons or mitochondria-inhibitor treated neurons are significant , and may account for the elevated mTORC1 activity . Certainly , there are other possible mechanisms; for instance , the effect might be due to decreased S6 and S6K phosphatase activity . Similar paradoxical observations have been described in previous studies ( Ng et al . , 2012; Nakai et al . , 2015 ) . In an extracellular matrix detachment model , Ng et al . ( 2012 ) demonstrated that after detachment AMPK is activated in both K-Ras V12-transformed and non-transformed mouse embryonic fibroblasts; interestingly , mTORC1 activity only decreased , in an AMPK-dependent manner , in transformed but not non-transformed fibroblasts . They further showed that this AMPK-mediated mTORC1 inhibition decreased protein synthesis , thus preserving ATP in the detached transformed fibroblasts and delaying cell death . As in our study , rapamycin or cycloheximide treatment had a significant cell survival effects in their model . Notably , in the transformed fibroblasts , the level of AKT Ser473 phosphorylation was markedly stronger than in non-transformed cells , reminiscent of the situation in NPCs and neurons . When neurons were treated with rotenone and antimycin-A , protein synthesis initially dropped to ~ 36% but gradually recovered to almost 80% , correlating with the increase in S6 and S6K phosphorylation , in spite of a continuous drop in ATP levels . In the MILS neurons , S6 and S6K phosphorylation were significantly higher than in BJ and H9 control neurons . Neuronal energy deficiency caused by mitochondrial dysfunction occurs in the presence of sufficient oxygen and ample nutrition , a situation that cells in vivo rarely encounter , and there is probably no selection pressure to evolve a proper response program . We believe that the recovery of protein synthesis in such situations is an improper response that further aggravates energy deficiency . Therefore , manipulating protein synthesis to match ATP production is beneficial for neurons . Hints exist in the work of others . Overexpression of 4E-BP , a negative regulator of protein synthesis , rescues the neuronal degeneration observed in Pink1 and Parkin fly mutants with mitochondrial defects ( Tain et al . , 2009 ) ; in light of our finding , an alternative explanation is that reduction of protein synthesis is energetically beneficial for neurons with mitochondrial dysfunction . Strikingly , pathogenic LRRK2 mutation in Parkinson's disease has recently been found to induce a large increase in protein synthesis in a Drosophila model , and , when treated with a low-dose of the anisomycin protein synthesis inhibitor , the locomotor deficits and dopamine neuron loss in mutant LRRK2 transgenic flies were rescued ( Martin et al . , 2014 ) . Notably , a recent genetic screen in yeast also revealed that downregulation of protein synthesis could rescue the growth of mutant cells with mitochondrial defects ( Wang and Chen , 2015 ) . As discussed in the Introduction , a cellular load of ATP6 T8993G greater than 90% causes Leigh syndrome , a severe form of infantile neurodegeneration , whereas a 70~90% load causes a less severe neurological disease called NARP syndrome with symptoms , such as neuropathy , ataxia , and retinitis pigmentosa , gradually developing with age . In contrast , carriers with a ~50% load are generally normal , only developing late-onset cone-rod dystrophy in their forties ( Porto et al . , 2001 ) . In a cybrid study , where patient platelets containing the T8993G mtDNA mutation were fused to human osteosarcoma cells devoid of mtDNA , mitochondrial ATP production was found to be negatively correlated with the mutation load in a nearly linear fashion ( Mattiazzi et al . , 2004 ) . In this study , we found that rapamycin treatment of T8993G neurons increased ATP level and improved their resistance to glutamate-induced neuronal fiber collapse , a process caused by decreased intracellular ATP ( Takeuchi et al . , 2005 ) . Therefore , a ~20% increase in ATP , the amount saved by rapamycin , which may seem small , could indeed have a significant effect on neuronal survival in Leigh or NARP syndrome patients . Leigh syndrome affects 1 in 40 , 000 newborns in the United States ( Darin et al . , 2001 ) . In one fourth of the cases , these mutations occur in mitochondrial DNA ( Finsterer , 2008; Pinto and Moraes , 2014 ) . The neurological degenerative phenotypes are usually apparent in newborns , but , currently , no effective therapy is available . Johnson et al . ( 2013b ) reported a dramatic therapeutic effect of rapamycin on a mouse model of Leigh syndrome . We also observed beneficial effects of rapamycin in human neurons treated with mitochondria inhibitors and in an iPSC-based disease model of maternally inherited Leigh syndrome . However , potential negative effects of rapamycin on neuronal development should not be neglected . A previous study found that focal infusion of rapamycin into dorsal hippocampus blocks axon fiber sprouting , but it should be noted that such a treatment is unable to reverse already established axon organization ( Buckmaster et al . , 2009 ) . Therefore , we suggest that long-term usage of rapamycin for newborns should be considered with caution , but could be tried on older NARP syndrome patients to delay disease progression , and short term rapamycin usage could be considered for younger Leigh patients in emergency situations , such as fever , which often drastically and irreversibly worsens the neurodegenerative symptoms ( Uziel et al . , 1997 ) . Besides these hereditary mitochondrial diseases , mitochondrial dysfunctions are frequently observed in neurodegenerative diseases ( Lin and Beal , 2006 ) . Elevated p70 S6 kinase activity has been documented in the brain tissue of Alzheimer's disease patients ( An et al . , 2003 ) . Therefore , a mild reduction in protein synthesis may be useful in other neurodegenerative diseases by increasing ATP level and simultaneously decreasing the workload of protein folding systems . We emphasize that , whether or to what extent , our observations from a cell culture-based mitochondrial disease model reflect the in vivo situation needs further investigation using animal models with various mitochondrial deficiencies . In particular , there are some critical questions that need to be addressed under in vivo condition , e . g . , the role of ATP deficiency in the neurodegenerative process; whether a reduction in protein synthesis can help balance neuronal energy expenditure and thereby delay neurodegeneration . Brain tissue is composed of mixed cell types , neurons and glial cells , which have distinct metabolic profiles and different responses to energy deficiency ( Bélanger et al , 2011; Almeida et al 2001 ) ; therefore , to study the changes in ATP levels in situ and in specific cell types , an effective approach preferably at the single cell level is required . An engineered ATP fluorescent biosensor is available for cell culture systems ( Tantama et al , 2013 ) , which could be adapted and introduced into animal models .
Cells were fixed in cold 4% paraformaldehyde in PBS for 10 min . iPSCs , NPCs and neurons were permeabilized at room temperature for 15 min in 0 . 2% TritonX-100 in PBS . Samples were blocked in 5% BSA with 0 . 1% Tween 20 for 30 min at room temperature . The primary antibodies and dilutions used were: goat anti-SOX2 ( Santa Cruz ) , 1:200; mouse anti-human Nestin ( Chemicon ) , 1:200; rabbit anti-βIII-tubulin ( Covance ) , 1:200; mouse anti-βIII-tubulin ( Covance ) , 1:200; rabbit anti-cow-GFAP ( Dako ) 1:200; mouse anti-MAP2AB ( Sigma ) , 1:200; secondary antibodies were Alexa donkey 488 and 568 anti-mouse , rabbit and goat ( Invitrogen ) , used at 1:1000 . Nuclear stainings were done with Hoechst ( Invitrogen ) . Cell lysates were prepared with lysis buffer containing 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 1% Triton X-100 , 2 . 5 mM sodium pyrophosphate , 1 mM β-glycerophosphate , 1 mM Na3VO4 , 1 μg/ml leupeptin . 1 mM PMSF was added immediately prior to use . The protein concentration was measured by DC protein assay ( Bio-Rad ) . The primary antibodies and dilutions were used as follow: glycolysis antibody sampler kits ( #8337&12866 , Cell Signalling ) used at 1:1000; the OXPHOS human WB Antibody cocktail , anti-CS and IDH2 used at 1:1000 ( Abcam ) ; anti-HSP60 and SUCLA2 ( Santa Cruz ) ; anti phospho AMPK T172 and AMPK , anti phospho ribosomal S6 Ser235/236 and S6; anti phospho mTOR Ser2481 and mTOR used at 1:1000 ( Cell Signaling ) . Generally , 20 μg of protein lysate were loaded on SDS-PAGE gel . Immunoblotting quantification was carried out on an Odyssey Imager ( Licor ) . The MILS patient and BJ fibroblasts were reprogrammed into iPSCs using the standard method described by Takahashi et al . ( 2007 ) . GM13411 primary fibroblasts derived from a male MILS syndrome patient were obtained from the Coriell Institute for Medical Research . BJ fibroblasts were from the ATCC . Fibroblasts were cultured in DMEM media supplemented with 10% FBS , 1x Glutamax , 5 ng/ml FGF2 . Fibroblasts from one well of a six-well dish were infected with retrovirus expressing OCT4 , SOX2 , KLF4 , and MYC , and after 2 days , were split onto a 10 cm plate containing 1 million mitotically-inactivated mouse embryonic fibroblasts ( mEFs ) . The growth medium was switched to DMEM/F12 supplemented with 20% knockout serum replacement , 1 mM L-glutamine , 0 . 1 mM non-essential amino acids , β-mercaptoethanol and 10 ng ml−1 FGF2 for the 21–28 days of reprogramming . hiPSC colonies were picked and cultured onto 24-well plates coated with inactivated mEFs . hiPSCs were split through mechanic passaging with a glass pipet at early passages , while at higher passages , hiPSC could be grown on Matrigel and enzymatically digested with dispase . Karyotyping analysis was performed by Cell Line Genetics ( Wisconsin , MD ) . The establishment of neural progenitor cells from iPSCs and neuronal differentiation were performed as previously described ( Brennand et al . , 2011 ) . hESC and iPSC lines were mainly maintained on Matrigel using mTeSR1 . For embryoid body formation , hESC and iPSC lines were cultured on a mitotically-inactive mouse embryonic fibroblast feeder layer in hESC medium , DMEM/F12 supplemented with 20% knockout serum replacement , 1 mM L-glutamine , 0 . 1 mM non-essential amino acids , β-mercaptoethanol and 10 ng ml−1 bFGF . Neural differentiation was induced as follows: hESCs grown on inactivated mEFs were fed N2/B27 medium without retinoic acid for 2 days , and then , colonies were lifted with collagenase treatment for 1 hr at 37°C . The cell clumps were then transferred to ultra-low attachment plates . After growth in suspension for 1 week in N2/B27 medium , aggregates form embryoid bodies , which were then transferred onto polyornithine ( PORN ) /laminin-coated plates and developed into neural rosettes in N2/B27 medium . After another week , colonies showing mature neural rosettes with biopolar neuroprogenitor cells migrating out from the colony border , were picked under a dissecting microscope , digested with accutase for 10 min at 37°C and then cultured on polyornithine ( PORN ) /laminin-coated plates in N2/B27 medium supplemented with FGF2 . For neuron differentiations , neuroprogenitor cells were dissociated with accutase and plated in neural differentiation media , 500 ml DMEM/F12 GlutaMAXTM , 1x N2 , 1X B27+RA , 20 ng/ml BDNF ( Peprotech ) , 20 ng/ml GDNF ( Peprotech ) , 200 nM ascorbic acid ( Sigma ) , 1 mM dibutyrl-cyclicAMP ( Sigma ) onto PORN/Laminin-coated plates . For one well of a 6-well plate , 200 , 000 cells/well were seeded; for one well of a 12-well plate , 80 , 000 cells were seeded . Neurons can be maintained for 3 months in a 5% CO2 37°C incubator . Total RNA was isolated using RNeasy kit ( QIAGEN ) . 500 ng of total RNA from each sample was used for cDNA synthesis by MMLV reverse transcriptase; and quantitative real-time polymerase chain reaction ( PCR ) was performed with SYBR Green Master Mix on ABI 7000 cycler ( Applied Biosystems ) and normalized to β-actin . Primer sequences were referred from qPCR primerDepot ( http://primerdepot . nci . nih . gov/ ) . Mitochondrial membrane potential was measured by flow cytometry of iPSCs , NPCs and neurons stained with TMRE ( Invitrogen ) . Cells were dissociated with accutase , spun down at 350 g for 10 min , and then , resuspended in PBS with 2% bovine serum albumin ( BSA ) loaded with 10 nM TMRE for 15 min at 37°C . The cells were washed again , filtered through a 250-µM nylon sieve and kept in PBS on ice . The TMRE signal was quantified using the FL2 channel of a Becton Dickinson FACScan . Each set of measurements included a control sample pretreated for 30 min with 20 μM of CCCP , a mitochondrial uncoupler , to abolish mitochondria membrane potential . Data were analyzed using FloJo; and the mean value was used to compare the mitochondria potential between BJ and T8993G cells . Similarly , cellular ROS level was measured by 10 μM CM-H2DCFDA ( Invitrogen ) staining for 30 min and detected in the FL1 channel . The OCR of iPSC , NPCs and neurons grown in Seahorse plates was measured using an extracellular Flux Analyzer ( Seahorse Bioscience ) , following the manufacturer’s instructions . After the measurement , cells were lysed in 60–100 µl lysis buffer with two “freeze and thaw” cycles on dry ice . Protein concentrations were determined by DC protein assay ( Bio-Rad ) . The OCR values were normalized by protein mass . For measurement of cellular ATP content , neurons were lysed directly on plates with protein extract buffer by two freeze-and-thaw cycles in dry ice . The ATP content was quantified by CellTiter-Glo Luminescent Cell Viability/ATP Assay kit ( Promega ) , and normalized by protein content measured by DC protein assay ( Bio-Rad ) . For measurement of secreted lactate levels , medium from iPSCs , NPCs and neurons was freshly changed and collected after 12 hr , and cells were frozen on the plate and lysed by two freeze-and-thaw cycles on dry ice . Medium lactate was measured using the Lactate Assay kit ( BioVision ) and normalized by total protein content . To avoid the medium change effect on mTOR signaling , 10 μl ( 100 μCi ) Express 35S protein labelling mix ( Perkin Elmer ) were added into NPCs or neurons grown in 12 well plate containing 1 ml NPC or neuronal growth medium . Cells were labelled for 2 hr and lysed on plate after two times of PBS wash . Twenty-five μl lysate were mixed with 5 μl 100% TCA , incubated on ice for 30 min , and spotted on Whatman 3MM filter paper . The filter papers were washed twice with cold 5% TCA and air-dried . The radioactivity was determined by scintillation counting . Neurons were grown in a 6-well plate . After growth in fresh medium for 12 hr , cells were washed quickly 3 times with cold PBS , and 0 . 45 ml cold methanol ( 50% v/v in water with 20 µM L-norvaline as internal standard ) was added to each well . Culture plates were transferred to dry ice for 30 min . After thawing on ice , the methanol extract was transferred to a microcentrifuge tube . Chloroform ( 0 . 225 ml ) was added , the tubes were vortexed and centrifuged at 10 , 000 g for 5 min at 4°C . The upper layer was dried in a centrifugal evaporator and derivatized with 30 µl O-isobutylhydroxylamine hydrochloride ( 20 mg/ml in pyridine , TCI ) for 20 min at 80°C , followed by 30 µl N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide ( Sigma ) for 60 min at 80°C . After cooling , the derivatization mixture was transferred to an autosampler vial for analysis . GC-MS analysis was performed in the Cancer Metabolism core at the Sanford-Burnham Medical Research Institute ( La Jolla , California ) . More details including the parameters of machine settings can be found in the publication from the center ( Scott et al . , 2011 ) . Comparisons were done by Student's t-test . Statistical analyses were performed using GraphPad Prism . | Living cells need to maintain an optimal balance between making new proteins and destroying older ones . Building proteins requires a supply of nutrients and appropriate levels of energy , and mammalian cells rely on a protein called mTOR to sense both nutrient availability and energy levels . Nutrients activate mTOR signaling to promote protein synthesis . In contrast , a lack of nutrients and low energy levels inhibit mTOR , which slows down protein synthesis to help the cell to conserve vital resources . The balance between protein synthesis and degradation is often perturbed in diseases that involve the progressive loss of nerve cells , and a drug called rapamycin – which inhibits mTOR signalling – can help treat this neurodegeneration in mice . Neurodegenerative diseases are also often linked to problems with the cellular structures called mitochondria that provide the cell with energy in the form of the chemical ATP . Previous research suggests that abnormal mitochondrial activity and energy deficiency could be a critical step that leads to neuron death in neurodegeneration . So far , the effect of rapamycin on energy deficiency in neurons has not been explored in detail . Zheng , Boyer et al . have now tested the therapeutic potential of rapamycin in a genetic disease called maternally inherited Leigh syndrome in which children suffer from severe neurodegeneration due to defects in their mitochondria . The experiments made use of neurons that could be grown in the laboratory and which faithfully mimicked the problems observed in maternally inherited Leigh syndrome patients . In some experiments , healthy neurons were treated with chemicals that inhibit ATP production . In other experiments , cells collected from a maternally inherited Leigh syndrome patient were coaxed into becoming neurons . Signaling via mTOR was enhanced in both kinds of neurons . Zheng , Boyer et al . then treated the defective neurons with rapamycin , which led to a significant rise in ATP levels . The production of proteins also slowed down . This could explain the observed rise in ATP levels , as making proteins consumes a lot of energy . Zheng , Boyer et al . propose that a mild reduction in protein synthesis may have the potential to treat neurodegeneration caused by defective mitochondria . Further work is needed to extend this analysis to animal models of neurodegenerative diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Alleviation of neuronal energy deficiency by mTOR inhibition as a treatment for mitochondria-related neurodegeneration |
Epigenetic modifiers are an emerging class of anti-tumor drugs , potent in multiple cancer contexts . Their effect on spontaneously developing autoimmune diseases has been little explored . We report that a short treatment with I-BET151 , a small-molecule inhibitor of a family of bromodomain-containing transcriptional regulators , irreversibly suppressed development of type-1 diabetes in NOD mice . The inhibitor could prevent or clear insulitis , but had minimal influence on the transcriptomes of infiltrating and circulating T cells . Rather , it induced pancreatic macrophages to adopt an anti-inflammatory phenotype , impacting the NF-κB pathway in particular . I-BET151 also elicited regeneration of islet β-cells , inducing proliferation and expression of genes encoding transcription factors key to β-cell differentiation/function . The effect on β cells did not require T cell infiltration of the islets . Thus , treatment with I-BET151 achieves a ‘combination therapy’ currently advocated by many diabetes investigators , operating by a novel mechanism that coincidentally dampens islet inflammation and enhances β-cell regeneration .
Acetylation of lysine residues on histones and non-histone proteins is an important epigenetic modification of chromatin ( Kouzarides , 2000 ) . Multiple ‘writers’ , ‘erasers’ , and ‘readers’ of this modification have been identified: histone acetyltransferases ( HATs ) that introduce acetyl groups , histone deacetylases ( HDACs ) that remove them , and bromodomain ( BRD ) -containing proteins that specifically recognize them . Chromatin acetylation impacts multiple fundamental cellular processes , and its dysregulation has been linked to a variety of disease states , notably various cancers ( Dawson and Kouzarides , 2012 ) . Not surprisingly , then , drugs that modulate the activities of HATs or HDACs or , most recently , that block acetyl-lysine:BRD interactions are under active development in the oncology field . BRDs , conserved from yeast to humans , are domains of approximately 110 amino-acids that recognize acetylation marks on histones ( primarily H3 and H4 ) and certain non-histone proteins ( e . g . , the transcription factor , NF-κB ) , and serve as scaffolds for the assembly of multi-protein complexes that regulate transcription ( Dawson et al . , 2011; Prinjha et al . , 2012 ) . The BET subfamily of BRD-containing proteins ( BRDs 2 , 3 , 4 and T ) is distinguished as having tandem bromodomains followed by an ‘extra-terminal’ domain . One of its members , Brd4 , is critical for both ‘bookmarking’ transcribed loci post-mitotically ( Zhao et al . , 2011 ) and surmounting RNA polymerase pausing downstream of transcription initiation ( Jang et al . , 2005; Hargreaves et al . , 2009; Anand et al . , 2013; Patel et al . , 2013 ) . Recently , small-molecule inhibitors of BET proteins , for example , JQ1 and I-BET , were found to be effective inhibitors of multiple types of mouse tumors , including a NUT midline carcinoma , leukemias , lymphomas and multiple myeloma ( Filippakopoulos et al . , 2010; Dawson et al . , 2011; Delmore et al . , 2011; Zuber et al . , 2011 ) . A major , but not the unique , focus of inhibition was the Myc pathway ( Delmore et al . , 2011; Mertz et al . , 2011; Zuber et al . , 2011; Lockwood et al . , 2012 ) . In addition , BET-protein inhibitors could prevent or reverse endotoxic shock induced by systemic injection of bacterial lipopolysaccharide ( LPS ) ( Nicodeme et al . , 2010; Seal et al . , 2012; Belkina et al . , 2013 ) . The primary cellular focus of action was macrophages , and genes induced by the transcription factor NF-κB were key molecular targets ( Nicodeme et al . , 2010; Belkina et al . , 2013 ) . Given several recent successes at transposing drugs developed for cancer therapy to the context of autoimmunity , it was logical to explore the effect of BET-protein inhibitors on autoimmune disease . We wondered how they might impact type-1 diabetes ( T1D ) , hallmarked by specific destruction of the insulin-producing β cells of the pancreatic islets ( Bluestone et al . , 2010 ) . NOD mice , the ‘gold standard’ T1D model ( Anderson and Bluestone , 2005 ) , spontaneously and universally develop insulitis at 4–6 weeks of age , while overt diabetes manifests in a subset of individuals beginning from 12–15 weeks , depending on the particular colony . NOD diabetes is primarily a T-cell-mediated disease , but other immune cells—such as B cells , natural killer cells , macrophages ( MFs ) and dendritic cells ( DCs ) —also play significant roles . We demonstrate that a punctual , 2-week , treatment of early- or late-stage prediabetic NOD mice with I-BET151 affords long-term protection from diabetes . Mechanistic dissection of this effect revealed important drug influences on both MFs and β cells , in particular on the NF-κB pathway . On the basis of these findings , we argue that epigenetic modifiers are an exciting , emerging option for therapeutic intervention in autoimmune diabetes .
T1D progresses through identifiable phases , which are differentially sensitive to therapeutic intervention ( Bluestone et al . , 2010 ) . Therefore , we treated NOD mice with the BET-protein inhibitor , I-BET151 ( GSK1210151A [Dawson et al . , 2011; Seal et al . , 2012] ) according to three different protocols: from 3–5 weeks of age ( incipient insulitis ) , from 12–14 weeks of age ( established insulitis ) , or for 2 weeks beginning within a day after diagnosis of hyperglycemia ( diabetes ) . Blood-glucose levels of insulitic mice were monitored until 30 weeks of age , after which animals in our colony generally do not progress to diabetes . I-BET151 prevented diabetes development , no matter whether the treated cohort had incipient ( Figure 1A ) or established ( Figure 1B ) insulitis . However , the long-term protection afforded by a 2-week treatment of pre-diabetic mice was only rarely observed with recent-onset diabetic animals . Just after diagnosis , individuals were given a subcutaneous insulin implant , which lowers blood-glucose levels to the normal range within 2 days , where they remain for only about 7 days in the absence of further insulin supplementation ( Figure 1C , upper and right panels ) . Normoglycemia was significantly prolonged in mice treated for 2 weeks with I-BET151; but , upon drug removal , hyperglycemia rapidly ensued in most animals ( Figure 1C , lower and right panels ) . The lack of disease reversal under these conditions suggests that β-cell destruction had proceeded to the point that dampening the autoinflammatory attack was not enough to stem hyperglycemia . However , there was prolonged protection from diabetes in a few cases , suggesting that it might prove worthwhile to explore additional treatment designs in future studies . 10 . 7554/eLife . 04631 . 003Figure 1 . I-BET151 inhibits diabetes and insulitis in NOD mice . Female NOD mice were treated with I-BET151 in DMSO ( 10 mg/kg daily ) or just DMSO from 3–5 weeks ( A and D ) or 12–14 weeks ( B and E ) of age . ( A and B ) Pre-diabetic mice . Hyperglycemia was monitored until 30 weeks of age . n = 10 per group . Blue shading = treatment window . ( C ) Recent-onset diabetic mice . Left: Individual blood-glucose curves . An insulin pellet was implanted subcutaneously within 1 day of diabetes diagnosis ( arrow ) . 2 days later , I-BET151 ( 10 mg/kg ) or DMSO was administered daily for 2 weeks ( shaded blue ) . Right: duration of normoglycemia . n = 7 or 11 . ( D and E ) Insulitis was visualized by H&E staining of paraffin sections . Left: representative histology . Middle: insulitis scores for individual mice . Grey = peri-insulitis; black = insulitis . The asterisk indicates no insulitis in any of the sections examined . Right: summary of the proportions of intact islets for individual mice . ( F ) Left: Total CD45+ cells from the spleen ( upper panel ) or pancreatic islets ( lower panel ) from mice treated with I-BET151 or DMSO as per Figure 1B , and analyzed at 14 weeks . n = 6 . ( G ) Summary data on the major immune-cell subsets as a fraction of CD45+ cells , from the spleen ( upper panels ) or pancreas ( lower panels ) . n = 5 or 6 . p values in A and B are from Gehan-Breslow-Wilcoxon tests and in C–G are from Student's t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 003 As a first step in dissecting the mechanisms of I-BET151 action , we examined its effect on insulitis . Analogous to the protocols employed above , NOD mice were treated with I-BET151 from 3–5 or 12–14 weeks of age , and their pancreas was excised for histology at 10 weeks ( 5 weeks being too early for quantification ) or 14 weeks , respectively . Drug treatment prevented effective installation of insulitis in the young mice ( Figure 1D ) and reversed established insulitis in the older animals ( Figure 1E ) . Next , we performed flow cytometric analysis of the pancreatic infiltrate . In this and subsequent mechanistic studies , we focused mainly on the 12–14-week treatment protocol because , of the successful regimens , it better models what might eventually be applied to humans . Consistent with the histological results , fewer total leukocytes ( CD45+ cells ) were found in the pancreas , but not the spleen , of mice administered I-BET151 from 12–14 weeks of age , vis a vis vehicle-only controls ( Figure 1F ) . The drop in pancreas-infiltrating cells in animals treated with the inhibitor was equally true of all populations examined ( encompassing the major lymphoid and myeloid subsets ) as their fractional representation within the bulk CD45+ compartment appeared to be unaltered in drug- vs vehicle-treated individuals ( Figure 1G ) . Given that NOD diabetes is heavily dependent on CD4+ T cells ( Anderson and Bluestone , 2005 ) , and that a few recent reports have highlighted an influence of BET-protein inhibitors on the differentiation of T helper ( Th ) subsets in induced models of autoimmunity ( Bandukwala et al . , 2012; Mele et al . , 2013 ) , we explored the effect of I-BET151 treatment on the transcriptome of CD4+ T cells isolated from relevant sites; that is , the infiltrated pancreas , draining pancreatic lymph nodes ( PLNs ) , and control inguinal lymph nodes ( ILNs ) . Microarray analysis of gene expression revealed surprisingly little impact of the 2-week treatment protocol on any of these populations , similar to what was observed when comparing randomly shuffled datasets ( Figure 2A ) . It is possible that the above protocol missed important effects on T cells because those remaining after prolonged drug treatment were skewed for ‘survivors’ . Therefore , we also examined the transcriptomes of pancreas-infiltrating CD4+ T cells at just 12 , 24 or 48 hr after a single administration of I-BET151 . Again , minimal , background-level , differences were observed in the gene-expression profiles of drug- and vehicle-treated mice ( Figure 2B ) . 10 . 7554/eLife . 04631 . 004Figure 2 . Little impact of BET-protein inhibition on CD4+ T cells in NOD mice . ( A ) Microarray-based transcriptional profiling of TCR+CD4+ cells sorted from pancreata , pancreatic lymph nodes ( PLNs ) and inguinal lymph nodes ( ILNs ) . Comparison plot of I-BET151- and DMSO-treated mice as per Figure 1B and analyzed at 14 weeks of age . Red , transcripts increased >twofold by I-BET151; blue , transcripts >twofold decreased . ( B ) Analogous plots of TCR+CD4+ cells sorted from the pancreas of mice given a single I-BET151 ( 10 mg/kg ) or DMSO injection , and analyzed 12 , 24 or 48 hr later . ( C ) Th1 , Th2 , Th17 or Treg signatures ( see ‘Materials and methods’ ) were superimposed on volcano plots comparing the transcriptomes of TCR+CD4+ cells from the pancreas of mice treated with I-BET151 or DMSO either as per Figure 1B and analyzed at 14 weeks of age ( upper panels ) or with a single injection and analyzed 24 hr later ( lower panels ) . Purple: over-represented signature transcripts; Green: under-represented signature transcripts . ( D and E ) Proportions of Treg ( D ) or Th17 ( E ) cells within the TCR+CD4+ population in the pancreas of I-BET151- or DMSO-treated mice . Left , representative cytofluorometric dot plots; right , summary data . n = 4–5 . p values in panel C are from the Chi-squared test ( the single significant value is shown; all others were not significant ) and in D–E from Student's t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 004 Signature analysis , wherein we superimposed existing Th1 , Th2 , Th17 or Treg gene-expression signatures on p-value vs fold-change ( FC ) volcano plots , failed to reveal statistically significant skewing within the transcriptomes of pancreatic CD4+ T cells from mice administered I-BET151 vs vehicle using either the 2-week or short-term protocols , with the possible exception of a slightly weaker Treg down-signature in I-BET151-treated animals ( Figure 2C ) . Yet , we found no differences in the fraction of Tregs in the pancreatic CD4+ T cell compartment , their Foxp3 expression levels or their display of CD25 ( Figure 2D ) . Nor , in contrast to a recent report on antigen + adjuvant induced autoimmune diseases ( Mele et al . , 2013 ) , could we demonstrate differences in the Th17 population—either the fraction of CD4+ cells expressing IL-17A or its expression level ( Figure 2E ) . It remains possible , however , that parameters we did not assay ( e . g . , splicing , microRNAs ) might have been different . To obtain a broader , unbiased view of the inhibitor's effect on the NOD pancreatic infiltrate , we undertook a transcriptome analysis of the bulk CD45+ cell population . The 2-week and short-term treatment protocols both resulted in sets of over- and under-represented transcripts ( Figure 3A , B and Supplementary files 1 , 2 ) . Interestingly , the set of decreased transcripts was evident sooner than the set of increased transcripts—the former troughing already at 12 hr , the latter still rising at 48 hr ( Figure 3B ) . Taking advantage of data-sets from the Immunological Genome Project ( ImmGen; www . immgen . org ) , which has profiled gene expression for over 200 immunocyte populations , we found the over-represented transcripts to be indicative of myeloid-lineage cells , in particular tissue-resident MFs ( Figure 3C ) . While it is possible that the changes in transcript levels reflect alterations in the relative representation of myeloid cell populations , we think rather that they resulted at least partially from gene induction or repression because they were so rapid and because only a fraction of the transcript set characteristic of any particular cell-type showed an altered level of expression . Gene-set enrichment analysis ( GSEA ) also highlighted transcriptional programs characteristic of MFs , the highest enrichment values being obtained for the eicosanoid , relevant nuclear receptor and complement pathways ( Figure 3D ) . All three of these pathways have been demonstrated to play a role in the resolution of inflammation , through multiple mechanisms , including the production of anti-inflammatory mediators and suppression of T cell responses ( Bensinger and Tontonoz , 2008; Ricklin et al . , 2010; Serhan , 2011; Lone and Tasken , 2013 ) . Parallel analyses on the set of pancreatic CD45+ cell transcripts under-represented in inhibitor-treated mice showed no striking enrichment across the ImmGen data-sets ( Figure 3E ) . And GSEA did not reveal any particular pathways to be significantly enriched . 10 . 7554/eLife . 04631 . 005Figure 3 . I-BET151 treatment promotes an MF-like , anti-inflammatory transcriptional program in pancreatic CD45+ cells . ( A and B ) A volcano plot comparing the transcriptomes of pancreatic CD45+ cells from mice treated with I-BET151 or DMSO as per Figure 1B . Red: transcripts increased >twofold; blue: transcripts decreased >twofold; numbers of modulated transcripts are indicated in the corresponding color . ( B ) Analogous plots for mice given a single injection of I-BET151 ( 10 mg/kg ) or DMSO only , and analyzed 12 , 24 or 48 hr later . ( C ) Cell-type distribution of the totality of transcripts whose expression was increased >twofold in panels A and B ( red ) . Expression data for and definition of the various cell-types came from ImmGen ( www . Immgen . org ) . Langerhans cells of the skin ( LC . SK ) have been re-positioned as per recent data ( Gautier et al . , 2012 ) . Expression values were row-normalized . ( D ) GSEA of the totality of transcripts increased in pancreatic CD45+ cells of mice treated with I-BET151 ( red dots in panels A and B ) . NES , normalized enrichment score . Representative genes showing increased expression on the right . ( E ) A plot analogous to that in panel C for the totality of transcripts >twofold under-represented in drug-treated mice ( blue dots in panels A and B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 005 We then focused specifically on the impact of I-BET151 on the NF-κB pathway in pancreatic leukocytes , with the following considerations in mind: ( 1 ) NF-κB is a critical player in many types of inflammation , exerting both pro- and anti-inflammatory influences ( Vallabhapurapu and Karin , 2009 ) ; ( 2 ) specifically , NF-κB has been implicated in T1D ( Rink et al . , 2012 ) ; ( 3 ) direct binding of a BET protein , Brd4 , to the RelA subunit of NF-κB has been documented ( Huang et al . , 2009; Zhang et al . , 2012; Zou et al . , 2014 ) ; and ( 4 ) BET inhibitors are known to influence the NF-κB-induced transcriptional program in MFs ( Nicodeme et al . , 2010 ) . Interestingly , both 2-week and short-term treatment of NOD mice with I-BET151 had a dichotomous effect on the NF-κB target genes expressed by pancreatic leukocytes ( Figure 4A , upper left panels , and Supplementary file 3; target genes extracted from [Gilmore , 2006] ) . Pathway analysis using Ingenuity showed that , amongst these NF-κB targets , pro-inflammatory genes , particularly those encoding proteins in the tumor necrosis factor ( TNF ) α-induced canonical NF-κB activation pathway , were under-represented; while anti-inflammatory loci , notably those specifying molecules implicated in signaling by nuclear receptor family members , that is , the peroxisome proliferator-activated receptor ( PPAR ) and liver-X-receptor ( LXR ) families , were over-represented subsequent to I-BET151 treatment ( Figure 4A , upper right ) . These findings were confirmed by examining expression of pathway signature genes derived from the Molecular Signatures Database ( www . broadinstitute . org/gsea/msigdb ) : again the TNF-induced NF-κB activation pathway was down-regulated and the two nuclear receptor pathways up-regulated ( Figure 4A , lower panel ) . There were no such effects on CD45+ cells isolated from the spleen or lymph nodes of the same mice . 10 . 7554/eLife . 04631 . 006Figure 4 . The NF-κB signaling pathway is a major focus of I-BET151's influence on NOD leukocytes . ( A ) Upper panels: The inhibitor's effect on NF-κB-regulated genes—defined as per http://www . bu . edu/nf-kb/gene-resources/target-genes . Left , relevant transcripts from pancreatic CD45+ cells of NOD mice treated long- or short-term with I-BET151 or DMSO . Red: over-represented; blue: under-represented . 2 wk: long-term , treatment as per Figure 1B; 48 hr: short-term , treatment with a single 10 mg/kg dose and analyzed 48 hr later . Right , signaling pathways represented by the enriched or impoverished transcripts in the data to the left , via Ingenuity pathway analysis ( www . ingenuity . com ) . Lower panels: Gene sets corresponding to the TNFα-induced canonical NF-κB pathway ( Schaefer et al . , 2009 ) or the PPAR and LXR pathways ( http://www . genome . jp/kegg/pathway/hsa/hsa03320 . html ) were retrieved from the Broad Institute's Molecular Signatures Database ( http://www . broadinstitute . org/gsea/msigdb ) , and their expression levels in CD45+ cells from pancreas of I-BET151- or vehicle-treated mice plotted . ( B ) 12-week-old NOD mice were injected once ip with BAY 11–7082 ( 10 mg/kg ) , sacrificed 24 hr later , and CD45+ cells from the pancreas isolated and transcriptionally profiled . A volcano plot comparing treatment with BAY 11–7082 and DMSO , with genes >twofold increased ( in red ) or decreased ( in blue ) by I-BET151 treatment ( pooled from all time-points of Figure 3A , B ) superimposed . ( C ) Effect of a single dose of 10 mg/kg BAY 11–7082 on insulitis in 12-week-old NOD mice , analyzed 24 hr after injection . Left: insulitis scores . Right: summary data for the fraction of islets with no infiltrate . Grey , peri-insulitis; Black , insulitis . ( D ) Suppression of in vitro T cell proliferation by cell populations isolated from the pancreas of I-BET151- or DMSO- treated mice ( as per Figure 1B ) . The CD11b+CD11c− ( top ) , CD11b−CD11c+ ( middle ) and TCRβ+CD4+CD25+ ( bottom ) fractions of CD45+ cells were sorted . To the left are representative plots of CFSE dilution; to the right are summary data quantifying division indices ( see ‘Materials and methods’ for details ) . p values in A and B are from the chi-squared test , and in C and D are from the Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 006 We evaluated to what extent an NF-κB inhibitor could mimic the effects of I-BET151 . Given its greater toxicity , a single dose of BAY 11–7082 ( which blocks IκBα phosphorylation ) was administered to 12-week-old NOD mice , and pancreatic CD45+ cells were isolated for transcriptome analysis 24 hr later . When the sets of genes over- or under-represented in the CD45+ population of I-BET151-treated mice ( drawn from Figure 3A , B ) were superimposed on a volcano plot of transcripts from CD45+ cells of animals given BAY 11–7082 vs vehicle , there was an impressive correspondence , particularly for the enriched transcripts ( Figure 4B ) . The match was not perfect , however , as some genes modulated by the BET-protein inhibitor were not affected by the NF-κB inhibitor and vice versa ( Figure 4B ) . Perhaps not surprisingly , then , BAY 11–7082 , even a single dose , could substantially clear NOD insulitis ( Figure 4C ) . ( Note that it was not possible to assess the effect of this drug on diabetes development because its toxicity precluded multiple administrations . ) We performed co-culture experiments to confirm the disease-dampening potential of pancreatic MFs from mice administrated BET inhibitor . As T cells are key orchestrators of NOD diabetes ( Anderson and Bluestone , 2005 ) , we examined the ability of various cell-types isolated from the pancreas of mice treated with I-BET151 or vehicle to impact an in vitro T cell proliferation assay , quantifying CFSE dilution of CD3/CD28-stimulated cells ( Figure 4D ) . The CD11b+ population from the pancreas of vehicle-treated mice was capable of repressing T cell proliferation , but the corresponding population from mice administered I-BET151 exhibited significantly more potent suppressive activity . Co-cultured CD11c+ cells did not appear to affect T cell proliferation whether isolated from control- or drug-treated mice . Treg cells isolated from the two types of mice were equally able to inhibit T cell proliferation . It was obviously of interest to see whether I-BET151 exerted similar effects on human MFs . CD14+ cells pooled from the peripheral blood of three donors were differentiated in culture; they were then incubated with I-BET151 or vehicle for 30 min , before stimulation with LPS for 4 hr; at which time , RNA was isolated for transcriptional profiling . A rank-order FC plot showed that the inhibitor greatly dampened the typical MF response to LPS ( Figure 5A ) . Focusing on the NF-κB pathway: most ( 32/43 ) of the target genes inhibited in murine CD45+ cells were also suppressed in human MFs ( Figure 5B ) . In particular , there was suppression of secondary , over primary , LPS-response genes ( Figure 5C ) , as was previously reported for BET inhibitor treatment of mouse bone-marrow-derived MFs ( Nicodeme et al . , 2010 ) . 10 . 7554/eLife . 04631 . 007Figure 5 . Effects of I-BET151 on human monocyte-derived MFs . Cultures of human MFs were differentiated from peripheral-blood-derived CD14+ cells , pre-cultured for 30 min with I-BET151 ( red bars ) or just DMSO ( blue bars ) , and stimulated for another 4 hr after the addition of LPS ( 100 ng/ml ) or vehicle only . Microarray data from two conditions , as indicated and detailed in ‘Materials and methods’ . ( A ) Distribution of FCs of I-BET151/DMSO for LPS-induced genes ( according to the data on human monocyte-derived MFs treated with LPS or just vehicle ) . ( B ) FCs of I-BET151/DMSO of the human orthologues of the set of murine NF-κB-regulated genes illustrated in Figure 4A . ( C ) Effects of I-BET151 on primary and secondary LPS-response genes ( defined as per [Hargreaves et al . , 2009; Ramirez-Carrozzi et al . , 2009] ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 007 We questioned whether the disease-protective effect of I-BET151 also entailed an influence on the target of the autoimmune attack , pancreatic islet β cells . As an initial approach , we compared global gene-expression profiles of flow-cytometrically purified β cells from NOD mice treated from 12–14 weeks of age with I-BET151 or vehicle . A set of about 300 transcripts was over-represented >twofold in mice given the inhibitor; and about 10-fold fewer were under-represented >twofold ( FDR < 5% ) ( Figure 6A and Supplementary file 4 ) . The subset of genes whose expression was increased most by the inhibitor encoded multiple members of the regenerating islet-derived ( Reg ) protein family , originally identified for their involvement in pancreas regeneration ( Figure 6B , in red ) ( Unno et al . , 1992; Huszarik et al . , 2010 ) ; several transcription factors ( TFs ) important for regeneration of β-cells ( in green ) ( Pagliuca and Melton , 2013; Shih et al . , 2013 ) ; and a number of proteins known to enhance insulin production ( in blue ) ( Hirayama et al . , 1999; Dai et al . , 2006; Winzell and Ahren , 2007 ) . Interestingly , the most highly induced subset of genes also included a number of loci usually associated with neural function , encoding , for example , survival factors ( Cntfr , Tox3 ) or promoters of synaptic development or function ( Snap25 , Lrrtm2 , Lgi1 ) . The top signaling pathways to emerge via GSEA were the IGF-1 , IGF/mTOR , PDGF and insulin-secretion pathways ( although , due to the relatively small number of genes involved , statistical significance was not reached ) . 10 . 7554/eLife . 04631 . 008Figure 6 . BET-protein inhibition promotes regeneration of islet β cells . ( A ) Pancreatic β cells were cytofluorometrically sorted from mice treated with I-BET151 or DMSO as per Figure 1B , and microarray-based transcriptional profiling performed . Red: transcripts increased >twofold by I-BET151; blue , transcripts >twofold decreased . ( B ) NOD β-cell transcripts increased by I-BET151 ranked by FC vis-à-vis DMSO treatment . Red , regenerating islet-derived ( Reg ) transcripts; green , transcripts encoding transcription factors important for β-cell differentiation and function; blue , transcripts encoding proteins that enhance insulin production . ( C , D , G ) Cytofluorometric quantification of EdU+ β cells from NOD ( D ) or NOD . Rag−/− ( G ) mice treated with I-BET151 or DMSO as in Figure 1B and injected with EdU during the last 24 hr n = 5–8 . Panel C shows the sorting strategy . ( E ) The set of red transcripts from panel A was superimposed on volcano plots comparing gene expression by β cells from NOD . Rag−/− mice treated as in Figure 1B with I-BET151 vs DMSO . ( F ) Relevant I-BET151-induced transcripts highlighted in panel B are situated on the volcano plot of panel E . ( G ) p values: *<0 . 05; **<0 . 01; *<0 . 001—from the Student's t test for panels D and G , and from the chi-squared test for panel E . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 00810 . 7554/eLife . 04631 . 009Figure 6—figure supplement 1 . Histologic analysis of β-cell proliferation in response to I-BET151 in NOD mice . NOD mice were treated with I-BET151 or DMSO as in Figure 1B . EdU was administrated during the last 24 hr . Frozen sections of pancreas were stained for insulin and EdU . Left , representative islet images; two for each condition; right , summary quantification . Small images in color are legends for insulin , EdU and DAPI , respectively . Red , EdU; green , insulin; blue , DAPI . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 00910 . 7554/eLife . 04631 . 010Figure 6—figure supplement 2 . Histologic analysis of β-cell proliferation in response to I-BET151 in NOD . Rag−/− mice . NOD . Rag−/− mice were treated with I-BET151 or DMSO as in Figure 1B . EdU was administrated during the last 24 hr . Frozen sections of pancreas were stained for insulin and EdU . Left , representative islet images , two for each condition; right , summary quantification . DOI: http://dx . doi . org/10 . 7554/eLife . 04631 . 010 The NF-κB pathway has been implicated in the death of β-cells associated with T1D , mainly the branches downstream of inflammatory cytokine and stress stimuli ( Cardozo et al . , 2001; Cnop et al . , 2005; Eldor et al . , 2006 ) . Several NF-κB-dependent transcripts previously reported to be induced in β cells by the inflammatory cytokines IL-1β and IFN-γ ( Cardozo et al . , 2001 ) were suppressed by I-BET151 , for example , Icam1 , 1 . 45-fold; Traf2 , 1 . 28-fold; Fabp5 , 1 . 23-fold . Overall , then , I-BET151 appeared to have a positive impact on β-cell regeneration and/or function . To determine whether BET-protein inhibition induced proliferation of β cells , we treated 12-week-old NOD mice for the usual 2 weeks with I-BET151 or vehicle , injected them with 5-ethyl-2-deoxyuridine EdU 24 hr before sacrifice , and quantified EdU incorporation by β cells , delineated by their high autofluorescence and strong staining with a fluorescently conjugated exendin peptide probe ( Figure 6C ) . Both the number and fraction of EdU+ β cells were significantly increased in I-BET151-treated mice ( Figure 6D ) . Analogous results were obtained using fluorescent microscopy to quantify EdU+ insulin-expressing β cells ( Figure 6—figure supplement 1 ) . Since it has been reported that T cells in the immune infiltrate can induce β-cell replication ( Sreenan et al . , 1999; Sherry et al . , 2006; Dirice et al . , 2013 ) , we repeated the preceding experiments on Rag1-deficient NOD mice in order to address the possibility of a secondary effect mediated through lymphocytes . Transcriptional analysis revealed that most of the β-cell transcripts over-represented >twofold by treatment of standard NOD mice with I-BET151 were also increased in inhibitor-injected NOD mice lacking T and B cells , and thereby devoid of insulitis ( Figure 6E; and the transcripts highlighted in panel 6B are precisely positioned in panel 6F ) . In addition , there was a clear induction of β-cell proliferation in the absence of insulitis ( Figure 6 , panel G and Figure 6—figure supplement 2 ) . We attempted to further evaluate the cell autonomy of I-BET151's effects on β cells by purifying islets from 12-week-old NOD mice and culturing them for 24 hr in the presence of inhibitor or just vehicle . No consistent differences were observed under the two culture conditions . Not surprisingly , then , a similar experiment using commercially available human islets failed to show a repeatable difference when they were cultured in the presence and absence of I-BET151 . Clearly , the islet isolation/culture conditions removed or destroyed a critical factor .
The studies presented here showed that treatment of NOD mice with the epigenetic modifier , I-BET151 , for a mere 2 weeks prevented the development of NOD diabetes life-long . I-BET151 was able to inhibit impending insulitis as well as clear existing islet infiltration . The drug had a dual mechanism of action: it induced the pancreatic MF population to adopt an anti-inflammatory phenotype , primarily via the NF-κB pathway , and promoted β-cell proliferation ( and perhaps differentiation ) . These findings raise a number of intriguing questions , three of which we address here . First , why do the mechanisms uncovered in our study appear to be so different from those proposed in the only two previous reports on the effect of BET-protein inhibitors on autoimmune disease ? Bandukwala et al . found that I-BET762 ( a small-molecule inhibitor similar to I-BET151 ) altered the differentiation of Th subsets in vitro , perturbing the typical profiles of cytokine production , and reducing the neuropathology provoked by transfer of in-vitro-differentiated Th1 , but not Th17 , cells reactive to a peptide of myelin oligodendrocyte glycoprotein ( Bandukwala et al . , 2012 ) . Unfortunately , with such transfer models , it is difficult to know how well the in vitro processes reflect in vivo events , and to distinguish subsidiary effects on cell survival and homing . Mele et al . reported that JQ1 primarily inhibited the differentiation of and cytokine production by Th17 cells , and strongly repressed collagen-induced arthritis and experimental allergic encephalomyelitis ( Mele et al . , 2013 ) . However , with adjuvant-induced disease models such as these , it is difficult to discriminate influences of the drug on the unfolding of autoimmune pathology vs on whatever the adjuvant is doing . Thus , the very different dual mechanism we propose for I-BET151's impact on spontaneously developing T1D in NOD mice may reflect several factors , including ( but not limited to ) : pathogenetic differences in induced vs spontaneous autoimmune disease models; our broader analyses of immune target cell populations; and true mechanistic differences between T1D and the other diseases . As concerns the latter , it has been argued that T1D is primarily a Th1-driven disease , with little , or even a negative regulatory , influence by Th17 cells ( discussed in [Kriegel et al . , 2011] ) . Second , how does I-BET151's effect , focused on MFs and β cells , lead to life-long protection from T1D ? MFs seem to play a schizophrenic role in the NOD disease . They were shown long ago to be an early participant in islet infiltration ( Jansen et al . , 1994 ) , and to play a critical effector role in diabetes pathogenesis , attributed primarily to the production of inflammatory cytokines and other mediators , such as iNOS ( Hutchings et al . , 1990; Jun et al . , 1999a , 1999b; Calderon et al . , 2006 ) . More recently , there has been a growing appreciation of their regulatory role in keeping diabetes in check . For example , the frequency of a small subset of pancreatic MFs expressing the complement receptor for immunoglobulin ( a . k . a . CRIg ) at 6–10 weeks of age determined whether or not NOD diabetes would develop months later ( Fu et al . , 2012b ) , and transfer of in-vitro-differentiated M2 , but not M1 , MFs protected NOD mice from disease development ( Parsa et al . , 2012 ) . One normally thinks of immunological tolerance as being the purview of T and B cells , but MFs seem to be playing the driving role in I-BET151's long-term immunologic impact on T1D . Chronic inflammation ( as is the insulitis associated with T1D ) typically entails three classes of participant: myeloid cells , in particular , tissue-resident MFs; lymphoid cells , including effector and regulatory T and B cells; and tissue-target cells , that is , islet β cells in the T1D context . The ‘flavor’ and severity of inflammation is determined by three-way interactions amongst these cellular players . One implication of this cross-talk is that a perturbation that targets primarily one of the three compartments has the potential to rebalance the dynamic process of inflammation , resetting homeostasis to a new level either beneficial or detrimental to the individual . BET-protein inhibition skewed the phenotype of pancreatic MFs towards an anti-inflammatory phenotype , whether this be at the population level through differential influx , efflux or death , or at the level of individual cells owing to changes in transcriptional programs . The ‘re-educated’ macrophages appeared to be more potent at inhibiting T cell proliferation . In addition , it is possible that MFs play some role in the I-BET151 influences on β-cell regeneration . The findings on Rag1-deficient mice ruled out the need for adaptive immune cells in the islet infiltrate for I-BET151's induction of β-cell proliferation , but MFs are not thought to be compromised in this strain . Relatedly , the lack of a consistent I-BET151 effect on cultured mouse and human islets might result from a dearth of MFs under our isolation and incubation conditions ( e . g . , [Li et al . , 2009] ) . Several recent publications have highlighted a role for MFs , particularly M2 cells , in promoting regeneration of β cells in diverse experimental settings ( Brissova et al . , 2014; Xiao et al . , 2014 ) , a function foretold by the reduced β-cell mass in MF-deficient Csf1op/op mice reported a decade ago ( Banaei-Bouchareb et al . , 2004 ) . Whether reflecting a cell-intrinsic or -extrinsic impact of the drug , several pro-regenerative pathways appear to be enhanced in β-cells from I-BET151-treated mice . Increased β-cell proliferation could result from up-regulation of the genes encoding Neurod1 ( Kojima et al . , 2003 ) , GLP-1R ( De Leon et al . , 2003 ) , or various of the Reg family members ( Unno et al . , 2002; Liu et al . , 2008 ) , the latter perhaps a consequence of higher IL-22R expression ( Hill et al . , 2013 ) ( see Figure 6B and Supplementary file 4 ) . Protection of β-cells from apoptosis is likely to be an important outcome of inhibiting the NF-κB pathway ( Takahashi et al . , 2010 ) , but could also issue from enhanced expression of other known pro-survival factors , such as Cntfr ( Rezende et al . , 2007 ) and Tox3 ( Dittmer et al . , 2011 ) ( see Figures 4 and 6B ) . Lastly , β-cell differentiation and function should be fostered by up-regulation of genes encoding transcription factors such as Neurod1 , Pdx1 , Pax6 , Nkx6-1 and Nkx2-2 . The significant delay in re-onset of diabetes in I-BET151-treated diabetic mice suggests functionally relevant improvement in β-cell function . In brief , the striking effect of I-BET151 on T1D development in NOD mice seems to reflect the fortunate concurrence of a complex , though inter-related , set of diabetes-protective processes . Lastly , why does a drug that inhibits BET proteins , which include general transcription factors such as Brd4 , have such circumscribed effects ? A 2-week I-BET151 treatment might be expected to provoke numerous side-effects , but this regimen seemed in general to be well tolerated in our studies . This conundrum has been raised in several contexts of BET-inhibitor treatment , and was recently discussed at length ( Shi and Vakoc , 2014 ) . The explanation probably relates to two features of BET-protein , in particular Brd4 , biology . First: Brd4 is an important element of so-called ‘super-enhancers’ , defined as unusually long transcriptional enhancers that host an exceptionally high density of TFs—both cell-type-specific and general factors , including RNA polymerase-II , Mediator , p300 and Brd4 ( Hnisz et al . , 2013 ) . They are thought to serve as chromatin depots , collecting TFs and coordinating their delivery to transcriptional start-sites via intra-chromosome looping or inter-chromosome interactions . Super-enhancers are preferentially associated with loci that define and control the biology of particular cell-types , notably developmentally regulated and inducible genes; intriguingly , disease-associated , including T1D-associated , nucleotide polymorphisms are especially enriched in the super-enhancers of disease-relevant cell-types ( Hnisz et al . , 2013; Parker et al . , 2013 ) . Genes associated with super-enhancers show unusually high sensitivity to BET-protein inhibitors ( Chapuy et al . , 2013; Loven et al . , 2013; Whyte et al . , 2013 ) . Second: although the bromodomain of Brd4 binds to acetyl-lysine residues on histone-4 , and I-BET151 was modeled to inhibit this interaction , it is now known to bind to a few non-histone chromosomal proteins as well , notably NF-κB , a liaison also blocked by BET-protein inhibitors ( Huang et al . , 2009; Zhang et al . , 2012; Zou et al . , 2014 ) . Abrogating specific interactions such as these , differing according to the cellular context , might be the dominant impact of BET inhibitors , a scenario that would be consistent with the similar effects we observed with I-BET151 and BAY 11–7082 treatment . Either or both of these explanations could account for the circumscribed effect of I-BET151 on NOD diabetes . Additionally , specificity might be imparted by different BET-family members or isoforms—notably both Brd2 and Brd4 are players in MF inflammatory responses ( Belkina et al . , 2013 ) . According to either of these explanations , higher doses might unleash a broader array of effects . Viewed in the context of recent reports , our data point to NF-κB as a direct target of I-BET151 . Traditionally , Brd4's impact on transcription has been thought to reflect its binding to histone acetyl-lysine residues , as a so-called ‘histone reader’ ( Muller et al . , 2011; Prinjha et al . , 2012 ) . Analogously , the influence of I-BET151 ( and like drugs ) on Brd4 function has generally been attributed to Brd4 interactions with acetylated histones ( Muller et al . , 2011; Prinjha et al . , 2012 ) . However , it is now clear that I-BET 151 directly targets Brd4's association with non-histone proteins as well . The best-studied example is NF-κB ( Chen et al . , 2005; Nowak et al . , 2008; Huang et al . , 2009; Zou et al . , 2014 ) . The two bromodomains of Brd4 cooperatively recognize RelA acetylated at the K310 position , and this interaction is blocked by drugs like I-BET151 and JQ1 . Of late , there has been substantial interest in treating individuals with , or at risk of , T1D with combination therapies . It would seem logical to design a combinatorial approach that targets two or more of the major players in disease—perhaps optimally , addressing elements of the innate immune system , adaptive immune system and islet target tissue . We have demonstrated that a single drug , the BRD blocker I-BET151 , has a potent effect on T1D in pre-diabetic NOD mice by coincidentally influencing MFs and β cells . Recently , another drug from the cancer world , the HDAC inhibitor , vorinostat , was reported to inhibit T1D , by a seemingly different mechanism impacting a multiplicity of cell-types ( Christensen et al . , 2014 ) . Thus , epigenetic modulators would seem to be exciting candidates to explore in human T1D patients .
NOD/Lt mice were bred under specific-pathogen–free conditions in our animal facility at the New Research Building of Harvard Medical School , cared for in accordance with the ethical guidelines of the Institutional Animal Care and Use Committee . Relevant studies were also conducted in accordance with GSK's Policy on the Care , Welfare and Treatment of Laboratory Animals . NOD . Cg-Rag1<tm1mom> mice were maintained in our lab's colony at Jackson Laboratory . For the evaluation of diabetes , mice were monitored until 30 weeks of age by measuring urine- and blood-glucose levels , as described ( Katz et al . , 1993 ) . Individuals with two consecutive measurements of a serum-glucose concentration above 300 mg/dl were considered diabetic . For the recent-onset cohorts: on the same day as diabetes diagnosis , individuals received a single subcutaneously implanted insulin pellet ( LinShin , Toronto , Canada ) , which lowers the blood-glucose level to the normal range within 2 days . For insulitis assessment , mice were euthanized , and their pancreas removed and fixed in 10% neutral-buffered formalin ( Sigma–Aldrich , St . Louis , MO ) . Paraffin-embedded sections were cut into 6 μm sections with 150 μm between steps , and were stained with hematoxylin and eosin ( H&E ) . Insulitis was scored as described ( Katz et al . , 1993 ) . The BET-protein BRD inhibitor ( I-BET151 , GSK1210151A ) was produced and handled as described ( Dawson et al . , 2011 ) . For long-term in vivo treatment , mice were ip-injected with I-BET151 dissolved in DMSO at a dose of 10 mg/kg daily for 2 weeks beginning at the designated age . The NF-κB inhibitor ( BAY 11–7082 ) was purchased from Sigma and dissolved in DMSO as a stock solution of 10 mg/ml . For in vivo treatment , BAY was injected ip at a dose of 10 mg/kg 24 hr before sacrificing mice for analysis . For immunocytes , all staining began by incubating with a mAb directed against FcγR ( 2 . 4G2; BD Pharmingen , San Diego , CA ) . mAbs against the following antigens were used: CD45 ( 30-F11 ) , CD4 ( RM4-5 ) , CD8 ( 53-6 . 7 ) , CD19 ( 6D5 ) , CD11b ( M1/70 ) , CD11c ( G418 ) , Ly6C ( AL-21 ) , F4/80 ( BM8 ) and IL-17A ( TC11-18H10 . 1 ) , all from BioLegend ( San Diego , CA ) ; and FoxP3 ( FJK-16s , eBioscience , San Diego , CA ) . For β cells , we followed the method described in ( Li et al . , 2009 ) with modifications . Briefly , the pancreas was perfused via the common bile duct with 5 ml of a solution of collagenase P ( 1 mg/ml , Roche ) ; and was then digested at 37°C for 15 min , followed by several steps of centrifugation and washing . On average , 80 islets were purified by hand-picking , and were then cultured in complete RPMI1640 medium with the fluorescent exendin-4 probe ( 100 nM ) . After 1 hr , islets were disrupted by treatment with trypsin-EDTA solution in a 37°C waterbath for 10 min . The single-cell suspensions containing β cells were analyzed by flow cytometry . The exendin-4 probe , EP12-BTMR-X , was synthesized as previously reported ( compound 17 in Table 1 of Clardy et al . , 2014 ) . Briefly , an azide-functionalized BODIPY TMR-X was conjugated to exendin-4 ( HGEGTFTSDLSPraQMEEEAVRLFIEWLKNGGPSSGAPPPS ) using microwave-assisted copper-catalyzed azide-alkyne Huisgen cycloaddition . The conjugate was purified by HPLC and characterized by MALDI-TOF mass spectrometry and an I125 binding assay . To quantify proliferation in vivo , we ip-injected mice with 1 mg EdU 24 hr before sacrifice , and single-cell suspensions were prepared , fixed and permeabilized using the eBioscience Fixation/Permeabilization set ( Cat# 00-5123 , 00-5223 ) . They were then stained using the EdU staining kit ( Click-iT EdU Flow Cytometry Assay Kits; Invitrogen , Carlsbad , CA ) , following the users' manual . Pancreata were collected and digested for 30 min at 37°C with collagenase IV ( 1 mg/ml ) and DNase I ( 10 U/ml ) . Single-cell suspensions were prepared and stained , and CD4+ or CD45+ cells were sorted into 500 ml TRIzol ( Invitrogen ) for RNA isolation . RNA was amplified in two rounds with MessageAmp aRNA ( Ambion ) and labeled with biotin ( BioArray High Yield RNA Transcription Labeling; Enzo , Farmingdale , NY ) , and the resulting cRNA was hybridized to MoGene 1 . 0 ST ( mouse ) or huGene 1 . 0 ST ( human ) arrays ( Affymetrix , Santa Clara , CA ) . Raw data were normalized by the robust multi-array average ( RMA ) algorithm implemented in the Expression File Creator module of the GenePattern genomic analysis platform ( Reich et al . , 2006 ) . For pathway analysis , GSEA was used , as described ( Subramanian et al . , 2005 ) . Gene-expression signatures of Th1 , Th2 and Th17 cells derived from ( Wu et al . , 2010 ) and ( Vehik and Dabelea , 2011 ) ; and signatures for Treg cells from ( Fu et al . , 2012a ) . For microarray analysis of islet β cells , NOD or NOD . Rag−/− mice were sacrificed , and collegenase P solution ( 0 . 5 mg/m ) administered intrapancreatically via the bile duct . Pancreata were than digested for 20 min at 37°C . Islets were hand-picked under a stereomicroscope and a single-cell suspension prepared by treating the isolated islets with trypsin-EDTA ( Life Technology , Grand Island , NY ) ( Brennand et al . , 2007 ) . β cells were cytofluorometrically sorted on cell size and autofluorescence ( Dirice et al . , 2013 ) . Transcript quantification and data analysis were as above . TCRβ+CD4+CD25− naïve T cells were sorted from spleen and labeled with 10 μmol/l CFSE ( Molecular Probes ) in RPMI 1640 at a concentration of 106/ml at 37°C for 20 min; then were washed , resuspended in complete culture medium ( RPMI1640 , 10% fetal calf serum , 2 mmol/l L-glutamine , penicillin/streptomycin , and 2-mercaptoethanol ) , and cultured at 1 × 105 cells/well in round-bottom , 96-well plates ( Corning , Corning , NY ) . T cells were activated with anti-CD3/CD28 beads ( at a ratio of 1:1 between cells and beads [Invitrogen] ) and IL-2 ( 20 U/ml ) . Pancreatic myeloid cells from I-BET151- , or DMSO-treated mice were sorted into CD11b+CD11clow/− or CD11blow/− CD11c+ fractions , and were added to cultured T cells at a ratio of myeloid:T cells of 1:4 . Similarly , CD4+CD25+ cells were sorted from I-BET151- , or DMSO-treated mice , and were added to cultured T cells at a ratio of 1:2 . Proliferation was measured by flow-cytometric analysis of CFSE dilution , and division indices were calculated by computing the weighted fractions of all the divisions as per ( D'Alise et al . , 2008 ) . Blood from three healthy volunteers was used to isolate buffy coats by Ficoll–Paque density gradient centrifugation . Human monocytes were purified by positive selection of CD14-expressing cells ( Miltenyi Biotec , UK ) . They were cultured in RPMI-1640 supplemented with 5% heat-inactivated fetal calf serum ( Hyclone , ThermoScientific , UK ) , 2 mM Glutamine ( Invitrogen , UK ) , 100 U/ml penicillin and 100 mg/ml streptomycin containing either 5 ng/ml GM-CSF ( R&D Systems , UK ) ( condition 1 ) or 100 ng/ml M-CSF ( R&D Systems , UK ) ( condition 2 ) and cultured for 5 days to generate MFs , respectively . After 5 days , MFs were treated with fresh medium containing either 0 . 1% DMSO or 1 μM I-BET151 for 30 min and then stimulated with 100 ng/ml LPS ( L4391 , Sigma , UK ) , or left unstimulated . Cells were harvested at 4 hr for extracting total RNA for transcriptional profiling . Human biological samples were sourced ethically , and their research use was in accord with the terms of the informed consents . For visualizing β-cell proliferation , we injected EdU ( 1 mg ) ip 24 hr before harvesting pancreata . Paraffin sections of pancreas 6 μm in thickness were prepared . Standard procedures were used for immunostaining ( Fu et al . , 2012b ) . Before the addition of primary mAbs , sections were blocked with 5% normal donkey serum ( Jackson ImmunoResearch , West Grove , PA ) , then incubated overnight at 4°C with anti-insulin ( Linco Research , St . Charles , MO ) , followed by FITC-AffiniPure Donkey Anti-Guinea Pig IgG ( Jackson ImmunoResearch ) . EdU staining was performed as above . Nuclei were stained with DAPI ( 4′ , 6-diamidino-2-phenylindole dihydrochloride ) . Images were acquired on an Axiovert 200M confocal microscope ( Zeiss , Peabody , MA ) with a xenon arc lamp in a Lambda DG-4 wavelength switcher ( Sutter Instrument ) , and were processed with Slidebook imaging software ( Intelligent Imaging , Denver , CO ) . Tests used for statistical analyses are described in the various figure legends ( GraphPad software v5 . 0; Prism , La Jolla , CA ) . p values of 0 . 05 or less were considered to be statistically significant . | The DNA inside a cell is often tightly wrapped around proteins to form a compact structure called chromatin . Chemical groups added to the chromatin can encourage nearby genes to either be switched on or off; and several enzymes and other proteins help to read , add , or remove these marks from the chromatin . If these chromatin modifications ( or the related enzymes and proteins ) are disturbed it can lead to diseases like cancer . It has also been suggested that similar changes may influence autoimmune diseases , in which the immune system attacks the body's own tissues . Drugs that target the proteins that read , add , or remove these chromatin modifications are currently being developed to treat cancer . For example , drugs that inhibit one family of these proteins called BET have helped to treat tumors in mice that have cancers of the blood or lymph nodes . However , because these drugs target pathways involved in the immune system they may also be useful for treating autoimmune diseases . Now Fu et al . have tested whether a BET inhibitor might be a useful treatment for type-1 diabetes . In patients with type-1 diabetes , the cells in the pancreas that produce the insulin hormone are killed off by the immune system . Without adequate levels of insulin , individuals with type-1 diabetes may experience dangerous highs and lows in their blood sugar levels and must take insulin and sometimes other medications . Using mice that spontaneously develop type-1 diabetes when still relatively young , Fu et al . tested what would happen if the mice received a BET inhibitor for just 2 weeks early on in life . Treated mice were protected from developing type-1 diabetes for the rest of their lives . Specifically , the treatment protected the insulin-producing cells and allowed them to continue producing insulin . The drug reduced inflammation in the pancreas and increased the expression of genes that promote the regeneration of insulin-producing cells . Diabetes researchers have been searching for drug combinations that protect the insulin-producing cells and boost their regeneration . As such , Fu et al . suggest that these findings justify further studies to see if BET inhibitors may help to treat or prevent type-1 diabetes in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2014 | Epigenetic modulation of type-1 diabetes via a dual effect on pancreatic macrophages and β cells |
Homologous recombination ( HR ) -mediated repair of DNA double-strand break ( DSB ) s is restricted to the post-replicative phases of the cell cycle . Initiation of HR in the G1 phase blocks non-homologous end joining ( NHEJ ) impairing DSB repair . Completion of HR in G1 cells can lead to the loss-of-heterozygosity ( LOH ) , which is potentially carcinogenic . We conducted a gain-of-function screen to identify miRNAs that regulate HR-mediated DSB repair , and of these miRNAs , miR-1255b , miR-148b* , and miR-193b* specifically suppress the HR-pathway in the G1 phase . These miRNAs target the transcripts of HR factors , BRCA1 , BRCA2 , and RAD51 , and inhibiting miR-1255b , miR-148b* , and miR-193b* increases expression of BRCA1/BRCA2/RAD51 specifically in the G1-phase leading to impaired DSB repair . Depletion of CtIP , a BRCA1-associated DNA end resection protein , rescues this phenotype . Furthermore , deletion of miR-1255b , miR-148b* , and miR-193b* in independent cohorts of ovarian tumors correlates with significant increase in LOH events/chromosomal aberrations and BRCA1 expression .
Double-stranded DNA break ( DSB ) s are deleterious for cell health and accurate repair of DSBs is critical for maintaining genome stability . Two major mechanistically distinct pathways , homologous recombination ( HR ) and non-homologous end joining ( NHEJ ) have evolved to repair DSBs ( Ciccia and Elledge , 2010; Chapman et al . , 2012b ) . HR requires an undamaged homologous DNA template to replace an adjacent damaged one with high fidelity ( San Filippo et al . , 2008 ) . By contrast , the untemplated NHEJ pathway is more error-prone as it rapidly processes and joins broken DNA ends ( Lieber , 2010 ) . There is tight regulation of the DSB repair pathways during the cell cycle as HR is restricted to the S/G2 phase and NHEJ is pre-dominant in G1 but has moderate activity throughout the cell cycle . The balance of HR and NHEJ proteins ( such as BRCA1 and 53BP1 ) involved in early steps of the two repair pathways are critical for pathway choice and the cell-cycle phase specific regulation of each pathway ( Bouwman et al . , 2010; Bunting et al . , 2010; Chapman et al . , 2012b ) . Initiation of HR via resection of broken DNA ends blocks NHEJ and could lead to unrepaired DSBs in G1 cells ( Helmink et al . , 2011; Escribano-Diaz et al . , 2013 ) . The physiological consequence of HR-mediated repair of a DSB in the G1-phase is the loss of heterozygosity ( LOH ) . The molecular mechanism of restricting HR to S/G2 phases of the cell cycle is of paramount importance in cancer biology . Ectopic HR leading to LOH may play a critical role in carcinogenesis as recessive oncogenic mutations are revealed and/or tumor suppressor function is lost ( Bishop and Schiestl , 2003 ) . The regulation of the HR pathway is also extremely relevant for cancer therapy . Chemical inhibitors of the DNA repair protein , poly ( ADP-ribose ) polymerase ( PARP ) , exhibit synthetic lethality in BRCA-deficient tumors that have a defective HR pathway ( Bryant et al . , 2005; Farmer et al . , 2005 ) . Although the molecular mechanism underlying this phenotype remains unresolved , lack of PARP activity in an HR-deficient scenario leads to the accumulation of DSBs in proliferating cells and this triggers apoptosis ( Helleday , 2011 ) . Epigenetic suppression of factors in the HR pathway occurs in sporadic breast and ovarian tumors , and collectively this phenotype has been described as ‘BRCAness’ ( Turner et al . , 2004 ) . Tumors with the ‘BRCAness’ phenotype are likely to respond to PARP inhibitors , and identifying factors ( such as microRNAs ) that induce the ‘BRCAness’ phenotype may enhance the clinical utility and scope of PARP inhibitors . MicroRNAs ( miRNAs ) are abundant small ( ∼20–22 nts ) non-coding RNAs that typically dampen gene expression at the post-transcriptional level ( Fabian et al . , 2010 ) and are aberrantly expressed in a variety of cancer cells ( Garzon et al . , 2009 ) . The regulation elicited by miRNAs is highly complex , given that a single miRNA can influence the expression of many mRNAs , and conversely a single mRNA is targeted by multiple miRNAs ( Bartel , 2009 ) . Due to this complexity the role of miRNAs in DSB repair remains largely unknown ( Chowdhury et al . , 2013 ) . Using the experimental system of in vitro hematopoietic cell differentiation , we discovered that miR-182 targets BRCA1 ( Moskwa et al . , 2011 ) . Over-expression of miR-182 in triple negative breast tumor lines ( in vitro and in vivo ) suppresses HR via down-regulation of BRCA1 and sensitizes cells to PARP inhibitors . This served as a ‘proof-of principle’ that miRNAs may influence HR-mediated repair of DSBs . In this study , we used PARP inhibitor sensitivity as a marker for HR deficiency to conduct a functional screen to identify miRNAs that down-regulate HR . Over-expression of seven miRNAs significantly reduced HR-mediated DSB repair . Based on their expression in a panel of breast and ovarian lines , we focused on characterizing the mechanism and physiological relevance of miR-1255b , miR-193b* , and miR-148b* . Despite lacking canonical binding sites , miR-1255b , miR-193b* , and miR-148b* associate with the BRCA1 , BRCA2 , and RAD51 transcripts and regulate their expression . Remarkably , the miRNA-mediated regulation of these genes is cell cycle dependent and inhibition of miR-1255b , miR-193b* , and miR-148b* leads to enhanced expression of BRCA1 , BRCA2 , and RAD51 specifically in the G1 phase . A functional consequence of inhibiting these miRNAs is a basal increase in DSBs in G1 cells . This phenotype can be reversed by silencing CtIP ( the BRCA1-associated DNA end resection factor ) which initiates HR-mediated DSB repair . Furthermore , deletion of these miRNAs in two independent cohorts of high-grade serous ovarian tumors correlates with increase in LOH .
To systematically identify miRNAs that impact PARP inhibitor sensitivity , we screened two commercially available miRNA mimic libraries ( Qiagen , Valencia , CA and Applied Biosystems , Grand Island , NY ) . miRNA mimics are 20–22 nt , chemically modified double-stranded RNA molecules designed to mimic endogenous mature miRNAs . The miRNA mimics from Qiagen and Applied Biosystems have proprietary chemical modifications that cause inactivation of the ‘passenger’ strand , allowing the ‘active’ strand to associate with target transcripts and regulate gene expression . The goal was to identify miRNA mimics that sensitize breast tumor cells , MDA-MB231 to the clinical PARP inhibitor , ABT888 ( Figure 1A , B ) . The miRNA mimics from Applied Biosystems impacted ABT888 sensitivity over a broad range , ∼10- to 12-fold difference between the mimics with the highest impact on viability and the control mimics . Whereas the viability range with the mimics from Qiagen was limited to 3- to 4-fold . BRCA2 siRNA served as a positive control for each plate and the mimic for miR-182 served as an internal control . BRCA2 is an important gene for repair of replication-induced DNA damage and depletion of BRCA2 has an impact on viability even in the absence of PARP inhibitors ( Figure 1—figure supplement 1 ) . miRNA mimics that caused significant loss of viability in the absence of PARP inhibitors ( equal or more than BRCA2 siRNA ) were not considered ( Figure 1—figure supplement 1 ) . The Applied Biosystems library yielded a rank ordered list of the top 13 miRNAs ( including miR-182 ) based on viability percentage that sensitized cells to ABT888 ( Figure 1C ) ( Z-score thresholds [Z < −2] ) . Prior to individual validation experiments , we utilized the results from the screen performed with the Qiagen library to confirm the impact of these miRNAs on ABT888 sensitivity . Our results show that 12 of the 13 miRNAs had at least a 25% impact on ABT888 sensitivity . Next , we confirmed the impact of these miRNAs on PARP inhibitor sensitivity using two clinical PARP inhibitors , ABT888 and olaparib ( Figure 1D , Figure 1E ) in two breast cancer cell lines , MDA-MB231 and 21NT ( Figure 1F ) . Together , these results validate our primary screen and suggest that these miRNAs may regulate the efficacy of HR-mediated repair of DSBs . 10 . 7554/eLife . 02445 . 003Figure 1 . miRNA screen for PARP inhibitor sensitivity . ( A ) Schematic of a gain-of-function screen using miRNA mimic libraries from Applied Biosystems and Qiagen to identify miRNAs that sensitize cells to the PARP inhibitor , ABT888 . ( B ) Scatter plot ( wells/plate ) of luminescence ( y-axis ) as a read-out for viability of each miRNA-transfected well ( grey circle ) in the presence of ABT888 ( 20 µM ) . The plates are numbered in the x-axis . Positive control ( BRCA2 siRNA , blue circles ) and negative controls ( control mimics , pink circles ) are shown . Scatter plot for untreated samples is shown in Figure 1—figure supplement 1 . ( C ) List of top miRNAs from the screen displayed in the order of % control viability along with Z-score . ( D ) Clonogenic survival assay to validate the impact of selected miRNAs on sensitivity to ABT888 . MDAMB231 cells were transfected with control miRNA mimics , indicated miRNA mimics , BRCA1 siRNA , or BRCA2 siRNA and treated with vehicle or ABT888 , before measuring colony formation . Curves were generated from three independent experiments and a representative image of colony formation with 1 µM ABT888 is shown in the inset . ( E and F ) Luminascence-based viability assay was performed in MDAMB231 cells with PARP inhibitor , olaparib ( E ) or in 21NT cells with ABT888 ( F ) . Cells were transfected with control miRNA , indicated miRNA mimics , BRCA1 siRNA , or BRCA2 siRNA and treated with vehicle or PARP inhibitor before ATP quantification . Curves were generated from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 00310 . 7554/eLife . 02445 . 004Figure 1—figure supplement 1 . miRNAs screen for PARP inhibitor sensitivity . Scatter plot ( wells/plate ) of luminescence ( y-axis ) as a read-out for viability of each miRNA transfected well ( grey circle ) in the absence of ABT888 . The plates are numbered in the x-axis . Positive control ( BRCA2 siRNA , blue circles ) and negative controls ( control mimics , pink circles ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 004 Sensitivity to PARP inhibitors directly correlates with HR deficiency , but it is plausible that other pathways may also contribute to this phenotype . To test whether expression of the 12 selected miRNAs impacts HR , we used a functional assay that assesses HR-mediated repair of an I-SceI-induced DSB ( Nakanishi et al . , 2005 ) . U2OS cells that have a single , stably integrated copy of the artificial recombination substrate DR-GFP were used for this assay . A chromosomal DSB was introduced in one of two tandem non-functional GFP genes by transient transfection of an I-SceI-encoding plasmid . A functional GFP gene can be reconstituted if the DSB is repaired by HR using the other partial GFP gene as a template . Using this assay , 11 of the 12 miRNAs significantly diminished HR-mediated DSB repair , which correlated with their impact on PARPi sensitivity . ( Figure 2A , B ) . We prioritized the miRNAs according to their impact on HR and selected 7 miRNAs that reduced HR-efficiency by ≥30% ( miR-1231 , miR-1255b , miR-148b* , miR-876-3p , miR-221* , miR-193b* , and miR-185* ) . RAD51 is involved in HR , and quantification of nuclear RAD51 foci has also been used to evaluate HR in various cell systems including primary breast tumors ( Willers et al . , 2009 ) . Consistent with our DR-GFP reporter assay , over-expression of all 7 miRNAs in MDA-MB231 cells had a significant impact on RAD51 foci ranging from a 50–75% reduction in foci formation ( Figure 2C ) . 10 . 7554/eLife . 02445 . 005Figure 2 . miRNAs sensitize cells to PARP inhibitors by targeting HR-mediated DSB repair . ( A and B ) Measurement of HR-mediated repair of an I-SceI induced site-specific DSB . Cells carrying a single copy of the recombination substrate ( DR-GFP ) were transfected with control miRNA mimic , indicated miRNA mimics , or BRCA2 siRNA before transfection with I-SceI or control vector . GFP positive cells were analyzed 48 hr later by flow cytometry ( FACS ) . Representative images of the FACS profile are shown in ( A ) , and the mean ± SD of six independent experiments is graphically represented in ( B ) . The dotted line represents the cut-off which was set at 70% of the control . ( C ) Analysis of HR-mediated repair by RAD51 focus formation . MDA-MB231 cells were transfected with control miRNA mimic , indicated miRNA mimics , or BRCA2 siRNA , stained for RAD51 ( green ) and 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( blue ) 6 hr after exposure to IR . The images were captured by fluorescence microscopy and RAD51 focus-positive cells ( with >5 foci ) were quantified by comparing 100 cells . Mean ± SD of three independent experiments is shown in right panel . * indicates p<0 . 05 . ( D ) γ-H2AX accumulation after treatment with ABT888 . Cells were transfected with control miRNA mimic , indicated miRNA mimics , or BRCA2 siRNA and treated with ABT888 ( 100 μM ) before evaluation of γ-H2AX by immunoblotting at indicated time points . Total H2AX served as loading control for these experiments . Images were quantified by ImageJ software and the mean ± SD of three independent experiments is graphically shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 005 γ-H2AX formation is a marker for DSBs , and PARP inhibitor treatment of BRCA-deficient cells cause accumulation of γ-H2AX ( Bryant et al . , 2005 ) . To assess whether the miRNAs mirror BRCA deficiency , we assessed γ-H2AX after treating cells with ABT888 ( Figure 2D ) . Transfection of mimics for the 7 miRNAs that impacted HR increased γ-H2AX levels 12 hr after treatment with ABT888 and this was maintained through 24 hr . Together , these results strongly suggest that miRNAs ( miR-1231 , miR-1255b , miR-148b* , miR-876-3p , miR-221* , miR-193b* , and miR-185* ) , impact the efficacy of DNA repair by impeding the HR-mediated repair of DSBs . All our results so far have been derived by artificially introducing miRNAs in cells . To explore the physiological relevance of miR-1231 , miR-1255b , miR-148b* , miR-876-3p , miR-221* , miR-193b* , and miR-185* , we first identified cells where these miRNAs are endogenously expressed . From the perspective of cancer biology , the HR pathway is extremely relevant for breast tumors , specifically triple negative breast cancer ( TNBC ) ( Lips et al . , 2011 ) , and clinical trials with PARP inhibitors are underway in TNBC patients ( O'Shaughnessy et al . , 2009 , 2011 ) . Therefore , we initially assessed the endogenous expression of the 7 miRNAs in a panel of TNBC lines . Of these seven miRNAs , only miR-1255b , miR-148b* , and miR-193b* had detectable and aberrant over expression in TNBC lines ( Figure 3A ) . Therefore , we focused on an in-depth analysis of how miR-1255b , miR-148b* , and miR-193b* impact the HR pathway in TNBC lines and also investigated the physiological relevance of this regulation . It is possible that the other 4 miRNAs ( miR-1231 , miR-876-3p , miR-221* , and miR-185* ) that impact HR may not be expressed in TNBCs but could have physiological relevance in other cell lineages ( expression of these miRNAs in other cell types is shown in Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 02445 . 006Figure 3 . miR-1255b , miR-193b* , and miR-148b* regulate PARP inhibitor sensitivity by regulating expression of HR factors in TNBCs . ( A ) miRNA expression profile in a panel of breast cancer lines . Endogenous expression of indicated miRNAs was quantified by qRT-PCR ( normalized to 5srRNA ) and represented relative to non-tumorigenic breast epithelial cell , HMEC . Expression of miR-1255b , miR-193b* , and miR-148b* were detected in these lines . Mean ± SD of four independent experiments is shown . ( B–D ) Expression of DDR genes is impacted by miR-1255b , miR-193b* , and miR-148b* . MDA-MB231 cells were transfected with control mimic or mimics for miR-1255b , miR-193b* , and miR-148b* and mRNA levels of predicted and prioritized DDR genes were analyzed by qRT-PCR using gene-specific primers and normalized to GAPDH . Mean ± SD of three independent experiments is shown ( B ) . ( C and D ) Cell lysates were then analyzed by immunoblot for factors which had ≥50% reduction in mRNA in cells transfected with miR-1255b , miR-193b* , and miR-148b* . Images were quantified by ImageJ software and the mean ± SD of three independent experiments is graphically shown , * indicates p<0 . 05 . ( E–G ) Interaction of target transcripts with miR-1255b , miR-193b* , and miR-148b* . MDA-MB231 cells were transfected with biotinylated-control mimics or biotinylated mimics for miR-1255b , miR-193b* , and miR-148b* as a single ( F ) or a combination ( G ) . The immunoprecipitated RNA was analyzed by qRT-PCR using gene-specific primers and normalized to GAPDH . Mean ± SD of three independent experiments is shown and statistical significance of enrichment of specific gene transcripts is indicated by * ( p<0 . 05 ) . The principle steps of the method are illustrated in Figure 3E . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 00610 . 7554/eLife . 02445 . 007Figure 3—figure supplement 1 . Expression of the excluded miRNAs . Endogenous expression of the indicated miRNAs that were excluded from selection was quantified by qRT-PCR ( normalized to 5srRNA ) and represented as relative expression . Mean ± SD of three independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 00710 . 7554/eLife . 02445 . 008Figure 3—figure supplement 2 . The effect of miRNAs on cell cycle . Cell cycle profile was examined after transfection of control , miRNA-1255b , miR-193b* , or miR-148b* and cell population of each phase is graphically represented . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 008 We rationalized that the 3 miRNAs identified in the TNBC lines above ( miR-1255b , miR-148b* , and miR-193b* ) are likely to influence HR by regulating the expression of genes involved in the DNA damage response ( DDR ) . We used a two-pronged approach to identify these genes . First , we used a collection of target prediction algorithms to identify genes targeted by these 3 miRNAs and compared it to the list of DDR proteins ( see the ‘Materials and methods’ ) . Second , we used a candidate-based approach using the prediction algorithm RNA22 to screen all the genes implicated in DDR for miRNA recognition element of miR-1255b , miR-148b* , and miR-193b* . This gave us a list of 18 proteins which have been previously implicated in HR-mediated repair of DSBs . We then assessed the impact of over-expressing miR-1255b , miR-148b* , and miR-193b* on the mRNA level of these 18 genes ( Figure 3B ) . There was at least a 50% reduction in the transcripts of EXO1 , MDC1 , BRCA1 , BRCA2 , BRIP1 , RAD51 , and PALB2 in cells over-expressing either miR-1255b , miR-148b* , or miR-193b* . We next determined the impact of these miRNAs on the protein levels of the putative targets . Over-expressing miR-1255b reduces the cellular levels of BRCA1 and BRCA2 , miR-148b* reduces RAD51 and miR-193b* reduces BRCA1 , BRCA2 , and RAD51 ( Figure 3C ) , and this impact is not related to an altered cell cycle ( Figure 3—figure supplement 2 ) . Although over-expression of miR-1255b , miR-148b* , and miR-193b* correlates with significant increase in the transcripts of certain DNA repair genes ( such as MRE11 and NBS1 for miR-1255b and miR-193b*; EXO1 for miR-193b* ) , there was no detectable increase in protein levels ( Figure 3C , Figure 3D ) . It is possible that the increase in transcript levels of the DNA repair genes is due to miRNA-mediated down-regulation of transcriptional repressors or mRNA-destablizing proteins , such as AUF1 . miRNAs typically suppress gene expression by direct association with target transcripts . To test for association of miR-1255b , miR-148b* , and miR-193b* with their respective target transcripts , we adopted a recently described method for capturing miRNA–mRNA complexes using streptavidin-coated beads from cells transfected with biotinylated forms of the miRNA mimics ( Orom and Lund , 2007; Lal et al . , 2011 ) . The amount of BRCA1 , BRCA2 , and RAD51 transcripts was measured in the pull-downs ( GAPDH served as a control transcript ) and the enrichment was assessed relative to pull-down with biotinylated control mimic ( Figure 3E–G ) . Strikingly , consistent with our previous results , miR-1255b selectively pulled-down BRCA1 and BRCA2 transcripts but not the RAD51 transcript ( Figure 3F ) . Conversely , miR-148b* pulled-down RAD51 transcript but not BRCA1 and BRCA2 transcripts , and miR-193b* pulled down all three transcripts . Furthermore , we used a combination of the biotinylated mimics to investigate overlapping targets , combining miR-1255b with miR-193b* and miR-148b*with miR-193b* ( Figure 3G ) . Relative to the individual counterparts , the miR-1255b/miR-193b*combination immunoprecipitated increased the amount of BRCA1 and BRCA2 mRNAs and the miR-148b*/miR-193b* combination immunoprecipitated increased RAD51 mRNA ( Figure 3G ) . We next scanned the BRCA1 , BRCA2 , and RAD51 transcripts for miR-1255b , miR-148b*or miR-193b* binding sites or miRNA recognition element ( MRE ) s . Canonical MREs are predicted based on the seed rule , that is , the target site within 3′ UTR forms Watson–Crick pairs with bases at positions 2 through 7 or 8 of the 5′ end of the miRNA ( Lewis et al . , 2005 ) . However , it is noteworthy that there is considerable evidence of ‘seedless’ or non-canonical association of miRNAs with target transcripts including sites in the coding sequence ( CDS ) ( Didiano and Hobert , 2006; Grimson et al . , 2007; Chi et al . , 2009; Lal et al . , 2009a; Hafner et al . , 2010; Shin et al . , 2010; Chi et al . , 2012; Loeb et al . , 2012 ) . None of the three transcripts BRCA1 , BRCA2 , and RAD51 had MREs with a canonical seed region in the 3′UTR for miR-1255b , miR-148b* , and miR-193b* . We have previously used the PITA algorithm to expand the search criterion and identify non-canonical MREs for specific miRNA/mRNA combinations ( Lal et al . , 2009a ) . The PITA algorithm identifies base pairing beyond the 5′end of the miRNA , allows G:U wobbles or seed mismatches and predicts MREs across the entire target transcript , not just the 3′UTR . Using this approach , we observed that consistent with our experimental data , there were putative non-canonical MREs for miR-193b* , miR-1255b , and miR-148b* in the transcripts of BRCA1 , BRCA2 , and RAD51 ( Figure 4A ) . Most of these sites had moderate to low evolutionary conservation ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 02445 . 009Figure 4 . Predicted miRNA recognition sites ( MREs ) of miRNAs and their impact on targets . ( A ) Predicted MREs were obtained from PITA ( http://genie . weizmann . ac . il/pubs/mir07/mir07_prediction . html ) and their mutants were generated by mutating nucleotides providing complementarity and G-U wobble to corresponding miRNAs . The region where MRE is located in the gene is indicated in the parentheses . CDS: coding sequence , 3′UTR: 3′ untraslated region . ( B ) Luciferase reporter assay to assess direct interaction of miR-1255b , miR-193b* , and miR-148b* with BRCA1 , BRCA2 , and RAD51 . Combinations of predicted miRNA recognition sites ( MREs ) for each putative target transcript of miR-1255b , miR-193b* , and miR-148b* were cloned into the luciferase reporter vector and transfected in MDA-MB231 cells along with the indicated miRNA mimics . Renilla luciferase activity of the reporter was measured 48 hr after transfection by normalization to an internal firefly luciferase control . Mean ± SD of three independent experiments is shown and statistical significance is indicated by * ( p<0 . 05 ) . ( C ) Luciferase reporter assay for individual MREs for each target of miRNAs was performed in the same way as described in Figure 4B . Mean ± SD of three independent experiments is shown and statistical significance is indicated by * ( p<0 . 05 ) . ( D ) Luciferase reporter assay with miR-1255b , miR-193b* , and miR-148b* ANTs . Combinations of predicted miRNA recognition sites ( MREs ) in the luciferase vector for each putative target transcript of miR-1255b , miR-193b* , and miR-148b* were transfected in MDA-MB231 cells along with the indicated miRNA ANTs . Renilla luciferase activity of the reporter was measured 48 hr after transfection by normalization to an internal firefly luciferase control . Mean ± SD of three independent experiments is shown and statistical significance is indicated by * ( p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 00910 . 7554/eLife . 02445 . 010Figure 4—figure supplement 1 . Conservation of predicted miRNA recognition sites ( MREs ) of miRNAs . Predicted MRE sequences in each miRNA target genes were aligned across different species . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 010 To verify further that BRCA1 , BRCA2 , and RAD51 are targets of miR-1255b , miR-148b* , and miR-193b* and to confirm that the interaction is mediated by the predicted MREs , we used the luciferase reporter assay which is a surrogate for target protein . The MREs were cloned in the 3′UTR of the luciferase gene , and expression monitored in cells transfected with mimics for miR-1255b , miR-193b* , and miR-148b* ( Figure 4A , B ) . As anticipated , there was significant decrease in luciferase activity , and this was ‘rescued’ by point mutations that disrupt base pairing between miR-1255b , miR-193b* , and miR-148b* and their corresponding MREs in BRCA1 , BRCA2 , and RAD51 ( Figure 4A , B ) . Analyzing all the MREs individually , we compared the relative impact of each MRE on luciferase activity ( Figure 4C ) . To confirm the interaction of endogenous miR-1255b , miR-193b* , and miR-148b* with specific MREs in the BRCA1 , BRCA2 , and RAD51 transcripts , we adopted a loss-of-function approach . We used miRNA inhibitors ( also known as antagomirs , ANTs ) that are single-stranded chemically enhanced oligonucleotides designed to irreversibly bind endogeneous miR-1255b , miR-193b* and miR-148b and suppress their activity . We estimated luciferase activity after inhibiting the miRNAs using antagomirs and , consistent with our previous results , found that inhibition of miR-1255b enhanced luciferase activity of the BRCA1 and BRCA2 reporter construct , inhibition of miR-148b* enhanced luciferase activity of the RAD51 reporter construct , and inhibition of miR-193b* enhanced luciferase activity of the BRCA1 , BRCA2 , and RAD51 reporter constructs ( Figure 4D ) . The specificity of the MREs was further validated as the mutant versions of the luciferase reporters were immune to the antagomirs ( Figure 4D ) . The luciferase reporter assays with MREs provide important information regarding the miRNA/mRNA association but have limited physiological relevance . To determine the functional significance of non-canonical MREs in the BRCA1 , BRCA2 , and RAD51 transcripts we generated expression constructs without the MREs by either deletion ( MREs in 3′UTR ) or mutation ( MREs in CDS ) of them . Next , MDA-MB231 cells were co-transfected with ( i ) miR-1255b and BRCA1 or BRCA2 expression plasmid lacking miR-1255b binding sites; ( ii ) miR-193b* and BRCA1 or BRCA2 or RAD51 expression plasmid lacking miR-193b* binding sites; ( iii ) miR-148b* and a RAD51 expression plasmid lacking miR-148b* binding sites . First , the BRCA1 , BRCA2 , and RAD51 expression constructs lacking the specific MREs completely restored the expression of these genes in the presence of the corresponding miRNA mimic further validating the predicted MREs ( Figure 5A , lower panel ) . Furthermore , in regard to ABT888 sensitivity , expression of BRCA1 or BRCA2 significantly ‘rescued’ the impact of miR-1255b , expression of BRCA1 or BRCA2 or RAD51 significantly ‘rescued’ the impact of miR-193b* , and expression of RAD51 significantly ‘rescued’ the impact of miR-148b* ( Figure 5A , upper panel ) . Together , these results strongly suggest that miR-1255b , miR-193b* , and miR-148b* influence HR-mediated repair of DSBs and PARP inhibitor sensitivity by regulating expression of BRCA1 , BRCA2 , and RAD51 . 10 . 7554/eLife . 02445 . 011Figure 5 . Impact of miRNAs on DSB repair in different phases of the cell cycle . ( A ) Rescue of the impact of miRNAs on ABT888 sensitivity . MDAMB231 cells were transfected with control miRNA or indicated miRNA mimics with or without target gene cDNAs ( lacking MREs ) and treated with vehicle or ABT888 , before viability assay by ATP quantification . Expression of each target protein is examined by immune blot . ( B and C ) Expression of miRNAs and target transcripts in synchronized cells . MDAMB231 ( B ) or MCF10A ( C ) cells were synchronized with mimosine and the relative amount of miR-1255b , miR-193b* , and miR-148b* or BRCA1 , BRCA2 , and RAD51 mRNA for G1- or S-phase was determined by qRT-PCR ( normalized to RNU1 or GAPDH , respectively ) . Mean ± SD of three independent experiments is shown and statistical significance is indicated by * ( p<0 . 05 ) . ( D–F ) Impact of inhibiting miRNAs on targets in G1 cells . MDAMB231 cells were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* as a single ( D ) or a combination ( F ) . Subsequently , the cells were synchronized with mimosine and BRCA1 , BRCA2 , and RAD51 mRNA was assessed by qRT-PCR ( normalized to GAPDH ) in the G1 and/or S-phase ( D and F ) . Cell lysates from G1 cells were analyzed by immunblot for BRCA1 , BRCA2 , and RAD51 ( E ) . Images were quantified by ImageJ software and the mean ± SD of three independent experiments is shown , * indicates p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 01110 . 7554/eLife . 02445 . 012Figure 5—figure supplement 1 . The impact of miRNA antagomirs ( ANTs ) on cell cycle progression . MDAMB231 cells or MCF10A cell ( data not shown ) were transfected with control antagomir or antagomirs for miR-1255b , miR-193b* , and miR-148b* for 48 hr . Subsequently , the cells were synchronized with 500 µM mimosine for 24 hr and collected at indicated times after release into growth media . The cell cycle distribution was analyzed by FloJo and representative images of the cell cycle profile from three independent experiments are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 012 Ectopic over-expression of miR-1255b , miR-193b* , and miR-148b* in cells down-regulates HR factors and influences PARP inhibitor sensitivity . However , the cellular function of endogenously expressed miR-1255b , miR-193b* , and miR-148b* remains unclear . To address this issue , we quantified their expression in cycling cells , specifically during the G1 to S transition . Consistent with previous reports ( Gudas et al . , 1996; Vaughn et al . , 1996; Chen et al . , 1997 ) , we observe that mRNA levels of BRCA1 , BRCA2 , and RAD51 are enhanced in the S-phase ( Figure 5B , C ) . miR-1255b , miR-193b* , and miR-148b* inversely correlate with their target transcripts , and are significantly down-regulated as cells move into the S-phase . This was observed both in the breast tumor line , MDA-MB231 ( Figure 5B ) and the non-tumorigenic breast epithelial line MCF10A ( Figure 5C ) . Antagonizing these miRNAs induces a specific increase in target transcripts ( Figure 5D ) and proteins ( Figure 5E ) in the G1-phase , that is transfection of the miR-1255b ANT causes significant increase in BRCA1 and BRCA2 and not RAD51 , miR-148b* ANT increases RAD51 and miR-193b* ANT increases BRCA1 , BRCA2 , and RAD51 in G1 cells ( Figure 5D , E , cell cycle profiles are shown in Figure 5—figure supplement 1 ) . Combined inhibition of miRNAs with overlapping targets causes a synergistic increase in target transcripts in G1 ( Figure 5F ) . There is a significant increase in BRCA1 and BRCA2 transcripts in cells co-transfected with miR-1255b ANT and miR-193b*ANT , and a significant increase in RAD51 mRNA in cells co-transfected with miR-148b*ANT and miR-193b*ANT . These results suggest that in proliferating cells miRNAs suppress HR in the G1 phase . There is emerging evidence that loss of NHEJ factors allows the initiation of HR in G1 cells and leads to the persistence of unrepaired DSBs ( Helmink et al . , 2011; Escribano-Diaz et al . , 2013 ) . Therefore , it is feasible that up-regulation of HR factors due to the inhibition of miR-1255b , miR-193b* , and miR-148b* in G1 cells may also de-stabilize the balance of the HR and NHEJ pathways and thereby impede DSB repair . To test this hypothesis , asynchronous cells were transfected with antagomirs for miR-1255b , miR-193b* and miR-148b* and stained with Cyclin A to visualize S/G2 cells and distinguish them from G1 cells; γ-H2AX was utilized as a marker for DSBs . Inhibition of miR-1255b and miR-193b* causes a moderate but statistically significant increase in basal γ-H2AX specifically in G1 cells ( Figure 6—figure supplement 1A ) . The impact of inhibiting the miRNAs on DSB repair in G1 cells is even more pronounced when these cells are exposed to IR ( Figure 6A , B ) . BRCA1 promotes end resection at DSBs via its interaction with CtIP ( Yun and Hiom , 2009 ) and that is the key step in blocking NHEJ . The significant increase in residual DSBs in cells transfected with antagomirs for miR-1255b , miR-193b* and miR-148b* is ‘rescued’ by silencing CtIP ( Figure 6A , B; silencing efficacy shown in Figure 6—figure supplement 1B ) , suggesting that increased levels of BRCA1 by miRNA inhibition in G1 promotes CtIP-mediated resection . 10 . 7554/eLife . 02445 . 013Figure 6 . Impact of inhibiting miRNAs on DSB repair . ( A–D ) Impact of inhibiting miRNAs on DSB repair in the G1 phase of MDA-MB231 cells . Cells were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* with or without 20 nM CtIP siRNA , exposed to IR ( 5 Gy ) and stained for γ-H2AX ( green ) ( A ) or RPA2 ( green ) ( C ) , cyclin A ( red ) and 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( blue ) . The images were captured by fluorescence microscopy and the number of γ-H2AX foci ( A ) or RPA2 foci ( C ) was calculated from 100 cells . Mean ± SD of three independent experiments is graphically represented ( B and D ) . * indicates p<0 . 05 . ( E–I ) Impact of inhibiting miRNAs on DSB repair in different phase of RPE-1 cells . RPE-1 cells expressing the Fucci system ( illustrated in Figure 6E ) were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* with or without 20 nM CtIP siRNA , exposed to IR ( 5 Gy ) and stained for γ-H2AX ( green ) and 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( blue ) . The images were captured by fluorescence microscopy and the number of γ-H2AX foci in G1 cells ( red , mKO2-Cdt1 ) was calculated from 100 cells ( F and G ) . The images were captured by fluorescence microscopy and the number of γ-H2AX foci ( green foci ) in S/G2/M cells ( green background , mAG-Geminin ) was calculated from 100 cells ( H ) . Mean ± SD of three independent experiments is graphically represented ( I ) . * indicates p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 01310 . 7554/eLife . 02445 . 014Figure 6—figure supplement 1 . The impact of miRNAs on DNA repair during cell cycle . ( A ) MDAMB231 cells were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* and stained for γ-H2AX ( green ) , cyclin A ( red ) and 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( blue ) . The images were captured by fluorescence microscopy and the number of γ-H2AX foci was calculated from 100 cells each in G1 phase and S/G2 phase . Mean ± SD of three independent experiments is shown in the right panel . * indicates p<0 . 05 . ( B ) MDAMB231 cells were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* with or without 20 nM CtIP siRNA . CtIP mRNA expression was assessed in control ANT ± CtIP siRNA transfected MDAMB231 cells by qPCR and normalized to GAPDH . * indicates p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 014 RPA2 foci mark the presence of single-strand DNA ( ssDNA ) , a consequence of end resection . Consistent with resection of DNA ends , and production of ssDNA , we observe a significant increase in RPA2 foci in the G1 phase of cells transfected with antagomirs for miR-1255b , miR-193b* and miR-148b* and again the depletion of CtIP negates the formation of RPA2 foci in these cells ( Figure 6C , D ) . To further confirm the cell cycle phase specificity of this phenotype using a different approach and a diploid cell line with relatively few genomic abnormalities , we utilized the Fucci system ( Sakaue-Sawano et al . , 2008 ) to visualize the G1 phase ( mKO2-Cdt1 , red fluoresence ) in hTERT-immortalized retinal pigment epithelial cell line ( RPE-1 ) cells ( Figure 6E–G ) . Fucci ( fluorescent ubiquitination-based cell-cycle indicator ) utilizes cell cycle-dependent degradation of Cdt1 and Geminin to mark G1 and S/G2/M cells by fusing the red ( mKO2 ) and green ( mAG ) fluorescent proteins to Cdt1 and Geminin , respectively ( illustrated in Figure 6E ) . Consistent with the previous results , inhibition of miR-1255b , miR-193b* , and miR-148b* in RPE-1 causes a CtIP-dependent increase in residual DSBs in G1 cells ( Figure 6F , G ) , but this effect is not observed in the S-phase ( Figure 6H , I ) . Together , these results strongly suggest that miR-1255b , miR-193b* , and miR-148b* play a role in preventing the initiation of deleterious HR-mediated DSB repair in the G1 phase . The two major mechanisms proposed to cause LOH are deletion of chromosomal fragments/chromosome loss and mitotic recombination between homologous alleles ( Lasko et al . , 1991 ) . Mitotic recombination is a consequence of HR-mediated repair of DSBs in the G1 phase . Recent studies demonstrate that homologous chromosomes are in close proximity at DSB sites in the G1 phase ( Gandhi et al . , 2012 , 2013 ) further supporting the notion that HR-mediated DSB repair is feasible in G1 cells and this in turn could lead to LOH . Based on our experimental data , the prediction would be that loss of miR-1255b , miR-193b* , and miR-148b* would correlate with enhanced LOH events that are occurring due to mitotic recombination . To test this hypothesis , we utilized the data from the Cancer Genome Atlas Research Network , 2011 ( TCGA , [2011] ) , where the processed LOH data for high-grade serous ovarian tumors is readily available . The HR-pathway and expression of BRCA1 and BRCA2 is extremely relevant in ovarian tumors , and we observed that there is detectable expression of miR-1255b , miR-148b* , and miR-193b* in a panel of ovarian tumor lines ( Figure 7A ) . The increased γ-H2AX foci correlating with inhibition of miR-1255b , miR-193b* , and miR-148b* is not observed in a BRCA1-mutant ovarian cell line UWB1 . 289 ( Figure 7—figure supplement 1 ) confirming the importance of BRCA1 in the phenotype induced by these miRNAs . 10 . 7554/eLife . 02445 . 015Figure 7 . Correlation of LOH with loss of miRNAs . ( A ) miRNA expression profile in a panel of ovarian cancer lines . Endogenous expression of indicated miRNAs was quantified by qRT-PCR ( normalized to 5srRNA ) and represented relative to non-tumorigenic ovarian epithelial cell , HIO-80 . Expression of miR-1255b , miR-193b* , and miR-148b* were detected in these lines . ( B and C ) Correlation of LOH with deletion of miRNAs in TCGA data set . Box plots show the frequency of ( B ) LOH or ( C ) somatic copy number amplification or deletion ( SCNΔ ) in the 418 high-grade serous ovarian tumors from TCGA that have no amplifications or deletions of any of these 3 miRNAs ( WT ) , against those with deletion of 1255b ( either −1 or −2 ) , 148b* or 193b* . The LOH events are >1 Mb . ( D ) Correlation of LOH with deletion of miRNAs in DF/HCC data set . Box plot shows the frequency of LOH in 47 high-grade serous ovarian tumors that have no amplifications or deletions of any of these 3 miRNAs ( WT ) , against those with deletion of miR-1255b ( either 1 or 2 ) , miR-148b* or miR-193b* . The LOH events are >1 Mb . ( E ) Correlation of BRCA1 expression with deletion of miRNAs in TCGA data set . Box plot shows expression levels of BRCA1 in the 418 high-grade serous ovarian tumors from TCGA that have no amplifications or deletions of any of these 3 miRNAs ( WT ) , against those with deletion of miR-1255b ( either 1 or 2 ) , miR-148b* or miR-193b* . In ( B–E ) , statistical significance was calculated using one tailed Mann Whitney's U test and significant differences ( p<0 . 05 ) indicated with asterisk . Width of the bars indicates the number of samples . The horizontal line indicates the median of the WT set . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 01510 . 7554/eLife . 02445 . 016Figure 7—figure supplement 1 . miRNA dependent regulation of DSB repair during cell cycle . Rescue of the impact of miRNA inhibition on DNA repair in ovarian cancer cells . UWB1 . 289 BRCA1-deficient ovarian cancer cells were transfected with control ANT or ANTs for miR-1255b , miR-193b* , and miR-148b* and stained for γ-H2AX ( green ) , cyclin A ( red ) and 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( blue ) . The images were captured by fluorescence microscopy and the number of γ-H2AX foci was calculated from 100 cells . Mean ±SD of 3 independent experiments is shown in the right panel . * indicates p <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 016 We compared the number of LOH events for a cohort ( 417 tumors ) of high-grade serous ovarian tumors that have deletions of miR-1255b , miR-148b* , and miR-193b* and those that are copy neutral at those loci . We observed that the number of LOH events is significantly higher in tumors with deletions of miR-148b* and miR-193b* relative to the tumors that are diploid copy neutral at these loci . The number of LOH events was also moderately higher ( One tailed Mann Whitney's U test; p<0 . 08 ) in tumors with deletion of miR-1225b compared to the copy neutral ones ( Figure 7B ) . It is noteworthy that mature miR-1255b is encoded from two distinct genomic loci ( miR-1255b-1 on chr4: 36427988-36428050 and miR-1255b-2 on chr1: 167967898-167967964 ) . In our deletion analysis , we correlated LOH events with deletion of either miR-1255b-1 or miR-1255b-2 , therefore the moderate correlation of LOH with loss of miR-1255b may be due to the presence of the second copy . The limited sample size with deletions of both miR-1255b-1 and miR-1255b-2 precludes any statistical analysis . Reflecting overall genomic instability , there is a significant increase in somatic amplifications and deletions in tumors with deletions in miR-1255b , miR-148b* , and miR-193b* ( Figure 7C ) . To validate our analysis in a second cohort , we obtained processed LOH and somatic copy number alteration data for 47 high-grade serous ovarian tumors , which had been processed at Dana-Farber/Harvard Cancer Center ( DF/HCC ) . The details of these samples have been previously described ( Wang et al . , 2012b ) . We compared the number of copy neutral LOH events between the samples that have deletions of miR-1255b , miR-148b* , and miR-193b*and those that are copy neutral at those loci ( Figure 7D ) . Consistent with the analysis of the TCGA data , the number of LOH events is higher in the tumors with deletions of miR-1255b , miR-148b* , and miR-193b* relative to the samples that are diploid copy neutral at these loci . The statistical significance was limited due to small size of the data set , but the patterns were consistent even with alternative cutoffs . In our analysis , we compared the copy neutral LOH events to specifically focus on LOH events that are likely to occur due to recombination between homologous alleles . Finally , in agreement with our experimental data , there is a significant increase in BRCA1 expression in the tumors with deletions in miR-1255b and miR-193b* ( Figure 7E ) , which we speculate is the underlying reason for the increase in LOH and overall genomic instability .
Contrary to other cellular processes and signaling pathways , there has been limited investigation of miRNAs in DSB repair ( Wan et al . , 2011; Chowdhury et al . , 2013 ) . In the few recent studies ( Hu et al . , 2010; Yan et al . , 2010; Wang et al . , 2011 , 2012a; Dimitrov et al . , 2013; Huang et al . , 2013; Neijenhuis et al . , 2013 ) , including ones from our group ( Lal et al . , 2009b; Moskwa et al . , 2011 ) , the focus has been the impact of a single miRNA on an individual DSB protein . A key conceptual limitation to this approach is that miRNAs typically target hundreds of transcripts and are more likely to have a profound impact on a pathway like the HR repair pathway by regulating several HR factors . Here , we systematically identified miRNAs influencing HR ( miR-1231 , miR-1255b , miR-148b* , miR-876-3p , miR-221* , miR-193b* , and miR-185* ) and observed that three of these miRNAs ( miR-1255b , miR-148b* , and miR-193b* ) regulate expression of BRCA1 , BRCA2 , and RAD51 . Intriguingly , they lack canonical binding sites in the 3′UTR of BRCA1 , BRCA2 , and RAD51 , but interact with these transcripts via non-canonical MREs . These results underline the importance of binding of miRNAs to their target transcripts that are not restricted to base pairing at ‘seed regions’ , and highlight the shortcomings of current prediction algorithms in identifying functional targets of miRNAs . The choice of DSB repair pathway is critical for maintaining genomic stability ( Chapman et al . , 2012b ) . Specifically 53BP1 and its associated factors ( RIF1 and PTIP ) promote NHEJ and counter BRCA1 thereby balancing the process of DSB repair between HR and NHEJ ( Bouwman et al . , 2010; Bunting et al . , 2010; Chapman et al . , 2012a; Callen et al . , 2013; Di Virgilio et al . , 2013; Escribano-Diaz et al . , 2013; Zimmermann et al . , 2013 ) . Loss of the 53BP1-associated protein RIF1 or the protein H2AX allows the BRCA1/CtIP-mediated resection of DNA ends in G1 cells impeding NHEJ and leading to the persistence of DNA lesions ( Helmink et al . , 2011; Escribano-Diaz et al . , 2013 ) . Conversely , it is likely that the ectopic increase in expression of BRCA1 and other HR factors in the G1 phase will also disrupt the balance of DSB repair . Our results represent an intriguing example of an organized , miRNA-driven physiological system for controlling the level of endogenous HR factors in the G1 phase of the cell cycle . Down-regulating the expression of HR factors , particularly , an HR-initiating factor like BRCA1 in the G1 cells is necessary for maintaining genomic stability . Inhibiting miR-1255b and miR-193b* allows increased BRCA1 expression in G1 cells leading to CtIP-mediated resection and a potential block in NHEJ . This will disrupt the balance of HR and NHEJ ( Figure 8 ) with significant increase in genomic instability . Furthermore , execution of HR-mediated repair in G1 cells would result in LOH which has the potential of revealing recessive oncogenic mutations . The loss of miR-1255b , miR-148b* , and miR-193b* largely correlates with increased genomic instability and a higher frequency of LOH , which is consistent with our experimental data . In some cases , the correlations are moderate and not statistically significant , and this may reflect the heterogeneity of primary tumors , limited sample size and the technical shortcomings of large sets of genomic data . These shortcomings notwithstanding , there is a clear pattern that emerges from our experimental data and clinical analysis that strongly suggests that miRNAs regulate the optimal expression of HR proteins in the course of the cell cycle and prevent ectopic activation of the HR pathway . A single copy loss of BRCA1 ( Konishi et al . , 2011 ) , BRCA2 ( Popova et al . , 2012 ) , or RAD51 ( Smeenk et al . , 2010 ) impacts genomic stability giving credence to the idea that a miRNA-mediated alteration in the cellular levels of these proteins may significantly impact the DNA repair process . Furthermore , for miR-1255b and miR-193b* , there is a moderate reduction in multiple HR factors , which further compounds their impact on the HR repair pathway . The underlying mechanism of cell cycle-dependent expression of miRNAs regulating HR factors remains to be investigated in future studies . 10 . 7554/eLife . 02445 . 017Figure 8 . A model of miRNA dependent regulation of DSB repair during cell cycle . Model of miRNA-dependent regulation of DSB repair during cell cycle . The balance of HR and NHEJ in dividing cells is crucial for efficient DSB repair . NHEJ is the preferred pathway in the G1 phase with 53BP1 and the Ku complex binding the broken DNA end . miRNAs ( such as miR-1255b and miR-193b* ) suppress HR factors , particularly BRCA1 , preventing end resection of the DNA lesions . However , when these miRNAs are inhibited or deleted it may disrupt the correct choice of DSB repair pathway . Ectopic over expression of BRCA1 will allow CtIP-mediated resection in G1 cells , preventing NHEJ . Furthermore , HR-mediated repair in G1 is detrimental to cell health as it would lead to LOH . In S-phase , DSBs are predominantly repaired by HR and down-regulation of miRNAs targeting BRCA1 , BRCA2 and RAD51 may be important in ensuring efficient HR-mediated DSB repair . Over-expression of miRNAs ( such as miR-1255b , miR-193b* , and miR-148b* ) targeting BRCA1 , BRCA2 and RAD51 in the S phase will impede various steps of HR , and the HR deficiency will sensitize these cells to PARP inhibitors . DOI: http://dx . doi . org/10 . 7554/eLife . 02445 . 017
BRCA1 MRE1 and BRCA2 MRE2 located in the gene coding sequence were mutated by site-directed mutagenesis . The primers used are as follows: BRCA1 MRE1-1-US , AAGTTTCTATCAGGCAAAGTATGGGCT , BR1_Res1-1-DS , AGCCCATACTTTGCCTGATAGAAACTT; BR1_Res1-2-US , CTAAAAGATGAAGAAACTATCAGGCAAA , BR1_Res1-2-DS , TTTGCCTGATAGTTTCTTCATCTTTTAG; BR2_Res2-US , AAATAGTCATATAAGGGCTCAGATGTTATTT , BR2_Res2-DS , AAATAACATCTGAGCCCTTATATGACTATTT . miRNA screen was carried out in triplicates using miRNA mimic libraries from Applied Biosystems ( Pre-miR miRNA Precursor Library , 885 miRNAs ) and Qiagen ( miScript miRNA Mimic Library , 875 miRNAs ) , both the libraries were based on miRbase 14 . MDA-MB231 cells were plated on 384-well plate at 1000 cells/well density and reverse transfected with 50 nM control miRNA , miRNAs , or BRCA2 siRNA using Lipofectamine RNAiMax transfection reagent before being treated with 20 µM ABT888 next day . In 3 days , cells were subjected to viability test using Celltiter Glo ( #G7572; Promega , Madison , WI ) . The average luminescence and Z-score was calculated . % viability control was calculated for each miRNA and the cut off was set at 75% control viability and Z-score of positive control , BRCA2 siRNA . Z-score was calculated by the following equation where x is luminescence value for each miRNA , μ is negative control mean , and σ is negative control standard deviation: z = ( x−μ ) /σ . Clonogenic assays and colorimetric assays with MDAMB231 or 21NT cells treated with PARP inhibitors were done as previously described ( Moskwa et al . , 2011 ) . We modified it to seed 500 cells per six-well plates for clonogenic assay and 1000 cells on 96-well plates for colorimetric assay . Surviving colonies of >1 mm diameter or live cells were quantified . HR assay was performed as described in Moskwa et al . ( 2011 ) . Immunofluorescence in MDAMB231 cells were done as previously described Lee et al . ( 2010 ) using RAD51 ( Santa Cruz , Dallas , TX ) , H2AX ( Cell Signaling , Danvers , MA ) , γ-H2AX ( Cell Signaling ) , or Cyclin A ( Santa Cruz ) . Total RNA was prepared and expression was analyzed by qRT-PCR as described in Moskwa et al . ( 2011 ) . Gene-specific primers used for qRT-PCR are as follows:Mre11A-F , TCTGCCTTTAGTGCTGAT GAC , R , GCTCTTCCTCTTTGAGACCC;RAD50-F , GTGCGGAGTTTTGGAATAGAG , R , GAGCAACCTTGGGATCGTGT;NBS1-F , CACTCACCTTGTCATGGTATCA , R , CCGTCCTGACAGATCAACATT;ATM-F , AAGGCTATTCAGTGT GCGAG , R , GGCTCCTTTCGGATGATGGA;ATR-F , CCAGCATTCTCCAGGTGACA , R , AGCCAGCATTCTCCAACCA T;ATRX-F , CAAGGTCTGCAAAGAAAGCAG , R , TGGAATCATCATTTTCATCTTCC;CtIP-F , GGCTTATGTGATCG CTGTGC , R , ACAGCATCAAGCAGCTGAGC;RPA2-F , TTCACAGGTCACTAT TGTGGG , R , GAACAAAAAGA GCCTGGTAGC;TOPBP1-F , TTTCCGTGCTGTGG TCTCAC , R , GAAACTCCAGGAC GTCCCAG;Abraxas-F , TGCA GGAGCATTTTTCAAACC , R , GTATGTCCACTGGTTTTAGCC;RAD52-F , TA CAATGGCTGGGCACACT C , R , TCCTTCCTTGCCTTCTCCAA;MDC1-F , GAGACATCTGAGGAAATACAAG , R , TCTTTCTGGTAGCAGTTTCTCA;BRCA1-F , AGGCAACTTATTGCAGTGTGG , R , ACTTTTCTGGATGCCTCTCAG;EXO1-F , AGCCAAAGGTGAACCTACTGA , R , AGCTTCCGAGACTTTCCCCT;RAD51-F , GGGTCGAGGTGAGCTTTCAG , R , GG GCGATGATATTTCCTCCA;BRCA2-F , GCCAAGTCATGCCACACATT , R , TGTGC CATCTGGAGTGCTT T;PALB2-F , TTGTTTGTCTCAGCAGGATCTC , R , CTTGGGTGTCATCTGTTCTTT G;BRIP1-F , ATGAACC AAGGAACTTCACGTC , R , CTGCTGTGTAGTTTCTAAGGGTC . We adopted unbiased target prediction algorithms for miRNA-1255b , miRNA-193b* , and miRNA-148b* and each algorithm predicted a list of targets as follows: Comparative target prediction analysis for miRNA-1255b , miRNA-193b* , and miRNA-148b*:Number of predicted targetsNumber of predicted targetsNumber of predicted targetsalgorithmURLmiR-1255bmiR-193b*miR-148b*TargetScanhttp://www . targetscan . org/134N/AN/AmiRWalkhttp://www . umm . uni-heidelberg . de/apps/zmf/mirwalk/micrornapredictedtarget . html9068731352miRDBhttp://mirdb . org/miRDB/131268351 The prediction list was compared with an updated list of Human DNA Repair Genes ( http://sciencepark . mdanderson . org/labs/wood/dna_repair_genes . html#Human%20DNA%20Repair%20Genes; Wood et al . , 2001 ) . However , there were no overlapping candidates from the predictions and the DDR genes . Therefore , we used a candidate-based prediction approach using RNA22 ( http://cm . jefferson . edu/rna22v2/ ) ( http://cbcsrv . watson . ibm . com/rna22 . html ) and PITA ( http://genie . weizmann . ac . il/pubs/mir07/mir07_data . html ) , to analyze the Human DNA Repair Gene list that resulted in a list of DDR genes predicted as targets of miRNAs of our interest . 18 of them are implicated in HR-mediated repair and they were examined in Figure 3B . The immunoblots were done as described previously ( Lee et al . , 2010; Moskwa et al . , 2011 ) with BRCA1 ( #OP92; Calbiochem , Billerica , MA ) , BRCA2 ( #ab16825-100; Abcam , Cambridge , MA ) , RAD51 ( #sc-8349; Santa Cruz ) , MDC1 ( #sc-27737; Santa Cruz ) , PalB2 ( #A301-246A; Bethyl , Montgomery , TX ) , BRIP1 ( #4578S; Cell Signaling ) , Exo1 ( #sc-19941; Santa Cruz ) , H2AX ( #2595S; Cell Signaling ) , and γ-H2AX ( #9718S; Cell Signaling ) antibodies and α-tubulin ( #T5168; Sigma , St . Louis , MO ) antibodies . MDAMB231 cells were plated at 0 . 3 × 106 cells/well on 6 well-plate overnight and transfected with biotinylated control miRNA ( C . elegans miRNA [Bi-cel-miR-67] ) or biotinylated miR-1255b , miR-193b* , or miR-148b* ( Sigma ) . The cells were harvested 6 hr after transfection in 700 µl lysis buffer ( 20 mM Tris-HCl ( pH 7 . 5 ) , 100 mM KCl , 5 mM MgCl2 , 0 . 3% IGEPAL CA-630 ) containing freshly added 300U RNaseOUT ( Invitrogen , Grand Island , NY ) and Protease Inhibitor Cocktail ( Roche , South San Francisco , CA ) and incubated on ice for 20 min . After centrifugation for 15 min at 10000×g , 4°C , a 50-ml aliquot of supernatant was taken as input for subsequent RNA extraction . The remaining supernatant was incubated with activated Streptavidin-Dyna beads ( Dynabeads M-280 Streptavidin , #112 . 05D; Invitrogen ) for overnight at 4°C . Reaction mixture was centrifugated to remove unbound material . The beads were washed five times with lysis buffer and treated with 10U DNase I in lysis buffer for 10 min at 37°C . The beads were washed and treated with 50 µg/100 µl Protease K in 10% SDS containing lysis buffer for 20 min at 55°C . After centrifugation , the supernatant was taken for RNA extraction using acid phenol-choloroform ( Applied Biosytems ) . RNA was subjected to qRT-PCR using gene specific primers . The analysis was done as follows: miRNA pull-down/control pull-down ( ‘A’ ) , miRNA input/control input ( ‘B’ ) ; fold enrichment = A/B . The wild type ( WT ) or mutant ( Mt ) miRNA recognition elements ( MREs ) of target genes were synthesized as oligo sequences , annealed and cloned in psiCHECK2 ( Promega ) downstream to Renilla luciferase . Luciferase assay in MDAMB231 cells using WT and Mt MRE constructs was done as described previously ( Moskwa et al . , 2011 ) . The oligonucleotide sequences are as follows:BRCA1 MRE1+2+3-F , TCGAAAGATGAAGTTTCTATCATCCAGAAGACCTACA TCAGGCCTTCATC CTGAAGCCAGGAGTGGAAAGGTCATCCC , R , GGCCGGGA TGACCTTTCCACTCCTGGCTTCAGGATGAAGGCCTG ATGTAGGTCTTC TGGA TGATAGAAACTTCATCTT; BRCA2 MRE1-F , TCGACAATCAACAAAATGGTCATCCA , R , GGCCTGGATGA CCATTTTGTTGATTG; BRCA1 MRE4+5+6-F , TCGAAGTGTGCAGCATTTGAAAACCCGAAGGCA AGATCTAGAGGGAACCCCTGAAGCTGGCCAACATGGTGAAACCCC , R , GGCCGGGGTTTCACCATGTTGGCCAGC TTCAGGGGTTCCCTCTAGATCTTGCC TTCGGGTTTTCAAATGCTGCACACT; BRCA2 MRE2-F , GGCCAGGGGTTA TATGACTATTTTTA , R , TCGATTGGCCAAG GTGGTGAAATCCC;RAD51 MRE1-F , TCGATTGGCCAAGGTGGTGA AATCCC , R , GGCCGGGATTTC ACCACCTTGGCCAA;RAD51 MRE2+3-F , TCGATCTTATGTTTCCAAGAGAACTAG AAGTCTCTACAAA AAATGCAGAACT , R , GGCCAGTTCTGCATTTTTTGTAGAGACTTCTAGTTCTCT TGGAAACATA AGA . The oligonucleotides for mutant MREs are as follows:Mt BRCA1 MRE1+2+3-F , TCGATTGATGAAGAA ACGAGCAGGCAGAAGTGGTA CAAGAGGCGAGCAGGCTGAAGCCTGGTGAGGAAACGAGTAGGC , R , GGCCG CCTACTCGTTTCCTCACCAGGCTTCAGCCTGCTCGCCTCTTGTACCACTTCTGCCTGCTCGTTTCTTCATCAA;Mt BRCA2 MRE1-F , TCGACAATGAACAAAATGGGCAGGCA , R , GGCCTGCCTGCCCATTTTGTTCATTG;Mt BRCA1 MRE4+5+6-F , TCGAAGTGAGCAGGAGTTTTAAAGCCGAAGGGTAG AGCGAGAGGGAACCCCTGAAGCAGTCGA ACAGGGTTTTAGGGC , R , GGCCGC;CCTAAAACCCTGTTCGACTGCTTCAGGGGTTCCCTCTCGCTCTACCCTTCGGC TTTAAAACTCCTGCTCACT;Mt BRCA2 MRE2-F , TCGAGTAAAATAGGCATTTTAGGGCT , R , GGCCAGCCCT AAA ATGCCTATTTTAC;Mt RAD51 MRE1-F , TCGATAGGGCAAGGAGTTTTAAGGCC , R , GGCCGGCCTT AAAACTCCT TGCCCTA;Mt RAD51 MRE2+3-F , TCGATCATATGTTGCCTTGAGAACTAGAAGACTCTACA TATTATCCTGAACT , R , GGCCAGTT CAGGATAATATGTAGAGTCTTCTAGTTC TCAAGGCAACATATGA . MDAMB231 or MCF10A cells were seeded at 0 . 2 × 106 cells/well and treated with 500 µM mimosine for 24 hr . The cells were washed and released into growth media and collected for FACS analysis and RNA extraction after indicated time intervals . For FACS , the cell were fixed in 70% ethanol , washed with PBS buffer , and analyzed in PI/RNase staining buffer ( BD PharMingen , San Jose , CA ) . MDAMB231 cells transfected with miRNA antagomirs were similarly synchronized 48 hr after transfection . The siRNA for CtIP was obtained from Dharmacon , Pittsburgh , PA ( #J-011376-06 ) . | The DNA in a cell is damaged thousands of times every day . One of the most serious types of damage involves something breaking both of the strands in the double helix . Such a double-strand break can delete genes or even kill the cell . In fact , conventional cancer therapy kills cancer cells by causing irreparable double-strand breaks . Conversely , a normal cell that is constantly exposed to DNA damaging agents can become a tumor if double-strand breaks are incorrectly repaired . An efficient and accurate double-strand break repair system needs to be in place to prevent this transformation . Therefore , an in-depth understanding of double-strand break repair and the factors involved are important for both gaining insight into the cause of cancer and to improve cancer therapy . Cells have evolved several different ways to detect and repair double-strand breaks . A method called homologous recombination , for example , uses an undamaged DNA molecule as a template that can be copied to make new DNA . Since it needs a readily available DNA template , this method only works in phases of the cell growth cycle where there are many copies of DNA—that is , in the post-DNA replication phases . In particular , homologous recombination does not work during the pre-replication , G1 phase . If homologous recombination is attempted during G1 , it will block the other methods employed by cells to repair broken strands of DNA . An important challenge is to understand how homologous recombination is restricted to particular parts of the cell cycle . Although certain proteins associated with the early stages of double-strand repair are thought to determine the type of DNA repair that occurs , the details of this process are not fully understood . One group of molecules that are thought to be involved are microRNAs , which normally limit the number of proteins produced from certain genes . However , since a single microRNA molecule can be associated with several proteins , and since a single protein can be associated with several microRNA molecules , it has proved difficult to establish the exact effects of a specific microRNA molecule . Choi et al . now show that seven microRNA molecules can control homologous recombination , and three microRNAs in particular restrict homologous recombination during the G1 phase of the cell cycle . If these microRNAs are inhibited during the G1 phase , which allows homologous recombination to start , and counter-intuitively more double-stranded breaks are seen . However , if a gene involved in starting homologous repair–called CtIP—is silenced while the microRNAs are inhibited , then the DNA breaks are repaired . Exactly , how the microRNA molecules produce different effects during different phases of the cell cycle will be need to be investigated by future studies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2014 | MicroRNAs down-regulate homologous recombination in the G1 phase of cycling cells to maintain genomic stability |
ER O-glycosylation can be induced through relocalisation GalNAc-Transferases from the Golgi . This process markedly stimulates cell migration and is constitutively activated in more than 60% of breast carcinomas . How this activation is achieved remains unclear . Here , we screened 948 signalling genes using RNAi and imaging . We identified 12 negative regulators of O-glycosylation that all control GalNAc-T sub-cellular localisation . ERK8 , an atypical MAPK with high basal kinase activity , is a strong hit and is partially localised at the Golgi . Its inhibition induces the relocation of GalNAc-Ts , but not of KDEL receptors , revealing the existence of two separate COPI-dependent pathways . ERK8 down-regulation , in turn , activates cell motility . In human breast and lung carcinomas , ERK8 expression is reduced while ER O-glycosylation initiation is hyperactivated . In sum , ERK8 appears as a constitutive brake on GalNAc-T relocalisation , and the loss of its expression could drive cancer aggressivity through increased cell motility .
GalNAc-type O-linked glycans are polysaccharides present on secreted and membrane-inserted proteins ( Tran and Ten Hagen , 2013 ) . Traditionally associated with mucin-like proteins , recent advances in mass spectrometric analysis have revealed O-glycosylation on hundreds of different proteins ( Steentoft et al . , 2013 ) . Recent results have also highlighted specific functional roles for O-glycans; for instance , in regulating the secretion of the phosphatemia regulator FGF23 ( Kato et al . , 2006 ) and in processing of the angiopoietin-like factor ANGPTL2 ( Schjoldager et al . , 2010 ) . O-glycans are synthesised through the step-wise action of various glycosylation enzymes , starting with the UDP-N-Acetyl-Alpha-D-Galactosamine:Polypeptide N-Acetyl-galactosaminyltransferases ( GalNAc-Ts ) , a large family of 20 different isoforms that catalyses the addition of N-Acetylgalactosamine ( GalNAc ) onto serine or threonine residues ( Bennett et al . , 2012 ) . The addition of GalNAc on proteins generates the Tn antigen , with antigenicity being lost upon the addition of other sugar residues . Earlier work demonstrated that carcinomas stain prominently with antibodies and lectins such as the Helix Pomatia Lectin ( HPL ) , which binds to Tn antigens ( Springer , 1983 ) . The high prevalence and specificity of this cancer glycophenotype is remarkable , with matching normal tissues and benign tumours expressing much lower levels . This increase in Tn levels is proposed to stem from a block or reduction in the activity of the main O-GalNAc-modifying enzyme , the Core 1 Galactosyl-Transferase ( C1GALT ) ( Ju et al . , 2002a , 2008b; Stanley , 2011 ) ; indeed , the loss of C1GALT in the high Tn-expressing T cell leukaemia Jurkat cell line has been reported ( Ju et al . , 2008a ) . In breast carcinoma , however , high Tn levels seem to be caused by a different mechanism: GalNAc-Ts are massively relocated from the Golgi apparatus to the endoplasmic reticulum ( ER ) with Tn staining largely located in the ER ( Gill et al . , 2013 ) . Further , in some cancer cells , O-glycosylation initiation in the ER has also been reported ( Egea et al . , 1993 ) . Trafficking of GalNAc-Ts to the ER can be stimulated by growth factors such as epidermal growth factor ( EGF ) and platelet-derived growth factor ( PDGF ) , with GalNAc-Ts active in the ER and GalNAc incorporation in proteins increasing after relocation ( Gill et al . , 2010 ) . It is surmised that glycosylation of ER-resident proteins likely explains this observed increase in Tn staining , as several of these proteins bear O-GalNAc in mass spectrometric analyses ( Steentoft et al . , 2013 ) . Although it is unclear which specific proteins are modified , O-glycosylation in the ER results in a marked stimulatory effect on cell adhesion and cell migration ( Gill et al . , 2013 ) . This suggests that ER O-glycosylation promotes the invasive and metastatic potential of malignant tumour cells . Tn levels are consistently higher in higher grade , more aggressive breast tumours . Conversely , ER-specific inhibition of O-glycosylation reduced drastically lung metastasis in a mice model ( Gill et al . , 2013 ) . GalNAc-Ts transport is stimulated by activated SRC tyrosine kinases and requires the COPI coat ( Gill et al . , 2010 ) . COPI is a multimeric protein complex required for the formation of transport carriers and functions in the retrograde traffic between the Golgi and the ER ( Beck et al . , 2009; Szul and Sztul , 2011 ) . COPI coat assembly is regulated by small GTPases of the Arf family and their regulator , the GTP exchange factor , GBF1; however , the regulation of COPI-coated carrier formation in response to extracellular signals is poorly understood . To better understand the mechanisms regulating Tn expression in cancer , we performed an RNAi screen targeting 948 genes presumed to be involved in signal transduction . We identified and validated 12 regulators , with a particular focus on the MAP kinase ERK8 ( alias MAPK15 ) , the most recently identified member of the MAPK family ( Abe et al . , 2002 ) . Unlike classical MAP kinases , ERK8 possesses an atypically long C-terminal domain and appears to constitutively auto-phosphorylate its Thr-X-Tyr motif ( Klevernic et al . , 2006 ) . Here , we find that a fraction of the ERK8 protein is localised at the Golgi where it specifically inhibits COPI vesicle formation and the export of GalNAc-Ts . The loss of ERK8 activity results in increased O-glycosylation and increased cell motility . We find that ERK8 expression is also frequently downregulated in lung carcinomas , which may partly explain the high Tn phenotype and invasiveness of these tumours .
We recently reported the results from a screen for regulators of Golgi morphology and organisation using various markers including fluorescently labelled HPL ( Chia et al . , 2012 ) . In the analysis presented in this study , we quantified Tn levels using HPL fluorescence intensity per cell ( Figure 1A ) . In a pilot screen using HeLa cells , which targeted 63 known players of membrane traffic ( Chia et al . , 2012 ) , we identified that knockdown of the SNARE gene Syntaxin 5 ( STX5 ) reproducibly induced a 6–7-fold increase in Tn levels relative to a non-targeting ( NT ) control ( GFP siRNA ) ( Figure 1A ) . This effect was presumably due to a defect in the balance in anterograde vs retrograde ER-to-Golgi traffic of GalNAc-Ts . Using STX5 and GFP siRNA as positive and negative controls , we then screened 948 signalling genes in search for regulators of O-glycosylation . We discarded the results for 134 siRNA pools that reduced the cell number to less than 20% of the control ( Figure 1B ) . Of the remaining siRNA pools , we identified numerous gene knockdowns that increased HPL levels significantly more than STX5 depletion . None of the gene depletions seemed to significantly reduce the basal levels of Tn in HeLa cells ( Figure 1C ) . The knockdown effects were reproduced in independent replicates ( Figure 1–figure supplement 1A ) , and the trend was mostly independent of the analysis algorithm used , although the fold increase was higher with one method than with the other ( Figure 1—figure supplement 1B ) . 10 . 7554/eLife . 01828 . 003Figure 1 . RNAi screening reveals 12 negative regulators of Tn expression . ( A ) Helix pomatia lectin ( HPL ) staining was analysed using the ‘Transfluor HT’ module of MetaXpress software ( Molecular Devices ) . A mask was generated for both HPL and nuclei ( Hoechst ) staining to classify the region of measurement ( lower panels ) . Scale bar: 30 µm . ( B ) Schematic overview of the screening process . Images from the RNAi screen in Chia et al . ( 2012 ) were quantified for HPL intensities . Non-targeting ( NT ) siRNA and Syntaxin 5 ( STX5 ) siRNA were used as negative and positive controls , resepectively . ( C ) Fold-change of HPL intensities normalised to NT siRNA treatment ( green dots ) and STX5 ( orange dots ) . Primary hits were selected based on a threshold of a nine-fold increase ( red dashed line ) and the final validated genes are labelled in red ( Hit genes ) . ( D ) Images from the screen of HPL staining in HeLa cells depleted of ERK8 . MannII-GFP labels the Golgi apparatus . Scale bar: 30 μm . ( E ) HPL staining in cells knockdown of ERK8 with a control siRNA or GalNAc-T1 and -T2 siRNA . Scale bar: 30 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 00310 . 7554/eLife . 01828 . 004Figure 1—figure supplement 1 . Helix Pomatia Lectin ( HPL ) stains reliably and specifically for Tn antigen . ( A ) Comparison of the HPL intensities between the two screen replicates . ( B ) Comparison of HPL intensities generated from different analysis algorithms from the HCSU ( high content screening unit ) application ( Chia et al . , 2012 ) and MetaXpress transfluor HT ( Molecular Devices ) . ( C ) Quantification of the Tn levels of ERK8 depletion with pooled and deconvoluted siRNAs ( dERK8-1 to 4 ) . ( D ) Hit validation using deconvoluted siRNA pools . HPL intensities were quantified in HeLa cells treated with each of the four individual duplex siRNAs from the pool for the 19 primary hits . A gene was validated if at least two unique siRNAs reproduced at least a 4 . 5-fold increase in HPL intensities . ( E ) Comparison of HPL and Vicia Villosa Lectin ( VVL ) staining intensities in ERK8 and MAP4K2 depletion . ( F ) Quantification of Tn levels upon co-knockdown ( coKD ) of each of the 12 validated Tn regulators with GalNAc-T1 and -T2 . ( G ) SDS-PAGE analysis of ERK8 expression levels in single and co-knockdown with GalNAc-T1 and -T2 . Values on graphs indicate mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to the non-targeting ( NT ) siRNA-treated cells . NS , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 004 To focus our analysis , we used a stringent cut-off of a ninefold increase in HPL staining intensity , which resulted in 19 genes ( Figure 1B , C ) . Depletion of one of these negative regulators—the Extracellular Signal Regulated Kinase 8 ( ERK8 ) —induced a particularly marked increase in Tn , ranging from 4–25-fold depending on the experimental design and RNAi reagent ( Figure 1D , Figure 1—figure supplement 1C , Figure 3—figure supplement 1A ) . It should be noted that knock-down was not optimised for the other 18 genes and thus Tn levels may reflect partly the extent of target depletion . To exclude the possibility of off-target effects in the 19 hits , we repeated the screen using the individual siRNAs that were used in each siRNA pool in the primary screen . For ERK8 , we observed that three out of the four single siRNAs significantly increased Tn levels above the NT control ( Figure 1—figure supplement 1C , D ) . Using a threshold of 4 . 5-fold increase for at least two independent siRNAs , 12 genes were considered validated ( Figure 1—figure supplement 1D ) . To verify that the effects observed were not specific to the detection method used , knockdown cells were also stained with a different lectin , Vicia Villosa Lectin ( VVL ) , which revealed a highly consistent pattern ( Figure 1—figure supplement 1E ) . To validate that these genes indeed led to the up-regulation in GalNAc protein O-glycosylation , we sought to downregulate the responsible enzymes . Although GalNAc-Ts represent a large family , the T1 and T2 isoforms are by far the most prevalent and represent most of GalNAc-T activity in HeLa cells ( Bennett et al . , 2012 ) . This is apparent in the almost complete loss of Tn levels when GalNAc-T1 and -T2 were depleted ( Figure 1E ) . Co-depletion of the two enzymes also reduced significantly the Tn increase from the knockdown of ERK8 as well as that from other Tn regulators ( Figure 1E , Figure 1—figure supplement 1F ) . This effect was not caused by inefficient ERK8 knockdown , as other co-knockdown experiments did not have such an effect . In addition , ERK8 levels were still significantly reduced in the triple knockdown configuration ( Figure 1—figure supplement 1G ) . Overall , our screen revealed several negative regulators of the O-GalNAc glycosylation process , which thus appears to be tightly controlled by signalling mechanisms . Two mechanisms are known to increase Tn levels: inhibition of O-GalNAc extension ( Ju et al . , 2008a , 2008b ) or relocation of GalNAc-Ts from the Golgi apparatus to the ER . The loss of expression of C1GALT or its molecular chaperone , COSMC , results in a failure to generate the subsequent Core 1 glycan structure ( the TF antigen ) and thus inhibits O-GalNAc extension , which is detectable as a loss in Peanut Agglutinin ( PNA ) lectin positive staining ( Swamy et al . , 1991 ) . Relocation of GalNAc-Ts from the Golgi apparatus to the ER , on the other hand , induces a modest but measurable increase in PNA staining ( Gill et al . , 2010 ) . To distinguish between the two possibilities , PNA staining was quantified upon depletion of each of the 12 Tn regulators . HeLa cells with a stable COSMC knockout which prohibits C1GALT activity , was used as a positive control and , as expected , completely abolished PNA staining . Comparatively , there was either no significant decrease or some increase in PNA staining following depletion of each of the Tn regulators ( Figure 2B , Figure 2—figure supplement 1A ) , with the most significant increase in PNA staining observed after ERK8 knockdown ( Figure 2A ) . Therefore , none of the Tn-regulating genes we identified appears to be required for core 1-forming activity and thus do not regulate Tn levels by inhibiting O-GalNAc extension . 10 . 7554/eLife . 01828 . 005Figure 2 . Tn regulators control Tn expression through GalNAc-T subcellular localisation . ( A ) Peanut Agglutinin ( PNA ) lectin staining in ERK8-depleted HeLa cells . Scale bar: 30 μm . ( B ) PNA lectin staining quantification after depletion of the 12 Tn regulators , using COSMC knockout HeLa cells as a positive control . ( C ) Co-staining for endogenous GalNAc-T1 and Golgi ( MannII-GFP ) and ER ( Calreticulin ) markers . Scale bar: 10 μm . ( D ) Co-localisation of the GalNAc-T1 and Calreticulin measured using Pearson’s correlation coefficient of the staining intensities of the two markers . Cells were analysed using MetaXpress Translocation-Enhanced analysis module . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to NT siRNA-treated cells . ( E ) A potential regulatory network of signalling proteins regulating GalNAc-T localisation . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 00510 . 7554/eLife . 01828 . 006Figure 2—figure supplement 1 . Tn regulators control GalNAc-T1 and -T2 localisation . ( A ) Peanut Agglutinin ( PNA ) lectin staining quantification after depletion of the 12 Tn regulators with a single siRNA from the pool used in the screen and in COSMC knockout cells . ( B ) Pearson's correlation coefficient between GalNAc-T1 and Calreticulin staining in cells knocked down with a single siRNA from the pool . ( C ) Pearson's correlation coefficient between GalNAc-T2 and Calreticulin staining in the depletion of each of the 12 Tn regulators . ( D ) Co-staining for endogenous GalNAc-T1 , medial Golgi ( MannII-GFP ) and trans Golgi ( TGN46 ) markers in NT , ERK8-KD and PI4KA-KD cells . Scale bar: 10 μm . ( E ) Pearson's correlation coefficient between MannII-GFP and Calreticulin staining in the depletion of each of the 12 Tn regulators . Values on graphs indicate mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to non-targeting ( NT ) siRNA-treated cells . NS , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 006 To evaluate whether the alternative mechanism of GalNAc-Ts relocation to the ER was implicated , the subcellular distribution of GalNAc-Ts was evaluated by immunofluorescence . In control cells , GalNAc-T1 localised exclusively in the perinuclear region , co-localising with the Golgi marker MannII ( Figure 2C ) . Upon ERK8 depletion , GalNAc-T1 distribution appeared more diffuse , co-localising with the ER marker Calreticulin ( Figure 2C ) , with MannII-GFP staining remaining mostly perinuclear ( Figure 2C ) . The MannII-GFP-positive structures were more fragmented in the knockdown cells than in the untreated cells; this finding is reminiscent of the effects of SRC activation ( Gill et al . , 2010 ) . Depletion of phosphatidylinositol ( PI ) 4-kinase ( PI4KA ) , another major regulator of Tn , also resulted in GalNAc-T1 redistribution ( Figure 2C ) . However , unlike ERK8 depletion , MannII-GFP appeared to also redistribute to a cytoplasmic pattern upon PI4KA depletion ( Figure 2C ) . This suggests that the entire content of the Golgi apparatus becomes redistributed to the ER , and is consistent with our previous findings , where PI4KA depletion induced a redistribution of TGN46 , a trans golgi marker ( Chia et al . , 2012 ) ( Figure 2—figure supplement 1D ) . Next , we used a quantitative approach to determine the extent of ER relocalisation that occurs in response to depletion of the other Tn-regulating genes , measuring the degree of co-localisation between GalNAc-T1 and Calreticulin staining . The Pearson’s correlation coefficient between these two markers was significantly increased following the knockdown of all 12 genes ( Figure 2D , Figure 2—figure supplement 1B ) . Most exhibited levels similar to that induced by Brefeldin A ( BFA ) treatment , which redistributes Golgi proteins to the ER ( Fujiwara et al . , 1988 ) . Comparatively , there was only slight perturbation of GalNAc-T1 localisation in the COSMC knockout cells . Similar trends were also observed for GalNAc-T2 staining ( Figure 2—figure supplement 1C ) . In contrast to the GalNAc-T staining , none of the signalling genes significantly affected the MannII-GFP distribution , apart from PI4KA depletion ( Figure 2—figure supplement 1E ) . Collectively , our results suggest that signalling genes influence Tn levels through the subcellular distribution of GalNAc-Ts and that , with the exception of PI4KA , they affect the trafficking of these enzymes specifically . To explore how the products of these genes might be functioning , we retrieved data pertaining to their subcellular localisation and protein–protein interactions from Protein Atlas , GeneCards and STRING ( Jensen et al . , 2009; Safran et al . , 2010; Uhlen et al . , 2010 ) . Three proteins—PI4KA , PKMYT1 and MAP4K2—have previously been reported to be localised at least partially at the Golgi apparatus ( Nakagawa , 1996; Ren et al . , 1996; Liu et al . , 1997 ) . PI4KA is proposed to generate phosphoinositol-4-phosphate ( PI4P ) , which is essential for recruiting membrane trafficking effectors to the Golgi ( De Matteis et al . , 2005 ) , including Vps74/GOLPH3 , which retains various glycosyltransferases at the Golgi through retrograde trafficking ( Wood et al . , 2009 ) . Although GalNAc-Ts have not been known to be regulated by Vps74/GOLPH3 , it represents a potential mechanism for their retention at the Golgi . PKMYT1 is required for the reassembly of the Golgi during telophase ( Nakajima et al . , 2008 ) . In addition , ERK8 has been reported to localise perinuclearly in A431 cells ( Uhlen et al . , 2010 ) , suggesting a potential Golgi localisation . Four other kinases—HIPK3 , TTK , MARK2 and DUSP7— interact with Golgi-associated proteins ( Dowd et al . , 1998; Colland et al . , 2004; Dou et al . , 2004; Sowa et al . , 2009; Cui et al . , 2010 ) . HIPK3 was also found to interact with Golgi structural protein GRASP65 , the Golgi-localised LIM kinase , and the ERK8 interactor , HIC-5 ( Colland et al . , 2004 ) . HIPK3 also interacts with PKMYT1 ( Wells et al . , 1999 ) . The MARK2 protein controls microtubule stability through phosphorylation of microtubule-associated proteins ( Yoshimura and Miki , 2011 ) and its interaction with the microtubule tracking protein , CLASP2 ( Sowa et al . , 2009 ) , suggests CLASP2 as a potential substrate . CLASP2 is involved in microtubule nucleation at the Golgi ( Miller et al . , 2009 ) and microtubules could be nucleated at the cis Golgi ( Rivero et al . , 2009 ) . Since Golgi-to-ER retrograde traffic depends on microtubule tracks ( Palmer et al . , 2005; Spang , 2013 ) , GalNAc-T relocation could depend on CLASP2-associated microtubules regulated by MARK2 . Two proteins—ERK8 and TGFBR2—interact with SRC ( Abe et al . , 2002; Galliher and Schiemann , 2007 ) . ERK8 activity was reported to increase in the presence of active SRC ( Abe et al . , 2002 ) and TGFBR2 is phosphorylated by SRC ( Galliher and Schiemann , 2007 ) . In addition , MAP4K2 , IKBKE and PPP6C are linked to the canonical NFkB pathway , suggesting that this pathway might control GalNAc-T localisation ( Shimada et al . , 1999; Chadee et al . , 2002; Eddy et al . , 2005; Stefansson and Brautigan , 2006 ) . Finally , several other interactions , either direct or with one intermediate , were found between the Tn-regulating genes . Altogether , this analysis suggests that the Tn-regulating genes are acting at the Golgi level , perhaps part of a regulatory network controlling the subcellular localisation of GalNAc-Ts ( Figure 2E ) . Further experiments are required to confirm the reality of this network and its precise connectivity . We next sought to understand the mechanistic basis of Tn level regulation by ERK8 . ERK8 displays high basal activity in resting cells and is not stimulated by growth factor activity ( Abe et al . , 2002; Klevernic et al . , 2006 ) but through auto-phosphorylation on residues Thr 175 and Tyr 177 . To test if its kinase activity is important , we selected the siRNA dERK8-4 for its potency ( Figure 3—figure supplement 1B ) and designed an siRNA-resistant ERK8 construct tagged with GFP ( GFP-ERK8-siR ) , as well as a kinase-inactive mutant counterpart ( GFP-ERK8-siR-T175A-Y177F ) . ERK8-depleted cells were then transfected with either wild-type or kinase-mutant ERK8 constructs 48 hr after siRNA treatment . HPL intensities were quantified for transfected ( GFP-expressing ) and non-transfected ( non-GFP-expressing ) cell populations , each comprising hundreds of cells . Non-transfected ERK8-depleted cells displayed a marked increase in HPL staining , whereas significantly lower HPL staining was observed in cells transfected with the wild-type GFP-ERK8-siR construct ( Figure 3A , B ) . Importantly , HPL levels remained almost similar to non-GFP-expressing ERK8-depleted cells when the cells were transfected with the kinase-inactive mutant ( GFP-ERK8-siR-T175A-Y177F ) indicating that kinase activity is important for the negative regulation of O-glycosylation ( Figure 3B ) . 10 . 7554/eLife . 01828 . 007Figure 3 . ERK8 regulates ER-localised O-glycosylation initiation . ( A ) Protein replacement by expression of siRNA-resistant wild-type ERK8 or the kinase inactive mutant in ERK8-depleted HeLa cells . Cells were stained for Helix pomatia lectin ( HPL ) and ERK8 . Scale bar: 30 µm . ( B ) Tn levels of non-targeting ( NT ) siRNA-treated and ERK8-depleted cells that were untransfected ( red bars ) or transfected with wild-type ERK8 ( blue bars ) or kinase-inactive mutant ERK8 ( orange bars ) . ( C ) Treatment with 5 µM ERK8 inhibitor Ro-31-8220 ( iERK8 ) over time and staining for Tn expression with HPL in HeLa cells . ( D ) Quantification of Tn expression after 5 µM iERK8 treatment . ( E ) SDS-PAGE analysis of ER-specific glycosylation reporter ( Muc-PTS ) expressed in HEK293T cells treated with vehicle or with 5 µM iERK8 for 3 . 5 hr . Muc-PTS was immunoprecipitated using HPL-conjugated agarose . ( F ) SDS-PAGE analysis of untreated , ERK8-depleted and inhibitor-treated cell lysates metabolically labelled using GalNAz-FLAG . Arrows point to bands with changed intensities ( G ) Tn staining after 5 µM iERK8 treatment for 3 hr followed by chase over time . Scale bar: 30 µm . ( H ) Quantification of Tn expression levels upon iERK8 treatment and washout . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to untransfected or ERK8 wild-type transfected cells in ( B ) and vehicle treated cells in ( D ) and ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 00710 . 7554/eLife . 01828 . 008Figure 3—figure supplement 1 . ERK8 regulates Tn expression through GalNAc-T relocation and not related to transcriptional or translational events . ( A ) Tn staining with Helix Pomatia Lectin ( HPL ) and quantification of cells expressing mirRNA ( Invitrogen ) for ERK8 . Scale bar: 10 μm . ( B ) SDS-PAGE analysis of ERK8 expression levels in cells treated with single siRNA dERK8-4 . ( C ) Quantification of Tn levels of ERK8 depletion in ER-localised Lec2-GFP cells and GFP-expressing cells . ( D ) Quantification of Tn levels in cells treated with 10 µg/ml transcription inhibitor α-amanitin and co-treated with 5 µM Ro-31-8220 ( iERK8 ) . ( E ) SDS-PAGE analysis of protein expression levels of O-glycosylation-initiating enzymes and molecular chaperones in ERK8-depleted cells . ( F ) GalNAc-T1 staining of cells treated and washout ( ‘Wash’ ) of 5 µM iERK8 at the indicated times . Scale bar: 30 µm . ( G ) Quantification of the Pearson's correlation coefficient of GalNAc-T1 and Golgi marker MannII-GFP with iERK8 treatment ( blue bars ) and compound washout ( red bars ) after a 4-hr treatment for the indicated wash times . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative non-targeting ( NT ) control ( A and C ) or vehicle-treated ( Control ) cells ( D and G ) . NS , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 008 Ro-31-8220 inhibits the kinase activity of ERK8 ( Klevernic et al . , 2006 ) . To see if this compound could recapitulate the effects of ERK8 depletion , cells were treated with 5 µM Ro-31-8220 for various durations . A significant increase in Tn was observed as early as 1 hr after treatment and peaked at a nearly eightfold increase after 3 . 5 hr as compared with basal levels ( Figure 3C , D ) . After that , we observed some cell death , which possibly explains the accompanying decrease in Tn levels . This result indicated that the increase in Tn staining is a relatively rapid phenomenon and that O-GalNAc glycosylated protein accumulation can be achieved in a few hours . Our findings also suggest that these changes were not caused by expression changes of the O-glycoproteins or the O-glycosylation machinery . Indeed , co-treatment of cells with Ro-31-8220 and a transcriptional inhibitor α-amanitin ( Chafin et al . , 1995 ) did not reduce Tn levels ( Figure 3—figure supplement 1D ) and protein levels of the enzymes and chaperones involved in the early O-glycosylation stages were unaffected in ERK8-depleted cells ( Figure 3—figure supplement 1E ) . The rapid effect of Ro-31-8220 suggests that ERK8 acts relatively directly on GalNAc-T traffic . GalNAc-T staining upon ERK8 depletion strongly suggests the relocalisation of these enzymes to the ER . To further confirm this , we used an ER-specific glycosylation reporter ( GFP-Muc-PTS ) , which contains a Pro-Thr-Ser ( PTS ) -rich sequence with up to 15 sites for GalNAc addition ( Gill et al . , 2010 ) . After pulldown with HPL-conjugated beads , we found a significant increase in the glycosylation of this reporter upon treatment with Ro-31-8220 for 3 . 5 hr ( Figure 3E ) . We also verified ER localisation using a stably expressed ER-localised GalNAc-T inhibitor described previously ( Gill et al . , 2013 ) . This inhibitor counteracted the increase in Tn levels observed upon ERK8 knockdown ( Figure 3—figure supplement 1C ) . Recently , several ER-resident proteins were shown to be O-glycosylated ( Steentoft et al . , 2013 ) . To determine the extent to which the proteome is modified upon ERK8 depletion , we metabolically labelled cells with a GalNAc sugar analogue , FLAG-GalNAz ( Laughlin and Bertozzi , 2007 ) . After 24 hr , cells depleted of ERK8 ( by siRNA or Ro-31-8220 ) exhibited substantial increases in GalNAz incorporation , as revealed by the presence of several bands on SDS-PAGE gels ( Figure 3F ) . This suggests that ERK8 controls the O-glycosylation status of several proteins , which probably includes ER residents . The effects of Ro-31-8220 offered the possibility to test how quickly Tn levels can return to baseline levels after drug washout . High Tn levels obtained after 3 . 5 hr of treatment decreased progressively with an approximately 2 . 5-hr half-life ( Figure 3G , H ) . Consistently , staining for GalNAc-Ts showed a similar trend , with a significant increase in enzyme localisation at the Golgi within 2 hr of washout ( Figure 3—figure supplement 1F , G ) . This suggests that reactivation of ERK8 slows the continuous retrograde flow of GalNAc-Ts and that anterograde traffic shifts their distribution back to the Golgi . These results show that GalNAc-T relocalisation is rapidly reversible and suggests that ERK8 provides a continual brake for GalNAc-T relocation at the Golgi . GalNAc-Ts are thought to regularly cycle between the ER and the Golgi apparatus ( Rhee et al . , 2005 ) . Thus , relocation of these glycosylation enzymes to the ER upon ERK8 depletion could result either from an enhanced export from the Golgi or an inhibition of exit from the ER . To address this , we first analysed the subcellular localisation of ERK8 protein by immunofluorescence and observed a predominantly cytosolic pattern in wild-type HeLa cells . However , prolonged permeabilisation ( 2 hr ) clearly revealed positive Golgi staining , suggesting that a fraction of ERK8 is associated with this organelle ( Figure 4A ) . This is consistent with the perinuclear pattern reported in A-431 cells by the Protein Atlas project ( Uhlen et al . , 2010 ) . 10 . 7554/eLife . 01828 . 009Figure 4 . ERK8 is dynamically localised at the Golgi . ( A ) High magnification of HeLa MannII-GFP expressing cells stained for endogenous ERK8 following cytosol extraction . ( B ) Cytosol-depleted cells treated with platelet-derived growth factor ( PDGF; 50 ng/ml ) for the indicated times and stained for ERK8 , Tn ( Helix pomatia lectin , HPL ) and the Golgi marker , TGN46 . Scale bar: 10 μm . ( C ) Pearson’s correlation coefficient between ERK8 and Golgi marker TGN46 in cells treated with PDGF for the indicated times . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to vehicle treated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 00910 . 7554/eLife . 01828 . 010Figure 4—figure supplement 1 . ERK8 is dynamically localised when SRC is increasingly activated at the Golgi . ( A ) MannII-GFP- and Src-8A7F-mcherry-expressing cells were treated with imidazole for the indicated times , cytosol-depleted and stained for ERK8 . Scale bar: 10 μm . ( B ) Pearson's correlation coefficient between ERK8 and Golgi marker MannII in cells treated with imidazole for the indicated times . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to vehicle treated cells . ( C ) Catalytically defective mutant Src-6N7F-mcherry-expressing cells were treated with imidazole for indicated times , cytosol-depleted and stained for ERK8 and the Golgi marker MannII-GFP . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 010 Next , HeLa cells were stimulated with 50 ng/µl of PDGF , resulting in an increase in HPL staining intensity between 30 min and 2 hr and a decrease in ERK8 at the Golgi apparatus ( Figure 4B ) . Using the Pearson’s Correlation coefficient of ERK8 and TGN46 staining , we found a 60% decrease after 2 hr , suggesting that ERK8 is displaced from the Golgi after cell stimulation ( Figure 4C ) . SRC is a key signal transducer between PDGF and GalNAc-T traffic . A mutant , inactive form of SRC ( Src-8A7F ) can be re-activated using imidazole ( Qiao et al . , 2006 ) . Using a HeLa cell line stably expressing Src-8A7F-mCherry , we observed a gradual decrease in ERK8 at the Golgi ( Figure 4—figure supplement 1A , B ) , whereas no change was observed with a catalytically defective SRC mutant ( Src 6N7F ) ( Figure 4—figure supplement 1C ) . This suggests that SRC activity regulates ERK8 localisation at the Golgi . Overall , our data indicate that ERK8 is dynamically localised at the Golgi apparatus where it likely controls GalNAc-T export . To test if the relocation of GalNAc-Ts in ERK8-depleted cells is dependent on COPI , we first expressed the dominant-negative mutant of Arf1 ( Q71L ) , which is unable to hydrolyse bound GTP ( Dascher and Balch , 1994 ) and found significant rescue of Tn levels in contrast with cells expressing wild-type Arf1 ( Figure 5—figure supplement 1A ) . ERK8-depleted cells were also treated with 50 nM of the GBF1 inhibitor Golgicide ( Saenz et al . , 2009 ) , which also provided significant rescue ( Figure 5—figure supplement 1B ) . Consistent with these results , combined knockdown of ERK8 and GBF1 almost completely reversed high HPL staining , further indicating that GBF1 is required for GalNAc-T relocation from the Golgi to the ER ( Figure 5A ) . Co-knockdown of ERK8 with Arf1 , -3 , -4 or -5 also reduced Tn levels by about 60% ( Figure 5B ) . Combined knockdowns of Arfs further increased the rescue , suggesting functional redundancy amongst the Arf proteins . By contrast , co-knockdown with Arf6 , which does not regulate COPI , did not affect Tn levels ( Figure 5B ) . These reductions in HPL levels were not due to reduced knockdown efficiencies of ERK8 , as similar effects were observed with increasing amounts of non-targeting ( NT ) siRNA added to the transfection mix ( Figure 5—figure supplement 1C ) . The efficiency and specificity was also verified by assaying protein expression of each gene ( Figure 5—figure supplement 1D ) . Collectively , these data indicate that the COPI trafficking machinery is essential for the ER relocation of GalNAc-Ts . 10 . 7554/eLife . 01828 . 011Figure 5 . ERK8 regulates COPI-dependent GalNAc-T traffic . ( A ) Co-knockdown of ERK8 with Arf1 or GTP exchange factor , GBF1 , and staining with Helix pomatia lectin ( HPL ) . NT , non-targeting . Scale bar: 30 μm . ( B ) Quantification of Tn levels upon ERK8 co-knockdown with Arf proteins and GBF1 . Grey bar indicates knockdown of ERK8 only . Blue bars indicate co-knockdowns . ( C ) SDS-PAGE analysis of total Arf and Arf1-GTP in cells treated with 5 µM Ro-31-8220 ( iERK8 ) . ( D ) Co-staining of Beta-COP ( COPB ) and GalNAc-T1 in cells treated with 5 µM iERK8 for 15 min . Transient tubular structures emanating from the Golgi appear stained for GalNAc-T1 and beta-COP ( arrowhead , second panel ) . Scale bar: 10 μm . ( E ) Effect of ERK8 depletion on GalNAc-T1 and KDEL receptor ( KDEL-R ) subcellular location . ( F ) Effect of expression of active SRC ( SrcE-mcherry , containing the E378G mutation ) or inactive SRC ( SrcK-mcherry , containing the K295M mutation ) on both proteins . Scale bar: 10 μm . ( G ) Visual scoring of KDEL-R and GalNAc-T1 redistribution from the Golgi in cells subjected to various treatment conditions . Cells were counted in each condition from three independent experiments ( NT control: 83 cells; ERK8 KD: 86; SrcE-mcherry: 42; SrcK-mcherry: 32 ) . ( H ) Temperature-sensitive vesicular stomatitis virus G glycoprotein ( VSVG-mcherry ) traffic to the Golgi induced by shift from restrictive to permissive temperature for 15 min in KDEL-R expressing cells stained for GalNac-T1 . Scale bar: 10 μm . ( I ) Visual scoring of KDEL-R and GalNAc-T1 relocation in VSVG expressing cells at 0 and 15 min after temperature shift . Cells were counted in each condition from three independent experiments ( 0 min , 44 cells; 15 min , 63 cells ) . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to NT siRNA-treated ( B and G ) cells and cells at 0-min timepoint ( I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 01110 . 7554/eLife . 01828 . 012Figure 5—figure supplement 1 . ERK8 regulated GalNAc-T traffic depends on the activity of COPI regulators . ( A ) Tn expression levels , as determined by Helix Pomatia Lectin ( HPL ) staining in cells expressing dominant-negative mutant Arf1 ( Q71L ) compared to wild-type Arf1 . GRASP55 labels the Golgi apparatus . Scale bar: 10 μm . ( B ) Quantification of Tn expression in ERK8-depleted cells upon treatment with 50 nM GBF1 inhibitor , Golgicide , for the indicated times . ( C ) Quantification of Tn expression in cells following co-knockdown with ERK8 and GFP or NT5 siRNA in amounts to equivalent to double ( GFP coKD ) , triple ( GFPx2 coKD ) and quadruple ( GFPx3 coKD ) siRNA transfection configurations . ( D ) SDS-PAGE analysis of protein levels of Arfs and GBF1 in single and double knockdown configurations . ( E ) Beta-COP ( COPB ) staining in control and ERK8-depleted cells . Scale bar: 10 μm . ( F ) Quantification of the relative numbers of COPI transport carriers ( COPB ) and Golgi fragments ( GM130 ) using the granularity measurement module in MetaXpress software . At least 30 cells ( non-targeting ( NT ) control , 37 cells; ERK8-KD , 34 ) were quantified for each treatment . ( G ) Staining of native COPI coatomer in cells treated with 5 μM Ro-31-8220 ( iERK8 ) over the indicated times . ( H ) Quantification of the relative numbers of COPI transport carriers using the granularity measurement module in MetaXpress software . Twenty-five or more cells ( Vehicle , 41 cells; 5 min iERK8 treatment , 25; 25 min iERK8 treatment , 57 ) were quantified for each treatment . ( I ) Staining of GalNAc-T1 ( left panel ) and KDEL receptor ( KDELR1 ) localisation ( right panel ) in HeLa KDEL-R1-GFP stable cell line depleted of ERK8 . ( J ) Quantification of total intensities of GalNAc-T1 and KDEL-R at the Golgi region demarcated by Giantin ( Golgi protein ) staining . Values on graphs indicate the mean ± SEM . **p<0 . 0001 , *p<0 . 05 by two-tailed unpaired t test , relative to vehicle treated ( B and H ) , single ERK8 knockdown ( C ) and NT siRNA-treated ( F and J ) cells . NS , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 012 A key activation step for the COPI coatomer is the exchange of GDP for GTP on Arf1 ( Antonny et al . , 2005; Beck et al . , 2009; Szul and Sztul , 2011 ) . Therefore , we assessed Arf1-GTP loading after ERK8 inhibition using Ro-31-8220 and found activation of Arf1 as early as 15 min and sustained for over 2 hr ( Figure 5C ) . To evaluate the effect of ERK8 depletion on COPI , we next stained cells depleted by siRNA for the Golgi marker GM130 and the COPI subunit beta-COP ( COPB ) . As previously observed with MannII-GFP , GM130 staining revealed significant Golgi fragmentation after ERK8 depletion ( Figure 5—figure supplement 1E ) . Interestingly , COPB staining was significantly more affected and found on small structures in the cytoplasm , suggesting enrichment on transport intermediates ( Figure 5—figure supplement 1E ) . Using a granularity measurement algorithm , we found a nearly 4-fold increase in distribution for COPB but only slightly more than a twofold increase for GM130 ( Figure 5—figure supplement 1F ) . When treated with the ERK8 inhibitor Ro-31-8220 , cells displayed a significant redistribution of COPI coatomer staining as early as 5 min after treatment ( Figure 5—figure supplement 1G , H ) . Furthermore , in cells inhibited by Ro-31-8220 for 15 min , numerous COPI-positive vesicular structures were clearly co-stained with GalNAc-T1 antibodies ( Figure 5D ) . In several instances , we could detect tubular structures emanating from the Golgi apparatus that stained positively for GalNAc-T1 as well as COPB , although not as homogenously along their length as for seen for GalNAc-T1 ( Figure 5D ) . A well-described cargo of COPI carriers in Golgi-to-ER retrograde traffic is the KDEL receptor ( KDEL-R ) ( Orci et al . , 1997 ) . KDEL-R trafficking can be induced by a wave of cargo or by SRC activation ( Bard et al . , 2003; Pulvirenti et al . , 2008 ) . However , we found that ERK8 depletion did not detectably affect KDEL-R distribution in cells where GalNAc-T relocation was extensive ( Figure 5E , Figure 5—figure supplement 1I ) . By contrast , expression of an active form of SRC ( E378G; SrcE ) similarly relocated GalNAc-T1 to the ER , as compared with inactive SRC ( K295M; SrcK ) but it also strongly affected KDEL-R-GFP distribution ( Figure 5F ) . Visual scoring of the localisation of KDEL-R and GalNAc-T1 revealed than more than 80% of the cells display relocation for both proteins in SrcE expressing cells . By contrast , while more than 70% of ERK8 depleted cells exhibit clear GalNAc-T redistribution from the Golgi , less than 20% show KDEL-R relocation ( Figure 5G ) . These results indicate that GalNAc-Ts and KDEL-R trafficking are differentially regulated . Consistent with this observation , when a wave of the temperature-sensitive mutant of the vesicular stomatitis virus G glycoprotein ( VSVG ) was induced by temperature shift , KDEL-R was relocated from the Golgi to the ER as previously reported ( Figure 5H; Pulvirenti et al . , 2008 ) , whereas GalNAc-Ts were not affected ( Figure 5H , I ) . Altogether , these results suggest that SRC stimulates both GalNAc-Ts and KDEL-R COPI-dependent retrograde traffic whereas ERK8 inhibits specifically the formation of transport intermediates containing GalNAc-Ts . O-glycosylation in the ER stimulates cell adhesion and cell migration and tends to induce a spindle-shaped morphology ( Gill et al . , 2013 ) . Interestingly , this morphology was also apparent in ERK8-depleted HeLa cells under phase contrast microscopy ( Figure 6A ) and after staining for the actin and tubulin cytoskeletons ( Figure 6B ) . 10 . 7554/eLife . 01828 . 013Figure 6 . ERK8 regulates cell migration through ER O-glycosylation . ( A ) Phase contrast images and ( B ) actin and tubulin staining of non-targeting ( NT ) siRNA-treated and ERK8-depleted cells . Scale bars: 100 μm in ( A ) and 10 μm in ( B ) . ( C ) Migration assay using scratch wound of cellular monolayer in NT siRNA-treated and ERK8-depleted cells . Scale bar: 100 μm . ( D ) Rate of wound closure ( area ) measured over 7 hr ( n = 4 experiments for each condition ) . Values on graphs indicate mean ± SEM . **p<0 . 001 , *p<0 . 05 by two-tailed unpaired t test . Red asterisks indicate t test between NT siRNA-treated and ERK8-depleted cells . Green asterisk indicates t-test between cells co-knockdown with ERK8 and GalNAc-T1 and -T2 ( ERK8+GalNAc-T1 & -T2 KD ) and cells co-knockdown with ERK8 and NT siRNA ( ERK8+NT control KD ) . Blue asterisk indicates t test between ERK8 knockdown in NGFP-expressing and ER-localised GalNAc-T inhibitor Lec2GFP cells . NS , not significant ( black vertical lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 01310 . 7554/eLife . 01828 . 014Figure 6—figure supplement 1 . ERK8 inhibits cell motility by controlling Tn expression on cell-surface O-glycoproteins . ( A ) Cell-surface staining using a specific Tn antibody ( green ) and Helix Pomatia Lectin ( HPL ) ( red ) on non-permeabilised ERK8-depleted cells . Scale bar: 10 μm . ( B ) Scratch wound assay of cellular monolayer of cells following co-knockdown with ERK8 and GalNAc-T1 and -T2 ( ERK8+GalNacT1 & T2 ) siRNA or co-knockdown with ERK8 and non-targeting ( NT ) siRNA ( ERK8+NT ) on fibronectin-coated plates . Scale bar: 100 μm . ( C ) Scratch wound assay of ERK8-depleted NGFP- and Lec2GFP-expressing cells . Scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 014 When tested on fibronectin-coated plates in a scratch-wound healing assay , ERK8-depleted HeLa cells migrated about twofold faster into the denuded area compared with NT siRNA-treated cells ( Figure 6C ) . This faster rate was constant over 7 hr ( Figure 6D ) , indicating that the faster wound closure is caused by faster cell migration and not enhanced reactivity to the initial wound . ERK8 knockdown also led to a dramatically higher cell surface Tn staining , with numerous Tn-positive protrusion structures ( Figure 6—figure supplement 1A ) . These Tn-bearing glycoproteins are likely to promote increased cell adhesion , as shown previously ( Gill et al . , 2013 ) . However , ERK8 has also been implicated in various other cellular processes . To verify that the increased cell motility was due to enhanced O-glycosylation , the scratch-wound healing assay was repeated in ERK8 and GalNAc-T1 and -T2 knocked down cells ( ERK8+GalNAcT1&2 KD ) . We found that these cells migrated significantly slower than ERK8 and ERK8+NT knockdown cells and were similar to NT control cells ( Figure 6D , Figure 6—figure supplement 1B ) . To further confirm the importance of ER O-glycosylation , we used the ER-localised GalNAc-T inhibitor , Lec2GFP . Cell migration rates induced by ERK8 depletion in Lec2GFP cells were significantly reduced compared with cells expressing only GFP ( NGFP cells ) ( Figure 6D , Figure 6—figure supplement 1C ) and , again , were rather similar to NT control cell migration . It is important to note that the Lec2GFP construct itself did not significantly slow cell migration in the absence of ERK8 depletion . Thus , collectively , our results indicate that ERK8 is a negative regulator of cell migration through inhibition of protein O-glycosylation in the ER . An initial goal of this study was to elucidate mechanisms for the increase in Tn levels frequently observed in tumours . Interestingly , ERK8 protein levels are reported to be fairly constant in normal breast tissue and benign tumours but drop significantly in malignant tumours , with a loss measured in approximately 50% , 80% and 100% of grade 1 , 2 and 3 tumours , respectively ( Henrich et al . , 2003 ) . Henrich et al . proposed that ERK8 stimulates the degradation of Oestrogen Receptor-alpha , suggesting a possible selective advantage for the loss of this MAPK ( Henrich et al . , 2003 ) . Interestingly , this trend is also consistent with the increased frequency and intensity of Tn staining ( Gill et al . , 2013 ) . To further examine this question , both ERK8 and Tn ( stained with VVL ) were co-labelled in a panel of 39 frozen tissue arrays comprising 5 normal and 34 invasive ductal breast carcinoma ( Figure 7—figure supplement 1A–C ) . Quantification of ERK8 levels of each tissue core was performed by measuring the area above a fixed threshold , normalised to the total area of the core represented by nuclei staining ( DAPI ) ( Figure 7—figure supplement 1D ) . Although levels varied considerably , more than half of the carcinoma cores ( 18/34 ) showed at least 50% lower expression of ERK8 ( Figure 7A ) . Tn levels also varied significantly but , in the large majority of samples , they were significantly higher in tumour samples as compared with normal cores ( Figure 7B , Figure 7—figure supplement 1C , D ) . In most cores , ERK8 and Tn levels appeared to show opposing trends , suggesting that the loss of ERK8 could partially drive high Tn expression . However , there was no clear correlation between the levels of both antigens ( Figure 7—figure supplement 1E ) . 10 . 7554/eLife . 01828 . 015Figure 7 . ERK8 is downregulated in human breast and lung carcinoma . ( A ) Quantification of ERK8 staining in human breast biopsies . Each point represents the staining of one tissue core normalised to the average staining of the normal tissue cores . ( B ) Quantification of Tn ( Vicia Villosa Lectin; VVL ) staining in human breast biopsies . ( C ) Quantification of ERK8 staining in human lung biopsies . ( D ) Quantification of Tn staining in human lung biopsies . *p<0 . 05 , **p<0 . 01 , ***p<0 . 0001 by two-tailed unpaired t test . ( E ) Co-staining VVL and ER marker Calnexin revealed extensive ER co-localisation of Tn in lung carcinoma ( FMC407: Core B3 ) , whereas Tn appeared as punctuate structures in the normal lung ( FBN406: Core C1 ) . Scale bar: 20 μm . ( F ) ERK8 and Tn staining in a lung adenocarcinoma core ( FMC407: Core B8 ) . Scale bar: 200 μm . ( G ) Close-up image of the core shown in ( F ) . Scale bar: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 01510 . 7554/eLife . 01828 . 016Figure 7—figure supplement 1 . ERK8 levels are frequently reduced in human lung carcinoma . ( A ) H&E staining of 40 breast biopsies on BRF404 slides purchased from US Biomax , Inc . Image is from Biomax website ( http://www . biomax . us/ ) . ( B ) ERK8 and ( C ) Vicia Villosa Lectin ( VVL ) were co-stained on the breast tissue cores . Red boxes indicate the positions of normal breast tissue cores . Note: Core E6 was absent on the slide , hence only 39 of the original 40 biopsies were included in the analysis . ( D ) Quantification method of ERK8 and VVL staining in the tissue cores . The proportion of tissue area above a fixed threshold , as highlighted by the masking , was measured . This was then normalised to the total core area represented by the nuclei ( DAPI ) staining . The fold change in ERK8 and VVL staining area with respect to ( wrt ) the average of all normal cores is presented below each example image . ( E ) Scatter plot of VVL and ERK8 staining in all 39 breast cores . ( F ) The same quantification method used in the breast array was adopted for the lung tissue array . Representative tissue core images and quantification results of the normal and cancerous lung cores are shown . ( G ) H&E staining of 23 lung biopsies on FBN406a and FMC407 slides purchased from US Biomax , Inc . ( H ) ERK8 and ( I ) VVL were co-stained on the lung cores . Note: Core B1 of FMC407 was absent on the slide , hence only 23 of the original 24 biopsies were included in the analysis . ( J ) Scatter plot of VVL and ERK8 staining in all 23 lung cores . DOI: http://dx . doi . org/10 . 7554/eLife . 01828 . 016 High Tn has also been reported in other tumour types , where oestrogen regulation is not thought to be critical . As ERK8 was previously found to be highly expressed in the lung ( Abe et al . , 2002 ) , we set out to explore the link between ERK8 and Tn levels in lung cancer with 23 lung biopsies containing 2 normal lung tissues , 14 squamous cell carcinomas , 6 adenocarcinomas and 1 small cell carcinoma . We found that ERK8 was clearly detectable in normal tissues but that the levels appeared markedly lower in all lung carcinomas ( Figure 7—figure supplement 1H ) , with an average 80% loss of expression , and a range of 40–90% ( Figure 7C ) . No specific trend in regard to the cancer type was noticed . Next , we quantified Tn levels ( Figure 7—figure supplement 1F ) and found a significantly higher expression of Tn in a majority of the samples , with a more than fourfold average higher expression ( Figure 7D , Figure 7—figure supplement 1I ) and a range of 2–10-fold; 18 cores ( 86% ) displayed higher than normal levels of Tn ( Figure 7D ) . We next explored if high Tn levels were linked to ER O-glycosylation , as in the case of breast carcinomas and indeed found that Tn staining co-localised extensively with the ER marker calnexin in the carcinoma samples . By contrast , Tn staining in the normal lung tissue was clearly more punctuate and reminiscent of a Golgi localisation ( Figure 7E ) . Thus , in most lung carcinoma tumours , ERK8 expression is lower while Tn levels are higher as compared with normal lung tissue . Furthermore , in samples with heterogeneous staining for Tn , we observed a reverse correlation between ERK8 and Tn staining ( Figure 7F , G ) . However , at the whole core level , the staining intensities of the two levels were not inversely correlated from sample to sample ( Figure 7—figure supplement 1J ) . The lack of a direct correlation indicates that , in human lung and breast tumours , ERK8 levels do not strictly control the levels of Tn . Given the number of regulators that have been already identified in this and previous studies , this is not surprising . Notwithstanding , our analysis indicates that ERK8 is frequently downregulated in lung and breast carcinomas , which probably facilitates the relocation of GalNac-Ts to the ER when other biochemical or genetic perturbations , such as SRC activation , are engaged .
Cellular levels of the Tn antigen vary dramatically in cancer cells , suggesting that O-glycosylation initiation and/or elongation is highly regulated . Indeed , our screen results reveal that several signalling molecules exert significant control over O-glycosylation initiation . In recent years , we have reported that this initiation step can be regulated through trafficking of the GalNAc-Ts between the ER and the Golgi ( Gill et al . , 2010 ) and that marked ER localisation explains the high Tn phenotype in at least 60% of human breast tumours ( Gill et al . , 2013 ) . High Tn levels can also arise from a loss or inhibition of the elongation of the O-GalNAc , a mechanism that has also been proposed to underlie the cancer phenotype ( Ju et al . , 2011 ) . In secondary screens , it appeared that most , if not all , signalling proteins affecting Tn levels regulate the subcellular localisation of GalNAc-Ts and not the elongation process . This obviously does not preclude the main elongation enzyme , C1GALT or its specific chaperone , COSMC ( alias C1GALT1C1 ) from being regulated in some conditions , but a significant inhibition of the activity of these proteins was not observed in our screen conditions . In contrast , the subcellular localisation of GalNAc-Ts appears to be a nodal point of control in a complex signalling network . Indeed , at least 12 independent negative regulators were identified and at least as many positive regulators , including the SRC family tyrosine kinases , are likely to be involved . This regulatory complexity suggests that perhaps signals of different origins are being integrated at the level of GalNAc-Ts traffic . ERK8 is one of the most potent regulators we identified . Multiple pieces of evidence indicate that it acts at the level of the Golgi by inhibiting the formation of GalNAc-T-containing COPI vesicles . Based on their genetic interaction profile , the other negative regulators appear to act at the same level as ERK8 . Consistently , two of these proteins , PKMYT1 and MAP4K2 , are also reported to localize at the Golgi ( Ren et al . , 1996; Liu et al . , 1997 ) , and several other regulators interact with Golgi-localised proteins . Together , this suggest that the incorporation of GalNAc-Ts in COPI vesicles is the key point of regulation of this potential regulatory network . This regulation point also reveals the existence of at least two different types of COPI-dependent Golgi-to-ER retrograde traffic carriers: one type of transport , GalNAc-Ts , is activated by growth factor stimulation and cancerous transformation and is repressed by ERK8; the other , KDEL-R , is activated by cargo protein traffic and is independent of ERK8 . However , both routes appear to be under the control of the SRC kinase family . A key step in understanding these differences will be to identify the relevant phosphorylation substrates for both SRC and ERK8 . The trafficking of GalNAc-Ts to the ER results in the glycosylation of multiple different substrates , as indicated by the metabolic labelling results . The precise identity of these substrates , as well as the functional effects of their glycosylation , remains to be established . Notwithstanding , the outcome of the relocation at the cellular level is clearly a significant stimulator of cell migration . Indeed , the stimulatory effect of ERK8 depletion is dependent on ER-localised O-glycosylation . These results are also consistent with our previous data based on the expression of an exogenous , ER-targeted form of GalNAc-T2 ( Gill et al . , 2013 ) . Thus , an interesting hypothesis is that the intensity of packaging of GalNAc-Ts into Golgi-derived COPI-coated vesicles could be a signalling integration point that sets the ‘motility potential’ of cells . In breast and lung cancer cells , this set-point appears constitutively high , as the relocation of GalNAc-Ts is extensive and frequent . The promotion of cell motility associated with ER-localised O-glycosylation appeared to favour the formation of lung metastases in a tail-vein injection-based assay ( Gill et al . , 2013 ) . Thus , how GalNAc-T relocation is stimulated in cancer cells has probably important medical implications . Our analyses suggest that multiple mechanisms are possible , including a decrease in ERK8 protein levels . However , the actual level of decrease required to stimulate relocation is not known and is anyway probably dependent on other cellular parameters . Additionally , Tn levels are not likely to depend only on the intracellular distribution of GalNAc-Ts . For instance , normal tissues with high levels of mucin expression , such as stomach , colon or kidney , tend to have higher endogenous Tn levels without clear evidence for relocation . This complexity probably contributes to the lack of direct correlation between ERK8 and Tn levels . In addition to promoting O-glycosylation , ERK8 depletion could also have other beneficial advantages for cancer cells . Indeed , ERK8 has been implicated in multiple , apparently unrelated , molecular processes , such as maintenance of genomic integrity ( Groehler and Lannigan , 2010 ) , regulation of telomerase activity ( Cerone et al . , 2011 ) , autophagy ( Colecchia et al . , 2012 ) and inhibition of nuclear receptor activity ( Henrich et al . , 2003; Saelzler et al . , 2006; Rossi et al . , 2011 ) . Recently , the Drosophila homolog Erk7 and human ERK8 were also shown to participate in the regulation of protein secretion during starvation through disassembly of ER exit sites ( Zacharogianni et al . , 2011 ) . Whether these different processes are somewhat linked through ERK8 or whether ERK8 is simply moonlighting in different , independent functions constitutes an interesting challenge for the future . In sum , our results suggest that initiation of O-glycosylation in the ER is under an elaborate regulatory control system of which ERK8 is a key player . This regulation sets the level of cellular motility and is frequently perturbed in cancer cells of breast and lung origins .
HeLa MannII-GFP was from Vivek Malhotra’s laboratory ( CRG , Barcelona ) and maintained in DMEM with 10% fetal bovine serum ( FBS ) . HeLa cells that were knockout of COSMC was obtained from U Mendel and H Clausen ( University of Copenhagen , Denmark ) . HeLa ER-Lec2-GFP and KDEL-R-GFP cell lines were generated by lentiviral infection of HeLa wild-type cells with ER-2Lec-GFP ( Gill et al . , 2013 ) and KDEL-R-GFP lentivirus and subsequently , FACS sorted to enrich for GFP-expressing cells . HEK293T cells were grown in DMEM supplemented with 15% FBS . All cells were grown at 37°C in a 10% CO2 incubator . Human MAPK15/ ERK8 ( NM_139021 ) was amplified from cDNA purchased from Origene ( #RG216589; Rockville , MD ) by PCR and cloned into entry vector pDONR221 ( Invitrogen , Life Technologies Corporation , Carlsbad , CA ) . The catalytically inactive ERK8 mutant construct was generated by introducing T175A and Y177F mutations using the QuikChange Site-Directed Mutagenesis Kit ( Stratagene , Amsterdam , The Netherlands ) . The entry vectors were subsequently cloned into pcDNA6 . 2-Nmcherry-DEST , a gateway compatible destination vector constructed by the replacement of the GFP tag with mCherry on pcDNA6 . 2-NGFP-DEST ( Invitrogen ) . Human ARF1 ( NM-001024228 ) -GFP expression clones were described previously ( Gill et al . , 2010 ) . miR-ERK8 and miR-GFP vector were generated from BLOCK- iT Pol II miR RNAi expression vector kits from Invitrogen . All constructs were verified by sequencing and restriction enzyme digests before use . Helix pomatia Lectin ( HPL ) conjugated with 647 nm fluorophore ( #L32454 ) , Alexa Fluor secondary antibodies , and Hoechst 33342 ( #H3570 ) were purchased from Invitrogen . siRNAs were obtained from Dharmacon ( Thermo Fisher Scientific , Wilmington , DE ) . OptiMEM was purchased from Invitrogen , and Hiperfect transfection reagent was purchased from Qiagen ( Valencia , CA ) . Anti-GalNAc-T1 , GalNAc-T2 and Tn hybridomas for immunofluorescence staining were a gift from U Mendel and H Clausen ( University of Copenhagen , Denmark ) . Anti-COPI coatomer ( targeting native coatomer ) was a gift from FT Wieland ( University of Heidelberg , Germany ) . Anti-GRASP55 was a gift from Vivek Malhotra’s laboratory ( CRG , Barcelona ) . Anti-GalNAc-T1 ( #sc-68491 ) for western blotting was purchased from Santa Cruz Biotechnology ( Dallas , TX ) . Anti-ERK8 antibody ( #HPA002704 ) was purchased from Sigma–Aldrich ( St Louis , MO ) . Anti-beta COP antibody ( #ab2899 ) , anti-Giantin ( #ab24586 ) , anti-C2GNT1 ( #ab38858 ) and anti-COSMC ( #ab93483 ) were from Abcam ( Cambridge , MA ) . Fluorescein-labelled Vicia Villosa Lectin ( VVL ) ( #FL-1231 ) and Rhodamine-labelled Peanut Agglutinin PNA ( #RL-1072 ) were from Vector laboratories Inc . ( Burlingame , CA ) . Ro-31-8220 ( #557521 ) and α-Amanitin ( #129741 ) was from Merck ( Rahway , NJ ) . Golgicide A ( GCA ) ( #G0923 ) was from Sigma–Aldrich . ERK8-specific siRNA sequences ( dERK8-1: 5′-GUAGUGGACCCUCGCAUUG-3′ , dERK8-2: 5′-AGAACGACAGGGACAUUUA-3′ , dERK8-3: 5′-GGAGAUACCUACUCAGGCG-3′ and dERK8-4: 5′-CCUAUGGCAUUGUGUGGAA-3′ ) were purchased from Thermo-fisher . siRNA transfection , immunofluorescence staining and imaging procedures were described in detail previously ( Chia et al . , 2012 ) . During automated image acquisition , four sites per well were acquired sequentially with a 20 × Plan Apo 0 . 75 NA objective on a laser scanning confocal high-throughput microscope ( ImageXpress Ultra , Molecular Devices , Sunnyvale , CA ) . Image analysis was performed using MetaXpress software ( version 3 . 1 . 0 . 89 ) . For each well , total HPL staining intensity and nuclei number was quantified using Transfluor HT application module in the software . Briefly , masking for both Cy5 ( HPL ) and DAPI ( Nuclei ) channels was generated by setting the mask dimensions and cut-off intensity above the background for each of the two channels ( Figure 1A ) and intensities were quantified in the area covered by the masking . Hundreds of cells were quantified and the averages per well were calculated . To compare HPL intensities between wells , HPL intensity per cell of each well was obtained by normalising the total HPL intensity ( ‘Integrated Granule Intensity’ of Cy5 channel ) with nuclei number of the well . The same procedures were performed for the quantification of PNA and GalNAc-T1 staining in the secondary screens . Statistical significance was measured using a paired t test assuming a two-tailed Gaussian distribution . For ERK8 inhibitor Ro-31-8220 treatment , HeLa MannII-GFP cells were seeded into imaging plates overnight , treated with 5 µM Ro-31-8220 in DMEM with 10% FBS for various durations and then fixed with 4% paraformaldehyde-4% sucrose in phosphate-buffered saline ( PBS ) followed by subsequent staining . For Golgicide A treatment , cells were treated with 50 nM of Golgicide A and fixed at different time points . Cells were seeded onto glass coverslips in 24-well dishes ( Nunc , Denmark ) . After the respective treatments , cells were fixed with 4% paraformaldehyde-4% sucrose in D-PBS , permeabilised with 0 . 2% Triton-X for 10 min and stained with the appropriate markers using the same procedure performed in the primary siRNA screen . To effectively observe ERK8 localisation at the Golgi , the cells were permeabilised with 0 . 2% Triton-X for 2 hr and stained with anti-ERK8 antibody diluted in 2% FBS in D-PBS overnight . For beta-COP ( COPB ) and COPI coatomer staining , cells were permeabilised with 0 . 05% NP40 for 5 min twice , washed with D-PBS twice for 5 min , blocked with 2% bovine serum albumin ( BSA ) for 1 hr at room temperature , and then stained with primary antibody diluted in 2% FBS in D-PBS overnight . Cells were mounted onto glass slides using FluorSave ( Merck ) and imaged at room temperature using an inverted confocal microscope ( IX81; Olympus Optical Co . Ltd , Tokyo , Japan ) coupled with a CCD camera ( model FVII ) either with a 60 × objective ( U Plan Super Apochromatic [UPLSAPO]; NA 1 . 35 ) or 100 × objective ( UPLSAPO; NA 1 . 40 ) under Immersol oil . Images were acquired at 100x magnification and processed using Olympus FV10-ASW software . HeLa KDELR-R-GFP expressing cells were transfected to express the temperature-sensitive mutant of vesicular stomatitis virus G glycoprotein ( VSVG-tsO45 ) tagged with mcherry at the C-terminus . For the VSVG pulse-chase experiment , cells were transferred to 40°C for 16 hr and then shifted to 32°C for various durations in the presence of 100 μg/ml of cycloheximide before fixation . Cells were imaged at 100x magnification and quantified by eye for KDEL-R and GalNAc-T subcellular localisation . Cells were seeded onto fibronectin-coated 35-mm plastic tissue culture dishes ( Ibidi GmbH ) and grown to confluence ( 16–24 hr ) . A wound was generated using a micropipette tip before washing to remove cell debris . Live phase contrast imaging was performed at 37°C using a Zeiss Axiovert microscope ( model 200M; Zeiss Microimaging; Thornwood , NY ) equipped with a CCD camera ( AxioCam HRc ) and a 20 × objective ( LD Plan-NEOFLUAR; 20 ×; N . A . 0 . 4 ) . Frames were acquired at 5-min intervals . Areas of wound invasion were calculated using ImageJ ( National Institutes of Health , Bethesda , MD ) . Frozen human tumour microarrays FBN406a and FMC407 were purchased from US Biomax , Inc . ( Rockville , MD ) . Briefly , the slides were dried and fixed in chilled in a 1:1 acetone:methanol solution for 10 min at room temperature . The slides were then washed three times with TBST and blocked with 10% goat serum-PBS for 30 min . Subsequent staining with ERK8 antibody ( 0 . 9 µg/ml ) , VVL-biotin ( 4 μg/ml ) and Hoescht ( 1:10 , 000 ) was performed overnight before staining with anti-rabbit Alexa Fluor 488 ( 1:1000 ) and Streptavidin-Alexa 594 ( 1:400 ) secondary antibodies for 30 min . Slides were counterstained with DAPI and then mounted ( Vectashield ) . The arrays were first automatically imaged ( using constant acquisition parameters ) using a 10 ×objective ( LD Plan-NEOFLUAR; 10 ×; N . A . 0 . 4 ) on a motorised stage coupled to a Zeiss inverted confocal microscope equipped with a CCD camera ( AxioCam HRc ) . Images of the cores were exported from Zeiss Zen2011 software to enable quantification of ERK8 and VVL staining in tumour cores . To quantify the levels of ERK8 or Tn ( VVL ) expression in a tissue core , the images were first converted to 8-bit images using ImageJ . The area above the threshold was set for background staining ( Threshold for ERK8 and VVL was 30 and DAPI was 40 ) and then quantified . The area of ERK8 and VVL was normalised to the total area of the core represented by nuclei ( DAPI ) staining . The values of each core were then normalised to the average area of the normal tissue cores . | The likelihood of an individual being able to recover from cancer depends on: where the cancer is within the body , how quickly the disease is detected and how quickly treatment is started . Cancers that have spread from their original location to another part of the body are particular challenging to treat , and cause the vast majority of cancer deaths every year . Treatments that can recognize and eradicate cancer cells , while leaving nearby healthy cells untouched , are still needed—and so there has been a lot of research into identifying the key differences between healthy cells and cancer cells . For several decades , researchers have been aware that cancer cells have more proteins coated with modified sugars on their cell surfaces than healthy cells . This is caused by the enzymes that add these sugars to the proteins relocating from one location within the cell , the Golgi apparatus , to another , called the endoplasmic reticulum . These specific ‘sugar-coated’ proteins are known to encourage cancer cells to migrate and invade new tissues , but the mechanisms that regulate the addition of these sugar molecules to proteins remains poorly understood . Now Chia et al . have discovered 12 molecules that regulate this process , including an enzyme called ERK8 that is found at the Golgi apparatus . ERK8 is shown to prevent the relocation of the sugar-adding enzymes from the Golgi to the endoplasmic reticulum , thereby restricting the production of sugar-coated proteins that help the cancer cells to spread within the body . By identifying 12 potential targets for new therapeutics aimed at preventing the spread of cancer , the work of Chia et al . could ultimately help to improve the chances of patients recovering from certain cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | ERK8 is a negative regulator of O-GalNAc glycosylation and cell migration |
Many recent models study the downstream projection from grid cells to place cells , while recent data have pointed out the importance of the feedback projection . We thus asked how grid cells are affected by the nature of the input from the place cells . We propose a single-layer neural network with feedforward weights connecting place-like input cells to grid cell outputs . Place-to-grid weights are learned via a generalized Hebbian rule . The architecture of this network highly resembles neural networks used to perform Principal Component Analysis ( PCA ) . Both numerical results and analytic considerations indicate that if the components of the feedforward neural network are non-negative , the output converges to a hexagonal lattice . Without the non-negativity constraint , the output converges to a square lattice . Consistent with experiments , grid spacing ratio between the first two consecutive modules is −1 . 4 . Our results express a possible linkage between place cell to grid cell interactions and PCA .
The system of spatial navigation in the brain has recently received much attention ( Burgess , 2014; Morris , 2015; Eichenbaum , 2015 ) . This system involves many regions , which seem to divide into two major classes: regions such as CA1 and CA3 of the hippocampus , which contain place cells ( O'Keefe and Dostrovsky , 1971; O'Keefe and Nadel , 1978 ) , vs . regions , such as the medial-entorhinal cortex ( MEC ) , the presubiculum and the parasubiculum , which contain grid cells , head-direction cells and border cells ( Hafting et al . , 2005; Boccara et al . , 2010; Sargolini et al . , 2006; Solstad et al . , 2008; Savelli et al . , 2008 ) . While the phenomenology of those cells is described in many studies ( Derdikman and Knierim , 2014; Tocker et al . , 2015 ) , the manner in which grid cells are formed is quite enigmatic . Many mechanisms have been proposed . The details of these mechanisms differ , however , they mostly share in common the assumption that the animal’s velocity is the main input to the system ( Derdikman and Knierim , 2014; Zilli , 2012; Giocomo et al . , 2011 ) , such that positional information is generated by the integration of this input in time . This process is termed 'path integration' ( PI ) ( Mittelstaedt and Mittelstaedt , 1980 ) . A notable exception to this class of models was suggested in a previous paper by Kropff and Treves ( 2008 ) ; and in a sequel to that paper ( Si and Treves , 2013 ) , in which they demonstrated the emergence of grid cells from place cell inputs without using the rat's velocity as an input signal . We note here that generating grid cells from place cells may seem at odds with the architecture of the network , since it is known that place cells reside at least one synapse downstream of grid cells ( Witter and Amaral , 2004 ) . Nonetheless , there is current evidence that the feedback from place cells to grid cells is of great functional importance . Specifically , there is evidence that inactivation of place cells causes grid cells to disappear ( Bonnevie et al . , 2013 ) , and furthermore , it seems that , in development , place cells emerge before grid cells do ( Langston et al . , 2010; Wills et al . , 2010 ) . Thus , there is good motivation for trying to understand how the feedback from hippocampal place cells may contribute to grid cell formation . In the present paper , we thus investigated a model of grid cell development from place cell inputs . We showed the resemblance between a feedforward network from place cells to grid cells to a neural network architecture previously used to implement the PCA algorithm ( Oja , 1982 ) . We demonstrated , both analytically and through simulations , that the formation of grid cells from place cells using such a neural network could occur given specific assumptions on the input ( i . e . zero mean ) and on the nature of the feedforward connections ( specifically , non-negative , or excitatory ) .
We initially considered the output of a single-layer neural network and of the PCA algorithm in response to the same inputs . These consisted of the temporal activity of a simulated agent moving around in a two-dimensional ( 2D ) space ( Figure 1A; see Materials and methods for details ) . In order to mimic place cell activity , the simulated virtual space was covered by multiple 2D Gaussian functions uniformly distributed at random ( Figure 1B ) , which constituted the input . In order to calculate the principal components , we used a [Neuron x Time] matrix ( Figure 1C ) after subtracting the temporal mean , generated from the trajectory of the agent as it moved through the place fields . Thus , we displayed a one-dimensional mapping of the two-dimensional activity , transforming the 2D activity into a 1D vector per input neuron . This resulted in the [Neuron X Neuron] covariance matrix ( Figure 1D ) , on which PCA was performed by evaluating the appropriate eigenvalues and eigenvectors . 10 . 7554/eLife . 10094 . 003Figure 1 . Construction of the correlation matrix from behavior . ( A ) Diagram of the environment . Black dots indicate places the virtual agent has visited . ( B ) Centers of place cells uniformly distributed in the environment . ( C ) The [Neuron X Time] matrix of the input-place cells . ( D ) Correlation matrix of ( C ) used for the PCA process . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 003 To learn the grid cells , based on the place cell inputs , we implemented a single-layer neural network with a single output ( Figure 2 ) . Input to output weights were governed by a Hebbian-like learning rule . As described in the Introduction ( see also analytical treatment in the Methods section ) , this type of architecture induces the output’s weights to converge to the leading principal component of the input data . 10 . 7554/eLife . 10094 . 004Figure 2 . Neural network architecture with feedforward connectivity . The input layer corresponds to place cells and the output to a single cell . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 004 The agent explored the environment for a sufficiently long time allowing the weights to converge to the first principal component of the temporal input data . In order to establish a spatial interpretation of the eigenvectors ( from PCA ) or the weights ( from the converged network ) we projected both the PCA eigenvectors and the network weights onto the place cells space , producing corresponding spatial activity maps . The leading eigenvectors of the PCA and the network’s weights converged to square-like periodic spatial solutions ( Figure 3A–B ) . 10 . 7554/eLife . 10094 . 005Figure 3 . Results of PCA and of the networks' output ( in different simulations ) . ( A ) 1st 16 PCA eigenvectors projected on the place cells' input space . ( B ) Converged weights of the network ( each result from different simulation , initial conditions and trajectory ) projected onto place cells' space . Note that the 8 outputs shown here resemble linear combinations of components #1 to #4 in panel A . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 005 Being a PCA algorithm , the spatial projections of the weights were periodic in space due to the covariance matrix of the input having a Toeplitz structure ( Dai et al . , 2009 ) ( a Toeplitz matrix has constant elements along each diagonal ) . Intuitively , the Toeplitz structure arises due to the spatial stationarity of the input . In fact , since we used periodic boundary conditions for the agent’s motion , the covariance matrix was a circulant matrix , and the eigenvectors were sinusoidal functions , with length constants determined by the scale of the box ( Gray , 2006 ) [a circulant matrix is defined by a single row ( or column ) , and the remaining rows ( or columns ) are obtained by cyclic permutations . It is a special case of a Toeplitz matrix - see for example Figure 1D] . The covariance matrix was heavily degenerate , with approximately 90% of the variance accounted for by the first 15% of the eigenvectors ( Figure 4B ) . The solution demonstrated a fourfold redundancy . This was apparent in the plotted eigenvalues ( from the largest to the smallest eigenvalue , Figure 4A and C ) , which demonstrated a fourfold grouping-pattern . The fourfold redundancy can be explained analytically by the symmetries of the system – see analytical treatment of PCA in Methods section ( specifically Figure 15C ) . 10 . 7554/eLife . 10094 . 006Figure 4 . Eigenvalues and eigenvectors of the input's correlation matrix . ( A ) Eigenvalue size ( normalized by the largest , from large to small ( B ) Cumulative explained variance by the eigenvalues , with 90% of variance accounted for by the first 35 eigenvectors ( out of 625 ) . ( C ) Amplitude of leading 32 eigenvalues , demonstrating that they cluster in groups of 4 or 8 . Specifically , the first four clustered groups correspond respectively ( from high to low ) to groups A , B , C & D In Figure 15C , which have the same redundancy ( 4 , 8 , 4 & 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 006 In summary , both the direct PCA algorithm and the neural network solutions developed periodic structure . However , this periodic structure was not hexagonal but rather had a square-like form . It is known that most synapses from the hippocampus to the MEC are excitatory ( Witter and Amaral , 2004 ) . We thus investigated how a non-negativity constraint , applied to the projections from place cells to grid cells , affected our simulations . As demonstrated in the analytical treatment in the Methods section , we could expect to find hexagons when imposing the non-negativity constraint . Indeed , when adding this constraint , the outputs behaved in a different manner and converged to a hexagonal grid , similar to real grid cells . While it was straightforward to constrain the neural network , calculating non-negative PCA directly was a more complicated task due to the non-convex nature of the problem ( Montanari and Richard , 2014; Kushner and Clark , 1978 ) . In the network domain , we used a simple rectification rule for the learned feedforward weights , which constrained their values to be non-negative . For the direct non-negative PCA calculation , we used the raw place cells activity ( after spatial or temporal mean normalization ) , as inputs to three different iterative numerical methods: NSPCA ( Nonnegative Sparse PCA ) , AMP ( Approximate Message Passing ) and FISTA ( Fast Iterative Threshold and Shrinkage ) based algorithms ( see Materials and methods section ) . In both cases , we found that hexagonal grid cells emerged in the output layer ( plotted as spatial projection of weights and eigenvectors: Figure 5A–B , Figure 6A–B , Video 1 , Video 2 ) . When we repeated the process over many simulations ( i . e . new trajectories and random initializations of weights ) we found that the population as a whole consistently converged to hexagonal grid-like responses , while similar simulations with the unconstrained version did not ( compare Figure 3 to Figure 5–Figure 6 ) . 10 . 7554/eLife . 10094 . 007Figure 5 . Output of the neural network when weights are constrained to be non-negative . ( A ) Converged weights ( from different simulations ) of the network projected onto place cells space . See an example of a simulation in Video 1 . ( B ) Spatial autocorrelations of ( A ) . See an example of the evolution of autorcorrelation in simulation in Video 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 00710 . 7554/eLife . 10094 . 008Video 1 . Evolution in time of the network's weights . 625 Place-cells used as input . Video frame shown every 3000 time steps up to t=1 , 000 , 000 . Video converges to results similar to those of Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 00810 . 7554/eLife . 10094 . 009Video 2 . Evolution of autocorrelation pattern of network's weights shown in Video 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 00910 . 7554/eLife . 10094 . 010Figure 6 . Results from the non-negative PCA algorithm . ( A ) Spatial projection of the leading eigenvector on input space . ( B ) Corresponding spatial autocorrelations . The different solutions are outcomes of multiple simulations with identical settings in a new environment and new random initial conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 010 In order to further assess the hexagonal grid emerging in the output , we calculated the mean ( hexagonal ) Gridness scores ( [Sargolini et al . , 2006] , which measure the degree to which the solution resembles a hexagonal grid [see Materials and methods] ) . We ran about 1500 simulations of the network ( in each simulation , the network consisted of 625 place cell-like inputs and a single grid cell-like output ) , and found noticeable differences between the constrained and unconstrained cases . Namely , the Gridness score in the non-negatively constrained-weight simulations was significantly higher than in the unconstrained-weight case ( Gridness = 1 . 07 ± 0 . 003 in the constrained case vs . 0 . 302 ± 0 . 003 in the unconstrained case . see Figure 7 ) . A similar difference was observed with the direct non-negative PCA methods ( 1500 simulations , each with different trajectories , Gridness = 1 . 13 ± 0 . 0022 in the constrained case vs . 0 . 27 ± 0 . 0023 in the unconstrained case ) . 10 . 7554/eLife . 10094 . 011Figure 7 . Histograms of Gridness values from network and PCA . First row ( A ) + ( C ) corresponds to network results , and second row ( B ) + ( D ) to PCA . The left column histograms contain the 60° Gridness scores and the right one the 90° Gridness scores . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 011 Another score we tested was a 'Square Gridness' score ( see Materials and methods ) where we measured the 'Squareness' of the solutions ( as opposed to 'Hexagonality' ) . We found that the unconstrained network had a higher square-Gridness score while the constrained network had a lower square-Gridness score ( Figure 7 ) ; for both the direct-PCA calculation ( square-Gridness = 0 . 89 ± 0 . 0074 in the unconstrained case vs . 0 . 1 ± 0 . 006 in the constrained case ) and the neural-network ( square-Gridness = 0 . 073 ± 0 . 006 in the constrained case vs . 0 . 73 ± 0 . 008 in the unconstrained case ) . All in all , these results suggest that when direct PCA eigenvectors and neural network weights were unconstrained they converged to periodic square solutions . However , when constrained to be non-negative , the direct PCA , and the corresponding neural network weights , both converged to a hexagonal solution . We investigated the effect of different inputs on the emergence of the grid structure in the networks' output . We found that some manipulation of the input was necessary in orderto enable the implementation of PCA in the neural network . Specifically , PCA requires a zero-mean input , while simple Gaussian-like place cells do not possess this property . In order to obtain input with zero-mean , we either performed differentiation of the place cells’ activity in time , or used a Mexican-hat like ( Laplacian ) shape ( See Materials and methods for more details on the different types of inputs ) . Another option we explored was the usage of positive-negative disks with a total sum of zero activity in space ( Figure 8 ) . The motivation for the use of Mexican-hat like transformations is their abundance in the nervous system ( Wiesel and Hubel , 1963; Enroth-Cugell and Robson , 1966; Derdikman et al . , 2003 ) . 10 . 7554/eLife . 10094 . 012Figure 8 . Different types of spatial input used in our network . ( A ) 2D Gaussian function , acting as a simple place cell . ( B ) Laplacian function or Mexican hat . ( C ) A positive ( inner circle ) - negative ( outer ring ) disk . While inputs as in panel A do not converge to hexagonal grids , inputs as in panels B or C do converge . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 012 We found that usage of simple 2-D Gaussian-functions as inputs did not generate hexagonal grid cells as outputs ( Figure 9 ) . On the other hand , time-differentiated inputs , positive-negative disks or Laplacian inputs did generate grid-like output cells , both when running the non-negative PCA directly ( Figure 6 ) , or by simulating the non-negatively constrained Neural Network ( Figure 5 ) . Another approach we used for obtaining zero-mean was to subtract the mean dynamically from every output individually ( see Materials and methods ) . The latter approach , related to adaptation of the firing rate , was adopted from Kropff & Treves ( Kropff and Treves , 2008 ) , who used it to control various aspects of the grid cell's activity . In addition to controlling the firing rate of the grid cells , if applied correctly , the adaptation could be exploited to keep the output's activity stable , with zero-mean rates . We applied this method in our system and in this case the outputs converged to hexagonal grid cells as well , similarly to the previous cases ( e . g . derivative in time , or Mexican hats as inputs; data not shown ) . 10 . 7554/eLife . 10094 . 013Figure 9 . Spatial projection of outputs’ weights in the neural network when inputs did not have zero mean ( such as in Figure 8A ) . ( A ) Various weights plotted spatially as projection onto place cells space . ( B ) Autocorrelation of ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 013 In summary , two conditions were required for the neural network to converge to spatial solutions resembling hexagonal grid cells: ( 1 ) non-negativity of the feedforward weights and ( 2 ) an effective zero-mean of the inputs ( in time or space ) . Under certain conditions ( e . g . , decaying learning rates and independent and identically distributed ( i . i . d . ) inputs ) , it was previously proved ( Hornik and Kuan , 1992 ) , using techniques from the theory of stochastic approximation , that the system described here can be asymptotically analyzed in terms of ( deterministic ) Ordinary Differential Equations ( ODE ) , rather than in terms of the stochastic recurrence equations . Since the ODE defining the converged weights is non-linear , we solved the ODEs numerically ( see Materials and methods ) , by randomly initializing the weight vector . The asymptotic equilibria were reached much faster , compared to the outcome of the recurrence equations . Similarly to the recurrence equations , constraining the weights to be non-negative induced them to converge into a hexagonal shape while a non-constrained system produced square-like outcomes ( Figure 11 ) . 10 . 7554/eLife . 10094 . 015Figure 11 . Numerical convergence of the ODE to hexagonal results when weights are constrained . ( A ) + ( B ) : 60° and 90° Gridness score histograms . Each score represents a different weight vector of the solution J . ( C ) + ( D ) : Spatial results for constrained and unconstrained scenarios , respectively . ( E ) + ( F ) Spatial autocorrelations of ( C ) + ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 015 Simulation was run 60 times , with 400 outputs per run . 60° Gridness score mean was 1 . 1 ± 0 . 0006 when weights were constrained and 0 . 29 ± 0 . 0005 when weights were unconstrained . 90° Gridness score mean was 0 . 006 ± 0 . 002 when weights were constrained and 0 . 8 ± 0 . 0017 when weights were unconstrained . A more detailed view of the resulting grid spacing showed that it was heavily dependent on the field widths of the place cells inputs . When the environment size was fixed and the output calculated per input size , the grid-spacing ( distance between neighboring peaks ) increased for larger place cell field widths . To enable a fast parameter sweep over many place cell field widths ( and large environment sizes ) , we took the steady state limit , and the limit of a high density of place cell locations , and used the fast FISTA algorithm to solve the non-negative PCA problem ( see Materials and methods section ) . We performed multiple simulations , and found that there was a simple linear dependency between the place field size and the output grid scale . For the case of periodic boundary conditions , we found that grid scale was S = 7 . 5sigma+0 . 85 , where sigma was the width of the place cell field ( Figure 12A ) . For a different set of simulations with zero boundary conditions , we achieved a similar relation: S=7 . 54sigma+0 . 62 ( figure not shown ) . Grid scale was more dependent on place field size and less on box size ( Figure 12H ) . We note that for very large environments , the effects of boundary conditions diminishes . At this limit , this linear relation between place field size and grid scale can be explained from analytical considerations ( see Materials and methods section ) . Intuitively , this follows from dimensional analysis: given an infinite environment , at steady state the length scale of the place cell field width is the only length scale in the model , so any other length scale must be proportional to this scale . More precisely , we can provide a lower bound for the linear fit ( Figure 12A ) , which depends only on the tuning curve of the place cells ( see Materials and methods section ) . This lower bound was derived for periodic boundary conditions , but works well even with zero boundary conditions ( not shown ) . 10 . 7554/eLife . 10094 . 016Figure 12 . Effect of changing the place-field size in fixed arena ( FISTA algorithm used; periodic boundary conditions and Arena size 500 ) ; ( A ) Grid scale as a function of place field size ( sigma ) ; Linear fit is: Scale = 7 . 4 Sigma+0 . 62; the lower bound , equal to 2π/k† , were k† is defined in Equation 32 in the Materials and methods section; ( B ) Grid orientation as a function of gridness; ( C ) Grid orientation as a function of sigma – scatter plot ( blue stars ) and mean ( green line ) ; ( D ) Histogram of grid orientations; ( E ) Mean gridness as a function of sigma; and ( F ) Histogram of mean gridness . ( G ) Gridness as a function of sigma and ( arena-size/sigma ) ( zero boundary conditions ) . ( H ) Grid scale for the same parameters as in G . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 016 Furthermore , we found that the grid orientation varied substantially for different place cell field widths , in the possible range of 0–15 degrees ( Figure 12C , D ) . For small environments , the orientation strongly depended on the boundary conditions . However , as described in the Methods section , analytical considerations suggest that as the environment grows , the distribution of grid orientations becomes uniform in the range of 0–15 degrees , with a mean at 7 . 5° . Intuitively , this can be explained by rotational symmetry – when the environment size is infinite , all directions in the model are equivalent , and so we should get all orientations with equal probability , if we start the model from a uniformly random initialization . In addition , grid orientation was not a clear function of the gridness of the obtained grid cells ( Figure 12B ) . For large enough place cells , gridness was larger than 1 ( Figure 12E–G ) . It is known that in reality grid cells form in modules of multiple spacings ( Barry et al . , 2007; Stensola et al . , 2012 ) . We tried to address this question of modules in several ways . First , we used different widths for the Gaussian/Laplacian input functions: Initially , we placed a heterogeneous population of widths in a given environment ( i . e . , uniformly random widths ) and ran the single-output network 100 times . The distribution of grid spacings was almost comparable to the results of the largest width if applied alone , and did not exhibit module like behavior . This result is not surprising when thinking about a small place cell overlapping in space with a large place cell . Whenever the agent passes next to the small one , it activates both weights via synaptic learning . This causes the large firing field to overshadow the smaller one . Additionally , when using populations of only two widths of place fields , the grid spacings were dictated by the size of the larger place field ( data not shown ) . The second option we considered was to use a multi-output neural network , capable of computing all 'eigenvectors' rather than only the principal 'eigenvector' ( where by 'eigenvector' we mean here the vectors achieved under the positivity constraint , and not the exact eigenvectors themselves ) . We used a hierarchical network implementation introduced by Sanger , 1989 ( see Materials and methods ) . Since the 1st output’s weights converged to the 1st 'eigenvector' , the network ( Figure 13A–B ) provided to the subsequent outputs ( 2nd , 3rd , and so forth ) a reduced-version of the data from which the projection of the 1st 'eigenvector' has been subtracted out . This process , reminiscent of Gram-Schmidt orthogonalization , was capable of computing all 'eigenvectors' ( in the modified sense ) of the input's covariance matrix . It is important to note though that , due to the non-negativity constraint , the vectors achieved in this way were not orthogonal , and thus it cannot be considered a real orthogonalization process , although , as explained in the Methods section , the process does aim for maximum difference between the vectors . 10 . 7554/eLife . 10094 . 017Figure 13 . Hierarchial network capable of computing all 'principal components' . ( A ) Each output is a linear sum of all inputs weighted by the corresponding learned weights . ( B ) Over time , the data the following outputs 'see' is the original data after subtration of the 1st 'eigenvector's' projection onto it . This is an iterative process causing all outputs' weights to converge to the 'prinipcal components' of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 017 When constrained to be non-negative , and using the same homogeneous 'place cells' as in the previous network , the networks' weights converged to hexagonal shapes . Here , however , we found that the smaller the 'eigenvalue' was ( or the higher the principal component number ) the denser the grid became . We were able to identify two main populations of grid-distance 'modules' among the hexagonal spatial solutions with high Gridness scores ( >0 . 7 , Figure 14A–B ) . In addition , we found that the ratio between the distances of the modules was −1 . 4 , close to the value of 1 . 42 found by Stensola et al . ( Stensola et al . , 2012 ) . Although we searched for additional such jumps , we could only identify this single jump , suggesting that our model can yield up to two 'modules' and not more . The same process was repeated using the direct PCA method , utilizing the covariance matrix of the data after simulation as input for the non-negative PCA algorithms , and considering their ability to calculate only the 1st 'eigenvector' . By iteratively projecting the 1st 'eigenvector' on the simulation data and subtracting the outcome from the original data , we applied the non-negative PCA algorithm to the residual data obtaining the 2nd 'eigenvector' of the original data . This 'eigenvector' now constituted the 1st eigenvector' of the new residual data ( see Materials and methods ) . Applying this process to as many 'outputs' as needed , we obtained very similar results to the ones presented above using the neural network ( data not shown ) . 10 . 7554/eLife . 10094 . 018Figure 14 . Modules of grid cells . ( A ) In a network with 50 outputs , the grid spacing per output is plotted with respect to the hierarchical place of the output . ( B ) The grid spacing of outputs with high Gridness score ( >0 . 7 ) . The centroids have a ratio of close to √2 . ( C ) + ( D ) Example of rate maps of outputs and their spatial autocorrelations for both of the modules . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 018
As a consequence of the requirements for PCA to hold , we found that the place cell input needed to have a zero-mean , otherwise the output was not periodic . Due to the lack of the zero-mean property in 2D Gaussians , we used various approaches to impose zero-mean on the input data . The first , in the time domain , was to differentiate the input and use the derivatives ( a random walk produces zero-mean derivatives ) as inputs . Another approach was to dynamically subtract the mean in all iterations of the simulation . This approach was reminiscent of the adaptation procedure suggested in the Kropff & Treves paper ( Kropff and Treves , 2008 ) . A third approach , applied in the spatial domain was to use inputs with a zero-spatial mean such as Laplacians of Gaussians ( Mexican hats in 2D , or differences-of-Gaussians ) or negative – positive disks . Such Mexican-hat inputs are quite typical in the nervous system ( Wiesel and Hubel , 1963; Enroth-Cugell and Robson , 1966; Derdikman et al . , 2003 ) , although in the case of place cells it is not completely known how they are formed . They could be a result of interaction between place cells and the vast number of inhibitory interneurons in the local hippocampal network ( Freund and Buzsáki , 1996 ) . Another condition we found crucial , which was not part of the original PCA network , was a non-negativity constraint on the place-to-grid learned weights . While rather easy to implement in the network , adding this constraint to the non-convex PCA problem was harder to implement . Since the problem is NP-hard ( Montanari and Richard , 2014 ) , we turned to numerical methods . We used three different algorithms ( Montanari and Richard , 2014; Zass and Shashua , 2006; Beck and Teboulle , 2009 ) to find the leading 'eigenvector' of every given temporal based input . As shown in the results section , both processes ( i . e . direct PCA and the neural network ) resulted in hexagonal outcomes when the non-negativity and zero-mean criteria were met . Note that the ease of use of the neural network for solving the positive PCA problem is a nice feature of the neural network implementation , and should be investigated further . We also note that while our network focused on the projection from place cells to grid cells , we cannot preclude the importance of the reciprocal projection from grid cells to place cells . Further study will be needed to ‘close the loop’ and simultaneously consider both of these projections at once . We note that similar work has noticed the relation between place-cell-to-grid-cell transformation and PCA . Notably , Stachenfeld et al . , ( 2014 ) have demonstrated , from considerations related to reinforcement learning , that grid cells could be related to place cells through a PCA transformation . However , due to the unconstrained nature of their transformation , the resulting grid cells were square-like . Furthermore , there has been an endeavor to model the transformation from place cells to grid cells using independent-component-analysis ( Franzius et al . , 2007 ) . We also note that there is now a surge of interest in the feedback projection from place cells to grid-cells , which is inverse to the anatomical downstream direction from grid cells to place cells ( Witter and Amaral , 2004 ) that has guided most of the models to-date ( Zilli , 2012; Giocomo et al . , 2011 ) . In addition to several papers from the Treves group , in which the projection from place cells to grid cells is studied ( Kropff and Treves , 2008; Si and Treves , 2013 ) , there has been also recent work from other groups as well exploring this direction ( Castro and Aguiar , 2014; Stepanyuk , 2015 ) . As far as we are aware , none of the previous studies noted the importance of the non-negativity constraint and the requirement of zero mean input . Additionally , to the best of our knowledge , the analytic results and insights provided in this work ( see Materials and methods ) are novel , and provide a mathematically consistent explanation for the emergence of hexagonally-spaced grid cells . Based on the findings of this work , it is possible to make several predictions . First , the grid cells must receive zero-mean input over time to produce hexagonally shaped firing patterns . With all feedback projections from place cells being excitatory , the lateral inhibition from other neighboring grid cells might be the balancing parameter to achieve the temporal zero-mean ( Couey et al . , 2013 ) . Alternatively , an adaptation method , such as the one suggested in Kropff and Treves , ( 2008 ) may be applied . Second , if indeed the grid cells are a lower dimensional representation of the place cells in a PCA form , the place-to-grid neural weights distribution should be similar across identically spaced grid cell populations . This is because all grid cells with similar spacing would have maximized the variance over the same input , resulting in similar spatial solutions . As an aside , we note that such a projection may be a source of phase-related correlations in grid cells ( Tocker et al . , 2015 ) . Third , we found a linear relation between the size of the place cells and the spacing between grid cells . Furthermore , the spacing of the grid cells is mostly determined by the size of the largest place cell – predicting that the feedback from large place cells is not connected to grid cells with small spacing . Fourth , we found modules of different grid spacings in a hierarchical network with the ratio of distances between successive units close to √2 . This result is in accordance with the ratio reported in Stensola et al . , ( 2012 ) . However , we note that there is a difference between our results and experimental results because the analysis predicts that there should only be two modules , while the data show at least 5 modules , with a range of scales , the smallest and most numerous having approximately the scale of the smaller place fields found in the dorsal hippocampus ( 25–30 cm ) . Fifth , for large enough environments our model suggests that , from mathematical considerations , the grid orientation should approach a uniform orientation in the possible range of 0–15° . This is in discrepancy with experimental results which measure a peak at 7 . 5° , and not a uniform distribution ( Stensola et al . , 2015 ) . As noted , the discrepancies between our results and reality may relate to the fact that a more advanced model will have to take into account both the downstream projection from grid cells to place cells together with the upstream projection from place cells to grid cells discussed in this paper . Furthermore , such a model will have to take into account the non-uniform distribution of place-cell widths ( Kjelstrup et al . , 2008 ) . In light of our results , we further asked what is special about the hexagonal shape which renders it a stable solution . Past works have demonstrated that hexagonality is optimal in terms of efficient coding . Two recent papers have addressed the potential benefit of encoding by grid cells . Mathis et al . , ( 2015 ) considered the decoding of spatial information based on a grid-like periodic representation . Using lower bounds on the reconstruction error based on a Fisher information criterion , they demonstrated that hexagonal grids lead to the highest spatial resolution in two dimensions ( extensions to higher dimensions were also provided ) . The solution is obtained by mapping the problem onto a circle packing problem . The work of Wei et al . , ( 2013 ) also took a decoding perspective , and showed that hexagonal grids minimize the number of neurons required to encode location with a given resolution . Both papers offer insights into the possible information theoretic benefits of the hexagonal grid solution . In the present paper , we were mainly concerned with a specific biologically motivated learning ( development ) mechanism that may yield such a solution . Our analysis suggests that the hexagonal patterns can arise as a solution that maximizes the grid cell output variance , under non-negativity constraints . In Fourier space , the solution is a hexagonal lattice with lattice constant near the peak of the Fourier transform of the place cell tuning curve ( Figures 15 and 16; see Materials and methods ) . To conclude , this work demonstrates how grid cells could be formed from a simple Hebbian neural network with place cells as inputs , without needing to rely on path-integration mechanisms .
We implemented a single-layer neural network with feedforward connections that was capable of producing a hexagonal-like output ( Figure 2 ) . The feedforward connections were updated according to a self-normalizing version of a Hebbian learning rule referred to as the Oja rule ( Oja , 1982 ) , ( 1 ) ΔJit=εt ( ψtrit− ( ψt ) 2Jit ) , where εt denotes the learning rate , Jit is the ith weight and ψt , rit are the output and the ith input of the network , respectively ( all at time t ) . The weights were initialized randomly according to a uniform distribution and then normalized to have norm 1 . The output ψt was calculated every iteration by summing up all pre-synaptic activity from the entire input neuron population . The activity of each output was processed through a sigmoidal function ( e . g . , tanh ) or a simple linear function . Formally , ( 2 ) ψt=f∑i=1nJit⋅rit , where n is the number of input place cells . Since we were initially only concerned with the eigenvector associated with the largest eigenvalue , we did not implement a multiple-output architecture . In this formulation , in which no lateral weights were used , multiple outputs were equivalent to running the same setting with one output several times . As discussed in the introduction , this kind of simple feedforward neural network with linear activation and a local weight update in the form of Oja’s rule ( 1 ) is known to perform Principal Components Analysis ( PCA ) ( Oja , 1982; Sanger , 1989; Weingessel and Hornik , 2000 ) . In the case of a single output the feedforward weights converge to the principal eigenvector of the input's covariance matrix . With several outputs , and lateral weights , as described in the section on modules , the weights converge to the leading principal eigenvectors of the covariance matrix , or , in certain cases ( Weingessel and Hornik , 2000 ) , to the subspace spanned by the principal eigenvectors . We can thus compare the results of the neural network to those of the mathematical procedure of PCA . Hence , in our simulation , we ( 1 ) let the neural networks' weights develop in real time based on the current place cell inputs . In addition , we ( 2 ) saved the input activity for every time step to calculate the input covariance matrix and perform ( batch ) PCA directly . It is worth mentioning that the PCA solution described in this section can be interpreted differently based on the Singular Value Decomposition ( SVD ) . Denoting by R the T×d spatio-temporal pattern of place cell activities ( after setting the mean to zero ) , where T is the time duration and d is the number of place cells , the SVD decomposition ( see Jolliffe , 2002; sec . 3 . 5 ) for R is R=ULA' . For a matrix R of rank r , L is a r×r diagonal matrix whose kth element is equal to lk1/2 , the square root of the kth eigenvalue of the covariance matrix RR' ( computed in the PCA analysis ) , A is the d×r matrix with kth column equal to the kth eigenvector of RR' , and U is the T×r matrix whose kth column is lk−1/2Rak . Note that U is a T×r dimensional matrix whose kth column represents the temporal dynamics of the kth grid cell . In other words , the SVD provides a decomposition of the place cell activity in terms of the grid cell activity , as opposed to the grid cell representation in terms of place cell activity we discussed so far . The network learns the spatial weights over place cells ( the eigenvectors ) as the connections weights from the place cells , and 'projection onto place cell space' ( lk−1/2Rak ) is simply the firing rates of the output neuron plotted against the location of the agent . The question we therefore asked was under what conditions , when using place cell-like inputs , a solution resembling hexagonal grid cells emerges . To answer this we used both the neural-network implementation and the direct calculation of the PCA coefficients . We simulated an agent moving in a 2D virtual environment consisting of a square arena covered by n uniformly distributed 2D Gaussian-shaped place cells , organized on a grid , given by ( 3 ) rit ( X ( t ) ) =exp ( − ( X ( t ) −Ci ) 22σi2 ) , i=1 , 2 , . . . , n where X ( t ) represents the location of the agent . The variables rit constitute the temporal input from place cell i at time t , and Ci , σi are the ith place cell’s field center and width , respectively ( see variations on this input structure below ) . In order to eliminate boundary effects , periodic boundary conditions were assumed . The virtual agent moved about in a random walk scheme ( see Appendix ) and explored the environment ( Figure 1A ) . The place cell centers were assumed to be uniformly distributed ( Figure 1B ) and shared the same standard deviation σ . The activity of all place cells as a function of time ( r ( t ) 1 , r ( t ) 2…r ( t ) n ) was dependent on the stochastic movement of the agent , and formed a [Neuron x Time] matrix ( r∈RnxT , with T- being the time dimension , see Figure 1C ) . The simulation was run several times with different input arguments ( see Table 1 ) . The agent was simulated for T time steps , allowing the neural network's weights to develop and reach a steady state by using the learning rule ( Equations 1 , 2 ) and the input ( Equation 3 ) data . The simulation parameters are listed below and include parameters related to the environment , simulation , agent and network variables . 10 . 7554/eLife . 10094 . 019Table 1 . List of variables used in simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 019Environment:Size of arenaPlace cells field widthPlace cells distributionAgent:Velocity ( angular & linear ) Initial position-------------------Network:# Place cells/ #Grid cellsLearning rateAdaptation variable ( if used ) Simulation:Duration ( time ) Time step------------------- To calculate the PCA directly , we used the MATLAB function Princomp in order to evaluate the n principal eigenvectors {q→k}k=1n and corresponding eigenvalues of the input covariance matrix . As mentioned in the Results section , there exists a near fourfold redundancy in the eigenvectors ( X-Y axis and in phase ) . Figure 3 demonstrates this redundancy by plotting the eigenvalues of the covariance matrix . The output response of each eigenvector q→k corresponding to a 2D input location ( x , y ) is ( 4 ) Φ ( x , y ) k=∑j=1nqkj exp ( − ( x−cxj ) 22σx2− ( y−cyj ) 22σy2 ) , k=1 , 2 , . . . , n where cxj and cyj are the x , y components of the centers of the individual place cell fields . Unless otherwise mentioned , we used place cells in a rectangular grid , such that a place cell is centered at each pixel of the image ( that is – number of place cells equals the number of image pixels ) . Projections between place cells and grid cells are known to be primarily excitatory ( Witter and Amaral , 2004 ) , thus if we aim to mimic the biological circuit , a non-negativity constraint should be added to the feedforward weights in the neural network . While implementing a non-negativity constraint in the neural network is rather easy ( a simple rectification rule in the weight dynamics , such that weights which are smaller than 0 are set to 0 ) , the equivalent condition for calculating non-negative Principal Components is more intricate . Since this problem is non-convex and , in general , NP-hard ( Montanari and Richard , 2014 ) , a numerical procedure was imperative . We used three different algorithms for this purpose . The first ( Zass and Shashua , 2006 ) named NSPCA ( Nonnegative Sparse PCA ) is based on coordinate-descent . The algorithm computes a non-negative version of the covariance matrix's eigenvectors and relies on solving a numerical optimization problem , converging to a local maximum starting from a random initial point . The local nature of the algorithm did not guarantee a convergence to a global optimum ( recall that the problem is non-convex ) . The algorithm's inputs consisted of the place cell activities’ covariance matrix , α - a balancing parameter between reconstruction and orthonormality , β – a variable which controls the amount of sparseness required , and an initial solution vector . For the sake of generality , we set the initial vector to be uniformly random ( and normalized ) , α was set to a relatively high value – 104 and since no sparseness was needed , β was set to zero . The second algorithm ( Montanari and Richard , 2014 ) does not require any simulation parameters except an arbitrary initialization . It works directly on the inputs and uses a message passing algorithm to define an iterative algorithm to approximately solve the optimization problem . Under specific assumptions it can be shown that the algorithm asymptotically solves the problem ( for large input dimensions ) . The third algorithm we use is the parameter free Fast Iterative Threshold and Shrinkage algorithm FISTA ( Beck and Teboulle , 2009 ) . As described later in this section , this algorithm is the fastest of the three , and allowed us rapid screening of parameter space . Performing PCA on raw data requires the subtraction of the data mean . Some thought was required in order to determine how to perform this subtraction in the case of the neural network . One way to perform the subtraction in the time domain was to dynamically subtract the mean during simulation by using the discrete 1st or 2nd derivatives of the inputs in time [i . e . from Equation 3 , ∆r ( t+1 ) =r ( t+1 ) −rt] . Under conditions of an isotropic random walk ( namely , given any starting position , motion in all directions is equally likely ) it is clear that E[∆r ( t ) ]=0 . Another option for subtracting the mean in the time domain was the use of an adaptation variable , as was initially introduced by Kropff and Treves , ( 2008 ) . Although originally exploited for control over the firing rate , it can be viewed as a variable that represents subtraction of a weighted sum of the firing rate history . Instead of using the inputs rit directly in Equation 2 to compute the activation ψt , an intermediate adaptation variable ψadpt ( δ ) was used ( δ being the relative significance of the present temporal sample ) as ( 5 ) ψadpt=ψt−ψ¯t , ( 6 ) ψ¯t= ( 1−δ ) ⋅ψ¯t−1+δψt . It is not hard to see that for i . i . d . variables ψadpt , the sequence ψ¯t converges for large t to the mean of ψt . Thus , when t→∞ we find that E[ψadpt]→0 , specifically , the adaptation variable is of zero asymptotic mean . The second method we used to enforce a zero mean input was simply to create it in advance . Rather than using 2D Gaussian functions ( i . e . [Equation 3] ) as inputs we used 2D difference-of-Gaussians ( all σ are equal in x and y axis ) : ( 7 ) rit ( X ( t ) ) =c1 , i exp ( − ( X ( t ) −Ci ) 22σ1 , i2 ) −c2 , i exp ( − ( X ( t ) −Ci ) 22σ2 , i2 ) , i=1 , 2 , . . . , n where the constants c1 and c2 are set so the integral of the given Laplacian function is zero ( if the environment size is not too small , then c1 , i/c2 , i ≈ σ2 , i/σ1 , i ) . Therefore , if we assume a random walk that covers the entire environment uniformly , the temporal mean of the input would be zero as well . Such input data can be inspired by similar behavior of neurons in the retina and the lateral-geniculate nucleus ( Wiesel and Hubel , 1963; Enroth-Cugell and Robson , 1966 ) . Finally , we implemented another input data type; positive-negative disks ( see Appendix ) . Analogously to the difference-of-Gaussians function , the integral over input is zero so the same goal ( zero-mean ) was achieved . It is worthwhile noting that subtracting a constant from a simple Gaussian function is not sufficient since at infinity it does not reach zero . In order to test the hexagonality of the results we used a hexagonal Gridness score ( Sargolini et al . , 2006 ) . The Gridness score of the spatial fields was calculated from a cropped ring of their autocorrelogram including the six maxima closest to the center . The ring was rotated six times , 30∘ per rotation , reaching in total angles of 30∘ , 60∘ , 90∘ , 120∘ , 150∘ . Furthermore , for every rotated angle the Pearson correlation with the original un-rotated map was obtained . Denoting by Cγ the correlation for a specific rotation angle γ , the final Gridness score was ( Kropff and Treves , 2008 ) : ( 8 ) Gridness = 12 ( C60+C120 ) −13 ( C30+C90+C150 ) . In addition to this 'traditional' score we used a Squareness Gridness score in order to examine how square-like the results are spatially . The special reference to the square shape was driven by the tendency of the spatial solution to converge to a rectangular shape when no constrains were applied . The Squareness Gridness score is similar to the hexagonal one , but now the cropped ring of the autocorrelogram is rotated 45∘ every iteration to reach angles of 45∘ , 90∘ , 135∘ . As before , denoting Cγ as the correlation for a specific rotation angle γ the new Gridness score was calculated as: ( 9 ) Square Gridness=C90−12 ( C45+C135 ) . All errors calculated in gridness measures are SEM ( Standard Error of the Mean ) . As described in the Results section , we were interested to check whether a hierarchy of outputs could explain the module phenomenon described for real grid cells . We replaced the single-output network with a hierarchical , multiple outputs network , which is capable of computing all 'principal components' of the input data while maintaining the non-negativity constraint as before . The network , introduced by Sanger , 1989 , computes each output as a linear summation of the weighted inputs similar to Equation 2 . However , the weights are now calculated according to: ( 10 ) ΔJijt=εt ( rjtψit−ψit∑k=1iJkjtψkt ) . The first term in the parenthesis when k=1 was the regular Hebb-Oja derived rule . In other words , the first output calculated the first non-negative 'principal component' ( in inverted commas due to the non-negativity ) of the data . Following the first one , the weights of each output received a back projection from the previous outputs . This learning rule applied to the data in a similar manner to the Gram-Schmidt process , subtracting the 'influence' of the previous 'principal components' on the data and recalculating the appropriate 'principal components' of the updated input data . In a comparable manner , we applied this technique to the input data X in order to obtain non-negative 'eigenvectors' from the direct nonnegative-PCA algorithms . We found V2 by subtracting from the data the projection of V1 on it , ( 11 ) X~=X−V1T ( V1⋅X ) . Next , we computed V2 , the first non-negative 'principal component' of X~ , and similarly the subsequent ones . In order to test the stability of the solutions we obtained under all types of conditions , we applied the ODE method ( Kushner and Clark , 1978; Hornik and Kuan , 1992; Weingessel and Hornik , 2000 ) to the PCA feature extraction algorithm introduced in pervious sections . This method allows one to asymptotically replace the stochastic update equations describing the neural dynamics by smooth differential equations describing the average asymptotic behavior . Under appropriate conditions , the stochastic dynamics converge with probability one to the solution of the ODEs . Although originally this approach was designed for a more general architecture ( including lateral connections and asymmetric updating rules ) , we used a restricted version for our system . In addition , the following analysis is accurate solely for linear output functions . However , since our architecture works well with either linear or non-linear output functions , the conclusions are valid . We can rewrite the relevant updating equations of the linear neural network ( in matrix form ) , ( see [Weingessel and Hornik , 2000] Equations 15–19 ) : ( 12 ) ψt+1=Q⋅Jt⋅ ( rt ) T , ( 13 ) ΔJt=εt ( ψt ( rt ) T−Φ ( ψt⋅ ( ψt ) T ) Jt ) . In our case we setQ=I , Φ=diag . Consider the following assumptions A typical suitable sequence is εt=1t , t=1 , 2… . For long times , we denote ( 14 ) E[ψt ( rt ) T]→E[J⋅r⋅rT]=E[J]⋅E[r⋅rT]=JΣ , ( 15 ) limt→∞E[ψψT]=E[J]⋅E[rrT]⋅E[JT]=JΣJT . The penultimate equalities in these equations used the fact that the weights converge with probability one to their average value , resulting from the solution of the ODEs . Following Weingessel and Hornik , ( 2000 ) , we can analyze Equations 12 , 13 under the above assumptions , via their asymptotically equivalent associated ODEs ( 16 ) dJdt=JΣ−diag ( JΣJT ) J , with equilibria at ( 17 ) JΣ=diag ( JΣJT ) J . We solved it numerically by exploiting the same covariance matrix and initializing with random weights J . In line with our previous findings , we found that constraining J to be non-negative ( by a simple cut-off rule ) resulted in a hexagonal shape ( in the projection of J onto the place cells space; Figure 11 ) . In contrast , when the weights were not constrained they converged to square-like results . From this point onwards , we focus on the case of a single output , in which J is a row vector , unless stated otherwise . In the unconstrained case , from Equation 17 any J which is a normalized eigenvector of Σ would be a fixed point . However , from Equation 16 , only the principal eigenvector , which is the solution to the following optimization problem ( 18 ) maxJ:JTJ=1 JΣJT would correspond to a stable fixed point . This is the standard PCA problem . By adding the constraint J≥0 we get the non-negative PCA problem . To speed up simulation and simplify analysis we make further simplifications . First , we assume that the agent’s random movement is ergodic ( e . g . , an isotropic random walk in a finite box as we used in our simulation ) , uniform and covering the entire environment , so that ( 19 ) JΣJT=Eψ2 ( X ( t ) ) =1|S|∫Sψ2 ( x ) dx , where x denotes location vector ( in contrast to X ( t ) , which is the random process corresponding to the location of the agent ) , S is the entire environment , and |S| is the size of the environment . Second , we assume that the environment S is uniformly and densely covered by identical place cells , each of which has the same a tuning curve r ( x ) ( which integrates to zero ) . In this case , the activity of the linear grid cell becomes a convolution operation ( 20 ) ψ ( x ) =∫SJ ( x' ) r ( x−x' ) dx' , where J ( x ) is the synaptic weight connecting to the place cell at location x . Thus , we can write our objective as ( 21 ) 1|S|∫Sψ2 ( x ) dx=1|S|∫S ( ∫SJ ( x' ) r ( x−x' ) dx' ) 2dx under the constraint that the weights are normalized ( 22 ) 1S∫SJ2 ( x ) dx=1 , where either J ( x ) ∈ℝ ( PCA ) or J ( x ) ≥ 0 ( 'non-negative PCA' ) . Since we expressed the objective using a convolution operation ( different boundary conditions can be assumed ) , it can be solved numerically considerably faster . In the non-negative case , we used the parameter free Fast Iterative Threshold and Shrinkage algorithm [FISTA ( Beck and Teboulle , 2009 ) ; in which we do not use shrinkage , since we only have hard constraints] , where the gradient was calculated efficiently using convolutions . Moreover , as we show in the following sections , if we assume periodic boundary conditions and use Fourier analysis , we can analytically find the PCA solutions , and obtain important insight on the non-negative PCA solutions . Any continuously differentiable function f ( x ) , defined over S≜[0 , L]D , a 'box' region in D dimensions , with periodic boundary conditions , can be written using a Fourier series ( 23 ) f^ ( k ) ≜1|S|∫Sf ( x ) eik⋅xdx , f ( x ) ≜∑k∈S^f^ ( k ) e−ik⋅x , where |S|=LD is the volume of the box andS^≜{ ( 2m1πL , … , 2mdπL ) } ( m1 , … , md ) ∈ℤD is the reciprocal lattice of S in k-space ( frequency space ) . Assuming periodic boundary conditions , we use Parseval’s identity , and the properties of the convolution , to transform the steady state objective ( Equation 21 ) to its simpler form in the Fourier domain , ( 24 ) 1|S|∫Sψ2 ( x ) dx=∑k∈S^|J^ ( k ) r^ ( k ) |2 . Similarly , the normalization constraint can also be written in the Fourier domain , ( 25 ) 1|S|∫SJ2 ( x ) dx=∑k∈S^|J^ ( k ) |2=1 . Maximizing the objective Equation 24 under this constraint in the Fourier domain , we immediately get that any solution is a linear combination of the Fourier components , ( 26 ) J^ ( k ) ={1 , k=k*0 , k≠k* . where ( 27 ) k*∈argmaxk∈S^r^ ( k ) , and J^ ( k ) satisfies the normalization constraint . In the original space , the Fourier components are ( 28 ) J ( x ) =eik·x+iϕ , where ϕ∈[0 , 2π ) is a free parameter that determines the phase . Also , since J ( x ) should assume real values , it is composed of real Fourier components ( 29 ) J ( x ) =12 ( eik*⋅x+iϕ+e−ik*⋅x−iϕ ) =2cos ( k*⋅x+ϕ ) . This is a valid solution , since r ( x ) is a real-valued function , r^ ( k ) =r^ ( −k ) and therefore -k*∈ argmaxk∈S^r^ ( k ) . In this paper we focused on the case where r ( x ) has the shape of a difference of Gaussians ( Equation 7 ) , ( 30 ) r ( x ) ∝c1 exp ( −‖x‖22σ12 ) −c2 exp ( −‖x‖22σ22 ) where c1 and c2 are some positive normalization constants , set so that ∫Sr ( x ) dx=0 ( see appendix ) . The Fourier transform of r ( x ) is also a difference of Gaussians ( 31 ) r^ ( k ) ∝exp ( −12σ12‖k‖2 ) −exp ( −12σ22‖k‖2 ) ∀k∈S^ , as we show in the appendix . Therefore the value of the Fourier domain objective only depends on the radius ‖k‖ , and all solutions k* have the same radius ‖k*‖ . If L→∞ , then the k-lattice S^ becomes dense ( S^→ℝD ) and this radius is equal to ( 32 ) k†=argmaxk≥0 exp ( −12σ12k2 ) −exp ( −12σ22k2 ) which is a unique maximizer , that can be easily obtained numerically . Notice that if we multiply the place cell field width by some positive constant c , then the solution k† will be divided by c . The grid spacing , proportional to 1k† , would therefore also be multiplied by c . This entails a linear dependency between the place cell field width and the grid cell spacing , in the limit of a large box size ( L→∞ ) . When the box has a finite size , k-lattice discretization also has a ( usually small ) effect on the grid spacing . In that case , all solutions k* are restricted to be on the finite lattice S^ . Therefore , the solutions k* are the points on the lattice S^ for which the radius ‖k*‖ is closest to k† ( see Figure 15B , C ) . 10 . 7554/eLife . 10094 . 020Figure 15 . PCA k-space analysis for a difference of Gaussians tuning curve . ( A ) The 1D tuning curve r ( x ) . ( B ) The 1D tuning curve Fourier transform r^ ( k ) . The black circles indicate k-lattice points . The PCA solution , k* , is given by the circles closest to k† , the peak of r^ ( k ) ( red cross ) . ( C ) A contour plot of the 2D tuning curve Fourier transform r^ ( k ) . In 2D k-space the peak of r^ ( k ) becomes a circle ( red ) , and the k-lattice S^ is a square lattice ( black circles ) . The lattice point can be partitioned into equivalent groups . Several such groups are marked in blue on the lattice . For example , the PCA solution Fourier components lie on the four lattice points closest to the circle , denoted A1-4 . Note the grouping of A , B , C & D ( 4 , 8 , 4 and 4 , respectively ) corresponds to the grouping of the 20 highest principal components in Figure 4 . Parameters: 2σ1=σ2=7 . 5 , L=100 . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 020 The number of real-valued PCA solutions ( degeneracy ) in 1D is two , as there are exactly two maxima , k* and −k* . The phase ϕ , determines how the components at k* and −k* are linearly combined . However , there are more maxima in the 2D case . Specifically , given a maximum k* , we can write ( m , n ) =L2πk* , where ( m , n ) ∈ℤ2 . Usually there are 7 other different points with the same radius: ( m , −n ) , ( −m , −n ) , ( −m , n ) , ( −n , −m ) , ( n , −m ) , ( −n , m ) and ( n , m ) , so we will have a degeneracy of eight ( corresponding to the symmetries of a square box ) . This is case of points in group B , shown in Figure 15C . However , we can also get a different degeneracy . First , if either m=±n , n=0 or m=0 we will have a degeneracy of 4 , since then some of the original eight points will coincide ( groups A , C and D in Figure 15C ) . Second , additional points ( k , r ) can exist such that k2+r2=m2+n2 , ( Pythagorean triplets with the same hypotenuse ) – for example , 152+202=252=72+242 . These points will also appear in groups of four or eight . Therefore , we will always have a degeneracy which is some multiple of 4 . Note that in the full network simulation , the degeneracy is not exact . This is due to the perturbation noise from the agent’s random walk as well as the non-uniform sampling of the place cells . Next , we add the non-negativity constraint J ( x ) ≥0 . As mentioned earlier , this constraint renders the optimization problem NP-hard , and prevents us from a complete analytical solution . We therefore combine numerical and mathematical analysis , in order to gain intuition as to why Our numerical results indicate that the Fourier components of any locally optimal 1D solution of non-negative PCA have the following structure: This structure suggests that the component at k* aims to maximize the objective , while the other components guarantee the non-negativity of the solution J ( x ) . In order to gain some analytical intuition as to why this is the case , we first examine the limit case that L→∞ and r^ ( k ) is highly peaked at k† . In that case the Fourier objective ( Equation 24 ) simply becomes 2|r^ ( k† ) |2|J^ ( k† ) |2 . For simplicity , we will rescale our units so that |r^ ( k† ) |2=1/2 , and the objective becomes |J^ ( k† ) |2 . Therefore , the solution must include a Fourier component at k† or the objective would be zero . The other components exist only to maintain the non-negativity constraint , since if they increase in magnitude , then the objective , which is proportional to |J^ ( k† ) |2 , must decrease to compensate ( due to the normalization constraint – Equation 25 ) . Note that these components must include a positive 'DC component' at k=0 , or else ∫SJ ( x ) dx∝J^ ( 0 ) ≤0 , which contradicts the constraints . To find all the Fourier components , we examine a solution composed of only a few ( M ) componentsJ ( x ) =J^ ( 0 ) +2∑m=1MJ^mcos ( kmx+ϕm ) . Clearly , we can set k1=k† , or otherwise , the objective would be zero . Also , we must haveJ^ ( 0 ) =−minx ( 2∑m=1MJ^mcos ( kmx+ϕm ) ) ≥0 . Otherwise , the solution would be either ( 1 ) negative or ( 2 ) non-optimal , since we can decrease J^ ( 0 ) and increase |J1| . For M=1 , we immediately get that , in the optimal solution , 2J^1=J^ ( 0 ) =2/3 ( ϕm does not matter ) . For M=2 , 3 and 4 a solution is harder to find directly , so we performed a parameter grid search over all the free parameters ( km , J^m and ϕm ) in those components . We found that the optimal solution ( which maximizes the objective |J^ ( k† ) |2 ) , had the following form ( 33 ) J ( x ) =∑m=−MMJ^ ( mk† ) cos ( mk† ( x−x0 ) ) , where x0 is a free parameter . This form results from a parameter grid search for M=1 , 2 , 3 and 4 , under the assumption that L→∞ and r^ ( k ) is highly peaked . However , our numerical results in the general case ( Figure 16A ) , using the FISTA algorithm , indicate that the locally optimal solution does not change much even if L is finite , and r^ ( k ) is not highly peaked . Specifically , it has a similar form ( 34 ) J ( x ) =∑m=−∞∞J^ ( mk* ) cos ( mk* ( x−x0 ) ) . Since J^ ( mk* ) is rapidly decaying ( Figure 16A ) , effectively only the first few components are non-negligible , as in Equation 33 . This can also be seen in the value of the objective obtained in the parameter scan ( 35 ) M1234|J^ ( k† ) |2160 . 23670 . 24570 . 2457 , where the contribution of additional high frequency components to the objective quickly becomes negligible . In fact , the value of the objective cannot increase above 0 . 25 , as we explain in the next section . And so , the main difference between Equations 33 and 34 is the base frequency , k* , which is slightly different from k† . As explained in the appendix , the relation between k* and k† depends on the k-lattice discretization , as well as on the properties of r^ ( k ) . 10 . 7554/eLife . 10094 . 021Figure 16 . Fourier components of Non-negative PCA on the k-lattice . ( A ) 1D solution ( blue ) includes: a DC component ( k=0 ) , a maximal component with magnitude near k† ( red line ) , and weaker harmonics of the maximal component . ( B ) 2D solution includes: a DC component ( k = ( 0 , 0 ) ) , a hexgaon of strong components with radius near k† ( red circle ) , and weaker components on the lattice of the strong components . White dots show underlying k-lattice . We used a difference of Gaussians tuning curve , with parameters 2σ1=σ2=7 . 5 , L=100 , and the FISTA algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 021 The 1D properties , described in the previous section , generalize to the 2D case in the following manner: k∈S^ | k=∑i=1Bnik* ( i ) ( n1 , . . . , nB ) ∈ℤB Interestingly , given these properties of the solution we already get hexagonal patterns , as we explain next . Similarly to the 1D case , the difference between ‖k* ( i ) ‖ and k† is affected by lattice discretization , and the curvature of r^ ( k ) near k† . To simplify matters , we focus first on the simple case that L→∞ and r^ ( k ) is sharply peaked around k† . Therefore , the Fourier objective becomes ∑i=1B|J^ ( k* ( i ) ) |2 , so the only Fourier components that appear in the objective are {k* ( i ) }i=1B , which have radius k† . We examine the values this objective can have . All the base components have the same radius . This implies , according to the Crystallographic restriction theorem in 2D , that the only allowed lattice angles ( in the range between 0 and 90 degrees ) are 0 , 60 and 90 degrees . Therefore , there are only three possible lattice types in 2D . Next , we examine the value of the objective for each of these lattice types: 1 ) Square lattice , in which k* ( 1 ) =k† ( 1 , 0 ) , k* ( 2 ) =k† ( 0 , 1 ) , up to a rotation . In this case , J ( x , y ) =∑mx=−∞∞∑my=−∞∞J^mx , mycos ( k† ( mxx+myy ) +ϕmx , my ) and the value of the objective is bounded above by 0 . 25 ( see proof in appendix ) . 2 ) 1D lattice , in which k* ( 1 ) =k† ( 1 , 0 ) , up to a rotation . This is a special case of the square lattice , with a subset of J^mx , myequal to zero , so we can write , as we did in the 1D caseJ ( x ) =∑m=−∞∞J^mcos ( k†mx+ϕm ) Therefore , the same objective upper bound , 0 . 25 , holds . Note that some of the solutions we found numerically are close to this bound ( Equation 35 ) . 3 ) Hexagonal lattice , in which the base components arek* ( 1 ) =k† ( 1 , 0 ) , k* ( 2 ) =k† ( −12 , 32 ) , k* ( 3 ) =k† ( −12 , −32 ) up to a rotation by some angle α . Our parameter scans indicate that the objective value cannot surpass 0 . 2 in any solution composed of only the base hexgonal components {k* ( m ) }m=13 and a DC component . However , taking into account also some higher order lattice components , we can find a better solution , with an objective value of 0 . 2558 . Though this is not necessarily the optimal solution , it surpasses any possible solutions on the other lattice types ( bounded below 0 . 25 , as we proved in the appendix ) . Specifically , this solution is composed of the base vectors {k* ( m ) }m=13 and their harmonicsJ ( x ) =J^0+2∑m=18J^mcos ( k* ( m ) ⋅x ) with k* ( 4 ) =2k* ( 1 ) , k* ( 5 ) =2k* ( 2 ) , k* ( 6 ) =2k* ( 3 ) , k* ( 7 ) =k* ( 1 ) +k* ( 2 ) , k* ( 8 ) =k* ( 1 ) +k* ( 3 ) . Also , J^0=0 . 6449 , J^1=J^2=J^3=0 . 292 , J^4=J^5=J^6=−0 . 0101 and J^7=J^8=−0 . 134 . Thus , any optimal solution must be on the hexagonal lattice , given our approximations . In practice , the lattice hexagonal basis vectors do not have exactly the same radius , and , as in the 1D case , this radius is somewhat smaller then k† , due to the lattice discretization , and due to that r^ ( k ) is not sharply peaked . However , the resulting solution lattice is still approximately hexagonal in k-space . For example , this can be seen in the numerically obtained solution in Figure 16B – where the strongest non-DC Fourier components form an approximate hexagon near k† , from the Fourier components A , defined in Figure 17 . 10 . 7554/eLife . 10094 . 022Figure 17 . The modules in Fourier space . As in Figure 15C , we see a contour plot of the 2D tuning curve Fourier transform r^ ( k ) and the k-space the peak of r^ ( k ) ( red circle ) , and the k-lattice S^ ( black circles ) . The lattice points can be divided into approximately hexgonal shaped groups . Several such groups are marked in blue on the lattice . For example , group A and B are optimal since they are nearest to the red circle . The next best ( with the highest-valued contours ) group of points , which have an approximate hexgonal shape , is C . Note that group C has a k-radius of approximately the optimal radius times 2 ( cyan circle ) . Parameters: 2σ1=σ2=7 . 5 , L=100 . DOI: http://dx . doi . org/10 . 7554/eLife . 10094 . 022 | Long before the invention of GPS systems , ships used a technique called dead reckoning to navigate at sea . By tracking the ship’s speed and direction of movement away from a starting point , the crew could estimate their position at any given time . Many believe that some animals , including rats and humans , can use a similar process to navigate in the absence of external landmarks . This process is referred to as “path integration” . It is commonly believed that the brain’s navigation system is based on such path integration in two key regions: the entorhinal cortex and the hippocampus . Most models of navigation assume that a network of grid cells in the entorhinal cortex processes information about an animal’s speed and direction of movement . The grid cell network estimates the animal’s future position and relays this information to cells in the hippocampus called place cells . Individual place cells then fire whenever the animal reaches a specific location . However , recent work has shown that information also flows from place cells back to grid cells . Further experiments have suggested that place cells develop before grid cells . Also , inactivating place cells eliminates the hexagonal patterns that normally appear in the activity of the grid cells . Using a computational model , Dordek , Soudry et al . now show that place cell activity could in principle trigger the formation of the grid cell network , rather than vice versa . This is achieved using a process that resembles a common statistical algorithm called principal component analysis ( PCA ) . However , this only works if place cells only excite grid cells and never inhibit their activity , similar to what is known from the anatomy of these brain regions . Under these circumstances , the model shows hexagonal patterns emerging in the activity of the grid cells , with similar properties to those patterns observed experimentally . These results suggest that navigation may not depend solely on grid cells processing information about speed and direction of movement , as assumed by path integration models . Instead grid cells may rely on position-based input from place cells . The next step is to create a single model that combines the flow of information from place cells to grid cells and vice versa . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Extracting grid cell characteristics from place cell inputs using non-negative principal component analysis |
A presynaptic adhesion G-protein-coupled receptor , latrophilin-1 , and a postsynaptic transmembrane protein , Lasso/teneurin-2 , are implicated in trans-synaptic interaction that contributes to synapse formation . Surprisingly , during neuronal development , a substantial proportion of Lasso is released into the intercellular space by regulated proteolysis , potentially precluding its function in synaptogenesis . We found that released Lasso binds to cell-surface latrophilin-1 on axonal growth cones . Using microfluidic devices to create stable gradients of soluble Lasso , we show that it induces axonal attraction , without increasing neurite outgrowth . Using latrophilin-1 knockout in mice , we demonstrate that latrophilin-1 is required for this effect . After binding latrophilin-1 , Lasso causes downstream signaling , which leads to an increase in cytosolic calcium and enhanced exocytosis , processes that are known to mediate growth cone steering . These findings reveal a novel mechanism of axonal pathfinding , whereby latrophilin-1 and Lasso mediate both short-range interaction that supports synaptogenesis , and long-range signaling that induces axonal attraction .
Correct wiring of the nervous system critically depends on both long-range diffusible cues and short-range contact-mediated factors which can be attractive or repulsive ( Chen and Cheng , 2009 ) . However , the relatively small repertoire of key molecules known to be involved in axon guidance or trans-synaptic adhesion cannot fully explain the complexity and specificity of synaptic connections . Indeed , new interacting partners and signal-modulating ligands are now being found for many well-established guidance factors ( Karaulanov et al . , 2009; Leyva-Díaz et al . , 2014; Söllner and Wright , 2009 ) . Furthermore , our novel findings demonstrate that at least one receptor pair can both mediate cell contacts and , unexpectedly , also act as a long-range signaling factor and its receptor . This trans-synaptic receptor pair consists of presynaptic latrophilin-1 ( LPHN1 ) and postsynaptic Lasso ( Silva et al . , 2011 ) . LPHN1 ( also known as ADGRL1 for Adhesion G-protein-coupled Receptor , Latrophilin subfamily 1 [Hamann et al . , 2015] ) is a cell-surface receptor that is expressed by all central neurons ( Davletov et al . , 1998; Ichtchenko et al . , 1999; Matsushita et al . , 1999; Sugita et al . , 1998 ) . An array of data indicates that LPHN1 is localized on axons , axonal growth cones and nerve terminals ( Silva et al . , 2011 ) . Activation of LPHN1 by its agonist , mutant latrotoxin ( LTXN4C ) , stimulates vesicular exocytosis ( Ashton et al . , 2001; Lajus et al . , 2006; Lelyanova et al . , 2009; Silva et al . , 2009; Tobaben et al . , 2002; Volynski et al . , 2003; Deák et al . , 2009 ) . LPHN1 knockout ( KO ) in mice leads to abnormal rates of embryonic lethality and psychotic phenotypes ( Tobaben et al . , 2002 ) , indicating the importance of LPHN1 in early development and in cognitive functions in adulthood . The second member of this receptor pair , Lasso , is a representative of teneurins ( TENs ) , large single-pass transmembrane proteins ( Baumgartner et al . , 1994; Levine et al . , 1994 ) . Lasso is the splice variant of TEN2 ( TEN2-SS ) ( Figure 1A ) that specifically binds LPHN1 in cell adhesion experiments ( Li et al . , 2018 ) . Given also that only Lasso is isolated by affinity chromatography on LPHN1 ( Silva et al . , 2011 ) , we will refer here to TEN2 that is able to bind LPHN1 as Lasso . All TENs possess a large C-terminal extracellular domain ( ECD ) containing a series of epidermal growth factor ( EGF ) -like repeats and other repeat domains ( Figure 1A ) . Inter-chain disulfide bridges mediate TEN homodimerization ( Figure 1B , left ) ( Feng et al . , 2002; Vysokov et al . , 2016 ) . Similar to Notch , during the intracellular processing of TENs , their ECDs are constitutively cleaved by furin at site 1 ( Figure 1A , B , left ) ( Rubin et al . , 1999; Tucker and Chiquet-Ehrismann , 2006; Vysokov et al . , 2016 ) . However , the cleaved ECD remains tightly tethered to the cell surface due to its strong interaction with the transmembrane fragment ( Figure 1B , middle ) ( Vysokov et al . , 2016 ) . TENs have been implicated in promoting axon guidance and neurite outgrowth ( Minet et al . , 1999; Rubin et al . , 1999; Antinucci et al . , 2013; Leamey et al . , 2007; Young et al . , 2013; Hor et al . , 2015 ) . For example , different TENs can mediate neuronal cell adhesion ( Boucard et al . , 2014; Rubin et al . , 2002; Silva et al . , 2011 ) . TEN2 and TEN4 , which are present on dendritic growth cones and developing filopodia , may be responsible for dendritic spine formation ( Rubin et al . , 1999; Suzuki et al . , 2014 ) , while substrate-attached TEN1 supports neurite growth ( Minet et al . , 1999 ) . However , a mechanistic insight into the role of TENs in axonal growth is still lacking . One possibility is that TENs , as bona fide cell-surface receptors , could bind other cell-surface molecules and thus mediate axonal pathfinding . TENs can form homophilic complexes ( Rubin et al . , 2002; Beckmann et al . , 2013 ) . However , TENs failed to mediate homophilic cell adhesion in direct experiments ( Boucard et al . , 2014; Li et al . , 2018 ) . In addition , homophilic interactions of a recombinant soluble TEN2 ECD with the cell-surface TEN2 inhibited ( rather than promoted ) neurite outgrowth ( Beckmann et al . , 2013; Young et al . , 2013 ) . By contrast , heterophilic interactions of TENs can promote synapse formation ( Mosca et al . , 2012; Silva et al . , 2011 ) . More specifically , heterophilic interaction between Lasso and LPHN1 , its strongest ligand ( Silva et al . , 2011; Boucard et al . , 2014 ) , consistently mediates cell adhesion ( Silva et al . , 2011; Boucard et al . , 2014; Li et al . , 2018 ) and is thought to facilitate synapse formation ( Silva et al . , 2011 ) . However , our surprising finding ( Vysokov et al . , 2016 ) that Lasso/TEN2 is partially released from the cell surface by regulated proteolysis ( at site 3; Figure 1B , right ) was inconsistent with a solely cell-surface function of Lasso . On the other hand , we found that the released Lasso fragment retained its ability to bind cell-surface LPHN1 with high affinity and induce intracellular signaling ( Silva et al . , 2011; Vysokov et al . , 2016 ) . Thus , it was possible that the released , soluble ECD of Lasso/TEN2 could act as a diffusible ( attractive or repulsive ) factor and mediate some of the TEN2 functions in neurite pathfinding described above . Therefore , we hypothesized that the binding of soluble Lasso to LPHN1 on distant neurites could trigger important changes in their growth . Here , we test this hypothesis using cultured hippocampal neurons . First , we show that developing neurons release a substantial proportion of Lasso ECD into the medium , while LPHN1 is concentrated on the leading edge of axonal growth cones . We then use microfluidic chambers to demonstrate that a spatio-temporal gradient of soluble Lasso attracts neuronal axons , but not dendrites , and that this process involves LPHN1 that is present on axonal growth cones . Using model cells expressing functional LPHN1 , and mouse neuromuscular preparations , we also show that LPHN1 activation by soluble Lasso causes intracellular Ca2+ signaling , which leads to increased exocytosis . This suggests a plausible cellular mechanism causing axons to turn in the direction of a gradient of soluble Lasso . Moreover , the LPHN1-Lasso pair illustrates a novel principle of chemical guidance whereby cell-surface receptors engage not only in short-range interactions , but also in long-range signaling , which can further contribute to the formation of complex neuronal networks .
We previously showed in model cell lines and in adult brain that Lasso is cleaved at several sites ( sites 1 , 2 , three in Figure 1A , B ) and is released into the extracellular environment in a regulated manner ( Vysokov et al . , 2016 ) . To test whether Lasso undergoes the same processing and release during neuronal development , we followed Lasso expression at different stages of neuron maturation in hippocampal cell cultures ( Kaech and Banker , 2006 ) . Soon after plating , embryonic ( E18 ) rat hippocampal neurons produced Lasso , which was detectable at 3 days in vitro ( DIV ) ( Figure 1C , D ) . A large proportion of Lasso ( ~90% ) was constitutively cleaved at site 1 during neuronal development in vitro ( Figure 1—figure supplement 1A ) . Increasing amounts of cleaved fragment also appeared in the medium at 7 and 14 DIV ( Figure 1D and Figure 1—figure supplement 1A , green ) , indicating a slow cleavage at site 3 . Thus , Lasso is fully cleaved at site 1 and partially released by regulated cleavage at site 3 not only in transfected immortalized cells , but also in developing neurons and in the postnatal rat brain ( Vysokov et al . , 2016 ) . We also examined the neuronal structures that could release soluble Lasso ECD . We found that large amounts of Lasso were present on dendrites and dendritic growth cones ( Figure 1—figure supplement 1B ) , while it was practically absent from axons and axonal growth cones ( Figure 1E ) . Since about 80% of Lasso was not normally released ( Figure 1D , Figure 1—figure supplement 1A ) , these data suggested that the compartments rich in Lasso , that is dendrites and dendritic growth cones , were the main source of the soluble Lasso fragment . As early as 3 DIV , the developing neurons also expressed LPHN1 , the high-affinity receptor for soluble Lasso ECD , and the amounts of LPHN1 continued to increase through all time points ( Figure 1B ) , in parallel with the increasing amounts of soluble Lasso ( Figure 1—figure supplement 1A ) . This correlation between the soluble Lasso and cell-surface LPHN1 further supported the idea of their likely interaction during neuronal development . Interestingly , in developing hippocampal neurons , LPHN1 was found concentrated in axons and especially in axonal growth cones , where it co-localized with synapsin ( Figure 1—figure supplement 1C , D , arrowheads ) . LPHN1 was also enriched in axonal varicosities , which were identified as en passant synapses by immunostaining for PSD-95 ( Figure 1—figure supplement 1D , asterisks ) . We then studied the expression of LPHN1 in growth cones in more detail by transfecting hippocampal neurons with GFP , which greatly simplified the identification and tracking of axons and axonal growth cones . All GFP-labeled axonal growth cones showed a clear enrichment of endogenous LPHN1 ( Figure 1F , G , I ) . Conversely , when LPHN1 expression was knocked down by shRNA ( delivered together with GFP in the same bicistronic vector ) , it clearly disappeared from the growth cones of transfected neurons , while the growth cones of non-transfected cells were not affected ( Figure 1—figure supplement 1E , arrow and arrowhead , respectively ) . We also discovered that endogenous LPHN1 expression within axonal growth cones was polarized in relation to the cone’s symmetry axis , such that one side of each growth cone contained on average 1 . 88 ± 0 . 22 fold more LPHN1 than the other ( Figure 1G , H ) . To assess whether this LPHN1 enrichment correlated with the direction of axonal growth , we traced the growth trajectories of a number of symmetrical growth cones and compared these with the distribution of LPHN1 . This analysis clearly demonstrated that LPHN1 polarization within the growth cones very strongly positively correlated with the direction of their turning ( Figure 1G , H ) . Moreover , in non-symmetrical growth cones , which had clearly started turning prior to fixation , LPHN1 expression had a bimodal distribution , being enriched not only near the ‘neck’ of a turning cone , but also close to its leading edge ( Figure 1—figure supplement 1F , G ) . Such leading-edge enrichment also extended into fine growth cone protrusions . Thus , filopodia and lamellipodia located on the leading edge of a growth cone ( Figure 1I , left , arrowheads ) showed a much higher amount of LPHN1 than the processes on the trailing edge of the growth cone ( Figure 1I , right ) . We concluded that LPHN1 expression within growth cones correlated positively with the global directionality of growth and with the fine structures that underpin the growth cone’s extension . Next , we tested the interaction between soluble Lasso and cell-surface LPHN1 . For these tests we expressed a shorter , constitutively secreted construct , Lasso-D ( Figure 2A , right ) in HEK293A cells and affinity-purified it ( Figure 2B ) . 100 nM Lasso-D was incubated with neuroblastoma cells stably expressing ( i ) LPHN1 , ( ii ) a chimeric construct LPH-82 containing ECD from EMR-2 used as a negative control , ( iii ) Lasso-A , or ( iv ) Lasso-FS ( Figure 2A , left ) . As expected , Lasso-D did not interact with LPH-82 ( Figure 2C , panel 4 ) . The lack of Lasso-D binding to Lasso-A and released fragment of Lasso-A binding to Lasso-FS ( Figure 2D , panels 2 , 3; Figure 2—figure supplement 1B ) was somewhat surprising , since homophilic interactions between membrane-bound and soluble TENs were reported previously ( Bagutti et al . , 2003; Beckmann et al . , 2013; Hong et al . , 2012; Rubin et al . , 2002; Boucard et al . , 2014 ) , but this could be due to a relatively low affinity of Lasso-Lasso interaction and relatively long washes employed in our protocol . On the other hand , and consistent with previous reports of high affinity between LPH1 and Lasso ( Silva et al . , 2011; Boucard et al . , 2014 ) , Lasso-D and the released fragment of Lasso-A bound strongly to cells expressing LPHN1 ( Figure 2C , panels 2 , three and Figure 2—figure supplement 1A ) . To verify that the soluble ECD of Lasso , when proteolytically released from the cell-surface as depicted in Figure 2A ( Lasso-A ) , could diffuse between individual cells and bind LPHN1 on distant cells , we co-cultured neuroblastoma cells stably expressing Lasso-A with cells stably expressing LPHN1 . When co-cultured at high density , these cells formed clusters , held together by LPHN1/Lasso-A intercellular adhesion complexes ( Figure 2E , panel 1 ) . In more sparsely plated co-cultures , the Lasso-A fragment was released into the medium , where it diffused and bound to cells expressing LPHN1 , but not to the wild type ( WT ) neuroblastoma cells ( Figure 2E , panel 2 , and Figure 2—figure supplement 1C ) . Interestingly , after binding Lasso , the LPHN1 staining appeared to concentrate in large patches , a pattern very different from LPHN1 distribution in control conditions ( Figure 2C , panel 1 ) ( see also below ) . These experiments suggest that ( i ) when Lasso is released into the medium as a result of its regulated cleavage , it retains its affinity for LPHN1 and ( ii ) on reaching distant LPHN1-expressing cells by diffusion , Lasso causes LPHN1 redistribution on the cell surface . We then asked whether the soluble Lasso ECD could similarly bind to LPHN1 in neurons and , more specifically , on axonal growth cones . To control for the specificity of Lasso binding to LPHN1 , this experiment was carried out on cultured hippocampal neurons from LPHN1 WT ( Adgrl1+/+ ) and LPHN1 KO ( Adgrl1-/- ) newborn mice ( P0 ) . Also , to unequivocally distinguish between the soluble and cell-surface Lasso , we used exogenous Lasso-D , which was detected using anti-FLAG antibody . As expected , in WT mouse neurons , LPHN1 was found mostly in axonal growth cones ( arrowheads ) and varicosities ( asterisks ) ( Figure 2—figure supplement 2A , green ) . The exogenous Lasso-D clearly bound to these structures ( Figure 2—figure supplement 2A , red; C ) , but in general did not interact with dendrites . By contrast , the axons and growth cones of LPHN1 KO neurons did not show specific LPHN1 staining and appeared unable to bind the soluble exogenous Lasso-D ( Figure 2—figure supplement 2B , C ) . These results indicated that released Lasso ECD could interact with LPHN1 on axonal growth cones . Based on the data above , we hypothesized that the interaction of released Lasso ECD with LPHN1 on axonal growth cones could represent one of the mechanisms that underlie the previously formulated , but so far unexplained , role of TENs in axonal pathfinding and brain patterning ( Antinucci et al . , 2013; Hor et al . , 2015; Leamey et al . , 2007; Young et al . , 2013 ) . To study this effect , we developed a new method of long-term exposure of hippocampal axons to stable gradients of Lasso using ‘microfluidic axon isolation devices’ ( MAIDs ) ( Figure 3A ) . The advantage of this method over conventional ligand-puffing was that the MAIDs enabled exposure of axons to long-term stable gradients of Lasso , which was critical for our assay . The device used here had two compartments , each consisting of two cylindrical wells connected by a ‘corridor’; a 150 μm-thick wall that separated the two corridors had multiple parallel microchannels ( 2–3 μm tall and 10 μm wide ) connecting the two compartments ( Figure 3A , middle ) . When neurons are plated in one of the compartments ( designated as the Somal Compartment ) , their neurites grow in all directions , but only the axons ( identified by NF-H staining ) readily penetrate the microchannels and cross into the empty , Axonal Compartment ( Figure 3A , right; 3B , C ) . While there is a large number of dendrites in the Somal Compartment ( identified by microtubule-associated protein 2 , MAP-2 , staining ) , only a few of them enter the Axonal Compartment and then terminate close to the wall ( Figure 3B , C ) . From the previously described physical characteristic of microfluidic chambers ( Zicha et al . , 1991 ) , we predicted that a concentration gradient across the microchannels in our devices could be established over time . This was modelled by adding TRITC-conjugated BSA to one compartment and visualizing the dye in the microchannels ( Figure 3D ) . We found that a gradient was formed within the first 24 hr and remained stable over several days ( Figure 3D , E ) . To test the functionality of the MAIDs for studying axonal guidance , we employed brain-derived neurotrophic factor ( BDNF ) known to act as an axonal chemoattractant ( Li et al . , 2005 ) . Rat hippocampal neurons were plated into the Somal Compartment , and at 3 DIV , when axons normally start entering microchannels , BDNF was added to the Axonal Compartment ( PBS was added to control cultures ) ( Figure 3F ) . After a further 5 DIV , we observed a 2 . 2-fold higher number of axons crossing into the Axonal Compartment in the presence of BDNF compared with the control ( Figure 3G , H ) . This effect was statistically significant ( Figure 3H ) . This proof-of-concept experiment confirmed that MAIDs could be used to study the long-term effects of chemoattractant gradients on axonal migration . We then used this methodology to study the reaction of LPHN1-expressing neuronal growth cones to a gradient of soluble released Lasso . Lasso-D was added to the Axonal Compartment ( Figure 4A ) , and the integrity of Lasso during the experiment was verified by Western blotting ( Figure 4B ) . Quantification of axons in Axonal Compartments by NF-H immunofluorescence ( Figure 4C , D ) revealed a statistically significant 1 . 5-fold increase in axonal growth induced by Lasso-D . Thus , soluble Lasso-D clearly functioned as an attractant of axonal elongation and/or steering . Since LPHN1 is present on axonal growth cones ( Figure 1 , Figure 1—figure supplement 1 ) , binds soluble Lasso ( Figure 2 , Figure 2—figure supplement 1 ) and is the strongest interacting partner of Lasso ( Boucard et al . , 2014; Silva et al . , 2011 ) , we hypothesized that LPHN1 may be involved in the observed Lasso-mediated attraction of axons ( Figure 4—figure supplement 1A ) . To investigate this , hippocampal cultures from LPHN1 KO or WT mice ( genotyping shown in Figure 4—figure supplement 1B ) were exposed to a gradient of Lasso-D added to the Axonal Compartment . The total amounts of neurites and cells in both compartments were quantified using the lipophilic membrane tracer DiO ( see Materials and methods for details ) . The results clearly demonstrated that the neurites from LPHN1-expressing ( WT ) hippocampal neurons crossed into the Lasso D-containing Axonal Compartment 5 . 5-fold more readily than the neurites from neurons lacking this receptor ( Figure 4E , left ) . Importantly , this effect was not due to a lower viability of LPHN1 KO neurons , because there was no difference between the KO and WT cells within the Somal Compartment ( Figure 4E , right ) . We also studied the behavior of axons in response to a spatio-temporal Lasso gradient in the corridor of the Axonal Compartment , by exposing axons to an increasing concentration of the attractant during the whole growth process . In order to achieve a stable increase in protein concentration over time , we seeded HEK293A cells stably expressing soluble Lasso-D ( untransfected HEK293A cells were used in control ) into the wells of the Axonal Compartment ( Figure 5A ) . The presence of secreted Lasso-D within the Axonal Compartments was verified at the end of each experiment ( Figure 5B ) , and the distribution of axons was quantified by NF-H immunofluorescence ( Figure 5C , D ) . In this experiment , we observed not only a significantly greater number of axons being attracted , but also axons growing deeper into the corridors of the Axonal Compartments ( Figure 5D ) . On the other hand , quantification of MAP-2 immunofluorescence demonstrated that released Lasso-D did not attract dendrites; in fact , there was a slight repulsive effect ( Figure 5E ) . Taken together , these experiments indicate that a gradient of the soluble Lasso fragment specifically induces axonal attraction . Soluble Lasso fragment also induced strong axonal fasciculation ( e . g . Figures 4C and 5C ) . This effect was quantified by measuring the width of axonal bundles at 100 μm from the separating wall , where axons grew mostly away from the wall rather than along it . Based on the average width of a single axon ( 1 µm ) , an average bundle contained 2–3 axons in control conditions , but more than five axons in the presence of 1 . 5 nM Lasso-D ( Figure 5F ) . Thus , Lasso fragment can induce axonal fasciculation in a concentration-dependent manner . In order to rule out the possibility that the observed effects of the released Lasso fragment were due to a general positive trophic effect ( e . g . an increase in axonal elongation speed ) , Lasso-D was added directly to cultures of hippocampal neurons . To visualize axons , neurons were transfected with GFP prior to plating and allowed to grow for 4 DIV , after which the longest neurites of GFP-positive neurons were traced and measured . We did not detect any increase in the length of neurites when neurons were exposed to Lasso-D ( Figure 5G , H ) . Taken together , these data demonstrate unequivocally that a gradient of the soluble fragment of Lasso acts as an axonal attraction cue without affecting their overall growth . To determine the downstream effects of the interaction between soluble Lasso ECD and LPHN1 , we used neuroblastoma cells stably expressing LPHN1 . It was reported previously that the signaling machinery downstream of LPHN1 in these cells is similar to that in neurons ( Silva et al . , 2009; Volynski et al . , 2004 ) . When the LPHN1-expressing neuroblastoma cells are stimulated by the known LPHN1 ligand and potent secretagogue LTXN4C , the N-terminal and C-terminal fragments ( NTF and CTF ) of LPHN1 undergo rearrangement ( as illustrated in Figure 6A , middle ) . In turn , this induces intracellular Ca2+ signaling which involves the activation of Gαq and phospholipase C ( PLC ) , and release of inositol 1 , 4 , 5-trisphosphate ( IP3 ) ( Silva et al . , 2009; Volynski et al . , 2004 ) . These observations suggested that Lasso might also affect the distribution of NTF and CTF of LPHN1 in the plasma membrane . Indeed , we noticed that soluble Lasso-D or Lasso-A caused the NTF to aggregate into patches on the surface ( Figure 2C , panel 2; Figure 2—figure supplement 1C ) . To test whether Lasso also causes a redistribution of the CTF required for intracellular signaling , we applied Lasso-D to LPHN1-expressing cells and followed the fate of both NTF and CTF . We observed a dramatic rearrangement of both LPHN1 fragments in the membrane , leading to the formation of large molecular aggregates also containing Lasso ( Figure 6C ) . Similar clustering of both LPHN1 fragments was also induced by LTXN4C , a strong LPHN1 agonist ( Figure 6D ) . On the other hand , an antibody recognizing the V5 epitope at the N-terminus of NTF only caused NTF clustering , but did not affect the distribution of CTF ( Figure 6A , right; Figure 6E ) . Thus , soluble Lasso ECD , which causes the association of the LPHN1 fragments , might be a functional agonist of LPHN1 , similar to LTXN4C . By analogy , this also indicated that the soluble Lasso fragment could induce signal transduction via the CTF of LPHN1 coupled to a G-protein . The effect of LTXN4C can be assessed by monitoring cytosolic Ca2+ ( Silva et al . , 2011; Volynski et al . , 2004 ) . We therefore investigated whether the soluble Lasso ECD could induce similar effects . LPH1-expressing neuroblastoma cells were stimulated with saturating concentrations of Lasso-D , LTXN4C ( positive control ) or buffer ( negative control ) , while cytosolic calcium levels were monitored using an intracellular Ca2+-sensing dye , Fluo-4 ( see Figure 7—figure supplement 1A for the scheme of experiment ) . Similar to LTXN4C , in the absence of extracellular Ca2+ , Lasso-D did not cause any Ca2+ signals in LPHN1-expressing NB2a cells ( Figure 7A ) . However , when extracellular Ca2+ was added to the cells , the rise in intracellular Ca2+ signal was significantly higher in the presence of the ECD of Lasso , compared to negative control ( Figure 7A ) . Thus , Lasso-D is able to cause intracellular Ca2+ signaling in LPHN1-expressing cells . One of the features of LTXN4C-induced effects ( such as Ca2+ signaling and neurotransmitter release ) is that they develop with a delay of ~20 min , which has been attributed to the time taken by the toxin to assemble the LPHN1 fragments together and cause its maximal activation ( Volynski et al . , 2004 ) . We predicted , therefore , that the rearrangement of the NTF and CTF induced by soluble Lasso ( Figure 6C ) should prepare the signaling machinery for stimulation by the toxin . To test this idea , we first treated the LPHN1-expressing cells with Lasso-D and then with LTXN4C ( Figure 7—figure supplement 1B ) . When Lasso-D was applied in the presence of 2 mM Ca2+ , it induced relatively short-lived intracellular Ca2+ signaling ( Figure 7B , right , prior to the blue arrowhead ) . However , when LTXN4C was then added , it triggered Ca2+ signaling after a shorter delay ( ~14 min ) , instead of the usual ~23 min ( Figure 7C ) . This additivity of effects is consistent with soluble Lasso inducing intracellular Ca2+ signaling via the same molecular mechanism as LTXN4C . Another well-known effect of LTXN4C is the burst-like release of neurotransmitters , linked to the elevated levels of cytosolic Ca2+ ( Lelyanova et al . , 2009; Volynski et al . , 2003 ) . As Lasso-D likewise increased intracellular Ca2+ concentration , it might also trigger such transmitter exocytosis . To test this hypothesis , we applied a previously characterized ( Silva et al . , 2011 ) , soluble , short C-terminal Lasso construct ( Lasso-G , Figure 1A ) to mouse neuromuscular preparations and recorded the spontaneous miniature end plate potentials ( MEPPs ) , which correspond to individual exocytotic events . We found that incubation with Lasso-G significantly increased MEPPs frequency from 1 . 61 ± 0 . 27 Hz in control to 3 . 83 ± 0 . 79 Hz in the presence of Lasso-G ( Figure 7D , E ) . However , this was much less than the effect of LTXN4C , which triggered massive secretion of neurotransmitter reaching 29 . 5 ± 4 . 1 Hz ( Figure 7F ) . To ascertain that both these effects were mediated by LPHN1 , we used neuromuscular preparations from LPHN1 KO mice . Interestingly , unstimulated LPHN1 KO motor neurons showed an increased MEPPs frequency compared to synapses from WT animals ( 3 . 33 ± 0 . 79 Hz in KO synapses ) . However , neither Lasso-G , nor LTXN4C had any effect on exocytosis in preparations lacking LPHN1 ( Figure 7E , F; 3 . 4 ± 0 . 68 Hz with Lasso-G and 3 . 8 ± 1 . 4 Hz with LTXN4C ) . In all the recordings , the mean amplitudes of MEPPs under any condition did not differ significantly ( Figure 7—figure supplement 1C ) , which indicated a purely presynaptic effect of the two LPHN1 agonists and of LPHN1 ablation . These results show that the soluble Lasso fragment can increase exocytosis at nerve terminals , and confirm the importance of LPHN1 in the observed effects of LTX and the ECD of Lasso . From the results reported here , we hypothesize that the soluble Lasso fragment , released by developing neurons , interacts with LPHN1 on axonal growth cones and nerve terminals . It then induces clustering of LPHN1 fragments and activation of downstream signaling , causing an increase in cytosolic Ca2+ and subsequent exocytosis . The latter two processes are known to be key regulators of axonal attraction ( Tojima et al . , 2011 ) . Thus , the ability of soluble Lasso to activate these processes on axonal growth cones could underpin the mechanisms by which it attracts axons .
This study provides evidence that Lasso ( a splice variant of TEN2 lacking a 7-residue insert in the β-propeller domain , TEN2-SS ) functions specifically as an attractant for axons expressing LPHN1 , and proposes a molecular mechanism for this effect . By using microfluidic devices to create long-term gradients of soluble proteins ( Figure 3 ) , we demonstrate that a gradient of soluble ECD of Lasso can act as an attractant for axons from hippocampal neurons ( Figures 4 and 5A–E ) . Importantly , growing hippocampal neurons in a medium containing a uniform concentration of Lasso had no effect on the length of their axons ( Figure 5G ) . This shows that Lasso plays an instructive role in the directionality , rather than the amount , of axonal growth . This is consistent with the effect of other axon attractants acting via similar mechanisms . For example , short-term exposure of axonal growth cones to gradients of BDNF stimulates IP3-induced Ca2+ release ( IICR ) that causes axonal attraction without an overall effect on neurite extension ( Li et al . , 2005 ) . One interesting observation from this project was the fasciculation of neurites in response to soluble Lasso/TEN2 ( Figure 5C , F ) . Fasciculation of axons is one of the major mechanisms of axonal navigation , for example in limb development ( Bastiani et al . , 1986 ) . While axonal fasciculation has not been previously linked to a soluble ECD of TEN , neurite bundling was actually observed in hippocampal cultures in response to TEN1 C-terminal peptide ( TCAP-1 ) ( Al Chawaf et al . , 2007 ) . Furthermore , knockdown of TEN1 in C . elegans resulted in de-fasciculation of the axons in the ventral nerve cord ( Drabikowski et al . , 2005 ) . Potential mechanisms of axonal bundling include actin reorganization induced by an LPHN1-mediated rise in cytosolic Ca2+ , other unknown interactions with cell adhesion molecules , or it could also be due to the divalent Lasso/TEN2 fragment crosslinking adjacent axons , thus promoting their parallel elongation . The soluble Lasso/TEN2 fragment could potentially have two membrane-anchored receptors: ( i ) TEN2 itself , as a homophilic ligand ( Bagutti et al . , 2003; Rubin et al . , 2002 ) , or ( ii ) LPHN1 , as a heterophilic ligand ( Boucard et al . , 2014; Silva et al . , 2011 ) . However , we have not observed TEN2 expression in growth cones of hippocampal axons ( Figure 1E ) , but found it to be abundant on dendrites ( Silva et al . , 2011 ) ( Figure 1E , Figure 1—figure supplement 1B ) . We also did not detect any appreciable binding of the released Lasso ECD to membrane-anchored Lasso ( Figure 2D , Figure 2—figure supplement 1B ) . In addition , homophilic interaction of Lasso/TEN2 actually has been reported to inhibit neurite outgrowth in neuroblastoma cells ( Beckmann et al . , 2013 ) , while we saw an opposite effect ( Figures 4 and 5 ) . Thus , the potential Lasso/TEN2 homophilic interaction could not explain the observed axonal attraction . On the other hand , we found strong expression of LPHN1 on the axonal growth cones of cultured hippocampal neurons ( Figure 1E–I , Figure 1—figure supplement 1C–F ) ( Silva et al . , 2011 ) . Importantly , the released soluble ECD of Lasso strongly bound to LPHN1 that was expressed on neuroblastoma cells or neuronal growth cones ( Figure 2 , Figure 2—figure supplements 1–2 ) . Furthermore , we found that deletion of LPHN1 precluded axonal attraction by Lasso ( Figure 4 ) , while it had no effect on neuronal cell bodies and dendrites in the Somal Compartment . These data strongly implicate LPHN1 in mediating Lasso-induced axon attraction . Our studies also reveal the likely mechanism that underlies the Lasso/LPHN1-induced axonal attraction . LPHN1 is a G-protein-coupled receptor ( GPCR ) that physically and functionally links to Gαq/11 ( Rahman et al . , 1999 ) . Activation of LPHN1 by its non-pore-forming agonist , LTXN4C , leads to aggregation of the NTF and CTF of LPHN1 ( Silva et al . , 2009; Volynski et al . , 2004 ) . This results in assembly of a functional GPCR , with subsequent activation of the downstream signaling cascade , which includes Gαq/11 , phospholipase C , production of IP3 and IP3-receptor-mediated release of Ca2+ from intracellular stores ( Capogna et al . , 2003; Lajus et al . , 2006; Volynski et al . , 2004 ) , thus inducing IICR . IICR is also regulated and enhanced by increased cAMP levels ( Tojima et al . , 2011 ) , and we previously demonstrated that activation of LPHN1 expressed in COS7 cells induces an increase in cAMP production ( Lelianova et al . , 1997 ) . In line with this , the recent study by Li et al . ( 2018 ) confirmed the ability of LPHN1 to regulate cAMP signaling . In that work ( Li et al . , 2018 ) , the cAMP signaling interference system was based on HEK293 cells expressing exogenous β2 adrenoceptor ( β2AR ) . Activation of β2AR by its agonist led to an increase in cAMP production , while a large excess of co-expressed LPHN1 interfered with β2AR signaling . This clearly suggests that LPHN1 uses the same cAMP signaling machinery as β2AR , and that when LPHN1 is not stimulated , it can titrate components of this machinery , decreasing their availability to β2AR . In agreement with the role of Lasso as a functional LPHN1 agonist , the binding of the released Lasso fragment to LPHN1 similarly causes the re-association of LPHN1 fragments ( Figure 6 ) and Ca2+ signaling ( Figure 7A–C ) . A rise in cytosolic Ca2+ concentration , in turn , can increase the rate of exocytosis , and we indeed observed enhanced acetylcholine release in mouse neuromuscular junctions in response to soluble Lasso ( Figure 7D–F ) . This response to Lasso was clearly mediated by LPHN1 , as it was not detected in neuromuscular preparations from LPHN1 KO mice ( Figure 7D–F ) . On the other hand , the effect of soluble Lasso on vesicular exocytosis was much weaker – and probably more physiological – than the massive effect of LTXN4C . In addition to Ca2+ regulation , Lasso binding to LPHN1 can induce cAMP signaling . Indirect evidence for this is provided by the cAMP signaling interference experiments mentioned above ( Li et al . , 2018 ) . When LPHN1 co-expressed with β2AR was stimulated for 24 hr with Lasso/TEN2 ( expressed on the same or opposite cells ) , this strongly decreased cAMP levels induced by β2AR activation . The most likely reason could be that , following an initial Lasso-induced LPHN1 activation , which normally subsides within 30 min ( Figure 7B ) , the continued LPHN1 stimulation led to massive heterologous receptor desensitization ( Kelly et al . , 2008 ) and inhibition of β2AR-mediated effect . Intriguingly , the effects of soluble Lasso resemble the well-known mechanism that underpins axonal attraction and consists of IP3 receptor-mediated local release of Ca2+ from intracellular stores , coupled with an increase in cAMP levels , that leads to increased exocytosis at the advancing edge of a growth cone ( Akiyama et al . , 2009; Qu et al . , 2002; Tojima et al . , 2011; Tojima and Kamiguchi , 2015 ) . Thus , when a gradient of soluble Lasso ECD approaches one side of an axonal growth cone , it may cause local activation of LPHN1 and its downstream signaling , ultimately leading to IICR . Local IICR in growth cones can induce an increase in vesicular exocytosis ( as observed in our experiments with Lasso-G , Figure 7 ) and the remodeling of actin filaments ( Tojima et al . , 2011 ) . The resulting augmented membrane delivery and actin-driven extension of filopodia at the edge facing a Lasso gradient would support the growth cone’s advance in this direction . Thus , based on all our data , we propose this chain of events ( summarized in Figure 8 ) as a likely mechanism for axonal attraction by soluble Lasso observed in this study . While TEN2 has been implicated in axon guidance in the visual pathway ( Young et al . , 2013 ) , here we report that it can also trigger axonal steering in developing hippocampal neurons , which is consistent with the strong expression of both Lasso/TEN2 and LPHN1 in the hippocampus ( Davletov et al . , 1998; Otaki and Firestein , 1999 ) . Furthermore , both proteins are expressed throughout the CNS , suggesting that this mechanism of soluble Lasso/LPHN1-mediated axonal attraction may apply widely across the brain , especially in such areas as the cortex , cerebellum , thalamus and spinal cord . Interestingly , the splice variant of TEN2 ( TEN2+SS ) , which contains the 7-amino acid insert in the β-propeller domain and cannot mediate cell adhesion via LPHN1 ( Li et al . , 2018 ) , might attract dendrites instead of axons , in contrast to Lasso ( TEN2-SS ) . Thus , in an artificial synapse formation experiment ( Li et al . , 2018 ) , HEK293 cells expressing TEN2+SS were seen covered by neurites from co-cultured hippocampal neurons that contained GABAA receptors . However , these processes did not show a proportionate accumulation of PSD-95 and thus probably represented en passant dendrites that were attracted to TEN2+SS cells , but unable to form mature inhibitory synapses with them . This could be a mechanism by which TEN2+SS could provide a substrate for the growth of dendrites searching for their ultimate target/s . Although the relative abundance of Lasso and TEN2+SS in the brain is unknown , these data suggest that various TEN isoforms could participate in distinct interactions , possibly with opposite results . High expression of LPHN1 and Lasso/TEN2 throughout the CNS , combined with their fundamental role in axon guidance , is consistent with lethal phenotypes observed in simpler organisms ( Langenhan et al . , 2009; Mosca et al . , 2012 ) . In knockout mice , however , the phenotype is less severe ( Tobaben et al . , 2002; Young et al . , 2013 ) ( Ushkaryov , to be published elsewhere ) suggesting that LPHN1 deletion is not completely penetrant , likely due to a compensatory effect of multiple LPHN and TEN homologs expressed in the mammalian brain . Indeed , LPHN1 can also weakly interact with TEN4 ( Boucard et al . , 2014 ) , and LPHN3 can interact with TEN1 ( O'Sullivan et al . , 2014 ) . Moreover , LPHN and TEN isoform expression patterns overlap ( Oohashi et al . , 1999; Sugita et al . , 1998; Zhou et al . , 2003 ) . This predisposition to compensation further raises the possibility that the mechanism of axonal guidance involving the interaction of soluble TEN2 with LPHN1 , described in this study , may occur between different members of the LPHN and TEN families . These observations provide evidence of further diversity of interactions and local specificity of developmental pathways for more accurate and plastic patterning of neural networks within the mammalian CNS .
All chemicals and reagents were purchased from Sigma-Aldrich , unless otherwise stated . Cell culture reagents were from PAA Laboratories or Thermo Fisher Scientific . Purified proteins: LTXN4C ( Volynski et al . , 2003 ) ; LTXN4C labeled with Alexa Fluor 647 ( Volynski et al . , 2004 ) ; Lasso-G ( Silva et al . , 2011 ) ; Lasso-D ( Silva et al . , 2011 ) were prepared in this laboratory; human BDNF was from R&D Systems ( 248-BD ) ; BSA-TRITC , from Thermo Fisher Scientific ( A23016 ) . The following antibodies were used in this work: Rabbit anti-NF-H ( Neuromics , RA22116 ) ; mouse anti-MAP-2 ( Neuromics , MO22116 ) ; mouse monoclonal anti-V5 ( clone SV5-Pk1 , AbD Serotec/Bio-Rad , MCA1360 ) ; rabbit anti-V5 ( Thermo Fisher Scientific , PA1-29324; RRID:AB_1961277 ) ; mouse monoclonal anti-myc ( clone 9E10 , Millipore , 05–419; RRID:AB_309725 ) ; chicken anti-myc ( Millipore , AB3252; RRID:AB_2235702 ) ; mouse anti-FLAG M2 ( Sigma-Aldrich , F3165; RRID:AB_259529 ) ; anti-FLAG M2 affinity gel ( Sigma-Aldrich , A2220 ) ; mouse anti-actinin ( Sigma-Aldrich , A7811 ) ; rabbit polyclonal anti-LPHN1-peptide ( PAL1 , ( Davydov et al . , 2009 ) ; rabbit polyclonal anti-LPHN1 NTF ( RL1 ) ( Davletov et al . , 1998 ) ; mouse anti-Lasso/TEN2 C-terminus ( TN2C , dmAb ) ( Silva et al . , 2011 ) ; sheep anti-TEN2 N-terminus ( TN2N , R and D systems , AF4578; RRID:AB_10719438 ) ; mouse anti-synapsin ( Santa-Cruz Biotechnology , sc-376623; RRID:AB_11150313 ) ; rabbit anti-PSD-95 ( Millipore , AB9708; RRID:AB_11212529 ) ; rabbit anti-Tau ( Synaptic Systems , 314 002; RRID:AB_993042 ) ; rabbit anti-GFP ( Thermo Fisher Scientific , A-11122; RRID: AB_221569 ) . The following cell lines were used: human embryonic kidney cells ( HEK293A , purchased from ECCC; RRID:CVCL_6910 ) ; mouse neuroblastoma cells ( NB2a , a kind gift from Dr . C . Isaac , Imperial College London; originally from ATCC and subsequently authenticated by ATCC using their proprietary methods . ; RRID:CVCL_0470 ) . Both cultures are mycoplasma-free , based on a mycoplasma test kit PlasmoTest ( Invivogen ) . A LPHN1 KO mouse ( strain AG148-2 , Adgrl1-/- ) was generated on the 129SvJ genetic background . Briefly ( details to be published elsewhere ) , the LPHN1 gene was isolated from a BAC clone containing a 36-kbp fragment of mouse genomic DNA . This was used to design a transfer vector for homologous recombination , containing a 13-kbp gene fragment of the LPHN1 gene , in which the intron between exons 1 and 2 was replaced with a neomycin gene/promoter cassette flanked by two loxP sequences . This insert disrupted the open reading frame in the mRNA transcribed from the resulting mutated LPHN1 gene . The transfer vector , carrying also a negative selection marker ( diphtheria toxin A-chain ) , was used to generate stably transfected 129Sv/J ES cell lines and chimeric mice , using standard transgenic techniques . Mice transmitting the inactivated LPHN1 gene through the germline were selected , inbred , back-crossed onto C57BL/6J background , and maintained at Charles River UK . LPHN1 gene disruption was confirmed by Southern blotting , PCR amplification using multiple primer pairs and Western blotting . The genotype of all animals used for breeding and tissue extraction was determined by PCR . All procedures ( breeding and Schedule 1 ) were approved by the University of Kent Animal Welfare Committee and performed in accordance with Home Office regulations and the European Convention for the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes . E18 hippocampi were obtained from rats ( BrainBits UK , Rhp ) . P0 hippocampi were prepared from P0 mice ( strains: C57BL/6J , Adgrl1+/+ , LPHN1 WT , or AG148/2 , Adgrl1-/- , LPHN1 KO ) . Flexor digitorum brevis muscle preparations were isolated from P21 male mice ( C57BL/6J or AG148/2 ) . The sequences of human Lasso ( Ten–2 ) mutants used in this study are available at GenBank: Lasso-FS ( JF784340 ) , Lasso-A ( JF784341 ) , Lasso-D ( JF784344 ) , GST-Lasso ( JF784347 ) . N- and C-terminally tagged rat LPHN1 ( termed also LPH-42 , MF966512 ) was described previously as V5-LPH-A ( Volynski et al . , 2004 ) . All cDNAs were subcloned into the pcDNA3 . 1 vector ( Thermo Fisher Scientific ) . A negative control plasmid , pLenti6 . 2-GW/EmGFP-miR ( Thermo Fisher Scientific , K4938-00 ) , was used for GFP expression , and the miRNA oligonucleotides listed below were cloned into this vector for LPHN1 knock-down experiments . Oligonucleotides for targeting LPHN1 mRNA were: LPHN1miR14T , ( TGCTGATAAAC AGAGCGCAGCACATAGTTTTGGCCACTGACTGACTATGTGCTGCTCTGTTTAT ) and LPHN1miR14B ( CCTGATAAACAGAGCAGCACATAGTCAGTCAGTGGCCAAAACTATGTGCT GCGCTCTGTTTATC ) . PCR primers for genotype analysis were: Neo Forward ( N255 , CGAGACTAGTGAGACGTGCTACTTCCATTTGTC ) ; LPHN1 Forward ( N425 , CTGACCCATA ACCTCCAAGATGATGTTTAC ) ; Neo/LPHN1 Reverse ( N424 , GATCTTGTCA TCTGTGCGCCCGTA ) . Human embryonic kidney ( HEK293A ) and rat neuroblastoma ( NB2a ) cell lines were cultured using standard techniques in DMEM with 10% heat-inactivated fetal bovine serum ( FBS , PAA Laboratories ) , at 5% CO2 and 37°C . Stable cell lines were generated using the Escort III transfection reagent and Geneticin selection ( Thermo Fisher Scientific ) . The positive cells were further enriched by fluorescence-assisted cell sorting ( FACSCalibur , BD Biosciences ) . All NB2a cell cultures contain proliferating , spindle-like cells and differentiated , neuron-like cells . We have not observed any difference in Lasso or LPHN1 expression between these two types of cell in stably transfected NB2a cultures . For increased expression of Lasso or LPH constructs , the complete medium was replaced with a serum-free DMEM ( for HEK23A cells ) or Neurobasal-A containing supplements ( for NB2a cells ) . Lasso-D was purified by immunoaffinity chromatography . Briefly , serum-free medium conditioned by HEK293A cells expressing Lasso-D was filtered through 0 . 2 µm filters and incubated with anti-FLAG M2 affinity gel overnight at 4°C . Lasso-D was then eluted with 20 mM triethylamine , neutralized with 1 M HEPES , dialyzed against PBS , sterile-filtered for use in cell culture and concentrated on sterile 30 kDa MWCO filtration units ( Vivaspin , GE Lifesciences ) . Medium above non-transfected cells was processed in the same manner and used as a negative control . Amount and purity of concentrated Lasso-D were assessed by SDS-PAGE and Coomassie staining . Activity was confirmed by measuring its binding to cell-surface or soluble LPHN1 constructs ( Silva et al . , 2011 ) . Hippocampal cultures were prepared from Sprague-Dawley E18 rat hippocampi ( BrainBits UK ) , according to the supplier’s instructions , or dissected from P0 AG148/2 mouse pups ( Adgrl1-/- , LPH1 KO ) under sterile conditions . Hippocampi were digested with 2 mg/ml papain in Ca2+-free Hibernate-A medium and dissociated in Hibernate-A medium with B27 supplement using fire-polished Pasteur pipettes . Cells were seeded in Neurobasal-A/B27 medium on poly-D-lysine-coated 13 mm coverslips at 5 × 104 cells/coverslip and maintained at 5% CO2 and 37°C . The medium was partially replaced at least once a week . Primary hippocampal neurons were transfected using Amaxa Rat Neuron Nucleofector Kit ( Lonza ) as described by the manufacturer . Briefly , dissociated cells were resuspended in Rat Neuron Neucleofector Solution with Supplement , then mixed with 3 µg of pcDNA6-GFP and electroporated in Nucleofector using the G-013 program . The transfected cells were resuspended in 500 µl of a recovery medium , containing a 1:3 mixture of Hibernate-A/B27 and Ca-free Hibernate-A ( BrainBits UK ) , and incubated at 37°C for 15 min . Cells were plated at a higher concentration to compensate for cell death . Next day , 0 . 8 nM Lasso-D was added to the medium ( PBS was added to control medium ) . At 4 DIV , the cultured hippocampal cells were fixed with 4% paraformaldehyde ( PFA ) , stained and visualized as described below in Image Analysis . To investigate axonal responses to chemoattractant gradients , MAIDs ( Figure 5 ) with 150 µm separation walls ( Xona Microfluidics LLC ) were prepared in accordance with the manufacturer’s guidelines ( Harris et al . , 2007a; Harris et al . , 2007b ) . Briefly , MAIDs were sterilized with ethanol , washed with sterile water and dried . To facilitate firm attachment of MAIDs , 22 × 22 mm coverslips ( VWR International ) were sonicated in water and ethanol , autoclaved , dried , then coated with 1 mg/ml poly-D-lysine overnight , washed , and dried overnight before the assembly . For neuronal cell culture in MAIDs , E18 rat hippocampi were dissociated as above . Neurons ( 1 . 5 × 105/10 µl ) were added to Somal Compartments and allowed to settle for 30 min . MAIDs were then filled with Neurobasal-A/B27 . After 3 DIV , the medium in Axonal Compartments was carefully replaced with medium containing soluble Lasso-D or with control medium . Alternatively , HEK293A cells stably expressing Lasso-D ( or untransfected ) were plated in the wells of Axonal Compartment . At 8 DIV , the cells were fixed and processed as described below . For diffusion modeling experiments , MAIDs were assembled as above and filled with PBS; then 0 . 1 mg/ml BSA-TRITC ( Thermo Fisher Scientific ) in PBS was added to Axonal Compartments without changing liquid level in any compartment ( to avoid creating a hydrostatic pressure in the microchannels ) . BSA-TRITC diffusion in MAIDs was monitored by time-lapse fluorescent imaging of all compartments for 5 days under an Axiovert fluorescent microscope ( Carl Zeiss ) equipped with a temperature- and humidity-controlling enclosure , and a Canon G5 camera . Fluorescence intensity profiles across the microchannels at multiple time points were generated in ImageJ ( NIMH , Bethesda; RRID:SCR_002285 , RRID:SCR_003070 ) and normalized to the fluorescence profile of 100 ng/ml BSA-TRITC forced into the microchannels and both compartments . Cells on coverslips or inside MAIDs were fixed for 10 min with 4% PFA ( for staining requiring SDS treatment to aid epitope retrieval , the fixative also included 0 . 1% glutaraldehyde ) . Cells were permeabilized with 0 . 1% Triton X-100 ( or 1% SDS for PAL1 and dmAb staining ) , washed , then blocked for 1 hr with 10% goat serum in PBS and incubated with primary antibodies in blocking solution ( dilutions used were: PAL1 , 3 ng/ml; dmAb , 1:300; anti-NF-H , anti-myc mAb , and anti-GFP , 1:1 , 000; anti-V5 , 1:2 , 000 ) for 1 hr at room temperature ( or overnight at 4°C with PAL1 and dmAb ) . The coverslips or MAIDs were then washed three times and incubated for 1 hr with secondary antibodies in blocking solution , followed by three washes . Coverslips were mounted using FluorSave mounting medium ( Calbiochem ) , while neurons in MAIDs were imaged within 4 hr after the washes . NB2a cells stably expressing LPH-42 were grown on poly-D-lysine-coated coverslips in DMEM , 10% fetal calf serum ( PAA Laboratories ) to 30–50% confluency and to test receptor clumping incubated at 0°C for 20 min in PBS with one of the three potential LPHN1 ligands: ( 1 ) 20 nM Lasso-D , ( 2 ) 2 nM Alexa Fluor 647-labeled LTXN4C ( Volynski et al . , 2004 ) , or ( 3 ) rabbit anti-NTF antibodies ( RL1 ) , followed by a 20 min incubation with Alexa Fluor 546-conjugated goat anti-rabbit IgG . In control , only the fluorescent secondary antibody was added for the last 20 min . The cells were then fixed for 10 min with 4% PFA in PBS , blocked with 10% goat serum in PBS , and subsequent procedures were designed to reveal the distribution of the three components of each assay ( NTF , CTF , and ligand ) . First , in all experiments , the V5 epitope on LPHN1 NTF was detected with a rabbit anti-V5 antibody ( 1 hr in blocking solution ) , followed by Alexa Fluor 488-conjugated goat anti-rabbit IgG and fixation . Subsequent staining depended on the ligand used: ( 1 ) Lasso-D was stained using a mouse anti-FLAG mAb and Alexa Fluor 546-conjugated goat anti-mouse IgG . For LPHN1 CTF detection , the cells were then permeabilized with 0 . 1% Triton X-100 , incubated with a chicken anti-myc antibody , fixed , blocked , and stained with Alexa Fluor 647-conjugated anti-chicken antibody . ( 2 ) With LTXN4C-induced patching , the cells were permeabilized , incubated with a mouse anti-myc mAb , fixed , blocked , and stained with an Alexa Fluor 546-conjugated anti-mouse IgG . ( 3 ) With RL1-induced patching ( and in controls ) , the cells were permeabilized , incubated with the chicken anti-myc antibody , fixed , blocked , and stained with Alexa Fluor 647-conjugated anti-chicken antibody . The primary antibodies were used at 1:1000 dilution; the secondary antibodies , 1:2000; the cells were washed three times with PBS after each stage . At the end , the cells were briefly fixed , blocked , washed , and mounted using FluorSave reagent ( Calbiochem , Cat . No . 345789 ) . Images of axons in MAIDs were acquired on an Axiovert 200M microscope ( Carl Zeiss ) using LD Plan-Neofluar 20x objective and Volocity-controlled camera , filters , shutter , and stage . Images were taken with a 5% overlap to facilitate stitching ( Perkin-Elmer; RRID:SCR_002668 ) . Blank images were subtracted to correct for optical artifacts . The images were stitched automatically and ‘despeckled’ , using a 3 × 3 median filter ( ImageJ ) . To correct for large illumination artifacts , background was subtracted in ImageJ using the ‘Subtract background’ plug-in , with a 100 µm window and the sliding paraboloid algorithm . Images of immunostained cells and neurons on coverslips ( other than for neurite tracing ) were acquired using an upright laser-scanning confocal microscope ( LSM-510 , Zeiss; RRID:SCR_014344 ) equipped with 40x or 100x oil-immersion objectives; 488 , 543 , and 633 nm lasers; and 505–530 , 560–615 , and >650 nm emission filters . Images for neurite tracing were acquired using Axio Observer . Z1 microscope ( Zeiss ) equipped with Hamamatsu ORCA-Flash 4 sCMOS camera , EC Plan-Neofluar 40x objective , Colibri 2 LED illumination and appropriate filters . To correlate the polarity of LPH1 expression and growth cone turning , GFP images of growth cones and preceding axons were traced using CorelTRACE X3 ( Corel , Canada ) . The obtained contour images were aligned along their median line , with all axons starting at the same point . The images were then flipped so that the higher LPHN1 staining was located in the right half of each growth cone . The trajectory of respective axons was then assessed: correlation was considered positive if the axon approached its cone from the right quadrant . To plot Jeffreys confidence intervals ( CI ) for a binomial distribution the standard formula was used: CI = p + z*sqrt ( p* ( 1 p ) /n ) , where z = 3 for confidence level CI = 0 . 9973 . For profiling of neurite growth within MAID Axonal Compartments , regions of interest encompassing the depth of the compartments , were selected , avoiding artefacts ( e . g . antibody aggregates or HEK cell bodies ) . The average fluorescence was determined as a function of distance ( see Figure 5A ) from the separation wall and binned over 100 µm intervals . Background fluorescence in the areas beyond 1200 µm from the wall ( that contained no axons ) was subtracted from all other fluorescence values , and the results were used for statistical analysis as described below . For axon fasciculation measurements in MAIDs , the width of each axon/bundle was determined in pixels at 100 µm from the separation wall and converted to µm . Neurite tracing of GFP-positive neurons was performed in ImageJ ( Schindelin et al . , 2012 ) using default settings in Simple Neurite Tracer plug-in ( Longair et al . , 2011 ) . The longest neurite for each cell was used as a single independent measurement ( data obtained from three independent cultures ) . Analysis of the co-localization of the NTF , CTF , and respective ligands in the plasma membrane was carried out using a method previously developed and tested ( Silva et al . , 2011 ) . Here , the confocal images were obtained near the middle of each cell ( optical plane , Z = 0 . 5 μm ) . For consistency , the recorded images were assigned false colors according to the detected protein , irrespective of the actual fluorescence wavelength used for detection . The fluorescence profiles for each protein along the cell’s perimeter were collected using ImageJ . Pearson’s correlation coefficient r was then calculated for the pairs of resulting profiles obtained from 4 to 7 independent experiments . In the representative images that were used in the Figures , the contrast and brightness were enhanced in the same manner as in respective control images . For experiments with LPHN1 KO and WT/HET cultures in MAIDs , the membranes of cell bodies and axons were labeled using 5 µM DiO ( Vybrant DiO , Life Technologies ) in Neurobasal-A , containing B-27 supplement and 0 . 005% Pluronic F-127 ( Sigma-Aldrich ) , which had been passed through a 0 . 2 µm filter . After 30 min incubation , the excess dye was carefully washed two times , and the cell bodies ( Somal Compartments ) and axons ( Axonal Compartments ) were solubilized in 1% Triton X-100 in PBS . The undiluted axonal and 10-fold diluted somal fractions were analyzed in microtiter plates using a Fluoroskan Ascent Fluorometer ( 485 nm excitation , 505 nm emission filters ) ( Thermo Fisher Scientific ) . In some experiments , 2 μL samples of lysates were individually measured using a NanoDrop ND-3300 Fluorospectrometer ( Thermo Fisher Scientific ) with the following settings: 470 nm Blue LED excitation , 500–700 nm emission spectrum , quantified at 504 nm . The levels of fluorescence were proportional to the amount of axons/cells bodies present in respective compartments . For Western Blot analysis of conditioned media , these were passed through 0 . 2 µm low protein-binding filters ( PALL , USA ) . The cells on coverslips were lysed in ice-cold RIPA buffer ( 1% sodium deoxycholate , 0 . 1% SDS , 1% Triton X-100; 10 mM Tris-HCl , pH 8; 140 mM NaCl ) , supplemented with protease inhibitors and 1 mM EDTA . To prepare samples for electrophoresis , the cell lysates and media were incubated at 50°C for 30 min with sample buffer containing 2% SDS and 100 mM DTT . The samples were separated on standard SDS-containing polyacrylamide gels , blotted onto polyvinylidene fluoride membrane ( Immobilon-P , IPVH00010 , Merck ) , blocked with 5% non-fat dry milk , incubated with primary antibodies diluted in 2% BSA for TN2N or 5% milk for all other antibodies ( dilutions used were: PAL1 , 1:500; dmAb , 1:1 , 000; TN2N , 1 µg/ml; actinin , 1:1 , 500; NF-H , 1:10 , 000 ) and respective horseradish-peroxidase conjugated secondary antibodies . The stained membranes were visualized by WestFemto chemiluminescent substrate kit ( Thermo Fisher Scientific ) and LAS3000 gel/blot documentation system ( FUJIFILM ) . Cytosolic Ca2+ concentration was monitored using Fluo-4 Ca2+ indicator ( the method was also described in ( Silva et al . , 2009; Volynski et al . , 2004 ) . The stably transfected NB2a cells expressing LPH-42 were pre-incubated in serum-free medium for 24 hr in 30 mm dishes . Then the cells were equilibrated for 20 min in physiological buffer ( in mM: NaCl , 145; KCl , 5 . 6; glucose , 5 . 6; MgCl2 , 1; EGTA , 0 . 2; HEPES , 15; pH 7 . 4; BSA , 0 . 5 mg/ml ) containing 2 . 5 µM Fluo-4 acetomethoxy ester ( Fluo-4-AM , Thermo Fisher Scientific ) and 10% Pluronic F–127 , washed and further incubated for 20 min for dye de-esterification . LPHN1-expressing cells were identified by staining with primary mouse anti-V5 mAb pre-labeled with Alexa Fluor 568 ( Zenon , Thermo Fisher Scientific ) . Images were acquired every 5 s under the LSM510 microscope using a 40x Achroplan water-dipping objective , 488 nm laser and a 505–550 nm band-pass emission filter . The following protocols were typically applied ( the addition times and final concentrations of the additives are indicated , see also Figure 7—figure supplement 1A and B ) . Protocol 1: 0 min , baseline recording; 5 min , 1 nM LTXN4C , 360 nM Lasso-D , or control buffer; 30 min , 2 mM Ca2+; 50 min , 1 nM wild-type α-LTX; 55 min , end . Protocol 2: 0 min , 2 mM Ca2+ , baseline recording; 5 min , 360 nM Lasso-D or control buffer; 30 min , 1 nM LTXN4C; 80 min , 1 nM α-LTX; 90 min , end . Ca2+ fluorescence of individual positive cells was quantified using the LSM510 software and normalized between the starting fluorescence and maximal fluorescence induced by α-LTX . MEPPs were recorded from isolated neuromuscular preparations by the method also used in ( Lelyanova et al . , 2009 ) . Flexor digitorum brevis muscles were dfrom male P21 mice ( C57BL/6J: Adgrl1+/+ or Adgrl1-/- ) , cleaned from connective tissue , fixed using entomological pins in Petri dishes pre-coated with Sylgard silicone polymer ( Dow Corning ) , and incubated in constantly oxygenated physiological buffer containing ( in mM ) : NaCl , 137; KCl , 5; MgCl2 , 1; EGTA , 0 . 2; glucose , 5 . 6; HEPES , 10; pH 7 . 5; tetrodotoxin ( Latoxan ) , 0 . 001 ) . Sharp electrodes with tip diameter <0 . 5 μm and 30–60 MOhm impedance were produced on a P-97 puller ( Sutter ) from borosilicate glass filament capillaries ( 1 . 5 mm; World Precision Instruments ) and filled with 5 M ammonium acetate . Spontaneous presynaptic activity ( based on MEPPs detection ) was recorded using a system consisting of an Axoclamp 2B pre-amplifier ( Axon Instruments ) in the current clamp mode , a secondary differential amplifier with a high-frequency filter ( LPF202A , Warner Instruments ) , a HumBug harmonic frequency quencher ( Quest Scientific ) , a Digidata 1322A digitizer ( Axon Instruments ) , and a microcomputer running AxoScope software ( Axon Instruments ) . The recorded traces were subsequently analyzed using MiniAnalysis software ( Synaptosoft Inc . ) . The data shown are the means ± SEM , unless otherwise stated . A Lilliefors test was applied to all data sets to assess normality in data distribution . Statistical significance was then determined using two-tailed heteroscedastic t-test , with Bonferroni correction in cases of multiple pair-wise comparisons . For non-normally distributed data , a Mann-Whitney test was applied . The axonal fluorescence curves obtained from image analysis in MAIDs were compared using n-way ANOVA algorithm ( MATLAB; RRID:SCR_001622 ) , where n reflected the number of factors involved in an assay ( treatment type , distance from the separation wall and batch number ) ( Vysokov , 2018 ) . To test for correlation in axonal fasciculation measurements , a Pearson correlation coefficient ( R2 ) and the p values ( to test the correlation hypothesis ) were calculated using MATLAB . Jeffreys confidence intervals were used to assess statistical significance of correlation between LPH1 enrichment and growth cone turning direction . Differences were considered significant if p<0 . 05 . The specific p and n values are indicated in corresponding figure legends or the following notation is used to denote statistical significance: NS ( non-significant ) , p>0 . 05; * , p<0 . 05; ** , p<0 . 01; *** , p<0 . 001 . The investigators were blinded to the identity of samples during data collection and analysis in all experiments involving LPHN1 KO . | The brain is a complex mesh of interconnected neurons , with each cell making tens , hundreds , or even thousands of connections . These links can stretch over long distances , and establishing them correctly during development is essential . Developing neurons send out long and thin structures , called axons , to reach distant cells . To guide these growing axons , neurons release molecules that work as traffic signals: some attract axons whilst others repel them , helping the burgeoning structures to twist and turn along their travel paths . When an axon reaches its target cell , the two cells join to each other by forming a structure called a synapse . To make the connection , surface proteins on the axon latch onto matching proteins on the target cell , zipping up the synapse . There are many different types of synapses in the brain , but we only know a few of the surface molecules involved in their creation – not enough to explain synaptic variety . Two of these surface proteins are latrophilin-1 , which is produced by the growing axon , and Lasso , which sits on the membrane of the target cell . The two proteins interact strongly , anchoring the axon to the target cell and allowing the synapse to form . However , a previous recent discovery by Vysokov et al . has revealed that an enzyme can also cut Lasso from the membrane of the target cell . The ‘free’ protein can still interact with latrophilin-1 , but as it is shed by the target cell , it can no longer serve as an anchor for the synapse . Could it be that free Lasso acts as a traffic signal instead ? Here , Vysokov et al . tried to answer this by growing neurons from a part of the brain called the hippocampus in a special labyrinth dish . When free Lasso was gradually introduced in the culture through microscopic channels , it interacted with latrophilin-1 on the surface of the axons . This triggered internal changes that led the axons to add more membrane where they had sensed Lasso , making them grow towards the source of the signal . The results demonstrate that a target cell can both carry and release Lasso , using this duplicitous protein to help attract growing axons as well as anchor them . The work by Vysokov et al . contributes to our knowledge of how neurons normally connect , which could shed light on how this process can go wrong . This may be relevant to understand conditions such as schizophrenia and ADHD , where patients’ brains often show incorrect wiring . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2018 | Proteolytically released Lasso/teneurin-2 induces axonal attraction by interacting with latrophilin-1 on axonal growth cones |
Centrioles are characterized by a nine-fold arrangement of microtubule triplets held together by an inner protein scaffold . These structurally robust organelles experience strenuous cellular processes such as cell division or ciliary beating while performing their function . However , the molecular mechanisms underlying the stability of microtubule triplets , as well as centriole architectural integrity remain poorly understood . Here , using ultrastructure expansion microscopy for nanoscale protein mapping , we reveal that POC16 and its human homolog WDR90 are components of the microtubule wall along the central core region of the centriole . We further found that WDR90 is an evolutionary microtubule associated protein . Finally , we demonstrate that WDR90 depletion impairs the localization of inner scaffold components , leading to centriole structural abnormalities in human cells . Altogether , this work highlights that WDR90 is an evolutionary conserved molecular player participating in centriole architecture integrity .
Centrioles and basal bodies ( referred to as centrioles from here onwards for simplicity ) are conserved organelles important for the formation of the centrosome as well as for templating cilia and flagella assembly ( Bornens , 2012; Breslow and Holland , 2019; Conduit et al . , 2015; Ishikawa and Marshall , 2011 ) . Consequently , defects in centriole assembly , size , structure and number lead to abnormal mitosis or defective ciliogenesis and have been associated with several human pathologies such as ciliopathies and cancer ( Gönczy , 2015; Nigg and Holland , 2018; Nigg and Raff , 2009 ) . For instance , centriole amplification , a hallmark of cancer cells , can result from centriole fragmentation in defective , over-elongated centrioles ( Marteil et al . , 2018 ) . Centrioles are characterized by a nine-fold radial arrangement of microtubule triplets , are polarized along their long axis , and can be divided in three distinct regions termed proximal end , central core and distal tip ( Hamel et al . , 2017 ) . Each region displays specific structural features such as the cartwheel on the proximal end , which is crucial for centriole assembly ( Nakazawa et al . , 2007; Strnad et al . , 2007 ) or the distal appendages at the very distal region , essential for membrane docking during ciliogenesis ( Tanos et al . , 2013 ) . The central core region of the centriole is defined by the presence of a circular inner scaffold thought to maintain the integrity of microtubule triplets under compressive forces ( Le Guennec et al . , 2020 ) . Using cryo-tomography , we recently showed that the inner centriole scaffold forms an extended helix covering ~70% of the centriole length and that is rooted at the inner junction between the A and B microtubules ( Figure 1A , B ) . This connection consists of a stem attaching the neighboring A and B microtubules and three arms extending from the same stem toward the centriolar lumen ( Le Guennec et al . , 2020; Figure 1A , B ) . The stem of the inner scaffold has been detected in Paramecium tetraurelia , Chlamydomonas reinhardtii and human centrioles , suggesting that it represents an evolutionary conserved structural feature . The molecular identity of some components of the inner scaffold has been uncovered using Ultrastructure Expansion Microscopy ( U-ExM ) , which allows nanoscale localization of proteins within structural elements ( Gambarotto et al . , 2019 ) . Notably , the centriolar proteins POC1B , FAM161A , POC5 and Centrin have been shown to localize to the inner scaffold along the microtubule blades in human cells ( Le Guennec et al . , 2020 ) . Moreover , these proteins form a complex that can bind to microtubules through the microtubule-binding protein FAM161A ( Le Guennec et al . , 2020; Zach et al . , 2012 ) . Importantly , a subset of these proteins has been shown to be important , such as POC5 for centriole elongation ( Azimzadeh et al . , 2009 ) as well as POC1B for centriole and basal body integrity ( Pearson et al . , 2009; Venoux et al . , 2013 ) . This observation highlights the role of the inner scaffold structure in providing stability to the entire centriolar microtubule wall organization . However , the exact contribution of the inner scaffold to microtubule triplets stability and how the inner scaffold is connected to the microtubule blade is unknown . We recently identified the conserved proteins POC16/WDR90 as proteins localizing to the central core region in both Chlamydomonas reinhardtii and human centrioles ( Hamel et al . , 2017 ) . Impairing POC16 or WDR90 functions has been found to affect ciliogenesis , suggesting that POC16/WDR90 may stabilize the microtubule wall , thereby ensuring proper flagellum or cilium assembly ( Hamel et al . , 2017 ) . Interestingly , POC16 has been proposed to be at the inner junction between the A and B microtubules ( Yanagisawa et al . , 2014 ) through its sequence identity with FAP20 , an axonemal microtubule doublet inner junction protein of Chlamydomonas reinhardtii flagella ( Dymek et al . , 2019; Ma et al . , 2019; Owa et al . , 2019; Yanagisawa et al . , 2014 ) . As the stem connects the A- and B-microtubules interface , these observations suggest that POC16/WDR90 may connect the inner scaffold to the microtubule triplet through this stem structure ( Figure 1C ) , thus ensuring integrity of the centriole architecture . In this study , using a combination of cell biology , biochemistry and Ultrastructure Expansion Microscopy ( U-ExM ) approaches , we establish that the conserved POC16/WDR90 proteins localize on the centriolar microtubule wall in the central core region of both Chlamydomonas and human cells . We further demonstrate that WDR90 is a microtubule-binding protein and that loss of this protein impairs the localization of inner scaffold components and leads to slight centriole elongation , impairment of the canonical circular shape of centrioles as well as defects in centriolar architecture integrity .
To test the hypothesis that POC16/WDR90 is a microtubule triplet component , we analyzed its distribution using U-ExM that allows nanoscale mapping of proteins inside the centriole ( Gambarotto et al . , 2019; Le Guennec et al . , 2020 ) . We observed first in Chlamydomonas reinhardtii isolated centrioles that the endogenous POC16 longitudinal fluorescence signal is restricted to the central core region as compared to the tubulin signal , which depicts total centriolar length ( Figure 1D–F ) . From top viewed centrioles , we measured the distance between both POC16 and tubulin maximal intensity signal from the exterior to the interior of the centriole and found that POC16 localizes precisely on the microtubule wall in the central core region of Chlamydomonas centrioles ( Figure 1M , N , average distance between POC16 and tubulin Δ = 0 nm +/- 8 ) . As a control , we could recapitulate the internal localization along the microtubule wall of POB15 , another central core protein ( Figure 1G–I and Figure 1M , N , average distance between POB15 and tubulin Δ = 12 nm +/- 7 ) as previously reported using immunogold-labeling ( Hamel et al . , 2017 ) . In human centrioles , the POC16 human homolog WDR90 localizes similarly to POC16 on the centriolar microtubule wall , demonstrating the evolutionary conserved restricted localization of POC16/WDR90 on microtubule triplets in the central core region of centrioles ( Figure 1J–L ) . Of note , POC16 and WDR90 display a punctate distribution that we hypothesize to be due to the poor quality of the antibody . Next , we compared the relative position of WDR90 from top view centrioles to previously described inner scaffold components ( Figure 1O–Q ) ( see Materials and methods ) . We found that while WDR90 precisely localizes to the centriolar microtubule wall ( Figure 1P , average distance between WDR90 and tubulin: Δ = 2 nm +/12 ) , POC1B , FAM161A , POC5 and Centrin signals were shifted toward the centriole lumen in comparison to the tubulin signal , as previously reported ( Figure 1P , Δ = 15 nm +/- 8; 22 nm +/- 5; 27 nm +/- 6 and 28 nm +/- 9 , respectively ) ( Le Guennec et al . , 2020 ) . These results demonstrate that WDR90 longitudinal distribution is similar to the inner scaffold components but its localization on the microtubule wall suggests that WDR90 is a component of the centriolar microtubule triplet of the central core region . Proteins of the POC16/WDR90 family consist of an N-terminal DUF667-containing domain ( domain of unknown function ) , similar to the ciliary inner junction protein FAP20 ( Figure 2—figure supplement 1A; Yanagisawa et al . , 2014 ) , followed by multiple WD40 repeats that form β-propeller structures ( Figure 2A and Figure 2—figure supplement 1B; Xu and Min , 2011 ) . First , we wanted to probe the evolutionary conservation of POC16/WDR90 family members as centriolar proteins . To this end , we raised an antibody against Paramecium tetraurelia POC16 and confirmed its localization at centrioles similarly to what we found in Chlamydomonas reinhardtii and human cells ( Figure 2—figure supplement 1C; Hamel et al . , 2017 ) . Further driven by its predicted similarity to the microtubule associated protein FAP20 ( Khalifa et al . , 2020 ) and the underlying hypothesis that POC16/WDR90 proteins might be joining A and B microtubules as well as by their precise localization on the microtubule wall ( Figure 1 ) , we first set out to understand the structural identity between the predicted structures of POC16-DUF667 domain to the recently published near atomic structure of FAP20 from flagella microtubule doublets ( Khalifa et al . , 2020; Ma et al . , 2019; Figure 2—figure supplement 2A–C ) . Strikingly , we observed high similarities between the two structures , suggesting similar biological functions at the inner junction . Moreover , we fitted POC16 model prediction into FAP20 cryo-EM density map and found a good concordance , further hinting for a conserved localization at the level of the microtubule triplet ( Figure 2—figure supplement 2D ) . Prompted by this result , we then tested whether POC16/WDR90 proteins , similar to FAP20 , can bind microtubules both in human cells as well as in vitro . To do so , we overexpressed the N-terminal part of WDR90 and POC16 comprising the DUF667 domain ( WDR90-N ( 1-225 ) and POC16 ( 1-295 ) , respectively ) fused to GFP in U2OS cells and found that this region is sufficient to decorate cytoplasmic microtubules ( Figure 2B and Figure 2—figure supplement 3A ) . We next tested whether overexpressing such a WDR90-N-terminal fragment could stabilize microtubules . To this end , we analyzed the microtubule network in cells overexpressing mCherry-WDR90-N after depolymerizing microtubules through a cold shock treatment ( Figure 2—figure supplement 3B–D ) . We found that while low expressing cells did not maintain a microtubule network , high expressing cells did . This suggests that WDR90-N can stabilize microtubules . In contrast , we observed that full-length WDR90 fused to GFP only anecdotally binds microtubules . This observation suggests a possible autoinhibition conformation of the full-length protein and/or to interacting partners preventing microtubule binding in the cytoplasm ( Figure 2—figure supplement 3E ) . Next , we determined whether different POC16/WDR90 N-terminal domains directly bind to microtubules in vitro and whether this function has been conserved in evolution . Bacterially expressed , recombinant POC16/WDR90 DUF667 domains from seven different species were purified and their microtubule interaction ability was assessed using a standard microtubule-pelleting assay ( Figure 2—figure supplement 1A and Figure 2C ) . We found that every POC16/WDR90 DUF667 domain directly binds to microtubules in vitro . This interaction was further confirmed using negative staining electron microscopy , where we could observe recombinant WDR90-N localizing on in vitro polymerized microtubules ( Figure 2E ) . We next investigated whether POC16/WDR90 DUF667 domain could also interact with free tubulin dimers , considering that closure of the inner junction between the A and B microtubules necessitates two microtubule/tubulin-binding sites as recently reported for FAP20 ( Ma et al . , 2019 ) . We observed that all POC16/WDR90 DUF667 orthologs directly interact with tubulin dimers , generating oligomers that pellet under centrifugation ( Figure 2D ) . We then tested whether the DUF667 domain could still interact with tubulin once bound to microtubules . We subsequently incubated either WDR90-N or POC16 ( 1-295 ) pre-complexed with microtubules with an excess of free tubulin and analyzed their structural organization by electron microscopy ( Figure 2E , F and Figure 2—figure supplement 3F , G ) . We observed an additional level of decoration due to the simultaneous binding of the DUF667 domains with tubulin and microtubules ( Figure 2E , F and Figure 2—figure supplement 3F , G ) . Furthermore , we revealed a 8 . 5 nm periodical organization of tubulin-WDR90-N oligomers on microtubules ( Figure 2G ) , similar to the recent high-resolution structure of the ciliary microtubule doublet showing that monomeric FAP20 interacts with both A- and B-microtubules every 8 nm at the inner junction ( Khalifa et al . , 2020; Ma et al . , 2019 ) . Due to its similarity , it is tempting to speculate that the DUF667 domain of POC16/WDR90 is also monomeric , however it is also possible that WDR90 forms a homodimer capable of interacting with the microtubules and tubulin . Based on these results , we concluded that POC16/WDR90 is an evolutionary conserved microtubule/tubulin-interacting protein with the capacity to connect microtubules , a functional prerequisite for an inner junction protein that simultaneously interacts with the A and B microtubules . We next assessed whether WDR90 recruitment at centrioles is correlated with the appearance of inner scaffold proteins during centriole biogenesis . In cycling human cells , centrioles duplicate only once per cell cycle during S phase , with the appearance of one procentriole orthogonally to each of the two mother centrioles . Procentrioles then elongate during the following G2 phase of the cell cycle , acquiring the inner scaffold protein POC5 that is critical for the formation of the central and distal parts of the nascent procentriole ( Azimzadeh et al . , 2009 ) . We followed endogenous WDR90 localization across the cell cycle by analyzing synchronized human RPE1 cells fixed at given time points and stained for either Centrin or HsSAS-6 , both early protein marker of duplicating centrioles ( Azimzadeh et al . , 2009; Strnad et al . , 2007; Figure 3A–F and Figure 3—figure supplement 1A , B ) . We found that while Centrin and HsSAS-6 are recruited as expected early on during procentriole formation in S phase ( 22 hr ) ( Strnad et al . , 2007 ) , WDR90 starts appearing only in early G2 when procentriole elongation starts ( 24 hr ) ( Figure 3A–F ) . Signal intensity analysis over the cell cycle further demonstrates that WDR90 appears on procentrioles in early G2 and reaches full incorporation by the end of G2 ( Figure 3G , H ) , similarly to the reported incorporation of the inner scaffold protein POC5 ( Azimzadeh et al . , 2009 ) . In addition , we noticed that besides its centriolar distribution , WDR90 localizes also to centriolar satellites , which are macromolecular assemblies of centrosomal proteins scaffolded by the protein PCM1 and involved in centrosomal homeostasis ( Drew et al . , 2017; Odabasi et al . , 2020; Figure 3—figure supplement 1C–D ) . Thus , we tested whether WDR90 satellite localization depends on the satellite protein PCM1 by depleting PCM1 using siRNA and assessing WDR90 distribution . We found that in absence of PCM1 , WDR90 is solely found at centrioles ( Figure 3—figure supplement 1E–H ) , demonstrating that WDR90 satellite localization is PCM1-dependent . Altogether , these data establish that WDR90 is a centriolar and satellite protein that is recruited to centrioles in the G2-phase of the cell cycle , during procentriole elongation and central core/distal formation , similarly to the recruitment of the inner scaffold protein POC5 . To better understand the function of WDR90 , we analyzed cycling human cells depleted for WDR90 using siRNA and co-labeled WDR90 with the early centriolar marker Centrin . As previously shown ( Hamel et al . , 2017 ) , WDR90 siRNA-treated cells showed significantly reduced WDR90 levels at centrosomes in comparison to control cells ( Figure 3—figure supplement 2A , C ) . Moreover , we observed an asymmetry in signal reduction at centrioles in WDR90-depleted cells , with only one of two Centrin-positive centrioles still associated with WDR90 in G1 and early S-phase ( 69% compared to 10% in controls ) and one of four Centrin-positive centrioles in S/G2/M cells ( 77% compared to 0% in controls , Figure 3—figure supplement 2B ) . As the four Centrin-positive dots indicate duplicated centrioles , this result suggests that the loss of WDR90 does not result from a duplication failure ( Figure 3—figure supplement 2B ) . We postulate therefore that the remaining WDR90 signal possibly corresponds to the mother centriole and that the daughter has been depleted from WDR90 ( Figure 3—figure supplement 2E ) , similarly to what has been observed for the protein POC5 ( Azimzadeh et al . , 2009 ) . We further conclude that WDR90 is stably incorporated into centrioles , in agreement with its possible structural role . We also noted that the intensity of the Centrin and POC5 signals were markedly reduced upon WDR90 siRNA treatment ( Figure 3—figure supplement 2D–K ) . Indeed , we found that only 39% of WDR90-depleted cells displayed 2 POC5 dots in G1 ( negative for HsSAS-6 signal ) in contrast to the 86% of control cells with 2 POC5 dots ( Figure 3—figure supplement 2H ) . Moreover , 68% of control cells had 2 to 4 POC5 dots in S/G2/M ( associated with 2 HsSAS-6 dots ) in contrast to 29% in WDR90-depleted condition ( Figure 3—figure supplement 2H ) . The HsSAS-6 signal was not affected in WDR90-depleted cells , confirming that initiation of the centriole duplication process is not impaired under this condition ( Figure 3—figure supplement 2G , J , L ) . Similarly , the fluorescence intensity of the distal centriole cap protein CP110 was not changed under WDR90-depletion in contrast to the Centrin signal reduction ( Figure 3—figure supplement 2M–O ) . To ascertain the specificity of this phenotype , we generated a stable cell line expressing a siRNA-resistant version of WDR90 fused to GFP in its N-terminus ( GFP-WDR90RR ) upon doxycycline induction . We found that expression of GFP-WDR90RR restores partially the Centrin and POC5 signals at centrioles ( Figure 3I–L ) . Taken together , these results indicate that the depletion of WDR90 leads to a decrease in Centrin and POC5 localization at centrioles but does not affect the initiation of centriole duplication nor the recruitment of the distal cap protein CP110 . To investigate the structural role of POC16/WDR90 proteins on centrioles , we initially turned to the previously studied Chlamydomonas reinhardtii poc16m504 and poc16m55 mutants ( Hamel et al . , 2017; Li et al . , 2016 ) . However , after backcrossing these two strains with a wild-type strain ( CC-124 ) , it was found that the poc16 mutation is unlinked to the motility phenotype of poc16m555 and unlinked to the ciliary assembly defect of poc16m504 previously reported ( personal communication from Prof . Susan Dutcher , Washington University in St . Louis ) . Further genetic characterization will be needed to study the phenotypes associated with poc16 mutations . Therefore , we decided to analyze WDR90 phenotype in human cells and asked whether WDR90 depletion might lead to a loss of inner scaffold components as well as to a centriole architecture destabilization . We tested this hypothesis by analyzing centrioles from WDR90-depleted U2OS cells using U-ExM ( Figure 4 ) . As expected , we observed a strong reduction of WDR90 at centrioles , with a reminiscent asymmetrical signal in one of the two mature centrioles ( Figure 4A , B ) . Unexpectedly , we found that WDR90-depleted centrioles exhibited a slight tubulin length increase ( 502 nm +/- 65 compared to 434 nm +/- 58 in controls ) , potentially indicative of a defect in centriole length regulation ( Figure 4C ) . In contrast , despite a slight decrease at the level of the central core , we did not observe , in neither of the conditions , any significant difference in centriole diameter at the proximal and very distal regions ( Figure 4D ) . A key prediction is that the inner scaffold is connected to the microtubule wall through the stem structure that may contain WDR90 . To test this , we next analyzed whether the localization of the four described inner scaffold components POC1B , FAM161A , POC5 and Centrin would be affected in WDR90-depleted cells . We found that the localization of these four proteins in the central core region of centrioles was markedly altered in WDR90-depleted daughter centrioles ( Figure 4E , F ) using CEP164 to label the mother centriole ( Figure 4—figure supplement 1A–C ) . Instead of covering ~60% of the entire centriolar lumen , we only observed a ~ 20% remaining belt , positive for inner scaffold components at the proximal extremity of the core region ( Figure 4E–G and Figure 4—figure supplement 1D , E ) , suggesting that their initial recruitment may not be entirely affected . Another possibility would be that incomplete depletion of WDR90 allows for partial localization of inner scaffold components . It should also be noted that Centrin , which displays a central core and an additional distal tip decoration ( Le Guennec et al . , 2020 ) , was affected specifically in its inner core distribution ( Figure 4E white arrow , Figure 4—figure supplement 1D , E ) . The discovery of the inner scaffold within the centriole led to the hypothesis that this structure is important for microtubule triplet stability and thus overall centriole integrity ( Le Guennec et al . , 2020 ) . In line with this hypothesis , we found that upon WDR90 depletion , 10% of cells had their centriolar microtubule wall broken , indicative of microtubule triplets fracture and loss of centriole integrity ( 15 out of 150 centrioles , Figure 5 , Videos 1 and 2 ) . The break occurred mainly above the remaining belt of inner scaffold components ( Figure 5A–D ) , possibly reflecting a weakened microtubule wall in the central and distal region of the centriole . We also noticed that the perfect cylindrical shape ( defined as roundness ) of the centriolar microtubule wall was affected with clear ovoid-shaped or opened centrioles seen from near-perfect top view oriented centrioles ( Figure 5E , F and Figure 5—figure supplement 1 , 95% of depleted centrioles in top view are affected ) , illustrating that loss of WDR90 and the inner scaffold leads to disturbance of the characteristic centriolar architecture . To assess whether WDR90 stability phenotype correlates solely with disturbance of inner scaffold proteins , we analyzed the distribution of the centriolar proteins FOP1 and CEP135 ( BLD10 ) as well as glutamylation ( PolyE ) , all known to be important for centriole stability ( Bayless et al . , 2012; Bayless et al . , 2016; Bobinnec et al . , 1998; Lin et al . , 2013; Matsuura et al . , 2004 ) . While CEP135 and glutamylation were not altered in WDR90-depleted cells ( Figure 4—figure supplement 1F–K ) , we found that FOP1 distribution was slightly disturbed at centrioles ( Figure 4—figure supplement 1L–N ) but still present , reinforcing our interpretation that the centriole breakage is probably due to the loss of the inner scaffold components . As the inner scaffold connects the microtubule triplet together , we wondered whether the remaining belt seen in WDR90 depleted cells could limit the phenotype of centriolar breakage . To test this hypothesis , we decided to co-deplete WDR90 with the inner scaffold protein POC5 . We first depleted POC5 alone using previously described siRNA ( Figure 6A , B; Azimzadeh et al . , 2009 ) . Consistently with WDR90 depletion , we found that the removal of the inner scaffold POC5 , which occurs mainly at daughter centrioles ( Figure 6—figure supplement 1A ) , led to a slight centriole elongation ( Figure 6C , D ) and resulted in 10% of broken centrioles ( Figure 6—figure supplement 2A , B; 4 out of 46 centrioles ) . We also confirmed that POC5 depletion leads to shorter procentrioles in metaphase as previously reported ( Azimzadeh et al . , 2009 ) but then become over elongated just after mitosis ( Figure 6—figure supplement 1B , C ) . We next assessed whether POC5 depletion would impair WDR90 distribution; however , we found this not to be the case , as WDR90 localization is not affected at centrioles upon POC5 depletion ( Figure 6E–H and Figure 6—figure supplement 1D ) . This result therefore indicates that WDR90 is upstream of POC5 . We next capitalize on this efficient POC5 depletion to co-deplete POC5 together with WDR90 ( Figure 6—figure supplement 1E–J ) . We found that the double siRNA led to a strong decrease of cell number as compared to WDR90 depletion alone , suggesting either an increase of cell mortality or a defect in cell cycle progression ( Figure 6—figure supplement 1J ) . As expected , we found that the remaining POC5 belt found in WDR90-depleted centrioles was completely removed ( Figure 6I ) . Moreover , centrioles appeared even further elongated under these conditions , indicating that the complete removal of POC5 further enhances the WDR90 phenotype ( Figure 6I , J ) . Structurally , we noticed beside the elongated centrioles about 30% , of abnormal centrioles in WDR90/POC5 depleted cells ( Figure 6—figure supplement 2A , C , D; 70 out of 260 centrioles ) , ranging from very short centrioles that seem to lack the entire core/distal region as well as centrioles with broken microtubule blades . We also noted a loss of centriole roundness ( Figure 6I , white arrow ) . Overall , these phenotypes support our prediction that depletion of inner scaffolds component strongly impairs centriole integrity . Collectively , we demonstrate that WDR90 is crucial to ensure inner core protein localization within the centriole core , as well as to maintain the microtubule wall integrity and the overall centriole roundness and stability ( Figure 5G ) .
What maintains centriole barrel stability and roundness is a fundamental open question . Centrioles are microtubule barrel structures held together by the A-C linker at their proximal region and a recently discovered inner scaffold in the central/distal region ( Le Guennec et al . , 2020 ) . The presence of such an extended scaffold covering 70% of the centriolar length has led to the hypothesis that this structure is important for maintaining centriole integrity ( Le Guennec et al . , 2020 ) . Our work demonstrates that POC16/WDR90 family proteins constitute an evolutionary conserved central core microtubule triplet component that is essential for maintaining the inner centriolar scaffold components in human centrioles . The depletion of WDR90 leads to centriolar defects and impairment of microtubule triplets organization resulting in the loss of the canonical circular shape of centrioles . We also found that this overall destabilization of the centriole can lead to microtubule triplet breakage . Whether this phenotype arises as a consequence of the loss of the inner scaffold or due to the destabilization of the inner junction of the microtubule triplet is still an opened question that should be addressed in the future . Moreover , although unlikely , we cannot exclude that fragile centrioles such as the ones found in WDR90-depleted cells could be affected and further distorted by the technique of expansion microscopy . We also demonstrate using expansion microscopy that POC16/WDR90 is a component of the microtubule triplet restricted to the central core region . In addition and based on the sequence and structural similarity to the DUF667 domain of FAP20 that composes the inner junction in flagella , we propose that POC16/WDR90 localizes at the inner junction of the A and B microtubule of the centriolar microtubule triplet . The fact that WDR90 localization is restricted to the central core region led us to hypothesize that another protein , possibly FAP20 as it has been previously reported at centrioles ( Yanagisawa et al . , 2014 ) , could mediate the inner junction between A- and B-microtubule in the proximal region of the centriole . Moreover , in POC16/WDR90 proteins , the DUF667 domain is followed by a WD40 domain sharing a similarity with the flagellar inner B-microtubule protein FAP52/WDR16 ( Owa et al . , 2019 ) leading us to propose that the WD40 domains of POC16/WDR90 might also be located inside the B-microtubule of the triplet . However , whether this is the case remains to be addressed in future studies . In addition , WDR90 is potentially not the only protein that forms the inner junction . Indeed , we and others also previously show that FAM161A ( Le Guennec et al . , 2020; Zach et al . , 2012 ) , similarly to WDR90 , is a microtubule-binding protein close to the inner microtubule wall of the centriole , raising the possibility that both might compose the stem and link the microtubule triplets to the inner scaffold . It will be interesting in the future to study whether these two proteins interact . Our work further establishes that WDR90 is recruited to centrioles in G2 phase of the cell cycle concomitant with centriole elongation and inner central core assembly . We found that WDR90 depletion does not impair centriole duplication nor microtubule wall assembly , as noted by the presence of the proximal marker HsSAS-6 and the distal cap CP110 . In stark contrast , WDR90 depletion leads to a strong reduction of inner scaffold components at centrioles , as well as some centriole destabilization . Although several examples of centriole integrity loss have been demonstrated in the past , the molecular mechanisms of centriole disruption are not understood . For instance , Delta- and Epsilon-tubulin mutants have been shown in several model organisms to affect centriole integrity ( Dutcher et al . , 2002; Dutcher and Trabuco , 1998; Garreau de Loubresse et al . , 2001; O'Toole et al . , 2003 ) with notably in human cells where Delta- and Epsilon-tubulin null mutant cells were shown to lack microtubule triplets and have thus unstable centrioles that do not persist to the next cell cycle ( Wang et al . , 2017 ) . Remarkably , these centrioles can elongate with a proper recruitment of the cartwheel component HsSAS-6 and the distal marker CP110 but fails to recruit POC5 , a result that is similar to our findings with WDR90-depleted cells . As Delta- and Epsilon-tubulin null human mutant cells can solely assemble microtubule singlets ( Wang et al . , 2017 ) , we speculate that WDR90 might not be recruited in these centrioles , as the A- and B-microtubule inner junction would be missing . As a consequence , the inner scaffold proteins may not be recruited , as already shown for POC5 , leading to the observed futile cycle of centriole formation and disintegration ( Wang et al . , 2017 ) . It would therefore be interesting to study the presence of WDR90 in these null mutants as well as the other components of the inner scaffold in the future . Our work also showed that WDR90 as well as POC5 depletion affects centriole length in human cells . Altogether , these results emphasize the role of these two proteins in overall centriole length regulation and suggest an unexpected role of the inner scaffold structure in centriole length control . It would be of great interest to understand if and how the absence of the inner scaffold can affect the length of the centriole without affecting distal markers such as CP110 , which remains unchanged in our experiments . It is very likely that the concomitant elongation of the centriole with the appearance of inner scaffold components in G2 can act on the final length of this organelle . Given the importance of centriole integrity in enabling the proper execution of several diverse cellular processes , our work provides new fundamental insights into the architecture of the centriole , establishing a structural basis for centriole stability and the severe phenotypes that arise when lost .
Tubulin at 10 µM was incubated with a slight molar ratio excess of each protein construct ( around 15 µM ) in MES buffer for 15 min on ice . After centrifugation at 13 , 000 x g at 4°C for 20 min , the supernatant and the pellet were analyzed by Coomasie stained SDS-PAGE . For simple decoration , Taxol-stabilized microtubules were nucleated as described ( Schmidt-Cernohorska et al . , 2019 ) and subsequently exposed to recombinant WDR90-N ( 1-225 ) in a 1:1 molar ratio for 30 min at room temperature . 5 µL of protein complexes solution were blotted on carbon square 300 mesh grids ( EMS ) and stained with Uranyl Acetate ( 2% ) for 3 then 30 s . For double decoration , in vitro microtubules were incubated with WDR90-N ( 1-225 ) in a 1:1 molar ratio for 5 min at room temperature prior to addition of 2X free tubulin for 30 min at room temperature . Negatively stained grids were prepared as above . For cryo-microscopy , 4 µL of double decorated microtubule were deposited on a Lacey Carbon film grid ( 300 Mesh , EMS ) , blotted manually for 2 s and plunge into liquid ethane using an homemade plunger . Electron micrographs were acquired on a Tecnai G2 Sphera electron microscope ( FEI Company ) and analyzed using ImageJ . GFP-WDR90-N ( 1-225 ) RR and GFP-WDR90 ( FL ) RR were cloned in the Gateway compatible vector pEBTet-eGFP-GW . Previously generated RNAi-resistant WDR90 DNA ( Hamel et al . , 2017 ) was used as template for PCR amplification . In brief , inserts were first subcloned in pENTR-Age-AGT using the restriction sites AgeI and XbaI . Second , a Gateway reaction was performed to generate the final expression plasmids pEBTet-GFP-WDR90-N ( 1-225 ) RR and pEBTer-GFP-WDR90 ( FL ) RR , which were sequenced verified prior to transfection in human cells . For transient expression , U2OS cells were transfected using Lipofectamine 3000 ( Life Technology ) . Protein expression was induced using 1 µg/mL doxycycline for 48 hr and cells were processed for immunofluorescence analysis . Cloning of the GFP-WDR90 construct used in Figure 2 was done as follows: WDR90 was cloned by nested RT-PCR using total RNAs extracted from human RPE1 cells . Three different fragments corresponding to aa . 1–578 , 579–1138 , 1139–1748 of WDR90 ( based on Genebank sequence NP_660337 ) were amplified and cloned separately using the pCR Blunt II Topo system ( Thermo Fisher Scientific ) . The full coding sequence was then reconstituted in pCR Blunt II by two successive cloning steps using internal Nru I and Sal I , introduced in the PCR primers and designed in order not to modify WDR90 aa sequence . WDR90 coding sequence was then cloned into a modified pEGFP-C1 vector ( Clontech ) containing Asc I and Pac I restriction sites . Cloning of POC16 ( 1-295 ) into the pXLG vector was performed as followed: the POC16 sequence synthetized by GeneArt using the E . coli codon usage ( described in Hamel et al . , 2017 ) was cloned into pXLG vector using NotI and BamHI restriction sites . U2OS cells were plated onto coverslips in a 6-well plate at 200 000 cell/well 24 hr prior transfection . For POC5 depletion , cells were transfected with 20 nM silencer select negative control siRNA1 ( 4390843 , Thermo Fisher ) and siPOC5 ( sequence Sense siPOC5-1: 5’ CAACAAAUUCUAGUCAUACUU 3’ and antisense: 5’ GUAUGACUAGAAUUUGUUGCU 3’ , adapted from Azimzadeh et al . , 2009 ) using Lipofectamine RNAimax ( Thermo Fischer Scientific ) . Medium was changed 4 hr post-transfection and cells were analyzed 48 hr post-transfection . For WDR90 depletion , cells were transfected with 10 nM of silencer select negative control siRNA1 and silencer select pre-designed siRNA s47097 using INTERFERin siRNA transfection reagent ( Polyplus ) . After 48 hr , medium was changed and cells were analyzed 96 hr post-transfection . For WDR90/POC5 depletion , cells were transfected with 10 nM of silencer select negative control siRNA1 and silencer select pre-designed siRNA s47097 using INTERFERin siRNA transfection reagent ( polyplus ) . Medium was changed at 48 hr prior transfection and cells were subsequently transfected with 20 nM silencer select negative control siRNA1 and siPOC5 using INTERFERin siRNA transfection reagent ( Polyplus ) . Cells were analyzed 48 hr after the second transfection . In U2OS:GFP-WDR90 ( FL-RR ) stable cell line , RNA-resistant protein expression was induced constantly for 96 hr using 1 µg/mL doxycycline . Cells grown on a 15 mm glass coverslips ( Menzel Glaser ) were pre-extracted for 15 s in PBS supplemented with 0 . 5% triton prior to iced-cold methanol fixation for 7 min . Cells were washed in PBS then incubated for 1 hr in 1% bovine serum albumin ( BSA ) in PBS-T with primary antibodies against WDR90 ( 1:250 , rabbit polyclonal , NovusBio NBP2-31888 ) ( note that the WDR90 antibody also decorates the border of the cell , reminiscent to focal adhesion pattern ) , Centrin ( 1:500 , mouse monoclonal , clone 20H5 , 04–1624 , Merck Millipore ) , POC5 ( 1:500 , rabbit polyclonal , A303-341A , Bethyl ) HsSAS-6 ( 1:100 , mouse monoclonal , sc-81431 , Santa Cruz Biotechnology ) , PCM1 ( 1:500 , rabbit polyclonal , sc-67204 , Santa Cruz Biotechnology ) , CP110 ( 1:500 , rabbit polyclonal , 12780–1 , Proteintech ) , GFP ( 1:500 , mouse monoclonal , ab1218 , Abcam ) , mCherry ( 1:500 , rabbit polyclonal ) or tubulin ( 1:500 , mouse monocolonal , ab7291 , Abcam ) . Coverslips were washed in PBS for 30 min prior to incubation with secondary antibodies ( 1:1000 ) for 1 hr at room temperature , washed again for 30 min in PBS and mounted in DAPCO mounting medium containing DAPI ( Abcam ) . The following secondary antibodies were used: goat anti-rabbit Alexa Fluor 488 IgG H+L ( 1:400 , A11008 ) and goat anti-mouse Alexa Fluor 568 IgG H+L ( 1:250 , A11004 ) ( Invitrogen , ThermoFisher ) . Imaging was performed on a Zeiss LSM700 confocal microscope or on a Leica Thunder DMi8 microscope with a PlanApo 63x oil immersion objective ( NA 1 . 4 ) and optical sections were acquired every 0 . 33 µm , then projected together using ImageJ . The following reagents were used in U-ExM experiments: formaldehyde ( FA , 36 . 5–38% , F8775 , SIGMA ) , acrylamide ( AA , 40% , A4058 , SIGMA ) , N , N’-methylenbisacrylamide ( BIS , 2% , M1533 , SIGMA ) , sodium acrylate ( SA , 97–99% , 408220 , SIGMA ) , ammonium persulfate ( APS , 17874 , ThermoFisher ) , tetramethylethylendiamine ( TEMED , 17919 , ThermoFisher ) , nuclease-free water ( AM9937 , Ambion-ThermoFisher ) and poly-D-Lysine ( A3890401 , Gibco ) . Monomer solution ( MS ) for one gel is composed of 25 μl of SA ( stock solution at 38% ( w/w ) diluted with nuclease-free water ) , 12 . 5 μl of AA , 2 . 5 μl of BIS and 5 μl of 10X phosphate-buffered saline ( PBS ) . For isolated Chlamydomonas basal bodies ( Klena et al . , 2018 ) , U-ExM was performed as previously described ( Gambarotto et al . , 2019 ) . Briefly , coverslips were incubated in 1% AA + 0 . 7% FA diluted in 1X PBS ( 1X AA/FA ) for 5 hr at 37°C prior to gelation in MS supplemented with TEMED and APS ( final concentration of 0 . 5% ) for 1 hr at 37°C and denaturation for 30 min at 95°C . Specifically , gels were stained for 3 hr at 37°C with primary antibodies against tubulin monobody AA345 ( 1:500 , scFv-F2C , Alpha-tubulin ) ( Nizak et al . , 2003 ) and POC16 ( 1:100 ) ( Hamel et al . , 2017 ) or POB15 ( 1:100 ) ( Hamel et al . , 2017 ) diluted in 2% PBS/BSA . Gels were washed 3 × 10 min in PBS with 0 . 1% Tween 20 ( PBST ) prior to secondary antibodies incubation for 3 hr at 37°C and 3 × 10 min washes in PBST . Gels were expanded in 3 × 150 mL ddH20 before imaging . Human U2OS cells were grown on 12 mm coverslips and processed as previously described ( Le Guennec et al . , 2020 ) . Briefly , coverslips were incubated for 5 hr in 2% AA + 1 . 4% FA diluted in 1X PBS ( 2X AA/FA ) at 37°C prior to gelation in MS supplemented with TEMED and APS ( final concentration of 0 . 5% ) for 1 hr at 37°C . Denaturation was performed for 1h30 at 95°C and gels were stained as described above . The following primary antibodies were used: tubulin monobodies AA344 ( 1:250 , scFv-S11B , Beta-tubulin ) and AA345 ( 1:250 , scFv-F2C , Alpha-tubulin ) ( Nizak et al . , 2003 ) , rabbit polyclonal anti-POC1B ( 1:250 , PA5-24495 , ThermoFisher ) , rabbit polyclonal anti-POC5 ( 1:250 , A303-341A , Bethyl ) , rabbit polyclonal anti-FAM161A ( 1:250 ) ( Le Guennec et al . , 2020 ) , mouse monoclonal anti-Centrin ( 1:250 , clone 20H5 , 04–1624 , Merck Millipore ) , rabbit polyclonal anti-CEP135 ( 1:250 , 24428–1-AP , Proteintech ) , rabbit polyclonal anti-PolyE ( 1:500 , AG-25B-0030 , AdipoGen ) , rabbit polyclonal anti-FGFR1OP ( FOP1 ) ( 1:250 , HPA071876 , Sigma Life Science ) , rabbit polyclonal anti-CEP164 ( 1:250 , 22227–1-AP , Proteintech ) rabbit polyclonal anti-WDR90 ( 1:100 , NovusBio NBP2-31888 ) . Specifically , as WDR90 staining is weak and dotty , incubation with anti-WDR90 antibodies was performed overnight at 37°C . Note that for the protein mapping in Figure 1 , the localisation of the proteins is relative to the epitopes detected by the antibodies used in this approach . The following secondary antibodies were used: goat anti-rabbit Alexa Fluor 488 IgG H+L ( 1:400 , A11008 ) and goat anti-mouse Alexa Fluor 568 IgG H+L ( 1:250 , A11004 ) ( Invitrogen , ThermoFisher ) . For each gel , a caliper was used to accurately measure its expanded size ( Exsize in mm ) . The gel expansion factor ( X factor ) was obtained by dividing Exsize by 12 mm , which corresponds to the size of the coverslips use for sample seeding . Thus , X factor = Exsize ( mm ) /12 ( mm ) . The table below shows the Exsize and X factor for all the gels used in this study . GelsiControl Exsize ( X factor ) siWDR90 Exsize ( X factor ) POC1B ( n = 1 ) 53 mm ( 4 . 42 ) 52 mm ( 4 . 33 ) POC1B ( n = 2 ) 49 mm ( 4 . 08 ) 50 . 5 mm ( 4 . 21 ) POC1B ( n = 3 ) 50 . 5 mm ( 4 . 21 ) 50 . 5 mm ( 4 . 21 ) FAM161A ( n = 1 ) 50 mm ( 4 . 16 ) 50 mm ( 4 . 16 ) FAM161A ( n = 2 ) 50 mm ( 4 . 16 ) 51 mm ( 4 . 25 ) FAM161A ( n = 3 ) 50 mm ( 4 . 16 ) 50 mm ( 4 . 16 ) POC5 ( n = 1 ) 51 mm ( 4 . 25 ) 50 . 5 mm ( 4 . 21 ) POC5 ( n = 2 ) 50 mm ( 4 . 16 ) 50 mm ( 4 . 16 ) POC5 ( n = 3 ) 50 . 5 mm ( 4 . 21 ) 49 mm ( 4 . 08 ) Centrin ( n = 1 ) 50 mm ( 4 . 16 ) 50 mm ( 4 . 16 ) Centrin ( n = 2 ) 50 mm ( 4 . 16 ) 50 mm ( 4 . 16 ) Centrin ( n = 3 ) 49 mm ( 4 . 08 ) 49 mm ( 4 . 08 ) Pieces of gels were mounted on 24 mm round precision coverslips ( 1 . 5H , 0117640 , Marienfeld ) coated with poly-D-lysine for imaging . Image acquisition was performed on an inverted Leica TCS SP8 microscope or on a Leica Thunder DMi8 microscope using a 63 × 1 . 4 NA oil objective with Lightening or Thunder SVCC ( small volume computational clearing ) mode at max resolution , adaptive as ‘Strategy’ and water as ‘Mounting medium’ to generate deconvolved images . 3D stacks were acquired with 0 . 12 µm z-intervals and an x , y pixel size of 35 nm . For centrioles counting , immunofluorescences were analyzed on a Leica epifluorescence microscope or on a Leica Thunder DMi8 microscope . For fluorescence intensity , maximal projections were used using Fiji ( Schindelin et al . , 2012 ) . Confocal centrosomal intensities were assessed using an area of 20 pixels on Fiji . For each experiment , control values were averaged and all individual measures for control and treated conditions were normalized accordingly to obtain the relative intensity ( A . U . ) . Normalized individual values were plotted on GraphPadPrism7 . Confocal centriolar intensities were assessed by individual plot profil ( 25 points ) on each pair of mature centrioles . For each experiment , the average ( Av ) of control values was calculated and all individual measures for control and treated conditions were normalized on Av to obtain the relative intensity ( A . U . ) . An average of all normalized measures was generated and plotted in GraphPadPrism7 . For U-ExM data , length coverage quantification was performed as previously published in Le Guennec et al . , 2020 . For top views , a measurement from the exterior to the interior of the centriole was performed on each microtubule triplet displaying a resolved signal for both tubulin and the core protein . For each tubulin measurement , the position ( x-value ) of the maximal fluorescence intensity of the core protein was aligned individually to the position of the respective tubulin maximal intensity . All individual values of distance were plotted and analyzed in GraphPadPrism7 . Measurements of diameter in siControl and siWDR90 conditions were performed on S-phase mature centrioles imaged in lateral view . Briefly , lines of 50 pixels thickness were drawn within the proximal , central and distal regions defined in respect with the position of inner core proteins POC5 and FAM161A . Proximal region was then defined as the portion of the centriole below staining of POC5 or FAM161A and the distal region as above . In the siWDR90 condition , proximal region was defined as below the remaining belt of POC5 of FAM161A , the core region was measured just above the remaining belt and the distal region as the last 100 nm of the centriole . The Fiji plot profile tool was used to obtain the fluorescence intensity profile from proximal to distal for tubulin and the core protein from the same line scan . Roundness was calculated on perfectly imaged top views of centrioles by connecting tubulin peaks on ImageJ . No statistical method was used to estimate sample size . The comparison of two groups was performed using a two-sided Student's t-test or its non-parametric correspondent , the Mann-Whitney test , if normality was not granted because rejected by Pearson test The comparisons of more than two groups were made using one- or two-way ANOVAs followed by post-hoc tests ( Holm Sidak’s multiple comparisons ) to identify all the significant group differences . N indicates independent biological replicates from distinct samples . Every experiment was performed at least three times independently on different biological samples unless specified . Data are all represented as scatter or aligned dot plot with centerline as mean , except for percentages quantifications , which are represented as histogram bars . The graphs with error bars indicate SD ( +/- ) and the significance level is denoted as usual ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 ) . All the statistical analyses were performed using Excel or Prism7 ( Graphpad version 7 . 0a , April 2 , 2016 ) . | Cells are made up of compartments called organelles that perform specific roles . A cylindrical organelle called the centriole is important for a number of cellular processes , ranging from cell division to movement and signaling . Each centriole contains nine blades made up of protein filaments called microtubules , which link together to form a cylinder . This well-known structure can be found in a variety of different species . Yet , it is unclear how centrioles are able to maintain this stable architecture whilst carrying out their various different cell roles . In early 2020 , a group of researchers discovered a scaffold protein at the center of centrioles that helps keep the microtubule blades stable . Further investigation suggested that another protein called WDR90 may also help centrioles sustain their cylindrical shape . However , the exact role of this protein was poorly understood . To determine the role of WDR90 , Steib et al . – including many of the researchers involved in the 2020 study – used a method called Ultrastructure Expansion Microscopy to precisely locate the WDR90 protein in centrioles . This revealed that WDR90 is located on the microtubule wall of centrioles in green algae and human cells grown in the lab . Further experiments showed that the protein binds directly to microtubules and that removing WDR90 from human cells causes centrioles to lose their scaffold proteins and develop structural defects . This investigation provides fundamental insights into the structure and stability of centrioles . It shows that single proteins are key components in supporting the structural integrity of organelles and shaping their overall architecture . Furthermore , these findings demonstrate how ultrastructure expansion microscopy can be used to determine the role of individual proteins within a complex structure . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2020 | WDR90 is a centriolar microtubule wall protein important for centriole architecture integrity |
Although high levels of 5-hydroxymethylcytosine ( 5hmC ) accumulate in mammalian neurons , our knowledge of its roles in terminal differentiation or as an intermediate in active DNA demethylation is incomplete . We report high-resolution mapping of DNA methylation and hydroxymethylation , chromatin accessibility , and histone marks in developing postmitotic Purkinje cells ( PCs ) in Mus musculus . Our data reveal new relationships between PC transcriptional and epigenetic programs , and identify a class of genes that lose both 5-methylcytosine ( 5mC ) and 5hmC during terminal differentiation . Deletion of the 5hmC writers Tet1 , Tet2 , and Tet3 from postmitotic PCs prevents loss of 5mC and 5hmC in regulatory domains and gene bodies , and hinders transcriptional and epigenetic developmental transitions . Our data demonstrate that Tet-mediated active DNA demethylation occurs in vivo , and that acquisition of the precise molecular properties of adult PCs require continued oxidation of 5mC to 5hmC during the final phases of differentiation .
Development of the mammalian brain requires generation of hundreds of millions of neuronal progenitors that differentiate into distinct cell types with refined functional properties . Although morphological and physiological maturation of most neurons occurs between mid-gestation and a few months or years after birth , the vast majority of CNS neurons must maintain a stable differentiated state for the life of the organism and remain sufficiently plastic to participate in novel behaviors . Studies of signaling molecules and transcriptional programs have identified many mechanisms that orchestrate critical steps in neurogenesis , cell type diversification , neuronal migration , axonal pathfinding , and differentiation . Although it has been established that epigenetic regulatory mechanisms are critical for proper development of all cell types , our knowledge of the precise roles of these mechanisms in neuronal differentiation , function , and vitality remains rudimentary . 5-Hydroxymethylcytosine ( 5hmC ) is produced from 5-methylcytosine ( 5mC ) by the Ten-eleven translocation dioxygenases ( Tet1 , Tet2 , Tet3 ) ( Iyer et al . , 2009; Tahiliani et al . , 2009 ) . It is present at approximately 10-fold higher levels in neurons than peripheral cell types ( Globisch et al . , 2010; Kriaucionis and Heintz , 2009 ) , and its distribution across the genome of adult neurons is cell specific and correlated with active gene expression ( Mellén et al . , 2017; Mellén et al . , 2012; Szulwach et al . , 2011 ) . The initial discovery that 5hmC accumulates within active gene bodies , coupled with the discovery that MeCP2 can bind probes containing 5hmC with high affinity , led to the proposal that MeCP2 binding within active genes facilitates their expression ( Mellén et al . , 2012 ) . Further studies demonstrating that MeCP2 binds 5hmC at high affinity in non-CG dinucleotides but does not bind to 5hmCG overturned this model ( Ayata , 2013; Gabel et al . , 2015; Mellén et al . , 2017 ) by revealing that accumulation of 5hmCG and the depletion of 5hmCH in gene bodies is correlated with less MeCP2 binding and increased expression ( Ayata , 2013; Gabel et al . , 2015; Mellén et al . , 2017 ) . Imaging of the binding and diffusion of single MeCP2 molecules in living neurons lacking Dnmt3a or Tet1 , Tet2 , and Tet3 demonstrate that its binding is exquisitely sensitive to the levels of both 5mC and 5hmC ( Piccolo et al . , 2019 ) . Given these data and the sensitivity of neuronal function to MeCP2 gene dosage ( Chahrour et al . , 2008; Nan and Bird , 2001 ) , relief of the repressive functions of MeCP2 through Tet-mediated conversion of high-affinity 5mCG-binding sites to low-affinity 5hmCG sites , which we have referred to as functional demethylation ( Mellén et al . , 2017 ) , provides an important mechanism for modulation of chromatin structure and transcription . In dividing cells , 5hmC serves as an intermediate in DNA demethylation because maintenance DNA methyltransferases do not recognize hemi-hydroxymethylated cytosines in order to reestablish methylation ( Wu and Zhang , 2017 ) . Consequently , 5hmC is lost passively and replaced by C due to replicative dilution . Loss of function studies in mouse embryonic stem cells ( ESCs ) ( Dawlaty et al . , 2013 ) and lymphocyte lineages Lio and Rao , 2019 have demonstrated that the role of Tet-mediated replicative DNA demethylation is to provide full accessibility to regulatory regions necessary for expression of genes required for differentiation . For example , at the activation-induced deaminase ( AID ) locus in B cells , Tet activity is required for demethylation and activation of enhancer regions that modulate AID expression to enable class switch recombination ( Lio et al . , 2019 ) . Tet-mediated replication-dependent passive demethylation is thought to be common in many dividing cell types , including progenitor cells in the developing nervous system ( MacArthur and Dawlaty , 2021 ) . Most neurons exit the cell cycle during mid-gestation and remain relatively simple and undifferentiated until birth . Following parturition , they initiate an elaborate program of differentiation that includes dramatic increases in size , morphological complexity , and connectivity . Since these neurons are postmitotic and remain so throughout life , removal of the repressive effects of 5mC by passive demethylation cannot occur . However , in addition to functional demethylation and passive demethylation , a third pathway for DNA demethylation , often referred to as active demethylation , has been proposed based on the finding that 5hmC can be further oxidized by Tet proteins to produce 5-formylcytosine ( 5fC ) and 5-carboxylcytosine ( 5caC ) . Removal of 5fC and 5caC by thymine DNA glycosylase ( TDG ) -dependent base excision repair ( BER ) provides another mechanism for 5hmC-mediated DNA demethylation ( He et al . , 2011 ) . Clear evidence that this pathway can operate in cultured cells has been presented , and detailed biochemical studies have delineated the mechanisms operating in TDG-BER DNA demethylation ( Weber et al . , 2016 ) . Active DNA demethylation is ideally suited for remodeling DNA methylation in postmitotic neurons . Evidence that continued accumulation of 5hmC is required in differentiating neurons and that it can participate in active demethylation in vivo is beginning to emerge . In studies of cerebellar granule cell development in ESC-derived GC cultures , primary GC cultures , and slice preparations , manipulations of Tet activity and 5hmC levels provide strong evidence that 5hmC is required for expression of axon guidance and ion channel genes , and that proper development of the GC dendritic arbor requires 5hmC ( Zhu et al . , 2016 ) . Demethylation of 5hmC-GFP transfected DNA fragments retrieved after several days in culture display reduced 5hmC levels at several sites as assessed by bisulfite sequencing ( BSSeq ) in HEK 293 cells and primary hippocampal neurons ( Guo et al . , 2011 ) . Changes in methylation levels at GGCC sites in the genome of hippocampal dentate gyrus neurons in response to electroconvulsive shock have been documented using the methylation-sensitive cut counting method that employs the methylation or hydroxymethylation-sensitive restriction enzyme HpaII and its methylation-insensitive isoschizomer Msp1 ( Guo et al . , 2011 ) . Global measurements of 5mC and 5hmC by mass spectroscopy in the hippocampus have shown that both decrease in response to the induction of seizure ( Kaas et al . , 2013 ) . And in germline Tet1 knockout mice , analysis of the Npas4 and c-Fos promoters has shown that Tet1 is required for their proper regulation in both the cerebral cortex and hippocampus ( Rudenko et al . , 2013 ) . Despite these observations , our knowledge of the roles of continued 5hmC accumulation in active DNA demethylation and postmitotic differentiation remains rudimentary . We report here single nucleotide resolution studies of DNA methylation and hydroxymethylation , transcription , chromatin accessibility , and H3K4me3 and H3K27me3 histone marks in postmitotic , differentiating Purkinje cells ( PCs ) . As PCs transition from relatively small , multipolar immature cells to fully elaborated large neurons with complex dendritic arbors and hundreds of thousands of synapses ( McKay and Turner , 2005 ) , 5hmC continues to accumulate , and DNA methylation and hydroxymethylation are reconfigured as epigenetic and transcriptional programs progress . Our data confirm previous studies of the relationships between transcription , DNA methylation , DNA hydroxymethylation , and chromatin organization ( Lister et al . , 2013; Mellén et al . , 2017; Tsagaratou et al . , 2017; Tsagaratou et al . , 2014 ) , and they reveal several novel modes of transcriptional activation and repression . Notably , we identify a class of developmentally induced PC-specific genes that are highly expressed and lose both 5mC and 5hmC in the final stages of PC differentiation . We report also studies of newly generated Tet1 , Tet2 , Tet3 PC-specific triple knockout ( Pcp2TetTKO ) mouse lines in which recombination is activated in the first postnatal week . These data demonstrate that postmitotic transcriptional and epigenetic maturation in PCs , including transcription of many ion channels and active demethylation of late expressed genes , requires continued oxidation of 5mC to 5hmC . Taken together , our data demonstrate that active demethylation occurs in select genes in postmitotic neurons , and that 5hmC plays an essential role in refining the transcriptional and epigenetic status of PCs during the final stages of differentiation .
PC progenitors complete their final divisions in the cerebellar primordium between e11 and e13 ( Butts et al . , 2014 ) and remain as a multilayered , simple migrating cell population until birth ( P0 ) . In the first postnatal week , they organize into a monolayer as the cerebellum enlarges and begin the transition from a multipolar primitive neuronal morphology to one of the largest neurons in the brain with a characteristic , highly elaborate planar dendritic arbor . As they begin the second postnatal week ( P7 ) , PCs undergo an important developmental transition that includes refinement of their climbing fiber input , formation of many thousands of parallel fiber synapses , and myelination of their axons ( Leto et al . , 2016; McKay and Turner , 2005 ) . The tremendous increase in synaptogenesis and connectivity continues until PCs attain their mature morphology and functions at 3–4 weeks of age . As these programs unfold , there is a global decrease of 5mCG and global increase of 5hmCG ( Figure 2—figure supplement 1A ) . To identify genes whose transcription is stable and those that are dynamically regulated during PC differentiation , we conducted differential expression analysis between P0 and adult Purkinje neurons , filtering for significance of p < 0 . 01 and log2 fold change of >2 in either direction ( Figure 2A–B , Supplementary file 1 ) . Using these criteria , there were 922 developmentally repressed genes ( down-regulated in adult ) enriched in categories involved in cell signaling and axon guidance pathways ( e . g . Grid1 , Pde1c; Figure 1F ) . We identified 432 genes with increased expression as PCs differentiate encoding proteins known to be important for the mature functions of PCs including calcium ion buffering and transport , regulation of the inositol triphosphate signaling pathway , and RNA splicing . As expected , similar analysis comparing P0 and P7 or P7 and adult data revealed genes overlapping with those identified in the overall P0 and adult comparative data ( Figure 2—figure supplement 1B-D ) , although additional categories of RNA metabolism , synaptic signaling , and ion transport are enriched in the P7 to adult analysis ( Figure 2—figure supplement 1C ) . As anticipated , the P7 profiles we have included in our analysis have been essential in assessing both the developmental course of epigenetic events studied here and the consequences of loss of 5hmC in PCs ( see below ) . Despite the many genes that are dynamically regulated during these developmental transitions , the majority of genes in PCs are either silent or constitutively expressed . Approximately half of the genes in the mouse genome ( e . g . genes encoding olfactory receptors ( Figure 1F ) , immunoglobulins , hemoglobin subunits , etc . ) are never expressed in PCs and remain heavily methylated and inaccessible . As previously reported for other neurons ( Mo et al . , 2015 ) , there is a second class of silent genes in PCs that are enriched in transcription factors expressed in very early embryos ( Hox clusters , other homeobox genes , etc . ) . These are completely demethylated and enriched for the H3K27me3 histone mark , indicating that they are repressed by the polycomb repressive complex ( Li et al . , 2018 ) . Actively transcribed genes whose levels vary little during PC differentiation , for example the calcium channel auxiliary subunit Cacnb4 ( Figure 1F ) , are also epigenetically stable . This is a large ( ~3000 ) and diverse class of genes that carry activating epigenetic marks that have been previously characterized: low levels of 5mCG , 5mCH , and 5hmCH; elevated levels of 5hmCG; they are ATAC accessible; and their promoters carry the activating histone mark H3K4me3 . As expected , the levels of these marks vary widely between genes and generally reflect the level of expression . As PCs mature , the conversion of 5mC to 5hmC continues to increase within the gene bodies of the most active constitutively expressed genes without a strong impact on expression ( Figure 2—figure supplement 1E ) . To identify mechanisms associated with developmental repression , we analyzed features previously associated negatively with transcription in the 922 genes whose expression decreases during PC differentiation . A consistent finding is that in most genes decreased expression is associated with a loss of promoter accessibility as assayed by assay for transposase-accessible chromatin sequencing ( ATACSeq ) ( Figure 2D and K–L , Figure 2—figure supplement 1E ) . These changes are not correlated with changes in DNA methylation or H3K27me3 occupancy ( Figure 2C , Figure 2—figure supplement 1E ) . Furthermore , 5hmCG accumulated over the genes bodies early in development is stable or continues to increase ( Figure 2C ) , suggesting that the presence of 5hmCG is not sufficient to maintain transcription . Interestingly , analysis of the repressive marks 5mCH and 5hmCH reveals at least two distinctly recognizable epigenetic patterns in genes that become repressed in PC development ( Figure 2F–J , Figure 2—figure supplement 1F-G ) . Repressed genes that accumulate 5mCH and 5hmCH over their gene bodies ( referred to as ‘highCHmod’ genes ) are associated with lower promoter accessibility ( Figure 2I ) and enhanced accumulation ( Figure 2H ) of H3K27me3 relative to those repressed genes with low levels of CH modification ( referred to as ‘lowCHmod’ genes ) ( Figure 2F–J , Figure 2—figure supplement 1F-G ) . Gene ontology ( GO ) analysis indicates that both sets of genes that are repressed as PC differentiation proceeds are associated with development , and that the genes that do not accumulate 5mCH and 5hmCH encode preferentially proteins related to cell morphogenesis and axonal projection ( Figure 2J ) . Although the data for highCHmod genes is consistent with transcriptional inhibition through both polycomb repressive complexes ( Li et al . , 2018 ) and MeCP2 binding ( Gabel et al . , 2015 ) , our data do not identify the mechanisms responsible for repression of the lowCHmod gene class that fail to accumulate repressive CH marks over the gene body as they mature . To discover the mode of repression for these genes will require close examination of intergenic regulatory domains and associated transcriptional repressors and epigenetic marks . We identified 432 genes that increase in expression as PCs differentiate ( Figure 2A , Supplementary file 1 ) . In general , these genes encode proteins known to be important for the mature functions of PCs , including calcium ion buffering and transport , regulation of the inositol triphosphate signaling pathway , and RNA splicing ( Figure 2B ) . The chromatin landscape of most genes whose expression increases during PC differentiation is similar to those that are constitutively transcribed ( e . g . Atp2a3 , Cep76; Figure 1G ) . Their promoters are highly accessible ( Figure 2D and K–L ) , they have high levels of H3K4me3 ( Figure 2E ) , their gene bodies have low levels of 5mCG and elevated levels of 5hmCG ( Figure 2C ) , and they do not accumulate 5mCH or 5hmCH ( Figure 2N ) . Our data reveal additional , surprising features that have not been documented in postmitotic cells . Thus , in a small subset of highly expressed genes ( e . g . Cep76 , Itpr1 , Mtss1 ) there is a profound loss of both 5hmCG and 5mCG during the terminal , postmitotic stage of PC differentiation ( Figures 1F and 3 ) . To investigate the apparent demethylation over this class of genes , we computationally identified DNA methylation valleys ( DMVs ) as previously described ( Jeong et al . , 2014; Mo et al . , 2015; Xie et al . , 2013 ) . In brief , DMVs were characterized by filtering undermethylated regions ( UMRs ) for length ( >5 kb ) and merging any regions within 1 kb ( Burger et al . , 2013; Figure 3—figure supplement 1A-B , Supplementary file 2 ) . This revealed multiple regions longer than 15 kb , an unusual finding given that the average length of DMVs is ~5 kb ( Figure 3—figure supplement 1D-E; Jeong et al . , 2014 ) . As expected from previous studies , a small number of these genes are inactive , fully demethylated , inaccessible , and covered by high levels of H3K27me3 repressive marks ( Figure 3A , Foxd1 , Figure 3—figure supplement 1C ) . These genes acquire their characteristics early in development , and their epigenetic features are stable . The majority of very large DMVs in PCs , however , arise during differentiation through loss of 5mC and 5hmC over active , highly expressed genes as PCs mature ( Figure 3 , Itpr1 , Mtss1 ) . In these genes , the timing of changes in DNA methylation , chromatin accessibility , and dynamic histone marks varies from gene to gene , but the overall progression of events associated with their expression is similar ( Figure 3 ) . At birth , these genes are actively transcribed ( Figure 3A ) , 5hmCG is enriched in the gene body ( Figure 3B ) and 5mCG is depleted ( Figure 3C ) , their promoters are accessible ( Figure 3D ) , and H3K4me3 is present over their promoters ( Figure 3E ) . These epigenetic properties resemble moderately expressed genes in many cell types . As differentiation proceeds and transcription increases , there is a gradual decrease in both 5mCG , 5hmCG , and H3K27me3 histone marks , and ATAC accessibility and H4K4me3 marks spread from the promoter into the gene body . To further support these data , we have included analysis of independent BSSeq experiments that detect the combined levels of 5mC and 5hmC ( Figure 3A , BSSeq ) . These data also reveal that hypo-DMRs ( differentially methylated regions ) present at P0 expand and fuse as PCs lose both methylation and hydroxymethylation to form the very large DMRs that are characteristics of these genes ( Figure 3A , hypo-DMR track ) . As shown in Figure 3B–E and Figure 3—figure supplement 1I-K , quantitation of these features demonstrates that that loss of DNA methylation and hydroxymethylation occurs over the entire gene body for these genes as H3K4me3 activating histone marks accumulate and ATAC accessibility increases . Interestingly , many of the active genes associated with broad DMVs are PC-specific and highly expressed ( Figure 3—figure supplement 1F ) . GO analysis indicates that they are involved in the inositol triphosphate/calcium signaling pathways , and a subset is associated with ataxia and autism ( Figure 3—figure supplement 1H ) . They do not accumulate modified cytosines in CpH context ( Figure 3—figure supplement 1G ) . These data provide strong evidence that DNA demethylation can occur in postmitotic neurons , and that it is enhanced in a specific class of genes that are very highly expressed and functionally important . Although our mass spectroscopic analysis of genomic DNA from differentiating PCs ( Figure 3—figure supplement 1L ) failed to detect 5fC or 5caC , their involvement as transient intermediates in the loss of 5mC and 5hmC cannot be ruled out because the small fraction of the genome covered by this gene class may preclude their detection ( Guo et al . , 2011; He et al . , 2011; Ito et al . , 2011 ) . Previous studies of enhancers and other regulatory sites have established that their activation is accompanied by increased ATAC accessibility and loss of DNA methylation ( Lio et al . , 2016 ) . To determine whether loss of DNA methylation can occur in putative regulatory sites in addition to large DMRs in postmitotic PCs , we used an established computational method ( Burger et al . , 2013 ) to identify small regions with statistically significant ATACSeq signal enrichment ( Figure 2K–N ) . We then employed differential accessibility analysis to divide those regions into two groups based the magnitude ( log2 fold change >4 ) and significance ( p < 0 . 01 ) of their changes during PC differentiation . Those that became more accessible between P0 and adult ( log2 fold change >4 ) experienced a ‘gain’ and those whose accessibility diminished during this time are characterized by a ‘loss’ . As anticipated , those ATAC peaks that gain accessibility lose both 5hmC and 5mC as cells progress from P0 to adult ( Figure 2N , Gain section ) whereas no significant changes in methylation at ATAC peak centers occur in sites that lose accessibility ( Figure 2N , Loss section ) . Although these ATAC peaks have not been identified as active enhancers , it is noteworthy that the transcription factor motifs found in these different classes of ATAC sites are distinct ( Figure 2—figure supplement 1H ) . These findings are consistent with prior studies indicating that enhancer activation is accompanied by enhanced ATAC accessibility and DNA demethylation ( Lio et al . , 2016 ) . In this case , however , DNA demethylation does not require cell division . Despite convincing evidence that Tet-mediated active DNA demethylation can occur in mouse ESCs through the TDG-BER pathway ( He et al . , 2011; Weber et al . , 2016 ) , evidence that this can occur in vivo has been difficult to obtain because such a proof requires loss or complete inhibition of all three Tet oxidases in a single cell type after cells have exited the cell cycle permanently . To provide this evidence , we generated PC-specific Tet1/Tet2/Tet3 triple knockout ( Pcp2TetTKO ) mouse lines ( Figure 4A ) . We chose to drive Cre recombinase expression using an engineered Pcp2 BAC employed previously to generate accurate Cre driver lines ( Gong et al . , 2007 ) because the onset of expression of the Pcp2 gene and the corresponding BAC vector occurs approximately 1 week after birth as the cerebellum enters its terminal phase of development and maturation . The Pcp2Cre BAC was introduced by pronuclear injection into ova from females carrying floxed alleles of all three Tet genes ( Figure 4—figure supplement 1A-B ) yielding several founder lines . PCR analysis of the floxed regions of each Tet gene in purified PC genomic DNA , and RNASeq analysis confirmed the deletion of exons from all three TET proteins ( Figure 4—figure supplement 1C-D ) . The lines chosen for analysis displayed no gross motor phenotype as assessed by rotarod performance , survival was normal , and recombination activity was specific to PCs ( Figure 4B–D , Figure 4—figure supplement 2A-D ) . The deletion of the Tet proteins did not affect the expression of Itpr1 significantly , and we confirmed that it could still be used for purification of PCs ( Figure 4—figure supplement 1E-F ) . Evaluation of the quality control metrics of the datasets showed strong correlation coefficients and enrichment for Purkinje markers ( Figure 4—figure supplement 1G-H ) . Cre recombinase activity , and thus recombination in the Tet genes , began at approximately P7 more than 2 weeks after PCs have completed their last division ( Figure 4—figure supplement 2A-D ) . These lines , therefore , allow assessment of the consequences of loss of Tet activity during the rapid phase of somatic and dendritic growth in postmitotic , differentiating Purkinje neurons . Loss of function studies of Tet1 , Tet2 , Tet3 and combinations thereof in ESCs and lymphocyte lineages have established that 5hmC plays a direct role in transcriptional regulation as a consequence of Tet-mediated replication-dependent loss of 5mC in enhancers , promoters , and gene bodies . To determine whether disruption of ongoing 5hmC accumulation in postmitotic PCs results in altered transcriptional regulation despite the inability to utilize passive DNA demethylation as a regulatory mechanism , we compared the transcriptome of adult Pcp2TetTKO mice with floxed Cre-negative littermates ( WT ) . Using p < 0 . 05 , log2 fold change of >0 . 5 in either direction to filter these data , we identified 721 genes whose transcription increases in the Pcp2TetTKO PCs and 548 genes with decreased levels of expression ( Figure 4E , Supplementary file 3 , Figure 4—figure supplement 2E-F ) . The large number of genes whose expression is impacted in the Pcp2TetTKO PCs is reminiscent of magnitude of changes documented in neurons in response to loss of MeCP2 ( Chahrour et al . , 2008 ) and other proteins involved nuclear organization and chromatin structure ( Hyun et al . , 2017 ) . GO analysis of these two classes of genes revealed differences in the biological categories represented ( Figure 4F ) . It is noteworthy that genes whose expression increases in the Pcp2TetTKO PCs are enriched for functions involved in terminal developmental events such as cell projection organization , synaptic transmission , and neuron differentiation . These include many ion channels , and receptors that should impact the fine-tuned physiology of PCs ( Figure 4E–F , dark blue dots , Supplementary file 3 ) , although given the complex firing patterns and action potential ( AP ) waveforms characteristic of PCs , it is difficult to predict the functional consequences of these changes . In contrast , genes whose expression is less in Pcp2TetTKO PCs are primarily involved in homeostasis and metabolic control . These are enriched in DNA transmembrane receptors and nucleic acid-binding proteins that may help regulate PC metabolism in response to external signals . To assess the loss of 5mC and 5hmC in differentiating PCs reflects a direct role of TET proteins in transcriptional regulation as a consequence of loss of 5mC in enhancers , promoters , and gene bodies , we first identified candidate regulatory domains by comparative analysis of P0 and adult ATACSeq data to detect changes in chromatin accessibility that occur during PC differentiation . We then focused on those regions associated with genes that are transcriptionally impacted in the Pcp2TetTKO . As shown in Figure 4G , PC regulatory sites that become more accessible as differentiation proceeds lose both 5mCG and 5hmCG between P7 and adult whether their expression increases or decreases in the Pcp2TetTKO cells . The increased 5mCG and 5hmCG evident over these sites in the Pcp2TetTKO relative to WT PCs indicates clearly that loss of both 5mCG and 5hmCG at these sites requires continued Tet oxidase activity . The requirement for continued 5hmC production in DNA demethylation is particularly evident from analysis of those very highly expressed genes that lose both 5mCG and 5hmCG over the gene body as differentiation proceeds ( Figure 3 ) . Genes with very large DMVs that are acquired early in development , for example , Hox cluster genes and Calb1 , are not impacted in the Pcp2TetTKO PCs ( Figure 4H , I ) . This is expected because DNA demethylation for these genes is evident before recombination and loss of Tet oxidase function occurs in the Pcp2TetTKO lines at approximately postnatal day 7 ( Figure 4—figure supplement 2A-D ) . In contrast , for those genes that acquire large DMVs between P7 and adult , for example Gpr63 and Grid2 , the loss of DNA methylation that is required for formation of the DMVs is strongly decreased ( Figure 4H , I ) . As shown in the metagene plots of DNA displaying 5mCG and 5hmCG levels of this class of genes over the gene body ( Figure 4I , right panel ) , the loss of both 5mCG and 5hmCG that occurs between P7 and adult is strongly impacted . Thus , the low levels of 5hmCG and 5mCG characteristic of this class of genes is not attained in the absence of Tet oxidase activity . Formation of large DMVs in PC genes differs in detail depending on the timing and rate of transcription and the size of the gene . Gpr63 is strongly activated during postnatal life to become one of the most actively transcribed genes in adult PCs and it becomes nearly completely demethylated over the entire 46 kb of its gene body . Grid2 is a very long gene ( ~1 . 44 Mb ) whose transcription also increases as differentiation proceeds . The Grid2 DMV also fails to develop fully in the Pcp2TetTKO ( Figure 4H ) . In this case , the large DMV develops over only the promoter and initial 5’ region of the gene . In these two cases , and in other genes of this class , the growth of the DMV is preceded by the accumulation of 5hmC , DNA demethylation is initiated at the 5’ end of the gene , and it spreads toward the 3’ end as differentiation proceeds ( Figure 4H ) . To understand better this variation in DNA demethylation in individual genes , we calculated the difference in the length of large DMVs over each gene in this class in the WT and Pcp2TetTKO PCs , and plotted these values as the negative length change ( Figure 4—figure supplement 2G , H ) . These data revealed a large variation in the length of the DMV that does not form in the Pcp2TetTKO PCs . We believe this reflects a complex relationship between the timing at which transcription is initiated , the rate of transcription over the entire gene body , the local activity of Tet oxidase within the gene as differentiation proceeds , and the timing of Tet oxidase loss from each cell following recombination at the Tet1 , Tet2 , and Tet3 loci . Taken together , these data demonstrate that loss of DNA methylation occurs over specific subsets of regulatory sites and transcription units in postmitotic Purkinje neurons . This loss of 5mCG requires continued production of 5hmC by Tet1 , Tet2 , and Tet3 . The simplest interpretation of these data is that Tet-mediated active demethylation can occur in neurons as a result of continued oxidation of 5hmC to 5fC and 5caC , followed by their removal through the BER pathway .
The data we have reported here advances our understanding of three major functions for 5hmC in the nervous system . Based on extensive studies of the requirements for Tet-mediated replication-dependent passive DNA demethylation in mESCs , the maturation of the germline , and developing lymphocyte lineages , it is probable that 5hmC is required to provide accessibility to important regulatory sites as neuronal progenitors exit the cell cycle and begin differentiation . The lack of DNA methylation we have observed at P0 over transcription factor genes ( Lhx5 , Lhx1 , Ldb ) that are expressed immediately after PCs exit from the cell cycle suggests that this may result from passive demethylation in dividing progenitors . Further support comes from the observation that DNA demethylation has been observed in comparisons of 5mC and 5hmC levels in the frontal cortex of fetal versus adult mouse and human brains , and the finding that a fraction of these loci retain their methylation status in Tet2-/- mice ( Lister et al . , 2013 ) . A second function discovered in mature neurons involves stable accumulation of 5hmCG within active genes which helps to reverse the repressive effects of MeCP2 in a process we refer to as functional demethylation ( Mellén et al . , 2017 ) . The finding that elevated 5hmC remains within genes that are repressed during PC differentiation ( Figure 2C ) adds to this model the fact that functional demethylation is not sufficient to maintain gene expression . Finally , we report here that Tet-mediated active DNA demethylation is required for proper expression of a subset of highly expressed PC-specific genes , adding a third important function for 5hmC in the nervous system . These findings place additional emphasis on investigation of ongoing functions of the TDG-BER pathway as elements of normal neuronal development rather than responses to accumulating DNA damage . Given these data and recent studies linking Tet mutations to neurodegenerative disease ( Cochran et al . , 2020; Marshall et al . , 2020 ) , it is evident that further exploration of each of these functions in the context of development , aging , and degeneration will continue enhance our understanding of 5hmC and the brain .
Wild-type C57BL6/J ( RRID:IMSR_JAX:000664 ) were obtained from Jackson Laboratories . Animals were maintained on a 12 hr light/12 hr dark cycle with food and water ad libitum . Animal protocols were approved by the Rockefeller University Institutional Animal Care and Use Committee , in accordance with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals . Tet1fl/fl::Tet2fl/fl::Tet3fl/fl animals were a gift from Anjana Rao . We used a previously characterized Pcp2-CRE BAC ( RP24-186D18 ) construct from the GENSAT project ( Gong et al . , 2007; Gong et al . , 2003 ) to create a conditional TKO in PCs . Pronuclear injections of the BAC construct into zygotes were performed at the Transgenic and Reproductive Technology Center at the Rockefeller University . Mice were decapitated at P0 and P7 , and the brains were dissected , and immersion fixed in 4% formaldehyde ( w/v ) overnight at 4°C . Adult mice were deeply anesthetized and then the brains were fixed by transcardiac perfusion with PBS followed by 4% formaldehyde . Brains were further fixed by immersion fixation in 4% formaldehyde overnight at 4°C . All brains were cryoprotected in 30% sucrose in PBS , embedded in OCT , and cut with a Leica CM3050 S cryostat into 20 µm sections . The sections were immediately mounted on slides and stored at –20°C . Antigen retrieval using sodium citrate buffer ( 10 mM sodium citrate , 0 . 05% Tween 20 , pH 6 . 0 ) was performed by heating the slides to 95–100°C and then 10 min incubation in the microwave at the lowest power . The slides were cooled off to RT for 1 hr . The slides were washed in PBS and blocked with 3% BSA in PBS with 0 . 1% Triton X-100 for 30 min at RT . Primary antibody incubation was performed overnight at RT , washed with PBS , incubated with secondary antibody for 1 hr at RT , washed with PBS , stained with DAPI ( 1:10000 ) for 10 min at RT and washed three times with PBS . Slides were cover-slipped with Prolong Diamond mounting media . Images were acquired using a Zeiss LSM700 confocal microscope using the same acquisition settings for all samples . Further image analysis was done using FIJI . Most data analysis was done in the R/Bioconductor environment ( Huber et al . , 2015 ) in RStudio ( https://www . R-project . org/ , http://www . rstudio . com/ ) . For general processing , data exploration , and visualization , we used the tidyverse array of packages , in particular ggplot and dplyr ( Wickham et al . , 2019 ) . | At birth , the mammalian brain contains tens of billions of neurons . Although the number does not increase much as the animal grows , there are many dramatic changes to their size and structure . These changes allow the neurons to communicate with one another , develop into networks , and learn the tasks of the adult brain . One way that these changes occur is by the accumulation of chemical marks on each neuron’s DNA that help dictate which genes switch on , and which turn off . One of the most common ways that DNA can be marked is through the addition of a chemical group called a methyl group to one of the four DNA bases , cytosine . This process is called methylation . When methylation occurs , cytosine becomes 5-methylcytosine , or 5mC for short . In 2009 , researchers found another modification present in the DNA in the brain: 5-hydroxymethylcytosine , or 5hmC . This modification appears when a group of proteins called the Tet hydroxylases turn 5mC into 5hmC . Converting 5mC to 5hmC normally helps cells remove marks on their DNA before they divide and expand . This is important because the newly generated cells need to be able to accumulate their own methylation marks to perform their roles properly . However , neurons in the brain accumulate 5hmC after birth , when the cells are no longer dividing , indicating that 5hmC may be required for the neurons to mature . Stoyanova et al . set out to determine whether mouse neurons need 5hmC to get their adult characteristics by tracking the chemical changes that occur in DNA from birth to adulthood . Some of the mice they tested produced 5hmC normally , while others lacked the genes necessary to make the Tet proteins in a specific class of neurons , preventing them from converting 5mC to 5hmC as they differentiate . The results reveal that neurons do not mature properly if 5hmC is not produced continuously following the first week of life . This is because neurons need to have the right genes switched on and off to differentiate correctly , and this only happens when 5hmC accumulates in some genes , while 5hmC and 5mC are removed from others . The data highlight the role of the Tet proteins , which convert 5mC into 5hmC , in preparing the marks for removal and demonstrate that active removal of these marks is essential for neuronal differentiation . Given the role of 5hmC in the development of neurons , it is possible that problems in this system could contribute to brain disorders . Further studies aimed at understanding how cells control 5hmC levels could lead to new ways to improve brain health . Research has also shown that if dividing cells lose the ability to make 5hmC , they can become cancerous . Future work could explain more about how and why this happens . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"genetics",
"and",
"genomics"
] | 2021 | 5-Hydroxymethylcytosine-mediated active demethylation is required for mammalian neuronal differentiation and function |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.